عنوان مقاله [English]
A servomechanism with undesirable nonlinearities, such as friction and backlash, is difficult to identify. These nonlinearities are the most significant drawbacks in high precision servo systems. Common identification
methods based on sinusoidal or binary excitation signals may be ineffective in the presence of backlash to determine the nonlinear friction.
The focus of the current study is to model the behavior of friction and backlash and to identify the exact kinetic coefficient of friction as well as backlash. To do this, system identification in the frequency domain is
proposed, utilizing frequency sweep as an excitation signal, to capture and construct the servo-system dynamics. As shown, the frequency sweep is better than the pseudo binary excitation signal, with respect to the power spectral
density. Moreover, it has no problem in the identification of nonlinear plants. The transfer function of the plant is identified based on the already proposed technique that uses the differences of inputs and outputs. To remove the effect of nonlinearity from the linear part, the plant is excited with two inputs. Then, the differences of these inputs and the differences of the corresponding outputs are used to identify the model. It can be shown that if the mplitudes of the inputs are increased, the effect of the nonlinear part is further reduced. This technique was already proposed for identification of the plant in the presence of friction, and is used in this paper for identification of a plant in the simultaneous presence of friction and backlash. The width of backlash is determined, based on the phase delay of the identified model, and the equivalent friction coefficient is determined based on the optimization.
For the first time, the CIFER user interface is employed for comprehensive analysis and transfer function modeling, regarding its unique capability of windowing. The results show that the current technique is
remarkably superior and could easily identify the system parameters. In addition, the robustness of the proposed method, with respect to the noisy output signals, is proven and it can outperform existing identification