Fault Diagnosis Research In Variable-frequency Motor Based On The LMD And The CZT Method

With the rapid development of power electronic technology, rotating electricalmachine is widespread used in the development of national economy. It is of great essenceto diagnose the fault of the rotor, the roller bearing and the gear case which are significantcomponents of variable-frequency motor. The fault signals of motor rotary mechanism aremostly in the forms of multi-component amplitude-modulation and frequency-modulation.Proper methods are expectably adopted to extract these fault features due to the nonlinearand non-stationary characteristics. Time-frequency analysis has become thebroad-accepted technique for it can provide frequency and time domain of the signalsimultaneously. Based on the research of local mean decomposition (LMD), a novelself-adaptive method combined with Chirp Z Transform (CZT) and LMDmarginal spectrum is proposed and applied to the fault diagnosis of motor rotor, rollerbearing and gear box.Initially the fault diagnosis research background of variable-frequency motor isintroduced and the failure mechanism of broken motor bars, rotor eccentricity, rollerbearing point corrosion and gear crack is analyzed in detail respectively. In addition, thedevelopment process of time-frequency analysis method is described in order to raise thelocal mean decomposition. Subsequently the cardinal principle and the algorithm designof the local mean decomposition is introduced at length. An amplitude-modulated andfrequency-modulated simulated signal is designed for the purpose of examining thefeasibility of the approach. On account of the low resolution problem existed in currentresearch, the CZT is proposed to correct and revise the instantaneous frequency toimprove partial resolution accuracy. Afterwards the concept of LMD spectrum and LMDmarginal spectrum is tentatively defined. The physical meaning of the LMD marginalspectrum is clear compared with the Fourier amplitude spectrum, which reflects theoverall energy distribution characteristics of the original signal.Relying on the rotating machinery vibration fault experimental platform namedQPZZ-II, the improved local mean decomposition method is applied to the simulatedsignals of broken motor bars, rotor eccentricity and actual signals of roller bearing pointcorrosion and gear crack. The experimental results indicate that improved novel methodproposed in the paper can successfully extract and demodulate the fault signal and the conjoint analysis of partial and integral characteristic information is realized. Inconclusion, the novel method can effectively applied in the fault diagnosis research ofvariable-frequency motor.