Hold It Right There
Simulation tackles the challenge of keeping fuel rods in place
It’s tough to keep a nuclear fuel rod from rattling in its cage.
At 12 feet long, the rod is barely wider than a crayon as it is buffeted by hot, swirling, pressurized water and bombarded by neutrons. Neutrons strike nuclei within the rod, forcing them to split apart in a reaction that brings the rod’s temperature above 1,000 degrees Fahrenheit.
The rod and more than 200 others are grouped in a bundle and held in place by as many as a dozen grids with strong, sturdy springs. Yet the frames and the springs sit in the same extreme environment as the rods. Eventually the springs lose their grip, the rods get banged around, and the cladding around the rods gets worn away.
If the rod is left in place long enough, products of the fission process taking place in the fuel can leak into the water. This is not good.
Calming the water is not the answer; it is turbulent to more effectively carry heat away from the reactor core and to steam generators, which produce steam to drive turbines that generate electricity. The water’s turbulence is so important that the grids have mixing vanes, little wings that jut out to keep the water stirred up.
One major challenge for the reactor’s operator, then, is to keep the rods safely immobilized inside the reactor as long as possible without risking a fuel leak. Researchers with the Consortium for the Advanced Simulation of Light-Water Reactors (CASL) are working to understand this challenge better by simulating it on Oak Ridge National Laboratory’s (ORNL’s) Titan supercomputer.
The technical term for damage done to rattling fuel rods is grid-to-rod fretting, or GTRF. CASL researchers recreate the problem on Titan with a computational fluid dynamics code called Hydra-TH.
“GTRF is the primary failure mechanism for fuel,” said CASL collaborator Mark Christon of Los Alamos National Laboratory (LANL). “The mixing vanes tend to generate a lot of turbulence, which leads to fluctuations in the pressure that make the rods vibrate.
“If they can design it out with better material or spacers, they can get a longer life from the fuel, as well as increased power output.”
The researchers at this point are focused on pressurized water reactors, or PWRs, the most common design for American power plants. It is distinguished from boiling water reactors, in which the water is not kept under pressure and is allowed to boil.
By using Hydra-TH, Christon and collaborators within CASL are able to model the thermal hydraulics of a PWR by dividing it into as many as 250 million separate cells.
Hydra (without the “TH”) is a toolkit developed at LANL, explained Jozsef Bakosi, another LANL researcher working on the project. In its original form it can focus on a wide variety of physical phenomena. Hydra-TH is a specific port of the code for analyzing thermal hydraulics in a nuclear reactor.
The code itself is not the only difficult piece of the simulation. The grid is an unstructured mesh, meaning the cells themselves are not all the same size and shape. In order to make the grid as true as possible to the actual surfaces within the reactor, it is composed of a custom arrangement of six-sided figures, or hexahedra. The increased accuracy helps the researchers better model the pressure and velocity of the water.
“When you break up the fluid volume into cells, those can be different-shaped elements,” Christon explained. “The order that those come in is random. You place the elements together to fit the geometry.”
“It’s hard to generate the mesh,” Bakosi added. “Since it’s very complicated and very unstructured, the effort needed is a nontrivial task. It used to take months to build a single mesh, but now we have a mesh generator that can create it in a week or two. This is a major accomplishment for the folks that generate the mesh. There’s a lot of time and money to be saved.”
Hydra-TH ran on Titan’s predecessor, Jaguar, before moving to the newer machine. In 2012 the team used nearly 22 million processor hours on Jaguar.
On Titan, the project has scaled up to 36,000 of the system’s nearly 300,000 CPU cores. Bakosi said the team is working with GPU maker NVIDIA to take full advantage of Titan’s GPU accelerators. Specifically, he said, they’re working with the company to adapt its linear algebra library for Hydra-TH.
“The solution for pressure is the most time-consuming part of the calculation,” he said. “This is an algorithm that takes 50 to 70 percent of the runtime. We would like to use NVIDIA’s library to take over that part. We’re using pretty advanced algorithms for the pressure solution. We want to take the algorithm and get it running completely on the GPUs. We’ll get a big speedup from that part of it.”
As it moves forward Hydra-TH will be used for problems other than GTRF, noted John Turner of ORNL, CASL’s lead for virtual reactor integration.
“There are multiple things that we need Hydra for,” he said. “You need a detailed flow to understand the heat transfer, the convection of the heat out of the rods in the fluid. Up to now there’s been much more crude or course flow models, and as long as you could validate to experiments, that was fine. But we’re getting into regimes where we need a more detailed description of the flow.”
Another problem, called CRUD, also needs Hydra’s capabilities. CRUD, short for “Chalk River Unidentified Deposits,” refers to a crust of boron that builds up on the outside of the fuel rods due to a phenomenon called “subcooled boiling.” In subcooled boiling, bubbles form on the rods even though the water is kept under pressure to prevent it from boiling.
“Imagine you put a pot of water on the stove, turn up the heat, and watch it as it comes close to boiling,” Turner explained. “If you stuck a thermometer in there it would be a long time before you really get a full boil. But way before then you’ll start to see bubbles on the bottom of the pan. And that’s subcooled boiling, because the bulk of the fluid is below the boiling point, but right there at the surface you’ve got conditions where you’ve got a little bit of boiling.”
Boron absorbs neutrons; reactor operators can therefore use it to help control the nuclear reaction. When it builds on the outside of fuel rods, however, it can interfere with the carefully controlled operation of the reactor.
“Boron is in the coolant,” Turner noted. “It’s part of the chemistry of the coolant, and it can collect where those bubbles form. Because boron is a strong neutron absorber, it affects the shape of the neutron flux in the reactor, and thus the power, so knowing where that subcooled boiling is happening is important.”
Hydra-TH is one of the components of CASL’s virtual nuclear reactor, called VERA for “Virtual Environment for Reactor Analysis.” The primary physical phenomena of this virtual reactor are thermal hydraulics (fluid flow and heat transfer) and neutron behavior (neutronics). Denovo, developed at ORNL, is a part of one of the neutronics capabilities in VERA.
“Coupling these codes is one of the big challenges,” Turner said, “for both numerical and software aspects. For instance, Denovo uses a Cartesian mesh, or block grid cells, while Hydra-TH uses unstructured meshes. Efficiently passing information back and forth between physics codes that use different types of meshes, while maintaining accuracy, can be a real challenge.”