Abstract: Since the speed characteristic of linear ultrasonic motor (LUSM) has strongly nonlinear property, it is difficult to derive an accurately mathematic model in theory. This problem is tried to be solved and a modeling method by identification is proposed in this paper. Firstly, the step responses of LUSM in different driving conditions were analyzed. It is proposed here that the model of LUSM can be simplified as one-order inertial pure lagging systems. The idea is to transform the effect of the driving frequency and phase difference of two-phase driving voltages into the changes of the parameters in the proposed model, such as time constants, gains and delays. Secondly, these parameters were obtained by least-squares regression, and dynamic nonlinear models of LUSM for frequency-speed control and phase-difference-speed control were proposed, which provide model basis for high accuracy control of LUSM. Finally, the validity of the proposed method is confirmed by experimental results.