Dresden 2026 – wissenschaftliches Programm

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

MA: Fachverband Magnetismus

MA 50: Bulk Materials: Soft and Hard Permanent Magnets

MA 50.6: Vortrag

Donnerstag, 12. März 2026, 16:15–16:30, POT/0151

Accelerating Sm-Fe-V Phase-Diagram Mapping via an Active-Learning Pipeline — •Aaron Dextre, Alex Aubert, Konstantin Skokov, Oliver Gutfleisch, and Pelin Tozman — Technische Universität Darmstadt

Sm1Fe12-based compounds are promising candidates to replace Nd-Fe-B, yet realizing their potential requires specific microstructures where grains are isolated by a low-melting-point phase. To locate the specific phase equilibria required for this coexistence, we developed an active-learning framework to accelerate the mapping of the Sm-Fe-V phase diagram. Integrating Neural Networks and Random Forests, our ensemble model iteratively directed the synthesis and annealing of target alloys to refine phase boundaries efficiently. After six cycles, we determined that the 1:12 phase is stable over a broader V range, while the target two-phase field is more confined than previously reported. These findings revise the Sm-Fe-V equilibria and demonstrate active learning*s utility in magnetic materials discovery.

Keywords: Permanent magnets; Active learning; Machine learning