Argonne National Laboratory

Argonne researchers receive HPC Innovation Excellence Award

November 18, 2014

A team of researchers from the Department of Energy’s (DOE) Argonne National Laboratory, in collaboration with scientists from the University of Illinois at Urbana-Champaign and three nuclear reactor design centers in the United States and Europe, received the HPC Innovation Excellence Award at SC’14, the supercomputing community’s premier annual event, Nov. 18.

The award, presented by International Data Corporation, recognizes noteworthy achievements by users of high-performance computing (HPC) technologies. The program’s main goals are to showcase return on investment and scientific success stories involving HPC; to help other users better understand the benefits of adopting HPC and justify HPC investments, especially for small- and medium-sized businesses; to demonstrate the value of HPC to funding bodies and politicians; and to expand public support for increased HPC investments.

The award-winning project involved development of high-fidelity simulation techniques for nuclear reactor thermal hydraulics. The project leads from Argonne’s Mathematics and Computer Science Division and Nuclear Engineering Division – Paul Fischer, Elia Merzari, Aleks Obabko, and Shashi Aithal – carried out high-performance simulations of liquid metal coolant passing through subassemblies of hexagonally arrayed fuel pins. The work was supported by the DOE Office of Nuclear Energy.

“Such simulations can increase the safety, reliability, and efficiency of nuclear reactors while reducing the overall cost per kilowatt-hour to the consumer,” said Aithal. “The challenges are enormous: one assembly may comprise several hundred pins. But the ability to predict mixing behavior can allow operation at higher power levels and can help in designing the overall reactor core.”

Key to the team’s success was use of the Argonne-designed computational fluid dynamics code Nek5000. Unlike most commercial codes that use finite-volume methods that are only second-order accurate and not highly scalable, Nek5000 is a higher-order spectral code that has scaled beyond 500,000 cores on Mira, the IBM Blue Gene/Q supercomputer operated at Argonne. Testing conducted at the Argonne Leadership Computing Facility indicate that Nek5000 can be used for truly predictive, nuclear fuel assembly simulations.

The Argonne team worked closely with three nuclear research groups – NRG (the Netherlands), SCK-CEN (Belgium), and TerraPower (USA) – to use high-fidelity, cost-saving simulations with Nek5000 in order to design and build the next generation of nuclear reactors.

“The thermal-hydraulic simulations using Nek5000 can reduce or even eliminate complex, inflexible and expensive testing, with a potential savings of over $1 million per year in testing costs,” said Aithal.  He emphasized that the ability to simulate large problems on leadership-class machines such as Mira at Argonne is critical.  

“With Nek5000, scientists can benchmark lower-fidelity design tools, thus making it an integral part in the plant design cycle,” Aithal said. “Nek5000 can also provide high-fidelity studies on smaller design subspace problems, thus reducing design margins and hence the associated cost.”