To solve the problem is aimed of monitoring large mechanical bearing early friction faults by Acoustic Emission (AE). On the basis of the theory of bearing friction model, the form of friction function between rotor and bearing was analyzed when friction fault occured, the friction between rotor and bearing was calculated through AE signal mean square root, which used for friction fault monitoring of the plain bearing. In view of the characteristics of weak information of friction fault in plain bearing, the amplitude of AE signal was envoloped to enhance information of friction fulse. The calculating formula of bearing friction was simplified by introducing bellshaped pulse function to identify friction fault pulse. Finally through experiment, the process of oil in bearing cutted off, rotor acceleration and bearing increased the load was simulated. Compared with the traditional AE features, this feature of bearing friction is more sensitive to fault of bearing friction, and the method is more effective, which is more suitable for condition monitoring of friction fault of the plain bearing.