AI Accelerates Nuclear Reactor Design and Licensing

A professor at Brigham Young University has discovered a method to expedite nuclear reactor design and licensing through the use of AI. Traditional processes require 20 years and $1 billion for licensing a design, followed by an additional five years and $5 to $30 billion for construction. AI has the potential to reduce this timeline by a decade, resulting in significant savings.

The design of nuclear reactors involves intricate physics, encompassing neutron movement, heat transfer, and materials. These elements are closely interrelated, rendering the process data-intensive and time-consuming. AI can alleviate this burden, facilitating quicker and more cost-effective designs.

Professor Memmott and his team have developed machine learning algorithms to forecast temperature profiles and optimize reactor parameters. Their AI model can geometrically optimize designs at a much faster pace than conventional methods. For instance, their AI took just two days to determine an optimal reactor shield design, a task that a real-world company took six months to complete.

Although human oversight remains crucial for final decisions and safety assurance, AI substantially cuts down on design time. With increasing energy demands, faster and more affordable nuclear reactor designs could be essential for meeting future power requirements without emissions.