Nuclear Energy Insight
—The fields of nuclear energy and high-performance computing research may seem worlds apart. But in the minds of top experts, they have the potential, when working together, to lead to significant breakthroughs.
Energy Secretary Steven Chu recognizes the potential for advancements in commercial reactor technology and uranium fuel. “The United States excels in [computer] simulation, and we can design new nuclear reactors and fuel rods that can increase the yield without the long development time you would normally need,” he said. “This has been done in airplane development. … We need to do this with nuclear reactors.”
The U.S. Department of Energy’s Nuclear Energy Modeling and Simulation Energy Innovation hub
is doing just that. The hub has awarded the Consortium for Advanced Simulation of Light Water Reactors (CASL)
project $122 million over five years to build a “virtual reactor” for predictive simulations. One of the simulator’s goals is to help reduce operational and new reactor design costs—in part by helping to target experiments that probe key phenomena.
“Modeling and simulation has been the mainstay of nuclear energy since its inception in the 1940s,” said Douglas Kothe, the consortium’s director. “We have not only an appreciation of what it can bring, but we use it day-to-day in workflow and decision-making on reactor design. ... The [nuclear energy] community relies on insights offered by modeling simulation capabilities.”
As part of its award from the Energy Department, CASL received 45 million “processor hours” on Oak Ridge National Laboratory's (ORNL)
Jaguar supercomputer. A processor hour represents a single computer processor running for one full hour.
is a Cray XT5 supercomputer that has 37,376 processors and occupies more than 6,000 square feet at the Oak Ridge Leadership Computing Facility in Tennessee.
“You configure your desktop or laptop to your specific needs, and we work with Cray to build a machine tailored for scientific applications,” said Jeff Nichols, associate laboratory director of the computing and computational sciences directorate at Oak Ridge National Laboratory.
With its processor hours, CASL plans to build a virtual reactor.
“Our focus is to develop and apply a [virtual reactor] core on key operational and safety ‘challenge problems’ important to industry,” said Kothe. “In the case of CASL, our initial focus these first few years is pressurized water reactor cores. We will use the virtual reactor to help solve some specific, impactful problems.”
CASL hopes to achieve three main goals:
- reduce the capital and operating costs of commercial reactors
- reduce the amount of high-level radioactive waste produced
- further enhance nuclear power’s safety.
In a virtual model that mirrors a real-life reactor, CASL can show the results of variables and predict what would happen in a real reactor.
One of the key experiments will involve simulating power uprates and reactor lifetime extensions—to show how fuels, materials and structures in the reactor would respond to increased stresses.
“Trying to push the reactor to high power can lead to fretting [when a vibrating fuel rod hits the grid that holds it in place, causing damage],” said Kothe. “From a simulation view, we think it’s a very challenging yet tractable problem. We are putting together some new higher-fidelity simulations that will allow us to understand why it happens and, through design changes, prevent it.”
New materials for fuels are another avenue of research. “Our materials calculations are getting to the point where they’re predictable, so we can do materials by design and form new materials that are more efficient,” said Nichols.
FINDING NEW FUEL
Jaguar has many other applications. The pharmaceutical industry has used Jaguar to pick combinations of chemicals that might make promising new drugs for trials in the real world.
“Why couldn’t we do the same thing with fuels in a nuclear reactor?” asked Nichols. “There are lots of possibilities for new fuels that would be much more efficient in terms of their burn and the waste they generate. This is where modeling and simulation can play a role.”
Kothe added that building a predictive virtual reactor is a long process that is just five months old. “We are in a very infant stage of a nominal five-year project. If we do well, we can potentially get another five-year award,” Kothe said. “Building a large, complex multi-physics simulation capability takes a long time.”
—Read more articles in Nuclear Energy Insight and Insight Web Extra.