Servomechanisms are always assumed and identified as linear systems. Actually, friction is a significant nonlinear factor in such systems. In this paper, friction compensation technique which is usually used as a part of control scheme is introduced in system identification process and it saves the nonlinear control effort in control scheme. Moreover, 3D friction model including position information is set up and the Kalman filter based radial basis function (RBF) network is designed to fit and compensate the nonlinear friction in servomechanisms. A motor drive servo system is set up as a test plant and is identified to show that the proposed method is simple and practical. Compared with the math model based friction fitting method, the proposed one realizes a much better fitting error result. Correspondingly, the identified linear system model is very close to the measured frequency response data as well.