A new patent granted to Oak Ridge National Laboratory researchers redefines how nuclear systems are validated, offering a faster path to next-generation nuclear energy.

Traditional nuclear system validation requires repetitive physical testing and subjective assessments that can delay timelines and increase project costs. ORNL's pioneering methodology applies entropy - a measure of uncertainty used in telecommunications - to identify which validation tests add new knowledge and where uncertainty still remains.

By using machine learning to identify high-value experiments, this approach can eliminate redundancies and define clear boundaries for AI and digital twin models, or virtual models that adapt with real-time data. This will support license applications and provide regulators with greater confidence in advanced reactors.

"This work establishes entropy as the Rosetta stone for developing trust in computational models," ORNL's Ugur Mertyurek said. "By identifying which data or experiments truly matter, we can cut licensing timelines down to a fraction of their current length while ensuring safety."

The patent was co-developed with Purdue University's Hany Abdel-Khalik.