This paper presents an evolutionary system identification strategy for mechatronics systems which include various nonlinearities. In the research, the saturation in power converters and the friction in mechanisms are considered as the nonlinear elements, some of which are generally difficult to detect and/or identify directly. The proposed scheme can evolutionally determine the structure of both linear and nonlinear elements in the system, where the observable input/output variables with time delay are used to express the linear components, and power series of the variables and nonlinear functions are used to express the nonlinear ones, respectively. In the evolutionary process, Genetic Algorithm (GA) is applied to optimize the combination of these variables and, as a result, the sub-optimal mathematical model for the whole system can be achieved. The effectiveness of the proposed scheme is verified by experiments using a 2-mass resonant vibration system.