Research On Motor Bearing Fault Diagnosis

With the constant improvement of technology, mechanical equipment faultdiagnosis draws more and more attention, while bearing fault diagnosis becomes moresignificant because it is the core device of mechanical equipment. So how to efficientlyuse the motor resources has became an important issue.This thesis analyses the bearing common fault mechanism at home and abroad,then we are clear that inner ring and outer ring fault is the major cause of bearing failure.According to this we design a inner ring and outer ring fault diagnosis platform, usingacceleration sensors for data acquisition and the technical scheme for bear faultdiagnosis technology research. Firstly, this paper studied the method of bearing faultdiagnosis. At the basis of wavelet denoising, empirical mode decomposition andclustering algorithm we proposed a combinative method which based on adaptivemorphological filtering, empirical mode decomposition energy entropy and clusteringalgorithm. This method firstly filter the noise based on kurtosis value, and then carry onthe empirical mode decomposition to the signal, calculate the energy entropy of eachfrequency band as the extracted feature vector. Finally using clustering algorithm toidentify the specific fault. The wavelet denoising technology need threshold as priorknowledge. But in practice different signals may need different threshold, it may causeexcessive denoising or lack denoising because of improper handling. This thesis adoptsthe adaptive morphological filtering technology and does not require priori knowledge.This make up the defects of the wavelet denoising. This article also uses theacceleration sensor, signal conditioning circuit, PCI data acquisition card and computerto acquisition vibration signals, and designed a PC software for bearing fault diagnosisbased on LabVIEW.Results show that the data acquisition system has a good data acquisitionprecision.and the combine method based on adaptive morphological filtering, empiricalmode decomposition energy entropy and clustering algorithm can overcome the defectsof wavelet denoising. So this method is an effective method for motor bearing faultdiagnosis.