Prof. Hajime Igarashi
He received the B.E. and M.E. degrees in electrical engineering and the Ph.D. degree in engineering from Hokkaido University, Sapporo, Japan, in 1982, 1984, and 1992, respectively. From 1995 to 1997, he worked with Prof. Arnulf Kost at the Technical University of Berlin with the support of the Humboldt Foundation.
He has been a Professor with the Graduate School of Information Science and Technology, Hokkaido University, since 2004. His research interests include computational electromagnetism, design optimization, and artificial intelligence (AI)-based design. He is the Vice President of the International Compumag Society. He is the author of “Topology Optimization and AI-based Design of Power Electronic and Electric Devices”, published by Academic Press in 2024.
Presentation: Case studies on topology optimization and its acceleration with machine learning
In this presentation, an overview of topology optimization will be provided, including a comparison of the NGnet (normalized Gaussian network) method based on population-based stochastic algorithm with the density and level-set methods based on gradient-based search. The focus will then be on the NGnet method, which is applied to the design of a permanent magnet motor for air mobility and microwave devices. The efficacy of this approach is demonstrated by the successful discovery of a novel electromagnetic structure, which has led to an AI-based patent. An improvement in the NGnet method will also be presented. I then present the approaches that utilize machine learning to accelerate the optimization process. The analysis reveals that the online method, which involves updating the surrogate model during the optimization procedure, is more appropriate for general-purpose optimization compared to the offline method.