Abstract: The vibration feature of bearing with fault is analyzed, and on the condition of the mathematical model with impact signal of bearing, the signal is pointed out to be non-stationary and have the wide frequency band, so the fault feature can’t be gotten with the amplitude spectrum. Referencing the property of rotating movement of bearing, the cyclostationarity of vibration signal is deduced, and the signal is almost cyclostationary. The demodulation function o the second-order cyclic statistics is introduced, the method of using filtering and the second-order cyclic statistics to identify the fault feature is derived. The vibration of bearing with fault is analyzed and showed the method proposed in the paper is effective.

Key words: Cyclostationary, Fault diagnosis, Filtering, Rolling bearing, Second-order cyclic statistics