Abstract: In order to solve the problems that the early fault vibration signals of rolling bearing
affected by the noise, the parameters of ensemble empirical mode decomposition (EEMD) are not
easy to be obtained. A method for fault bearings of rolling bearings based on improved EEMD and
spectrum kurtosis (SK) was presented. Firstly, obtaining the amplitude coefficient of added noise by
extracting the high frequency information, the number of ensemble members is obtained by
calculating the expectation error, the fault signal was decomposed into several intrinsic mode function
(IMF) by improved EEMD, the IMFs were reconstructed based on kurtosis criterion. Then, the central
frequency and bandwidth of a band-pass filter were determined with spectral kurtosis. Last, the
filtered signal was analyzed by using energy operator demodulation spectrum. The results
demonstrate that the proposed method can not only solve the problems such as losing fault
information and leaving noise due to the mode mixing in the process of denoising, but also extract the
fault frequency accurately.

Key words: rolling bearing, ensemble empirical mode decomposition, spectrum kurtosis, fault diagnosis