Wavelet Analysis Based Study Of Detection Of Incipient Faults Of Asynchronous Motor

This dissertation focuses on the study of detection method of incipient faults ofasynchronous motor based on wavelet analysis.Asynchronous motor is an important rotating machine widely used in powersystem, and its running state has great effect on the safe operating of the powersystem. After some period, the incipient faults in motor will develop into suddenserious faults, and the present relay protection equipments can only cut off the powerof faulted motor in case of further damage. However, there is still a possibility forthe electric power production to be interrupted unexpectedly, and a possibility for thecostly asynchronous motor to be terribly damaged. Therefore, there is significantsocial and economic benefits to do the research on how to detect the incipient faultsin asynchronous motor as soon as possible.Unsteady signal is often included in the current signal of faulted synchronousmotor, such as the start-up period, the sudden change signal of when faults happen instator windings, and complicated vibration signals of faulted motor axis, etc. Thetraditional analysis method of Fourier transform is suited to process steady signal,but restricted for unsteady signal. Employing Fourier transform for analyzingunsteady signals, big error or even serious faulted diagnosis may occur. However,wavelet transform has excellent time-frequency localization ability, and very suitedto do the analysis of unsteady signals. Based on this thinking, wavelet transform isintroduced in this dissertation to detect the incipient faults of asynchronous motor.The research work mainly focuses on the wavelet analysis detection method theincipient faults of rotor, stator and axis, and as an actual fruit, a set of waveletintelligent diagnosis system of incipient faults of asynchronous motor has beendeveloped.On the basis of introducing the basic theory of wavelet transform, thecharacteristic of time-frequency window is deeply studied in this dissertation. Fromthe view of the time-frequency window, some common principles to be complied forchoosing a wavelet function are presented for actual application, also, theamplitude-frequency characteristic and phase-frequency characteristic in frequencydomain of common wavelet functions are studied. Deep study has been done on the speciality of the rotor fault characteristiccomponent in start-up period of asynchronous motors, the wavelet ridge detectionmethod of faulted rotor in start-up period is presented, so the accurate diagnosis ofsuch fault is realized. It is demonstrated by the actual analysis that the new detectionmethod has no special demand on the load quantity of the motor. Even for a motorwhich runs with null load, the faulted rotor with only one broken bar can be detectedaccurately. The reasons of high misdiagnosis rate of faulted rotor under steady running statehave been intensively studied, and the relevant improved methods have been putforward, moreover, the simultaneous monitoring of stator voltage signal is alsoincluded into the detection of faulted rotor under steady condition. On the foundationabove, a novel method based on frequency-distance comparing to detect rotor fault ispresented. It is the first time that wavelet analysis has been introduced into the onlinedetection area of faulted motor rotor under steady state, and the wavelet algorithm todetect the rotor fault characteristic is presented. As the result, the diagnosis accuracyhas been improved greatly. A novel wavelet analysis method on the basis of the symmetry measurement ofthree-phase current is presented to detect the incipient inter-turn shorts fault of theasynchronous motor. Since wavelet transform is an analysis method based onfrequency band extraction, it shows itself excellent robust against the frequencydisturbance of the power net. Wavelet transform modulus maximum method isdiscussed at large to detect the sudden change signal of the stator windings fault. Theanalysis shows that wavelet transform modulus maximum method can not only detectthe sudden change signal when a fault occurs in the stator windings, but can localizethe time point when the sudden change happens. An in-depth analysis is performed to study the frequency subband criss-crossphenomenon which comes about in signal decomposition with wavelet packet, andthe reason is found to be the continuous signal decomposition with w2 nand w2n+1...