Deep Learning for Magnetics Design in Power Converters
Project Description
Magnetic components are critical elements in high-frequency power converters, strongly affecting power density, efficiency, and dynamic performance. Conventional magnetics design often relies on expert intuition and iterative simulations, making optimization across multiple objectives challenging and time-consuming. This project will explore artificial intelligence assisted methods for magnetic component modeling, optimization, and automated design exploration, aiming to improve design efficiency for next-generation high-density power converters. Data base of magnetic components will be provided.
Supervisor
WANG, Ping
Quota
2
Course type
UROP2100
UROP3100
UROP4100
Applicant's Roles
1. Conduct a literature review on magnetic component design methods and recent AI-assisted optimization techniques, and summarize key design trade-offs.
2. Develop simulation or data-driven models for magnetic component analysis, and investigate AI methods for automated optimization.
3. Validate the proposed approach through case studies and performance comparison in terms of loss, power density, or efficiency.
2. Develop simulation or data-driven models for magnetic component analysis, and investigate AI methods for automated optimization.
3. Validate the proposed approach through case studies and performance comparison in terms of loss, power density, or efficiency.
Applicant's Learning Objectives
1. Gain a fundamental understanding of magnetic component design and key trade-offs in high-frequency power converters.
2. Develop practical skills in simulation, modeling, and applying AI methods for power magnetics design optimization.
3. Acquire research experience in AI-assisted design methodology and performance evaluation for power electronics applications.
2. Develop practical skills in simulation, modeling, and applying AI methods for power magnetics design optimization.
3. Acquire research experience in AI-assisted design methodology and performance evaluation for power electronics applications.
Complexity of the project
Moderate