Abstract: Aiming at the problem that the traditional decomposition algorithm is not effective in extracting the weak fault feature information of rotor rubbing against the background of strong noise, a fault diagnosis method based on Fourier decomposition algorithm (FDM) noise reduction and fast spectral kurtosis analysis is proposed. Firstly, the complete decomposition of the rotor rubbing fault vibration signal is realized by using the FDM algorithm, and the component signals that can retain the fault characteristic information of the original vibration signal to the greatest extent are selected according to the criterion of signal kurtosis value. At the same time, the candidate signal components are reconstructed to obtain the noise-reduced fault vibration signal. Then, the reconstructed fault vibration signal is analyzed by the fast spectral kurtosis algorithm. With the advantage of the fast spectral kurtosis algorithm in the extraction of useful information bandwidth channel information of the vibration signal, the effective extraction of the rotor rubbing fault feature information is realized. The experimental analysis results show that the fault diagnosis method can not only realize the efficient noise reduction of the vibration signal of the rotor rubbing fault test sample, but also has more advantages in the extraction of fault feature information.