Time-Efficient Multi-Physics Optimization Approaches for the Design of Synchronous Reluctance Motors

The design of synchronous reluctance machines involves a multiphysics optimization with a consequent number of design parameters. To lower the optimization time, it is common to split the problem into subproblems, i.e., to optimize successively the flux barriers for the electromagnetic performance, and then the ribs for the structural integrity of the rotor. This however leads in general to a suboptimal design, because the cross-coupling between design parameter subsets (e.g., magnetic and mechanical) is this way ignored. In this study, different splitting optimization strategies have been implemented and evaluated by comparing not only the electromagnetic performance reached by the optimized designs, but also the computation time. Results show that the optimization time can indeed be significantly lowered by performing a magnetic optimization followed by a mechanical optimization, with little impact on the mean torque output. Pre-dimensioning analytically the radial ribs within the magnetic optimization further reduces the optimization time and improves the performance of the design. Finally, performing an additional iteration on the approach leads to a mean torque and torque ripple very close to the ones obtained by carrying the global optimization, with however half the optimization time.