Design of dynamic periodic adaptive learning controller for long-term cogging effect compensation

Cogging effect is a serious disadvantage of the permanent magnetic synchronous motor (PMSM), and cogging force is a position-dependent periodic disturbance. In our previous work [1], a dual high-order periodic adaptive learning compensation (DHO-PALC) method for state-dependent periodic disturbance was presented, where the long term stability issue was not addressed. Due to the impact of the high frequency components, when applying the DHO-PALC for a long time, the tracking errors may grow, and the system may become even unstable eventually. This phenomenon also appears in many other practical motion control systems using ILC or RC strategies. In this paper, in order to achieve long term stability, we propose a dual high-order dynamic adaptive learning compensation (DHO-D-PALC) method for cogging effect. In this method, stored information of more than one previous periods are included for both the composite tracking error and the estimate of cogging force. Particularly, since we use a dynamic learning control law to update the current estimate of cogging, the long term stability can be guaranteed. Extensive simulation results are included to demonstrate, 1) high-order in composite tracking error offers faster convergence, 2) high-order in cogging estimate better accommodates the case of varying reference signal, 3) dual high-order scheme has the potential of much better performance over the conventional first-order scheme, and 4) the introduction of dynamic learning updating scheme helps achieving the long term stability of the adaptive learning controller.