Abstract: Some time-frequency distributions belonging to the Cohen class, the spectrogram, smoothed pseudo Wigner-Ville distribution, Choi-Williams distribution and cone kernel distribution, are introduced and their performance are discussed. From a theoretical point of view, the interference cross－terms attenuation and time-frequency concentration are discussed. Meanwhile some corresponding performance also is researched. Aimming at gearbox condition monitoring and fault diagnosis, these distributions have been applied on some simulations and signals measured on a two—stage reduction gearbox. The results have shown that the Choi-Williams distribution provides a high sensitivity to various of short duration in non-stationary signal, which is available to extract the time-varying information. The distribution may provide a powerful tool for the detection of gear damage, especially for the early damage detection.

Key words: Fault diagnosis, Signal processing, Time-frequency distribution