A team of researchers in BYU’s Department of Chemical Engineering recently discovered a new way to design nuclear power structures using artificial intelligence.
Chemical engineering professor Matthew Memmott explained that in order to get a license to build a nuclear reactor, the design must be approved by the U.S. Nuclear Regulatory Commission. The biggest challenge? A nuclear reactor cannot be built until it receives approval, so a team of engineers must work on calculations for around 20 years and spend about a billion dollars just to get that license to build it.
However, Memmott and his team have researched a much more practical solution — using machine learning.
“The idea is to shorten it, make it safer, cheaper and faster to get the nuclear power, rather than take 20 years to get the license,” professor Matthew Memmott said.
Since so much of those 20 years is spent doing calculations, the research team thought about how machine learning can take large amounts of data and calculate it much faster that humans can.
Memmott shared how the team tested this idea by attempting to recreate a local nuclear company’s shield with an AI algorithm they created. The algorithm managed to almost perfectly match the company’s shield in just two days.
“It was slightly less effective, but it was also cheaper and lighter,” Memmott said. “They were like, this is amazing because within a couple of days we were able to do the same work that it took a team of engineers six months to do.”
The team recently published their findings in a paper on using AI to optimize nuclear shielding, focusing on improving the efficiency and cost-effectiveness of the designs using machine learning. The research process was a quick one, with results completed in March and the paper accepted for publication in May.
Memmott explained that engineers will ultimately have to make the design decisions, but AI can help with directing in the right direction.
“I am very optimistic,” graduate student Edward Mercado said. “Machine learning has proved to be a very valuable tool for design and optimization. My hope is that a similar machine learning framework can be used to optimize nuclear fuel consumption so that less waste is generated.”
Mercado is a chemical engineering student at BYU who helped with running simulation models to provide feedback for the algorithm to find the best geometric design.
“AI or machine learning has the potential to greatly reduce the computational costs and requirements involved in the reactor design process,” Mercado said. “If we can get simulations done in a fraction of the time, that means we can accelerate the development of next generation nuclear reactors to provide clean and reliable energy.”
In terms of next steps, Memmott shared that their AI algorithm could also be applied to electricity grid, to optimize renewable energy production.
“Applying AI to other areas is something we’re going to expand to see if we can figure out what else can be improved,” Memmott said.
The research team also has a project with the Pacific Northwest National Laboratory to use machine learning to optimize nuclear waste vitrification.
“By optimizing the amount of waste that can be loaded into each canister, it saves years from the waste processing,” John Hedengren, another chemical engineering professor involved in this research, said. “At $2 billion/year to operate the vitrification facility in Washington state, there is a big incentive to improve the efficiency.”
Memmott said he wants people to recognize that nuclear energy is not something to fear, but actually has potential for solving significant energy problems.
“People are worried about the waste and the cost, but things like that are being solved,” Memmott said. “Shift the focus away from, ‘Nuclear is the expensive, not desirable option’ to, ‘With the right tools, we can do it really well.’”