Simulations help drive innovation, but real-world data still key for nuclear

Computer simulations are increasingly important tools for nuclear power innovators, but even the most sophisticated machine learning techniques must lean on real-world data when building untested advanced reactors, experts at leading national laboratories say.

Advanced Test Reactor at Idaho National Laboratory (INL). (Source: INL)

Related Articles

Traditionally, researchers have used empirical models to simulate nuclear energy systems, though this requires a vast amount of experimental data, and especially designed systems and conditions that can be costly and time consuming.

Most of the previous generation of nuclear power stations were designed and built using paper-and-pencil calculations based on real-world observations, but advances in computer processing power, and modelling and simulation tools, means today a different approach can be applied.

However, hopes that a new, advanced nuclear plant, operating outside the parametric range of previous plants, can be designed, built, and tested solely through historical data, and artificial intelligence and machine learning tools – an idea rooted in the program to maintain the U.S.’s aging nuclear arsenal – are likely to be dashed when faced with regulators’ demands.  

“Simulations augment the value of the data from test reactors but cannot be a replacement for a test reactor,” Executive Director for the Versatile Test Reactor (VTR) Project at the Idaho National Laboratory (INL) Kemal Pasamehmetoglu.

INL’s VTR – which was stripped of funding this year by Congress but, Pasamehmetoglu hopes, will be reinstated at a later date – is earmarked to begin operations in the latter part of the decade and aims to provide intense fast spectrum neutron fluxes that are used to simulate prototypical conditions or conduct accelerated neutron damage irradiation studies for many advanced reactor designs.  

“What the simulations can do is enhance the data, and influence the type of experiments and data, we get from the test reactor, and that can accelerate the time period between testing and research, and commercialization. However, simulations will not replace the need for the basic data in the foreseeable future,” he says.

Improved modelling and a data-driven understanding of how the reactors work over long periods have brought enormous improvements in the efficiency of older plants and life-extensions that have sometimes doubled the initial life expectancy of the plant.

Source: U.S. Nuclear Regulatory Commission

Modelling the nuclear arsenal

For modelling and simulation to be useful tools, they must also submit to rigorous validation and verification, and uncertainty quantification (VVUQ), says Dave J Kropaczek,  Director of Nuclear Industry Technology Innovation at Oak Ridge National Laboratory (ORNL).

Much of the idea that nuclear technology could be examined and researched using only historical data and computational models has arisen from the Accelerated Strategic Computing Initiative (ASCI), Kropaczek says.

The U.S. Department of Energy (DOE) established the ASCI in 1995 to ensure the safety and reliability of the country’s nuclear weapons arsenal while fully adhering to the Comprehensive Test Ban Treaty.

“The ASCI initiative was formed to assure the reliability of the nuclear stockpile, as nuclear weapons decay over time and, in the absence of testing, can I use simulation to fill those gaps?” says Kropaczek.

However, people make mistakes in coding and two analysts using the same code with different inputs and model choices for a complicated first-of-a-kind application often produce different answers, he says.

In the absence of experiments, it is these uncertainties that will prompt regulators to insist on better VVUQ.

“You can get there with modelling and simulation, but you’ve got to do your homework, you have to do testing, and then you propagate uncertainties, you bound those uncertainties, and that’s what the regulator wants to see,” he says.

“When the regulator sees you’ve done your homework, you’ve showed them your code, the physics models, the geometry meshing, the input and model uncertainties, all those have been reviewed, and you’ve told them the amount of radiation that can be released under the worst-case, severe accident scenario and assure public safety, then the regulator will be satisfied.”

Back to first principles

There have been major technological advances made since the first nuclear reactor, the Chicago Pile-1, was assembled on a rackets court under the stands of the Alonzo Stagg Field football stadium in the University Chicago in 1942, and nuclear engineers now have no need to go back to first principles when designing a new reactor.

There remain some uncertainties in the nuclear cross section, but neutron transport in a reactor is something that can be predicted with a high level of confidence.

The flow of fluids in a reactor and its coolant circuit is less predictable and thermal hydraulics are less mappable than neutron transport, even with the most cutting-edge simulations.

“When it comes to behavior of materials and fuels, we’re even less close to be able to do a first principle model,” says Hussein S. Khalil, Director of Argonne National Laboratory (ANL) Nuclear Engineering research program.

“To understand the challenge, you should keep in mind the quantum mechanical basis for material properties … there’s a lot of progress being made in calculating material properties, based on quantum mechanics, but we’re not there yet.”

Taking previously used designs for non-light water systems and reiterating them for the present day is also not guaranteed without testing since the final goals, such as economical generation of electricity, would not be reflected the original design, Khalil says.

“When you’re facing the need to develop a system that can be economical at producing electricity or heat, you can’t just rebuild a system that was developed for experimentation or testing purposes," he says. 

"You really have to pay attention to optimizing the design so it is compact for its power capacity, that it uses readily available materials, it can be manufactured, it can be efficiently operated and maintained … in the end it’s about being able to produce energy cost effectively to compete with other energy forms.”

By Paul Day