Fault Diagnosis And Performance Degradation Assessmant Of Rolling Bearing

Rolling bearings are widely used in large rotating equipment such as generators,turbines,gearboxes and so on.They are considered as the most critical machine elements of almost all rotating machineries.However,rolling bearings are one of the most easily damaged parts in mechanical equipment because the harsh working environments,such as vibration,shocks,variable load operation and high temperature.Therefore,fault diagnosis and performance evaluation of rolling bearings is very important in promising the reliable operation of machinery.This paper mainly studies the fault diagnosis and performance degradation of bearing.Some effective characteristic index are explored to identify the type of bearing fault and damage degree.The specific work is as follows:(1)In order to solve the difficulty in identifying different faults and damage of rolling bearings,this paper proposes the Ensemble empirical mode decomposition(EEMD)and the fast approximate entropy(FAE)method to extract the bearing fault feature.The FCM clustering method is applied to establish the bearing condition model which can verify the validity of fast approximate entropy index in representing different fault states of bearings.(2)The traditional time-domain features in characterizing bearing performance degradation has shortcoming.This paper reasonably selects the effective performance index.EEMD Fast Approximate Entropy Method are used to further analyse the vibration data from run-to-failure test of rolling bearing.And RMS is used to replace the weak FAE component which is obtained by EEMD,and combine the RMS of original signals with other FAE to obtain high dimension degradation indicators.(3)In order to solve the "dimension disaster" problem in high-dimensional performance indicators,the nonlinear feature dimension reduction algorithm--locally linear embedding(LLE)is utilized to extract the significant feature of indicators and obtain the bearing degradation evaluation index,which can diagnose the performance degradation of the bearing in early stage and effectively reflect the degradation stages.Compared with the LE algorithm,LLE has a better effect in extracting the non-linear features and has a good Robustness.