Abstract: To reduce life cycle cost and to improve flight safety,the detection and diagnosis of gearbox faults is of vital importance.Impulsive vibration signals in gearbox are often associated with impulse resulted from faults, and thus,impulsive signals can be used as the symptoms of faults.In order to characterize fault signals,many approaches were used to seek salient characteristic signals in time-frequency domain.Wigner distribution is a common time-frequency representation.However,the fault characteristic signals are always baffled by signals from other vibration sources and the inherent cross-terms in bilinear operation of the Wigner distribution.To minimize the effect of these disturbances,Wigner distribution was combined with empirical mode decomposition.Rather than calculating Wigner distribution directly,we present a method using empirical mode decomposition as a preprocessor.The vibration signal is decomposed into a sum of intrinsic mode functions.Only the Wigner distributions of the intrinsic mode functions which relate to the gear meshing vibration are calculated.This approach was applied to the diagnosis of a gearbox test data,the results show that effectiveness of diagnosing is significantly improved.