The team, led by ECE Assistant Professor Baoyun Ge, won for their innovative magnetic prediction model using analogies in physics.
(text and background only visible when logged in)
A team led by Georgia Tech School of Electrical and Computer Engineering (ECE) Assistant Professor Baoyun Ge took home first place at the IEEE Power Electronics Society 2025 MagNet Challenge.
On the team with Ge were ECE Ph.D. students Piyush Chauhan and Yuanhao Mo, H. Milton Stewart School of Industrial and Systems Engineering (ISyE) Assistant Professor Xiaochen Xian, ISyE Ph.D. student Dongmin Li, and General Motors engineer Le Chang.
They won over 39 other teams from around the world for innovations on mechanical analogy for magnetic systems and outstanding performance.
The competition was sponsored by IEEE Power Electronics Society, Nvidia, Texas Instruments, Wurth Electronik, ITG Electronics, pSemi and hosted by Princeton University and Dartmouth College. It had student teams develop software algorithms to learn from existing training data and compete on unknown testing data with the goal of improving power magnetics modeling.
Using analogies, the team developed a physics‑informed prediction model that draws parallels between magnetic and mechanical behavior, enabling the model to explicitly learn and represent the effects of saturation, hysteresis, eddy currents, and displacement currents.
The team was able to greatly reduce the prediction error for five testing materials across a wide range of switching frequency (50kHz to 800kHz) and operating temperatures (25 C to 70 C).
“Accurately modeling the behavior of power magnetics has been a problem for more than 100 years since Steinmetz’s time in General Electric,” Ge said. “The approach we used has strong potential to be integrated with circuit simulation software as well as finite element software to facilitate high-fidelity virtual prototyping of power converters, electric motors, and transformers.”
The team has also submitted two provisional patents on the proposed technology.
According to the MagNet Challenge handbook, magnetic components contribute over 30 percent of the cost and over 30 percent of the loss in almost all power converters. The performance of magnetic components is an important bottleneck in the development of high-performance power electronics.
Circuit simulation tools have greatly accelerated the integrated circuit design process, and numerical field simulation tools have enhanced our understanding of sophisticated component geometries. Despite great progress in simulation tools, the necessary progress in the modeling and design of power magnetics is lagging.
Magnetic materials are highly nonlinear, and a large variation exists in the magnetic geometries due to the manufacturing process. Although physical theory can explain the phenomena involved in the core loss, it cannot predict it with useful accuracy for practical materials.
Existing magnetic material modeling tools are either too simple and thus, not accurate enough, or are reliant on experimental measurements that can only be performed after design and fabrication.
The project was completed over 10 months, starting in March 2025. The team accepted the award during the IEEE Applied Power Electronics Conference in San Antonio, Texas, on March 24, 2026.
Related Content
Datta Wins IEEE Andrew S. Grove Award
The professor was honored for his contributions to and leadership in high-performance and energy-efficient logic transistor research.
Sarkar Receives Top Honor for Memory Research at PRISM Review
The Ph.D. student earned the Best Demonstration Award for developing a 3D ferroelectric memory design that enables high-density, low-power storage for future AI systems.