A study of bispectral analysis in gearbox condition monitoring is presented. The theory of bispectrum and quadratic phase coupling (QPC) is first introduced, and then equations for computing bispectrum slices are obtained. To meet the needs of online monitoring, a simplified method of computing bispectrum diagonal slice is adopted. Industrial gearbox vibration signals measured from normal and tooth cracked conditions are analyzed using the above method. Experiments results indicate that bispectrum can effectively suppress the additive Gaussian noise and chracterize the QPC phenomenon. It is also shown that the 1-D bispectrum diagonal slice can capture the non-Gaussian and nonlinear feature of gearbox vibration when crack occurred, hence, this method can be employed to gearbox real time monitoring and early diagnosis. A study of bispectral analysis in gearbox condition monitoring is presented. The theory of bispectrum and quadratic phase coupling (QPC) is first introduced, and then equations for computing bispectrum slices are obtained. To meet the needs of online monitoring, a simplified method of computing Experiments results indicate that that bispectrum can effectively suppress the additive Gaussian noise and chracterize the QPC phenomenon. It is also shown in the 1- D bispectrum diagonal slice can capture the non-Gaussian and nonlinear feature of gearbox vibration when crack occurred, hence, this method can be employed to gearbox real time monitoring and early diagnosis.