Regensburg 2025 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 31: Data-driven Materials Science: Big Data and Worksflows

MM 31.11: Vortrag

Donnerstag, 20. März 2025, 17:45–18:00, H10

Machine Learning-Assisted Design of Magnetic Materials: Predicting Properties for not purely ternary Nd₂Fe₁₄B — •Manuel Enns, Daniel Urban, Wolfgang Körner, and Christian Elsässer — Fraunhofer IWM, Wöhlerstraße 11, 79108 Freiburg, Germany

Nd₂Fe₁₄B-based hard-magnetic materials are widely used for strong permanent magnets. Their re-use and recycling after the end of the magnet’s life cycle opens the question of the degradation of the magnetic properties due to the incorporation of unintentional impurity elements originating from the recycling procedures. In this talk, we present a data-mining and machine-learning (ML) approach using kernel-based learning methods to predict the influence of impurity atoms in Nd₂Fe₁₄B-based materials. The magnetic-property data used for training and testing the ML model were obtained by a combinatorial high-throughput screening (HTS) using density-functional theory calculations. We demonstrate that our ML approach can accurately predict the saturation magnetization, the uniaxial anisotropy constant, and the formation energy for Nd₂Fe₁₄B with impurities added by recycling.

Keywords: Machine Learning; Permanent Magnets; Data-Mining