An Identification Method Based on Model Merging for Servo System

The identification of electromechanical plant is an important part of the control performance tuning in the servo system. In engineering, it is common to build the rigid model of the electromechanical plant, and the high order dynamics such as resonance and lag are identified by closed-loop frequency sweep experiment. In the rigid model identification experiment, it is not possible to identify the high order dynamics of the electromechanical plant, and in the closed-loop frequency sweep experiment, due to the PI controller, the identified model in low frequency is inaccurate. Therefore, to optimize the two problems, this paper proposes an identification method based on model merging. The rigid model identification experiment is first carried out to identify the rigid model. And in the follow-up closed-loop frequency sweep experiment, the rigid model is used to weight the response data to identify the high order dynamic model. With the known rigid model and high order dynamic model, the electromechanical plant can be obtained by merging both of them. Simulation and experiment show that the proposed method can identify the electromechanical plant well in the full frequency bands.