Research On Vibration Signals For Rotary Machinery Condition Monitoring And Fault Diagnosis

Condition monitoring and fault diagnosis of rotary machinery equipment to ensure the safeoperation of the equipment has important practical significance and economic meaningful. Rotarymachinery vibration signal contains abundant information can represent the running status of theequipment. Application of theories with signal processing technology to extract informations fromvibration signals is most crucial problem for the reliability and accuracy in the mechanicalconditionmonitoring andfaultdiagnosis. Themainworkis as follows:(1) The basic theory of support vector machine (SVM) and wavelet packet decompositionmethod for rotary machinery fault diagnosis are studied in chapter2. Considering it is difficult toextract effective fault features from fault signal. Proposed method based on SVM and waveletpacket decomposition for fault diagnosis. With different bearings fault status as experimentaobjects, each state of bearing vibration signal decomposition by wavelet packet, signal energy invarious frequency bands and decomposed signal ratio of the total energy as a feature vector torepresent the equipment running status. As the input of the SVM fault classifier and fault detectionand diagnosis comes ture sueeessuflly. Finally, all kinds of factors which influence classifierpropertyoffaultclassifier areanalyzed deeply.(2) Research on the method based on neighborhood rough set attribute importance for faultfeature selection. In the process of mechanical fault diagnosis, Fault classifier efficiently if theinput of information with high quality. Combined with gear cracks experiment, fault featureextraction based on time-domain frequency-domain and hilbert transform, Select the optimalfeaturesubsetas inputoftheSVMfaultclassifierto recognitioncracks andcondition monitoring.(3) Import principal component analysis (PCA) of information fusion to fault diagnosis ofrotary machinery. Fusion the fault features of bearings and gearbox based on PCA, featuresdimensionality reduction as feature fusion, make the feature subset efficiently. The mothed offeaturefusionnotonly improvediagnosticaccuracybutalso shorten theoperationtime.