Vibration signal's time-frequency analysis and comparison for a rotating machinery

It is important to extract fault features in fault diagnosis of a machinery.The traditional signal analysis can not be satisfactory for a non-stationary vibration signal whose statistical properties are time-variant.To deal with a non-stationary signal,time-frequency analysis techniques are widely used.Here,experiment data of a typical vibration fault signal were analyzed with different methods,such as,short time Fourier transformation(STFT),Wigner-Ville distribution(WVD),Wavelet transformation(WT) and Hilbert-Huang Transformation(HHT).Comparing between these methods,it was demonstrated that the time-frequency resolutions of STFT and WVD were inconsistent,they were easy to work out cross terms or make the signal weaker;WT had a distinct time-frequency distribution,but it brought about redundant components;HHT time-frequency analysis could detect components with lower energy,and give a true and distinct time-frequency distribution.Therefore,HHT is a very effective tool to diagnose early faults of a rotating machinery,and it provides new means for state detection and fault diagnosis.