Research On Fault Diagnosis Of Rotating Machinery Method Based On EEMD Algorithm

Rotating machinery is the key equipment in industrial production. And therunning state of rotating machinery concerns enterprises’ safety production andeconomic benefits. Therefore, it has an important realistic significance to monitor itscondition and diagnose its faults. When rotating machine equipments have faults, itwill appear abnormal vibration. Because vibration signal has abundant informationwhich is related to rotating machinery, so it is an effective method to judge faultsaccording to the analysis result of vibration signal. Most of the rotating machine faultvibration signals have nonlinear and non-stationary characteristics. The traditionalsignal analysis methods could not be satisfy with the demand of the vibration signal,so it is necessary to choose an appropriate signal processing method to analyzenonlinear and non-stationary signal.Because the time-frequency analysis method can provide both the time andfrequency information, so it is widely used in rotating machine fault diagnosis field.This dissertation introduces frequently used time-frequency analysis methods, forexample，STFT、Wigner-Ville distribution and Wavelet Transformation etc. Thosemethods all have shortcomings and lacks in processing nonlinear and non-stationarysignals. Hilbert-Huang transformation is a self adaptive time-frequency analysismethod which proposed by N. E. Huang. It is more prominent to show the signal partcharacteristics compared with the traditional time-frequency analysis methods forprocessing nonlinear and non-stationary signals. But the empirical mode decompo-sition has the mode mixing problem, so Huang proposed a noise-assisted dataanalysis method--ensemble empirical mode decomposition, EEMD for short.This dissertation mainly researches the fault diagnosis of rotating machinerymethod based on EEMD algorithm. First of all, we should study the theory of theEEMD algorithm and analyze mode mixing problem. Aiming at improving parameterSettings in the EEMD algorithm, the energy standard deviation method added white noise is proposed. At the same time, the effective IMFs are extracted by using thecorrelation coefficient method. The method could effectively eliminate the falsemode function of the EEMD decomposition results.It introduces fault vibration theory and fault signal characteristics of the rollingbearing and gearing in the rotating machine in this dissertation. According to rollingbearing and gearing fault model, the effective of the algorithm is simulated andverified. The improved EEMD algorithm is applied to process the actual rollingbearing and gearing vibration signals. Besides that the Hilbert envelope spectrummethod based on the improved EEMD algorithm and the singular value entropy ofthe IMF are also applied to the fault diagnosis. Both the time-frequency spectrumanalysis interface and the improved EEMD algorithm diagnosis system interface aredesigned based on MATLAB. The results appeared in the interface could be moreintuitive to observe and analyze.