The Research Of Fault Diagnosis Of Rolling Bearing Based On EMD

Rolling bearing is the common component in rotating machinery. Its running state can influence the performance of the whole machine directly. Injury fault on rolling bearing, such as spalling, pitting, defects, scratches etc., can quickly intensify and cause the inactivation of the rolling bearing and be the greatest harm. Therefore, the monitoring and diagnosis of injury fault are the main work of fault diagnosis of rolling bearings. Because of the low SNR (Signal-to-Noise Ratio) characteristics of fault vibration signal, the diagnosis of early-stage injury fault has been a great problem in this area.In this thesis, an analysis of the fault mechanism and vibration characteristics of roller bearing is made first. And a detailed discussion concerning about the characteristics and development of injury fault is also stated. Then, a newly emerging non-linear, non-stable signal decomposition method, "EMD" is introduced. And two key issues of the application of EMD: Intrinsic Mode Function Criterion and Pseudo-component Identification, is particularly analyzed. In order to extract the characteristics of early-stage injury fault, based on the modulation characteristics of fault vibration signal, the feature extraction method of energy operator demodulation is proposed. In this way, the end-point problem of Hilbert transform demodulation is effectively avoided. After the smoothing processing of energy operator demodulation, a satisfactory demodulation effect can be achieved. Now, the effectiveness of the method has been proved through the diagnosis examples. Finally, the signal de-noising method based on EMD is proposed, because the background noise of rolling bearing fault vibration signal is very serious. Through the correlation analysis of intrinsic mode function and original signal, pseudo-components are removed, and noise components are identified through analyzing self-correlation of intrinsic mode function, the signal is reconstructed for the purpose of de-noising. The fault diagnosis of rolling bearing based on EMD de-noising is done; the results show that, the minimal active component can be extracted through this method. The diagnosis of multiple faults and fault location can be realized through the combination of EMD de-noising and demodulation method.