Iterative learning control for smooth operation of permanent magnet synchronous motors

    Student thesis: Doctor of Philosophy (PhD) - CDU

    Abstract

    Permanent Magnet Synchronous Motors (PMSM) have many advantages over other types of motors and are used in many applications. However, undesirable torque ripples are associated with these motors, caused by design, manufacturing imperfections or measurement inaccuracies. These torque ripples occur as periodic functions of the rotor position. While there are several control schemes to minimise torque ripple, such as using observers or pre-compensation, Iterative Learning Control (ILC), an adaptive control method capable of reducing torque ripples that are periodic in nature, has not been extensively investigated to date and may be a suitable method to significantly reduce torque ripple for PMSMs.

    Various methods of ILC are described in literature, such as the Single Channel First Order ILC (SCFO-ILC), which uses information from the previous cycle within a certain frequency range for the iterative learning process; Multi-Channel ILC(MC-ILC) which uses multiple channels; Higher Order ILC (HO-ILC), which uses information from more than one previous cycle; and adaptive ILC in which the learning gains vary with the error. Most ILC schemes used in PMSM control are the Proportional type ILC (P-ILC) or its variations. Although other types of ILC schemes are available, the effectiveness of these schemes in minimising torque ripple for PMSM are not described in detail in literature.

    Simulations and experimental results of this thesis showed that the various ILC schemes were able to suppress the major torque ripple harmonics of PMSM. Of the four categories of ILC, P-type ILC with a forgetting factor has the widest learnable band while D-type ILC and MC-ILC has the narrowest. MC-ILC has the lowest Torque Ripple Factor (TRF), fastest convergence and is relatively robust to parameter variations. Together with HO-ILC and adaptive ILC, they have lower TRF and faster convergence compared to SCFO-ILC.

    Two new ILC schemes were proposed in this work: Multi-Channel Higher Order ILC and Multi-Channel Adaptive ILC. Compared to SCFO-ILC, Multi-Channel Higher Order ILC converges faster and has a lower TRF. However, it is not robust to parameter variations. Multi-Channel Adaptive ILC on the other hand is robust, has a low TRF and converges the fastest among other ILC schemes. It can therefore be concluded that the Multi-Channel Adaptive ILC, developed in this thesis, is a suitable ILC scheme for PMSMs to minimise torque ripple.


    Date of Award16 Mar 2017
    LanguageEnglish
    Awarding Institution
    SupervisorFriso De Boer (Supervisor)

    Cite this

    Iterative learning control for smooth operation of permanent magnet synchronous motors
    Yeo, K. (Author). 16 Mar 2017

    Student thesis: Doctor of Philosophy (PhD) - CDU