ExaSMR simulation toolkit advances nuclear reactor design

Possible choices to carbon-producing calories assets are turning into ever extra crucial as local weather exchange presentations its results at the Earth and in our day by day lives. Even supposing fossil fuels nonetheless generate a lot of the electrical energy in the USA, utilities are an increasing number of including renewable assets comparable to wind and sun to their calories portfolios.

In 2021, 20% of the rustic’s electrical energy got here from renewables vs. 61% from fossil fuels. However through 2050, each are projected to offer 44% as renewables proceed to surge and fossil fuels decline.

The place does that depart out of date nuclear calories?

For the previous twenty years, fission reactors have produced a just about unchanging portion of the country’s electrical energy: round 20%. However that share may get started expanding quickly as new design approaches and reactor applied sciences promise to turn into the nuclear energy business.

The arrival of small modular reactors (SMRs) and complex reactor ideas (ARCs) indicators a brand new technology of fission energy. Not like maximum industrial nuclear reactors nowadays, SMRs are considerably smaller and use standardized designs, thereby lowering development prices and manufacturing time. In the meantime, ARCs discover new applied sciences to provide fission energy extra successfully and safely. Each efforts use pc simulations to expect the viability of proposed designs and to toughen them. However working such fluid dynamics and neutron delivery fashions will also be computationally not easy and dear, thus restricting their use through business.

Exascale SMR (ExaSMR), which is a high-performance computing (HPC) tool venture supported through the U.S. Division of Power’s (DOE’s) Exascale Computing Venture (ECP), targets to make large-scale nuclear reactor simulations more straightforward to get right of entry to, inexpensive to run, and extra correct than the present state-of-the-art.

“Via correctly predicting the nuclear reactor gasoline cycle, ExaSMR reduces the collection of bodily experiments that reactor designers would carry out to justify the gasoline use. Largely, that is what simulation is purchasing firms: a predictive capacity that tells you ways sure options will carry out in order that you do not want to bodily assemble or carry out as many experiments, which might be tremendously pricey,” mentioned Steven Hamilton, ExaSMR venture chief and R&D scientist within the HPC Strategies for Nuclear Programs Team at DOE’s Oak Ridge Nationwide Laboratory (ORNL).

The ExaSMR venture is operating to give you the nuclear business’s engineers with the highest-resolution simulations of reactor programs so far and in flip assist advance the way forward for fission energy.

Coupling physics codes right into a extra robust complete

Nuclear reactors generate electrical energy through splitting uranium nuclei to liberate calories in gasoline rods. Water is heated through this calories liberate and turns into scorching sufficient to become steam that spins electricity-producing generators. ExaSMR integrates essentially the most dependable pc codes to be had for modeling the other physics of this operation, thereby making a toolkit that may expect a reactor design’s whole fission procedure. This toolkit contains Shift and OpenMC for neutron particle delivery and reactor depletion and NekRS for thermal fluid dynamics.

Even supposing these types of codes are already neatly established in science and business, the ExaSMR workforce has given them an entire HPC makeover. For the previous 6 years, researchers from ORNL, Argonne Nationwide Laboratory (Argonne), the Massachusetts Institute of Generation, and Pennsylvania State College (Penn State) had been optimizing the codes for the brand new technology of GPU-accelerated, exascale-class supercomputers, comparable to ORNL’s Frontier and Argonne’s upcoming Aurora.

OpenMC’s construction has been led through Paul Romano, and demanding GPU-optimization paintings for Aurora has been performed through John Tramm; Romano and Tramm are computational scientists at Argonne. Shift was once at first authored through Thomas Evans, workforce chief for ORNL’s HPC Strategies for Nuclear Programs Team, and is now optimized for Frontier.

Each codes use Monte Carlo strategies—computational ways that use vast numbers of random samples to calculate the possible results of fashions—to simulate how neutrons that transfer during the nuclear reactor have interaction with isotopes, comparable to uranium, and motive the fission occasions that create warmth within the reactor’s gasoline rods. The 2 codes additionally fashion how those isotopes evolve through the years, which predicts the reactor’s existence span.

NekRS—a computational fluid dynamics solver advanced through Elia Merzari, affiliate professor of nuclear engineering at Penn State—necessarily describes how the water will transfer and behave when heated through the reactor’s gasoline cylinders. The ExaSMR workforce’s ENRICO (Exascale Nuclear Reactor Investigative Code), additionally advanced through Romano, allows OpenMC and NekRS to engage.

