| Rolling bearing is one of the most basic machinery units and it is widely used inmechanical equipment. It is also one of components that easy to loss and a part ofthe mechanical failure is caused by rolling bearing according to statistics. Thereforethe research of fault diagnosis for rolling bearing is significant.Firstly, wavelet theory is introduced in feature extraction of rolling bearing inthe thesis. Wavelet packet decomposition (WPD) is used in signal analysis of rollingbearing and the energy values of frequency bands are extracted as feature values tocompose the feature vector. A data processing method named random mean method(RMM) is proposed to optimize the feature extract process. The feature vectorsbefore and after optimized are compared and the result shows the RMM is benefitfor feature extraction of rolling bearing.Secondly, the feature recognition method is studied and support vector machine(SVM) is introduced in the fault diagnosis of rolling bearing. Optimal modelparameters of the classifier are searched through cross-validation method. The rawfeature vectors and optimal feature vectors are used in classifier building respectively.Test results of the two classifiers are compared, and the comparation shows theclassifier constructed by optimal feature vectors has higher classification accuracy.Thirdly, the life prediction method of rolling bearing is studied. Amplitude ofsensitive frequency spectrum is taken as feature value, and the feature values duringthe full life cycle of rolling bearing are calculated based on RMM, then the changingtrend of feature values is got. Campared to the prediction method by traditional timedomain parameters, the method proposed in the thesis is more accurarely and timely.Finally, the online monitoring and fault diagnosis system of rolling bearing isdesigned. Devices are chosed to build the hardware system according to operationalrequirements. Real-time vibration signal acquision, data storage, curve display,feature extraction, fault diagnosis and other functions are included in it. The faultdiagnosis function is designed based on the method proposed in the thesis. Faultsimulation experiment of rolling bearing is conducted and the system designed in thethesis is used in the experiment to verify its capability. The result shows the systemperformance well and it has high accuracy in the fault diagnosis of rolling bearingwhich is significant in guarantee the rolling bearing in mechanical equipmentoperating well. |
