Abstract: It’s known that auto-correlation technique is effective in extracting periodical signals from random noises. In the case of fault monitoring of rolling element bearing, we can’t acquire the fault information directly from the original signal because of the difference of signal phases. And the signal is shown as the wide band random signal in auto-correlation function. In this paper, the signal is pre-processed and the results are proved effective. Moreover, by taking the auto-correlation function we can obtain the determined and comparable samples. This is very important for establishing the data base of running condition and for detecting the faults.

Key words: auto-correlation function, wave shape factor, crest factor, impulse factor, kurtosis factor