“What we are doing in ExaSMR is a coupled physics simulation between the neutron delivery and the fluid dynamics—you could have those two physics codes which might be speaking from side to side to one another,” Hamilton mentioned. “The neutron delivery is telling you the place the warmth is generated. That warmth turns into a supply time period for the fluid dynamics calculation. The fluid dynamics tells you what temperature is as a consequence of that warmth supply. After which you’ll be able to modify the parameters within the simulation till each the neutron delivery and the fluid dynamics are in settlement.”

ExaSMR’s talent to correctly fashion in excessive decision the entire reactor procedure—thus predicting how a lot warmth the reactor’s fission occasions will produce, the facility of the reactor to switch that warmth to energy turbines, and the existence expectancy of all the gadget—supplies engineers with key insights to make sure the protection and potency in their reactor designs.

Making plans forward to steer clear of stumbling blocks

When the ECP and the ExaSMR venture began in 2016 to arrange tool apps and gear for exascale programs, the ones supercomputers did not exist but—now not even on paper. The workforce was once challenged with figuring out methods to easiest optimize codes for programs that had been years clear of being finalized.

“At the start of the venture, we did not even know precisely what the architectures of the exascale machines would seem like,” Hamilton mentioned. “It was once no doubt a problem to design our codes whilst taking a look forward with self belief that we’d be able to run successfully at the upcoming programs.”

The workforce confronted now not handiest the duty of coupling those separate codes for his or her new use-case situation of large-scale, high-fidelity reactor simulations but in addition the problem of adapting them to new computing architectures with but unknown processors. This uncertainty supposed pushing the bounds of compilers and tool applications through trying out use instances that had been a long way past what the tool have been examined for on the time—and it all started an ongoing procedure of continuing conversation.

“It required us to engage and iterate with the {hardware} distributors and the corporations that produce the tool to be sure that their merchandise can take care of our use instances. We have now researchers who’ve been in virtually day by day touch with people who find themselves writing compilers for the machines and looking to establish problems and enforce options which might be had to collect and run our codes,” Hamilton mentioned.

Luck eventually

The workforce’s interplay with distributors and builders paid off with really extensive enhancements within the strategies and algorithms utilized by the codes, yielding vast beneficial properties in functionality. With its preliminary runs on Frontier, ExaSMR blew previous the workforce’s speedup targets for its codes.

Shift carried out SMR simulations on as much as 8,192 nodes of Frontier and concerned simulating over 250 billion neutron histories in line with iteration. The functionality accomplished in those simulations is greater than 100× that of the baseline simulations carried out at the Titan supercomputer (i.e., the USA’s maximum robust supercomputer in 2016) and greater than double the functionality growth function of fifty× from Titan to Frontier.

NekRS carried out SMR simulations on as much as 6,400 nodes of Frontier, together with the most important reactor fluid-flow simulation carried out so far with over 1 billion spatial parts. The height functionality on Frontier displays a greater than 125× growth over corresponding baseline simulations carried out on Titan.

What is forward for ExaSMR?

Partnering with Westinghouse, which is a manufacturer of industrial nuclear energy era, the ExaSMR workforce implemented for a DOE Workplace of Complex Medical Computing Analysis Management Computing Problem grant. Westinghouse desires to judge the affect of the usage of higher-enrichment gasoline than what’s these days used of their reactors. Working ExaSMR on Frontier will permit them to accomplish high-fidelity simulations to expect how various kinds of fuels would carry out if utilized in a these days working reactor.

Likewise, Hamilton desires to use ExaSMR to present ARC applied sciences being explored within the energy business, comparable to the ones being advanced as a part of the DOE Workplace of Nuclear Power’s Complex Reactor Demonstration Program. This system works with industrial firms to assist accelerate the demonstration of complex reactors through offering preliminary investment. Two such reactors are slated for near-term deployment through 2027: X-energy’s Xe-100 pebble-bed reactor and TerraPower’s Natrium sodium-cooled rapid reactor. 5 further designs from Kairos, Westinghouse, BWX Applied sciences, Holtec Global, and Southern Corporate are ramping up for longer-term deployment.

Hamilton foresees ExaSMR turning into an indispensable software for firms which might be getting into a brand new technology of nuclear energy.

“Quite a lot of firms are exploring various kinds of reactor designs nowadays, and the high-performance, high-fidelity simulations that we are growing have a large number of interesting options for designers,” Hamilton mentioned. “It is not going, within the close to long term, that we’re going to have sufficient self belief in simulations that they’d absolutely exchange experiments, but when we will be able to scale back the collection of experiments which might be carried out, then there will also be massive monetary beneficial properties for those firms.”