Composite Friction Torque Compensation Strategy for PMSM Based on Piecewise Modeling

The performance degradation of Permanent Magnet Synchronous Motors (PMSMs) is caused by nonlinear friction torque during low-speed operation, a phenomenon known as low-speed crawling. To address this issue, a composite compensation strategy based on a piecewise modeling approach is proposed in this paper. To accurately characterize the friction behavior, a dynamic learning rate neural network (DLRNN) is utilized for the ultra-low-speed region, while an optimal 5th-order polynomial model is established for the low-speed region. Building on this piecewise modeling, a composite control strategy with friction compensation is studied. This strategy employs a friction compensator for the feedforward compensation of the deterministic friction torque components. Furthermore, a reduced-order extended state observer (RESO) is designed to estimate and compensate for residual friction errors and random disturbance torques, thereby achieving high-precision motor control. Finally, experimental results on a surface-mounted PMSM validate that the proposed strategy effectively improves tracking accuracy and suppresses torque ripple near zero-speed, all while remaining within the computational constraints of standard industrial control hardware.