Abstract: As one of the most important features of the rolling element bearings with a localized defect/fault, the impulse characteristic is behaved as a series of side slopes, with the main impulse energy, located at two sides of the nature frequency in the spectrum, which is not convenient for the diagnosis. While the impulse series shared the similar character in the time and frequency domain, the auto-correlation method (SAC) is introduced to extract impulse character from the frequency domain and then used to analysis the frequency response of the vibration signals caused by a defect on the outer or inner race surface. Transferred from the sides of the nature frequency, the side slopes are present as harmonics of the characteristic frequency. When the defect located on the inner race, side slopes caused by the load modulation also remains around the harmonics. This method is also applied on the 6220 type rolling element bearing fault diagnosis. Leaving out of account of designing the nature band-pass filter, the extracting procedure is more convenient and simpler. It’s shown that fault character extracted by the proposed method is more clearly and believable than methods, such as：envelope analysis, cepstrum analysis, and the auto-correlation method in the time domain, etc, when applied on the inner race fault analysis with its iterative property. SAC is proved to be valid and effective which will be valuable for the engineering application.

Key words: Envelope analysis, Feature extracting, Localized defect/fault, Rolling element bearing, Spectrum auto-correlation