Fusion of Position Estimation Techniques for a Swing Servo by a Permanent-Magnet Synchronous Machine

This article presents a sensorless motor drive scheme by a fusion of the electrical model-based estimation and the mechanical model-based observer with respect to the position servo with a heavy load. The signal injection is applied to extract the position information from the magnetic saliency. However, in the low-speed and heavy-load scenarios, the position observability degenerates in the electrical model due to the core saturation. Nevertheless, the mechanical motion states, including the rotor position, can always be adequately extracted from the mechanical motion model. The extended Kalman filter (EKF) is adopted to fuse the information from the mechanical motion perception and the signal injection-based estimation so that the speed estimate quality gets significant improvement especially when the core saturation occurs. A disturbance-rejection mechanism is introduced to resist the parameter inaccuracy and inconstancy of the EKF model. The fused estimates are applied with a model predictive control in the speed loop to realize the servo control. The sensorless scheme is validated in a heavy-load pendulum servo bench, 200% of the machine rating torque, driven by a surface-mounted permanent-magnet synchronous machine.