A fault feature classification method for Rotor startup process based on FRFT-PCA

The amplitude and phase obtained by traditional non-stationary signal analysis can only be used to qualitative evaluate the running state of the rotor, consequently, it is difficult to make pattern classification and quantitative classification. Aiming at solving this problem, the paper proposed a method named FRFT-PCA with holographic parameters for the fault feature classification of a startup rotor. In this paper Fractional Fourier Transform (FRFT) is used to extract the primary frequency component related to the rotating speed; the quantitative indicators of rotor vibration status are obtained by building a characteristic dataset with holographic parameters, the Principal Component Analysis (PCA) is used to find out the most optimal essential feature or the fusion feature which shows the pattern of the rotor fault status; finally the quantitative classification of the rotor fault can be obtained. And furthermore, simulation experiments of several different typical faults and even same rotor fault at different degrees are performed, the result shows that this method works well for the fault feature quantitative classification of a startup rotor.