An improved empirical mode decomposition（ EMD） method of fault diagnosis is presented

Which combined with the wavelet packet denoising

the correlation coefficient principle

the wavelet denoising with selecting different db vanishing moments and the EMD decomposition. The first

the fault signal is pretreated by wavelet packet denoising; then the fault signal is processed by the EMD decomposition to get a series of IMF

by calculating the correlation coefficient of each IMP with the original signal

the effective sets which need to repeat denoising are obtained. According to the mutation strength of the effective concentration IMF by selecting db wavelet with different vanishing moments

the noise reduction is repeated; Finally the signal is reconstructed and the power spectrum is generated. The experimental results show that this method can remove the noise which mixed in the fault signal

improve the signal to noise ratio

tell the faults of the gear and the bearing from the normal condition.