Robust topology optimization of interior permanent magnet synchronous motor for torque ripple reduction under current uncertainty

Abstract

This paper presents a robust topology optimization (RTO) method for an interior permanent magnet synchronous motor (IPMSM) under current uncertainty. Torque ripple issues in IPMSMs are typically addressed using topology optimization (TO), but real driving conditions contain uncertain harmonic components in the excitation current due to inverter control. Deterministic TO methods cannot ensure ripple performance, as they overfit to simulations and lack robustness to current uncertainties. To address this, a Chebyshev interval method-based RTO is proposed, offering a non-statistical approach that effectively handles uncertainties, including those from inverter control. Phase current data under space vector voltage pulse width modulation (SVPWM) control was obtained from an IPMSM driving system model with an inverter. A reduction technique is then used to transform a 3-dimensional phase current uncertainty into a 2-dimensional direct-quadrature (DQ)-axis current, leveraging IPMSM characteristics. This technique can avoid the challenging issue faced by existing statistical RTO methods of needing to model the constantly changing distribution of current uncertainties. The DQ-axis current uncertainty was modeled using maximum likelihood estimation, defining the interval parameters as confidence intervals. A Chebyshev metamodel was constructed with Chebyshev nodes as interpolation points. Compared to phase current metamodel, the proposed metamodel demonstrated computational efficiency while maintaining accuracy. Finally, topology optimization was performed by approximating the worst-case scenario using the metamodel's inclusion function. The proposed RTO method was validated on an 8-pole 48-slot traction motor, confirming that the optimal design effectively reduced torque ripple and remained robust under various control frequencies.