Could AI Speed Up the Design of Nuclear Reactors?

A professor at Brigham Young University "has figured out a way to shave critical years off the complicated design and licensing processes for modern nuclear reactors," according to an announcement from the university. "AI is teaming up with nuclear power."The typical time frame and cost to license a new nuclear reactor design in the United States is roughly 20 years and $1 billion. To then build that reactor requires an additional five years and between $5 and $30 billion. By using AI in the time-consuming computational design process, [chemical engineering professor Matt] Memmott estimates a decade or more could be cut off the overall timeline, saving millions and millions of dollars in the process - which should prove critical given the nation's looming energy needs.... "Being able to reduce the time and cost to produce and license nuclear reactors will make that power cheaper and a more viable option for environmentally friendly power to meet the future demand...." Engineers deal with elements from neutrons on the quantum scale all the way up to coolant flow and heat transfer on the macro scale. [Memmott] also said there are multiple layers of physics that are "tightly coupled" in that process: the movement of neutrons is tightly coupled to the heat transfer which is tightly coupled to materials which is tightly coupled to the corrosion which is coupled to the coolant flow. "A lot of these reactor design problems are so massive and involve so much data that it takes months of teams of people working together to resolve the issues," he said... Memmott's is finding AI can reduce that heavy time burden and lead to more power production to not only meet rising demands, but to also keep power costs down for general consumers... Technically speaking, Memmott's research proves the concept of replacing a portion of the required thermal hydraulic and neutronics simulations with a trained machine learning model to predict temperature profiles based on geometric reactor parameters that are variable, and then optimizing those parameters. The result would create an optimal nuclear reactor design at a fraction of the computational expense required by traditional design methods. For his research, he and BYU colleagues built a dozen machine learning algorithms to examine their ability to process the simulated data needed in designing a reactor. They identified the top three algorithms, then refined the parameters until they found one that worked really well and could handle a preliminary data set as a proof of concept. It worked (and they published a paper on it) so they took the model and (for a second paper) put it to the test on a very difficult nuclear design problem: optimal nuclear shield design. The resulting papers, recently published in academic journal Nuclear Engineering and Design, showed that their refined model can geometrically optimize the design elements much faster than the traditional method. In two days Memmott's AI algorithm determined an optimal nuclear-reactor shield design that took a real-world molten salt reactor company spent six months. "Of course, humans still ultimately make the final design decisions and carry out all the safety assessments," Memmott says in the announcement, "but it saves a significant amount of time at the front end.... "Our demand for electricity is going to skyrocket in years to come and we need to figure out how to produce additional power quickly. The only baseload power we can make in the Gigawatt quantities needed that is completely emissions free is nuclear power." Thanks to long-time Slashdot reader schwit1 for sharing the article.

Read more of this story at Slashdot.