Alexander Kovacs (Department of Integrated Sensor Systems, University for Continuing Education, Krems, Austria)
Titre : Machine Learning in Materials Discovery: A Case Study on Nd2Fe14B Magnet Materials
Institut Néel, Salle E424 (Salle Louis Weil)
Résumé : Rare-earth elements, including Neodymium, Terbium, and Dysprosium, are indispensable components of permanent magnets critical for various applications such as hybrid and electric vehicles and offshore wind turbine generators. The increasing demand for these elements has raised concerns about their future supply. In response, we employ machine-learning techniques to identify alternative magnetic materials with reduced Neodymium and without Terbium or Dysprosium content.
This seminar will show how the different length scales are important for the application of Nd2Fe14B in motors.
Our research has led to the development of machine-learning methods that support materials design by integrating physical models, bridging the length scale gap in Nd2Fe14B permanent magnets, spanning from atomistic details to micrometer-sized granular microstructures.
Throughout the seminar, we will …
• Showcase a machine learning model with the capability to predict critical material properties for permanent magnetic materials, such as saturation magnetization and magnetocrystalline anisotropy constants.
• Demonstrate a shallow neural network that accurately predicts the coercive field of individual magnetic grains based on their geometric and compositional descriptions.
• Explore future research avenues and acknowledge the ongoing challenges in the field of machine learning for magnet materials.
Join us in this seminar to explore contributions of machine learning in the discovery of advanced magnet materials.
