Abstract: On the basis of analyzing gearbox signals, one novel method of detecting early bearing faults based on cyclic-statistics is presented. Then one-dimensional and two-dimensional cyclostationary features of faults are put forward and they can be used to separate useful signals from noise and detect weak faults easily. In the end the proposed method is used to detect early bearing faults in one helicopter gearbox and the results demonstrate that it is superior to power spectrum density (PSD) and demodulation analysis and can be used to extract early fault features.

Key words: helicopter gearbox, fault detection, cyclic-statistics, cyclic frequency, spectrum correlation density function(SCDF)