| As power equipment, motors have wide application in every walk of life, it is of greatpractical significance to diagnose and rule out fault in time and accurately. In this paper, themotorfaultisdiagnosed by thefollowingaspects.(1) This paper describes in detail the common motor failure mechanisms, and on thisbasis, collect the motor vibration data, which include stator winding fault, anchor bolts loose,normal, rotor misalignment, rotor imbalance, broken rotor bars and rotor warpage,analyzeand preprocessthesedata.(2) Selecting crest factor, kurtosis, pulse factor, margin factor and form factor as thefeature quantity, Simple judgment of the unknown motor fault is carried out by judging thespace moment between the values of the parameters of the diagnostic sample to the standardvalue.(3) On the basis of simple diagnosis, using wavelet packet technique decompose thattime-domain data previously to obtain characteristic bands energy value corresponding.Usecharacteristic bands energy value as training and testing samples for the " one to one"support vector machines,"one to many "support vector machines. Among these ways,"oneto one" support vector machines exist defects in use, that the results are not unique, so usingdirected acyclic graph method to improve it. In order to compare the diagnostic effect ofSVM,thetrainingsamples andthetest samplearetrainedandtested in themodel ofPNN.(4) Using the above four classification model to diagnose the fault diagnosis samplesand compared the performance parameters of four kinds of classification models, to obtainthe optimal classification model for the directed acyclic graph improved "one on one"supportvectormachinemodel.In summary, the time domain parameters diagnosis method is suitable for simplediagnosis of motor, and its calculation is small. In the practical application can determine thegeneral situation of the motor by the method, then accurately diagnosis by using model of the directed acyclic graph improved "one on one" support vector machine. |
PDF Full Text Request