April Novak, a doctoral candidate in UC Berkeley’s Department of Nuclear Engineering, works with Asst. Professor Rachel Slaybaugh to develop modeling software that helps validate the safety and feasibility of advanced pebble-bed reactor (PBR) designs, and supports the licensing process for PBRs. She notes that software modeling is essential to the practicalities and economics of nuclear reactor development: “it’s generally a lot cheaper to do heavy computing than heavy experiments.” Novak’s work on Pronghorn, a software that models dispersion of heat in pebble-bed reactors, won first place in the advanced reactor systems category at the annual Innovations in Nuclear Technology R&D Awards earlier this year.
A native of the Chicago metropolitan area, Novak earned her undergraduate degree in Nuclear Engineering at the University of Illinois Urbana-Champaign (UIUC). “I decided to come to Berkeley because I love the department,” she said. “I love the focus on the future that I experience among the professors, the focus on developing new reactor technologies that haven’t been developed before, that can solve a lot of challenges that we have in the field. It’s a great environment for innovation, and for pushing the field forward.”
Modeling pebble-bed reactors
Pebble-bed reactors differ significantly from conventional, water-cooled nuclear reactors. Where conventional reactors employ bundled fuel rods, each containing a stack of pellets of fissionable material, pebble-bed reactors are fueled by small particles of fissionable material enclosed in a much larger, tennis-ball sized sphere (“pebble”) of pyrolytic graphite. PBRs, unlike their water-cooled predecessors, are considered “passively safe.” The reactors are designed to handle very high temperatures, and the high stability of the fuel results in less radioactive releases at high temperatures. Four PBRs have been built and operated to-date in Germany and China; none of the approximately 100 nuclear reactors currently operating in the United States is a PBR.
Novak’s work focuses almost exclusively on PBRs, which she identifies as “one of the most challenging systems to model in nuclear engineering, because of the huge range in length scales” -- from fuel particles ~1mm in diameter, to the whole reactor system, a massive industrial facility. Modeling such a complex system “requires a lot of computational effort,” Novak explains, “unless you can make some sophisticated simplifying assumptions. So half my work is developing those models, to predict [the behavior of] these systems without requiring supercomputers.”
Because temperature and density greatly affects neutron transport -- the sub-atomic activity that occurs during nuclear fission -- the ability to couple computational simulations of nuclear transport with simulations of thermal hydraulics is essential to determining the safety of a reactor’s design. Because it’s rare to be proficient in understanding and modeling both, these two types of computational simulations are developed by distinct groups of nuclear engineers. Therefore, Novak says, “the other part of my work is allowing software developed by different people to be combined for multi-physics simulations … by developing the interface software to allow these codes to work together.”
Heat-dispersion modeling and validation
Novak’s dissertation work combines the development of software and validation of the computational models. Validating software by comparing its predictions with empirical data generated by experiments assures that the software accurately predicts behavior of real-world systems. Novak redeveloped Pronghorn, the software that models dispersion of heat in pebble-bed reactors, from scratch while an intern at the Idaho National Laboratory (INL), and now works with others who have joined the project to refine and validate the codes.
The SANA facility in Germany was used to conduct heat transfer experiments between 1994-96, generating data on how a high-powered, non-nuclear element’s heat is dispersed through a bed of pebble materials constructed to model a high temperature, gas-cooled reactor core. One way Pronghorn has been validated by Novak and her colleagues is by comparing the software’s prediction of how heat will disperse in a facility whose modeled characteristics resemble the SANA facility, with measurements taken in the mid-1990s of heat dispersion in the actual test bed. This comparison has validated that Pronghorn accurately models the dispersion of heat in a PBR several hours after shutdown, and in doing so helps engineers designing a nuclear reactor to assess the “passive safety” of a reactor in the wake of a controlled or an unintentional shutdown of power-generating capacity and coolant flow -- that is, whether the reactor’s heat can be removed safely rather than cause a meltdown in the reactor core.
Codes that run on HPCs vs. on laptops
When Novak began her work at Berkeley, she developed software used to help design experimental facilities for nuclear tests on a scale far smaller (and therefore far less expensive) than an industrial-scale nuclear power plant. Novak compares these smaller-scale tests to design in the aviation industry, where airplane components are tested in a wind tunnel rather than in actual flight. For this work, essential to limiting the expense of reactor design, she utilized the Savio HPC cluster extensively.
Pronghorn, on the other hand, is intended for fast iterations run by industrial engineers in commercial design applications, at roughly the midpoint of a design process, prior to stages that require full-scale, high-fidelity engineering simulation of a PBR. In these applications, software needs to be able to run on a single, well-provisioned desktop computer rather than on an HPC cluster; and so Novak’s development and testing of the software could be performed on her laptop, or on a small cluster maintained by her advisor, Prof. Slaybaugh. In this work, Novak has taken advantage of the expertise and advice of Berkeley Research Computing (BRC) consultants when she has encountered a compilation problem, or a problem getting software libraries to load. BRC consulting is “extremely useful for researchers,” Novak explains, in order to quickly resolve issues like these.
Because “Pronghorn has been selected by the Nuclear Regulatory Commission as one of their analysis tools for gas-cooled pebble-bed reactors,” Novak says, “my time is devoted to providing the industry and the regulatory body with software they can use to analyze these systems.” Her dissertation will focus on multi-physics coupling of thermal hydraulics codes (Pronghorn) and existing neutronics codes to look at how changes in core layout, choice of fuel cycle, and other design choices are expected to affect reactor operation -- results that will offer preliminary suggestions for future PBR designs. In this dissertation work, she will utilize Savio to run high-fidelity stochastic neutron transport codes (Monte Carlo simulations), which will generate reduced-order parameters used as inputs to her multi-physics analyses.
The Berkeley Research Computing Program staff look forward to supporting Novak in her future work, and invite other campus researchers who might benefit from Research IT’s service offerings to contact us at research-it@berkeley.edu.
Pictured above: Pronghorn simulation results for the Mark 1 Pebble Bed Fluoride-Salt-Cooled High-Temperature Reactor (PB-FHR) designed by the nuclear engineering department at UC Berkeley. The core is an annular pebble bed contained within two graphite structures. Pictured on the left is the fluid temperature in the core and on the right the pressure.