Titolo della tesi: Multiphysics Design of High-Power Density and High-Speed Axial Flux Permanent-Magnet Synchronous Machines
The rapid electrification of transportation and the push toward energy-efficient industrial systems have created unprecedented demands on electric machine performance, particularly in terms of power density, torque density, compactness, and thermal robustness. Among various motor topologies, Axial Flux Permanent Magnet (AFPM) machines have emerged as highly promising candidates , owing to their short magnetic path, high torque-to-volume ratio, and suitability for space-constrained applications such as in-wheel drives, eVTOL propulsion systems, and modular direct-drive system. In the design of modern, high-performance AFPM machines, engineers are challenged by a highly restrictive set of design constraints and performance requirements, which often conflict with each other. These include maximizing power and torque density, minimizing axial length, managing thermal loads, and maintaining mechanical integrity under high-speed operation. As a result, the likelihood of identifying a feasible solution is significantly reduced, if such a solution exists at all. This complexity is further exacerbated by the inherently multiphysics nature of AFPM machines, where electromagnetic, thermal, and mechanical interactions are tightly coupled. Conventional design methodologies, whether heuristic, sequential, or optimization-based, have proven inadequate in addressing these interdependencies from the earliest stages of the design process. A possible solution is therefore to change the way the high-performance AFPM machines are approached. To this end, this thesis proposes a new methodology for the design of high-speed AFPM machines, based on the introduction of an analytical stage prior to the Finite Elements stage, where the ‘Actual Design Space’ (ADS) is determined, i.e., the ‘space wherein the final design can be found’. This ADS serves as a rigorously filtered subset containing only configurations that satisfy all essential physical constraints and operational requirements. By identifying this space analytically at the outset, the designer gains early insight into the feasibility of the problem, avoids excessive reliance on numerical simulation, and ensures that the design process proceeds efficiently. This shift in methodology not only reduces design cycle time and complexity but also enhances robustness and reliability in the final machine configuration. The process to determine the ADS begins with a rigorous identification of the independent design variables that govern the machine’s geometry, material choices, and operational parameters. Subsequently, internal relationships and external performance constraints are introduced one by one, with the purpose of systematically eliminating all infeasible candidates from the initial design domain. Through this sequential filtering, only those configurations that satisfy all critical criteria are retained. At the end of this process, the ADS is attained—characterized by containing only feasible and physically meaningful design solutions. As can be observed, this methodology is in- herently multiphysics in nature, as it handles thermal, mechanical, and electromagnetic constraints simultaneously, from the very beginning of the design phase. The insight provided by the ADS enables designers to 1) confirm the feasibility of the design problem (ensuring a non-empty ADS), and 2) apply targeted FEA within this refined space to finalize and validate optimal configurations. This structured and computationally efficient approach marks a significant improvement over traditional workflows that are either too general or lack physical insight. The thesis concludes with the practical implementation and validation of the proposed methodology through the design of a high-performance AFPM machine for traction applications. The application of the ADS methodology is demonstrated through the synthesis of realistic machine geometry and its evaluation under no-load and loaded conditions using FEA. The results demonstrate that the ADS methodology offers a powerful, transparent, and computationally efficient framework for addressing the increasingly demanding design constraints of next-generation electric machines.