A method for time-frequency feature extraction from vibration signal based on Hilbert-Huang Transform

Based on the Hilbert-Huang Transform (HHT), a method for time-frequency feature extraction from vibration signals was introduced into fault diagnosis of rotors. Firstly, the Empirical Mode Decomposition (EMD) was implemented on vibration signals measured by sensors. As a result, a set of components with different time scales, i.e. Intrinsic Mode Function (IMF), was extracted. Then, the Hilbert Transformation (HT) was applied to every IMF. Finally, time-frequency spectrum of vibration observation was constructed by all the transformations, from which nonlinear and nonstationary tendency embedded into vibration data was clearly indicated. Experiment on a rotor with two faults, i.e. unbalance and loose foundation showed that the proposed HHT based feature extractor can effectively analyze and represent nonlinear and nonstationary features excited by different faults, which lays an important foundation for fault diagnosis of rotors.