An ensemble empirical mode decomposition (EEMD) fault feature identification method based on wavelet threshold denoising has been proposed aiming at the problems of vibration impact and noise when the rotor is running, which is easy to mask the fault features in the vibration signal. The wavelet threshold denoising method was used to preprocess the vibration signal, and then the intrinsic modal function (IMF) obtained after the set of EEMD was selected according to the Kurtosis principle.The fault characteristics of the timefrequency analysis of the obtained results were analyzed. The results show that this method can effectively suppress the noise and extract the characteristics of rotor fault.