Automated Identification of the Origin of Energy Loss in Non-Oriented Electrical Steel by Feature-Extended Ginzburg–Landau Free-Energy Framework

This paper presents an interpretable, physics-informed machine learning framework based on a feature-extended Ginzburg–Landau free-energy model for analyzing energy loss mechanisms in non-oriented electrical steel.

Using a combination of simulation data and persistent topological features, the study offers a new method to automatically identify the origin of loss components, with relevance to industrial material development and magnetic systems research.

Authors & Affiliations:

• Michiki Taniwaki, Ryunosuke Nagaoka, Ken Masuzawa, Shunsuke Sato, Alexandre Lira Foggiatto, Chiharu Mitsumata, Takahiro Yamazaki — Tokyo University of Science
• Ippei Obayashi — Okayama University
• Yasuaki Hiraoka — Kyoto University
• Yasuhiko Igarashi, Yuta Mizutori — University of Tsukuba
• Hossein Sepehri-Amin — National Institute for Materials Science (NIMS)
• Tadakatsu Ohkubo — National Institute for Materials Science (NIMS)
• Hisashi Mogi — Nippon Steel
• Masato Kotsugi — Tokyo University of Science

Recommended citation: Taniwaki, M., Nagaoka, R., Masuzawa, K., Sato, S., Foggiatto, A. L., Mitsumata, C., Yamazaki, T., Obayashi, I., Hiraoka, Y., Igarashi, Y., Mizutori, Y., Sepehri-Amin, H., Ohkubo, T., Mogi, H., & Kotsugi, M. (2025). "Automated Identification of the Origin of Energy Loss in Non-Oriented Electrical Steel by Feature-Extended Ginzburg–Landau Free-Energy Framework." Scientific Reports, 15, 357. https://doi.org/10.1038/s41598-025-00357-z
Download Paper