A research team led by Professor Daio Sekiya of the Graduate School of Information Science and Technology at Chiba University and Assistant Professor Kisuke Akaki of the Faculty of Engineering at Tokyo University of Science announced that they have developed a new method for the automatic design of high-frequency power supply circuits using AI (Artificial Intelligence) and machine learning.
NewsChiba UniversityTokyo University of ScienceDaisuke SekiyaKiku Akia
Chiba University and Others Develop Method for Automatic Design of High-Frequency Power Supply Circuits Using AI
This article is a translation. Read the Japanese original
In electric vehicles (EVs) and data centers, there is a demand to achieve both device miniaturization and energy savings. While increasing the frequency of circuits contributes to miniaturization, it has posed challenges such as increasing losses in semiconductors and coils, thereby reducing efficiency.
The method developed this time combines "multiphysics," which integrates different physical phenomena such as electricity, magnetism, and heat, with "multi-objective optimization," which handles multiple objectives simultaneously. By performing exploration and optimization via AI, the method proposes designs suitable for the intended use.
When the research team verified this with 1kW-class / 1MHz LLC resonant converter circuits, they achieved a maximum power conversion efficiency of 96.1% in an efficiency-oriented design. Furthermore, they confirmed that in a design emphasizing miniaturization, the volume of magnetic components could be reduced by approximately 41% while maintaining high efficiency.
Sources
- Designing power supply circuits tailored to purpose using AI, Chiba University and others: Enabling "just right" designs （Editorial Select: Fundamental design by AI (2026-09-16), 2026-08-31）