Context. Iron losses are a key limitation in the design of electrical machines and drives. The push towards compact, high-power-density drives has raised operating frequencies to several hundred hertz, sharply increasing these losses, yet industry still lacks tools that accurately predict their level and distribution.
Project. Starting in November 2026, a 3-year research programme jointly funded by FNRS (Belgium) and DFG (Germany) brings together two complementary groups. The ACE group at ULiège specializes in computational electromagnetics and finite element methods; it originated the energy-based (EB) hysteresis model, now used worldwide, develops the open-source tools Gmsh and GetDP, and has recently developed neural-network-based homogenization techniques for laminated cores (arXiv:2607.14321). The IEM group at RWTH Aachen specializes in electrical machines and magnetic materials, and operates a state-of-the-art laboratory for standard and special magnetic measurements.
Objectives. The project will extend the EB hysteresis model to account for anisotropy and rotational losses. It will further develop neural surrogate and homogenization methods to enable fast and accurate iron loss predictions within design optimization loops. Finally, it will build experimentally validated material libraries for electrical steels, in the form of neural networks directly usable in finite element simulations.
Profile. Candidates should hold a PhD in computational science, applied mathematics, electrical engineering, physics or a related field. We are looking for a strong background in numerical methods for PDEs (finite elements in particular) and/or scientific machine learning, together with solid programming skills (Python, C++, PyTorch or JAX). Knowledge of electromagnetics is an asset but not a requirement. Fluency in English is expected.
How to apply. Please send a CV and the names and emails of two referees to cgeuzaine at uliege dot be. Review of applications begins immediately and continues until the position is filled.
