Abstract: A fault recognition method based on Hilbert-Huang Transform (HHT) and the fuzzy C-means clustering is proposed. This method takes advantages of HHT in processing nonlinear and non-stationary signals. The vibration acceleration signals collected from the axial piston pump housing are processed by HHT. First, the signals are treated by means of the Empirical Mode Decomposition (EMD) and the Ensemble Empirical Mode Decomposition (EEMD) respectively. Then, the Intrinsic Mode Function (IMF), which is most sensitive to faults, is chosen according to the short-term maximum entropy spectrum analysis, and the sensitive IMF is re-decomposed. Third, the fault characteristic vector is constructed based on the characteristic energy of the local marginal energy spectrum. Finally, the faults are recognized using the fuzzy C-means clustering algorithm. The results show that, compared with EMD, the interactive time based on EEMD is reduced greatly, and the fault recognition accuracy is substantially improved by the proposed recognition method.