Summary:
-
AI in Nuclear Design: Brigham Young University professor Matt Memmott has developed a method using AI to streamline the complex design and licensing processes of modern nuclear reactors. Traditionally, designing and licensing a new nuclear reactor in the U.S. takes about 20 years and $1 billion, with an additional five years and $5-$30 billion to build it. Memmott’s AI approach could reduce the overall timeline by a decade or more, significantly cutting costs.
-
Efficiency and Optimization: Memmott’s research demonstrates that AI can replace some of the extensive thermal hydraulic and neutronics simulations required in reactor design. By training machine learning models to predict temperature profiles and optimize geometric reactor parameters, AI can expedite the design process. Memmott and his team built and tested a dozen machine learning algorithms, refining them to handle preliminary data sets effectively.
-
Proven Results: The AI models showed remarkable efficiency. In one instance, an AI algorithm designed an optimal nuclear-reactor shield in two days, a task that took a molten salt reactor company six months. This research, published in the journal Nuclear Engineering and Design, underscores AI’s potential to handle massive design problems more swiftly than traditional methods. While final design decisions and safety assessments remain human tasks, AI’s role at the front end could be transformative, aiding in meeting future electricity demands with faster and cheaper nuclear power development.
!