Motor State Prediction and Friction Compensation for Brushless DC Motor Drives Using Data-Driven Techniques

Abstract

Friction compensation has become crucial for robust, dependable, and accurate position and velocity control of motor drives. Large position inaccuracies and vibrations caused by non-characterized friction may be amplified by stick-slip motion and limit cycles. In order to find the governing equations of motor dynamics, which also describe friction, this research applies two data-driven methodologies. Specifically, data obtained from a Brushless DC (BLDC) motor is subjected to the Sparse Identification of Nonlinear Dynamics with control (SINDYc) technique and low-energy data extraction from time-delayed coordinates of motor velocity to determine the underlying dynamics. Next, the identified nonlinear model was compared to a linear model without friction and a nonlinear model that contained the LuGre friction model. The optimal friction parameters for the LuGre model were determined using a nonlinear grey box model estimation approach using the collected data. The three validation datasets taken from the BLDC motor were then used to validate the resultant innovative nonlinear motor model with friction characteristics. Over 90% accuracy in predicting the motor states in all input excitation signals under consideration was demonstrated by the innovative model. Additionally, when compared to a system that was identical but used the LuGre model, a model-based feedback friction compensation technique demonstrated a relative improvement in performance.