Aiming at the characteristics of vibration signal of slewing bearings in wind turbines， such as low frequency, nonstationary, nonlinearity and weak, a new fault diagnosis method for the bearings is proposed． Based on probability density analysis of kurtosis is to segment signal for analysis, to count each section of the kurtosis index value, finally the characteristic of the vibration signal could be gotten indirectly by analyzing the change trend of these indexes. The experiments show that the method can accurately determine whether the wind turbine bearings is occured faulty, which is very effective in fault diagnosis.