Abstract: In view of the harsh working environment of large-scale wind turbines,the fault feature information of their generator bearings is difficult to detect due to strong background noise and other interference.In this paper,a new enhanced combination gradient morphological filter（ECGMF） is proposed to detect the fault feature information of rolling bearings.Combined with the characteristics of improved basic morphological operators,the proposed method can not only preserve the fault characteristics of the signal,but also suppress the noise interference.To overcome the scale instability in selecting structure elements based on kurtosis criteria and signal-tonoise ratio,the characteristic frequency intensity coefficient(C_CFI) is adopted in this paper to select the optimal structure element.Simulation and experimental results show that the proposed method can effectively extract the fault feature information of wind turbine bearing.Compared with other morphological filters,the results show that the proposed method is effective and advantageous.