Research In Locomotive Bearing Fault Diagnosis Method Based On Signal Modal Decomposition

With the increasing speed of the locomotive,the demand on the safety and reliability of the locomotive is higer and higher.As the main part of locomotive running part,the health state of the bearing affects the safety of the locomotive seriously.Therefore,it is of great significance to carry out fault diagnosis for locomotive bearings.In order to monitor the condition of locomotive bearing effectively and fault diagnosis for it,this paper studies the feature extraction method based on signal modal decomposition,the main research contents are as follows:The empirical mode decomposition(EMD)and the ensemble empirical mode decomposition(EEMD)are studied firstly,their characteristics analyzed through numerical simulation.The signal is decomposed into a series of simple components by EMD,but there are modal aliasing phenomenon in their decomposed components.After adding white noise to original signal,the signal is decomposed according to EMD,the components got by EEMD is the average of multiple decomposition results.The EEMD overcomes the modal aliasing problems of the EMD,but the white noise makes the energy of signal before and after decomposition not conserved,the method cannot achieve signal reconstruction perfectly.In order to select the useful components that represent the characteristic information of the original signal,the method of selecting the main components with the correlation coefficient as the index is studied.Firstly,by calculating the correlation coefficient between each component and the original signal,the main components with high signal correlation are selected,and the interference of useless components is eliminated.On the basis of the above,a method of kurtosis spectrum analysis based on Frequency Slice Wavelet Transform(FSWT)is studied.Firstly,the FSWT transform is used to extract the useful components,the kurtosis spectrum is obtained based on the results of time-frequency transform,and the characteristic frequency band is determined according to the spectral peak value of kurtosis.To solve the problems such as EEMD cannot be completely reconstructed,combined with the signal characteristic extraction method of kurtosis spectrum and FSWT,the complete ensemble empirical mode decomposition method is analyzed.First,the different modal components completely separated by complete ensemble empirical mode decomposition,signal feature can be extracted by the method of kurtosis spectrum and FSWT.The fault characteristic frequency is extracted from the signal in strong noise background,the locomotive bearing fault is identified.The residual noise and spurious components in the component obtained by the complete empirical mode decomposition method.In order to solve this problem,an improved algorithm is studied,the difference of the current residual and local mean as modal components,using modal components of white noise instead of the original white noise,this method avoids the decomposition results of the residual noise and false component,get closer to the real component of the original signal.Through the analysis of locomotive bearing vibration signal,the improved method can improve the effectiveness of fault feature extraction.