Fischer will use the grant to try to create models simple enough for engineers to be able run on desktop computers to analyze the flow of fluids through what’s known as the upper plenum in liquid metal-cooled reactors. The upper plenum is a space above the reactor core through which heated reactor coolant flows. Analyzing how fluid moves through the space involves modeling velocity, pressure, and temperatures at every possible point within it – potentially trillions of points.
“Where is this buoyantly driven coolant going to go, how is it going to behave?” Fischer asked. “That is the question we’re trying to answer. While we understand the equations at the microscale, we don’t know how to answer this question at the macroscale level because the systems are too large to solve directly.”
Fischer is an expert in both computing and fluid dynamics. The DOE work will allow him to make use of an ongoing project for which he was the chief architect, NEK5000. Fischer and a colleague won an R&D 100 Award in 2016 for the scalable computational fluid dynamics code.
Fluid flow is critical to a nuclear reactor’s function. It both cools the reactor and transfers the heat being generated that ultimately spins a turbine to generate electricity.
But thermal transport is turbulent. The greater that turbulence, the greater the friction involved, increasing the energy needed to move the fluid.
The calculations required to analyze turbulent flow through the upper plenum of a reactor would require mathematical operations of 10 to the 23rd power, forcing engineers to draw on computing power they don’t have, Fischer said.
“Design engineers have to do things on their work stations, or on clusters. They are not going to use exascale machines,” he said, referring to the coming generation of supercomputers, operating at exaflop (or 1,000 petaflop) pace.
And even if they had the computing power they needed, Fischer added, “It would be an expensive calculation.”
What Fischer and his co-investigators plan to do is to develop models that will bridge the gap between those next-generation exascale computers and work-station-accessible models. The team will use “snapshots” (velocity and temperature distributions) generated by Nek5000 on supercomputers to build reduced models of the governing equations. With sufficient preprocessing and careful mathematics, models can be built with just a few hundred unknowns that can readily be computed on a laptop or even a phone.
“So the solutions can be computed very fast, the design engineers can get their answers very fast,” he said.
Fischer will lead a team of investigators that includes Professor Yassin Hassan, the head of the Department of Nuclear Engineering at Texas A&M University; Mechanical Engineering Professor Anthony Patera of the Massachusetts Institute of Technology; Tommaso Taddei, who is a post-doctoral researcher in computer science at the Universite’ Pierre et Marie Curie; Gert Van den Eynde, a researcher at SCK-CEN, the Belgium Nuclear Research Center; and Milorad Dzodzo, an engineer in Westinghouse Electric Company’s Engineering Analysis Group.