Incipient Fault Detection And Health Condition Assessment Of Rolling Element Bearing

Rolling element bearing is one of the indispensable components in rotating machines.It is used to transmit power and motion to equipment at a low friction in rotary machines in practice.It is vulnerable to fault since the uneven distribution of circumferential loads,the repeated contact between the rolling element and the inner or outer race,and the atrocious working conditions,such as high speed,heavy load,or prolonged time.Accidental faults in operating bearings result in performance degradation or breakdown of the whole equipment,which will lead to enormous economic costs and casualties to hum an beings.For weakening the safety hazards of the equipment and making an optimal maintenance plan in time,the bearing is generally monitored by the vibration sensor installed in the housing.However,the fault feature is easily buried by the background noise in many cases and difficult to identify.Therefore,the detection and assessment of health condition stages are investigated in this dissertation based on the vibration signals of bearings.The main contents of the dissertation are listed as follows:(1)The new fault feature extraction methods are derived from the essence of the Empirical Mode Decomposition(EMD)and Multi-Resolution Singular Value Decomposition(MRSVD).Based on the cyclostationarity of the vibration signal of the bearing under the constant speed,two Intrinsic Mode Functions(IMFs)with the maximal kurtosis or correlation coefficient in EMD are selected and used to detect the fault of bearings;the impulse fault component of vibration signals can be enhanced by the weight operators based on the kurtosis and correlation coefficient of IMFs of EMD in Weight EMD(WEMD);the optimal decomposed levels of MRSVD can be determined adaptively by the kurtosis and correlation coefficient criterion.The MDDCsMRSVD is derived from the essence of the MRSVD based on the Maximum Difference of Detail Components(MDDCs)in two decomposed levels.The experiment results show that the fault feature information can be extracted by the four derived methods,respectively,and the extracted fault ability of MDDCs-MRSVD is best.(2)The change rule of bearing performance degradation in rate are revealed based on the change rate of envelope spectrum features of bearings.The real-time Mahalanobis Distance(MD)is proposed to reflect the relative change of data in time domain based on the principle of abnormal data detected by MD.Based on multiple envelope spectrum features of bearing,the Real-time MD with Cumulative Sum(RMD-CUMSUM)is obtained by the CUMSUM(Cumulative Sum)for the tendency of the performance degradation of bearings;the Rate of Real-time MD with CUMSUM(RRMD-CUMSUM)is constructed to reflect the change rule of performance degradation in rate.It is a result that the changing trends are different under different health condition stages,and the changing trend is similar under the same health condition stages in RRMDCUMSUM,and the change rule of bearing performance degradation in rate can be used to assess the health condition stages of bearing.(3)The real-time monitoring indicators of health condition stages of bearings are constructed based on the change rule of performance degradation rate and the diversity of the performance degradation features of vibration signals in bearings,respectively.A quickly identified indicator of the initial fault time and a real-time assessed indicator of health condition stages are proposed based on the change rule of performance degradation rate of bearing,respectively.The diversity of features is measured by the gradient in the amplitude direction of the features matrix of the vibration signal.Moreover,another monitoring indicator constructed by their standard deviation of the gradients is utilized to assess the performance degradation stages of bearings based on the Pauta criterion in real time.Meanwhile,considering that the bearings fault is generated gradually by the fatigue,the accurate initial fault time will be detected by the envelope analysis and backtracking strategy with the interval-halving principle.It is a result that the proposed methods can identify accurately the initial fault time and give a warning,recognize the health condition stages of bearings in real time,and the monitoring indicator constructed by the diversity of features is better than the monitoring indicator constructed by the change rule of performance degradation rate of the bearing with the compound fault in case.(4)The double impulse response of defect in bearings is extracted based on the proposed WEMD and Fast Independent Component Analysis(ICA).The low-frequency response is produced when a rolling element rolling into the defect of bearings,and the high-frequency response is generated when a rolling element rolling out the defect of bearings.The low-frequency component of vibration signal can be reduced by the WEMD.The high-frequency component and low-frequency component in double impulse phenomenon of defect fault of bearings can be extracted separately by Fast ICA based on the raw signal and the signal processed by the WEMD.Then,the defect size can be estimated by the relationship between the double impulse and the defect size.