World's most advanced supercomputers decode nuclear reactor turbulence

Researchers at Argonne National Laboratory are leveraging cutting-edge supercomputing power to improve nuclear reactor safety by accurately modeling turbulent fluid flow—a critical factor in heat transfer and gas mixing within reactors. Using advanced open-source computational fluid dynamics (CFD) tools, Nek5000 (CPU-based) and its GPU-optimized successor NekRS, the team can simulate complex turbulent behaviors, such as hydrogen gas mixing in containment structures, which is vital for preventing accidents like the 2011 Fukushima disaster. Their models demonstrated high accuracy in the international PANDA blind benchmark, successfully predicting gas flow patterns without prior experimental data. This breakthrough has attracted the attention of the U.S. Nuclear Regulatory Commission (NRC), which is collaborating with Argonne to apply these simulations for verifying complex containment geometries where traditional methods fall short. By transitioning simulations to the powerful Aurora supercomputer and integrating AI and machine learning through the DOE’s NEAMS program, Argonne aims to drastically reduce computation times and enhance predictive capabilities. This approach not only