Deep Learning for Magnetics Design in Power Converters
Project Description
Integrated magnetic structures are critical for compact and efficient power delivery in modern AI processors, data centers, and high-frequency converters. However, their performance strongly depends on geometry, parasitics, thermal effects, and manufacturing constraints, making conventional design methods difficult to scale. This project will investigate intelligent and automated approaches for integrated magnetics modeling, optimization, and design exploration, with emphasis on high-density power conversion applications. A database of magnetic structures and simulation results will be provided for training and validation.
Supervisor
WANG, Ping
Quota
2
Course type
UROP2100
UROP3100
UROP4100
Applicant's Roles
1. Review recent developments in integrated and planar magnetics, including coupled inductors, matrix transformers, and automated electromagnetic design techniques.
2. Build electromagnetic or surrogate modeling frameworks for predicting magnetic performance, and investigate optimization strategies for geometry and layout generation.
3. Evaluate the proposed methods using selected converter applications and compare different magnetic structures in terms of size, thermal behavior, coupling characteristics, and manufacturability.
2. Build electromagnetic or surrogate modeling frameworks for predicting magnetic performance, and investigate optimization strategies for geometry and layout generation.
3. Evaluate the proposed methods using selected converter applications and compare different magnetic structures in terms of size, thermal behavior, coupling characteristics, and manufacturability.
Applicant's Learning Objectives
1. Understand the operating principles and design considerations of integrated and planar magnetic components in high-frequency power converters.
2. Develop hands-on experience in electromagnetic simulation, surrogate modeling, and automated magnetic structure optimization.
3. Gain research experience in intelligent magnetics synthesis and multi-objective evaluation for advanced power electronics systems.
2. Develop hands-on experience in electromagnetic simulation, surrogate modeling, and automated magnetic structure optimization.
3. Gain research experience in intelligent magnetics synthesis and multi-objective evaluation for advanced power electronics systems.
Complexity of the project
Moderate