Pushing Creativity Further in Electrical Machine Design Using Topology Optimization

The sizing of electrical machines, at the heart of the energy transition (electrification of transportation, electricity production), generally relies on engineers' experience. One selects an existing structure that looks the most appropriate to meet the specifications and then optimizes it on a limited set of geometric parameters. This approach is necessarily biased by the designers' choices (initial structure, parameterization, modeling assumptions), which can leave out innovative and potentially better structures. In order to avoid this pitfall, non-parametric approaches such as topology optimization can be conducted.

This type of methodology is, in fact, capable of finding surprising structures without any information on the initial geometry. Initially used in mechanical engineering, topology optimization has gradually spread to all fields of physics and is now the subject of active research. This talk aims to present the main principles of this method, as well as some of its applications related to rotating electrical machines. The difficulties are multiple, both theoretical (local optima, non-existence of an optimal solution without regularization) and practical: several materials have to be considered (iron, air, conductors, magnets), as well as several physics (magnetostatics, mechanics...). Despite these pitfalls, we will show that it is possible to design a complete electrical machine in just a few minutes without any initial information.