At Argonne National Laboratory, researchers are trading in old-school approximations for raw supercomputing power, proving that the secret to a safer carbon-free future lies in mastering the math of chaos.
Researchers are advancing nuclear safety by using high-performance computing to model turbulent flow — the chaotic movement of fluids and gases that governs heat transfer and gas mixing within a reactor.
The team utilizes these open-source computational fluid dynamics (CFD) codes. While Nek5000 is CPU-based, the newer NekRS is optimized for GPUs, enabling faster, more complex simulations.
These tools have been specifically adapted to predict critical issues, such as hydrogen behavior in containment structures, a major safety concern following the Fukushima accident.
Argonne proved the accuracy of their models in an international PANDA experiment, where they successfully predicted flow outcomes without prior access to experimental data.
Simulating turbulence
Turbulence is everywhere. It’s the swirl in your morning coffee and the buffeting on an airplane wing.
Inside a nuclear containment building, however, it’s a matter of national security. During rare accidents — like the 2011 disaster at Fukushima — the way hydrogen gas mixes with air can determine the integrity of the entire facility.
It impacts how heat is transferred and how gases mix within a reactor. But older computer models often struggle to provide the precision needed for complex scenarios.
They “smooth out” the chaos, potentially missing the tiny, violent eddies that change how heat moves.
To solve this, the Argonne team is using two powerhouse tools: Nek5000 and its high-speed younger sibling, NekRS.
How do you trust a digital simulation with a real-world disaster? You take a test where you don’t know the answers.
The Argonne team participated in an international “blind benchmark” known as PANDA. Given only the shape of a tank and its starting conditions, they had to predict exactly how the gas would flow before seeing the actual experimental results.
And the team nailed it.
This research not only helps predict rare accident scenarios, such as hydrogen mixing, but also accelerates the approval of next-generation reactor designs by reducing the need for costly physical experiments.
Nuclear safety
This success caught the eye of the U.S. Nuclear Regulatory Commission (NRC), specifically targeting complex containment geometries where traditional tools often fail.
Through this collaboration, NRC staff worked directly with Argonne researchers to master these simulation techniques to verify complex fluid-dynamics data.
“Regulators need tools they can trust, especially for rare scenarios that are hard to test,” said Aleksandr Obabko, an Argonne computational scientist.
It could ensure that they have the most accurate tools available to evaluate nuclear safety and rare-accident scenarios.
The mission is now moving to Aurora, one of the most powerful supercomputers ever built.
Through shifting simulations from CPUs to blazing-fast GPUs, the team is shrinking weeks of calculation into days.
Beyond enhancing safety, Argonne’s simulations offer a significant economic advantage by replacing expensive physical experiments with digital twins, thereby accelerating the regulatory approval process for new reactor designs.
As part of the DOE’s Nuclear Energy Advanced Modeling and Simulation (NEAMS) program, the team is now looking to integrate AI and machine learning to further boost predictive power and speed.
This approach aims to modernize the entire nuclear industry, making it easier for regulators and commercial partners to adopt the next generation of clean energy technology with confidence.
Source: Interesting Engineering
Scientists May Have Found the “Holy Grail” of Quantum Computing
World’s most advanced supercomputers decode nuclear reactor turbulence to advance safety