Minimization of eddy currents in permanent magnets of an electric machine with shape derivatives
Alessio Cesarano, Peter Gangl
TL;DR
This paper tackles minimizing eddy-current losses in the magnets of an interior permanent magnet synchronous machine (IPMSM) by incorporating time-dependent eddy-current effects through a magneto-quasi-static PDE and a time-stepping formulation. It advances a gradient-based, shape-derivative optimization framework that couples forward time-dependent state equations with backward-in-time adjoint equations, enabling the computation of domain sensitivities for designing rotor air-pocket shapes. Through a scalarized objective that combines average eddy-current dissipation and average torque, the method achieves a ~17% reduction in eddy losses and a torque increase, demonstrating the practical viability of time-dependent, shape-optimized IPMSMs. Limitations related to topology changes and manufacturing feasibility motivate future work on mechanical-stiffness constraints and time-periodic problem formulations, as well as potential remeshing strategies to sustain optimization progress.
Abstract
In this work we deal with the shape optimization of an electric machine considering time-dependent effects such as eddy currents. The considered electric machine is an interior permanent magnet synchronous machine and we minimize the average dissipated power due to the eddy currents in the magnets over a period of time corresponding to a rotation, while at the same time maximizing the average torque. Our approach is based on the computation of the shape derivative which -- beside the computation of a time discretization of time-dependent state problem -- also involves solving a a time discretization of a time-dependent adjoint problem. The challenge of this problem is related to the dependency of each one of the N time steps of the adjoint problem on two different time steps, due to the use of finite difference in the calculation of eddy current losses.
