Axial Flux Machine — Surrogate-Based Topology Optimization
Humanoid robots require many electric actuators, increasing the demand for permanent magnets that provide high torque density. YASA axial-flux machines can use this material efficiently, but their three-dimensional flux paths make electromagnetic simulations computationally expensive. In my master's thesis, I developed a surrogate-model-based optimization process to reduce the number of required 3D finite-element simulations. A data-driven constraint cascade progressively narrows the design space to the relevant region without relying on predefined heuristics. Read More














