Geunhyeong Lee
Scientist at FRIB, Michigan State University
AI-driven design for thermal-fluid systems
My vision is to advance thermal-fluid systems beyond conventional design limits, connecting AI-driven discovery with fabrication and experimental validation for nuclear systems.
Research in focus
BiSTO
Binary Surrogate-based Topology Optimization
The Solid pointer identifies the outer shell first, then the black solid region as it develops in the cutaway. The Fluid pointer identifies the blue fluid region.
The animation shows design iterations, not elapsed physical time.
Solid-fluid design to minimize temperature — rotating beam dump example.
I developed BiSTO, a binary topology optimization framework designed for multiphysics applications. It combines AI-guided design search with physics-based simulation.
- AI search
- Simulation
- Design update
Research Approach
-
AI-driven Optimization
Reinforcement learning and surrogate modeling for shape and topology optimization
-
Multiphysics Modeling
Thermal-fluid and electromagnetic analysis of heat exchangers and liquid-metal systems
-
Fabrication & Validation
Design for 3D printing and experimental testing of thermal-fluid components