Abstract: To solve the problem that it is difficult to identify tooth crack in the high-speed stage of wind turbine transmission,this paper presents a method of tooth crack identification based on GBPNN. An analytical model of time-varying mesh stiffness considering the propagation path and direction of root cracks was constructed,and a multi-degree-of-freedom dynamic model of wind turbine highspeed stage were considered to establish. This paper analyzes the influence of root crack on dynamic characteristics,and the failure mode is identified by GBPNN. The results show that the gear fault vibration periodic shock signals will be transmitted along the transmission shaft,but the flexibility of transmission shaft will cause obvious attenuation of its amplitude;by using GBPNN and combining the peak,kurtosis,statistical matrix parameters and variance of the vibration acceleration at each shaft node,the gear root crack pattern can be effectively identified.