Online Identification of Servo System Mechanical Model Using Dynamic Weighted Recursive Instrumental Variable Least Squares

During the prolonged operation of servo systems, various factors such as equipment aging and variations in working conditions, inevitably cause the model parameters to deviate, leading to a degradation in control system performance. To ensure that the servo control system maintains superior dynamic performance under uncertain conditions, it must possess the ability to monitor the system model in real-time and adaptively adjust the controller parameters. For this purpose, accurate identification of the key parameters in the system model serves as the foundation for optimizing the performance of the speed or position loop. To address this challenge, the present study employs a discretized mechanical motion equation of the servo system based on zero-order hold, while fully considering the impact of viscous friction coefficients and load torques on the identification results. By applying the Dynamic Weighted Recursive Instrumental Variable Least Squares method to the discrete-time model, this study successfully achieves the concurrent identification of dominant mechanical model parameters. To validate the effectiveness and practicability of the proposed identification method, detailed simulation was conducted in the MATLAB environment, and then experimental verification was performed on an STM32-based servomotor test-bench. Both simulation and experimental results demonstrate that the identification method presented in this study possesses high accuracy and good real-time performance, capable of reflecting changes in system parameters in real-time, providing strong technical support for parameter self-tuning in servo control systems.