AI/ML Tested in Nuclear Reactor Control
A Nuclear Energy Agency task force is spotlighting the use of AI and machine learning in nuclear engineering. Purdue’s PUR-1 reactor is being used as a testbed for AI-driven control, anomaly detection, and predictive maintenance, offering a parallel for deploying AI in regulated, safety-critical aerospace environments.
- The PUR-1 reactor at Purdue University became the first and only nuclear reactor in the United States to be licensed by the Nuclear Regulatory Commission (NRC) with a fully digital instrumentation and control system in 2019. This upgrade from legacy analog systems enables the collection of real-time data from over 2,000 signals at one-second intervals, providing the high-fidelity data needed for machine learning models. - A key component of the project is a "digital twin" of the PUR-1, an AI-powered simulation that receives live data from the physical reactor's sensors. This allows for the safe testing of algorithms without impacting the reactor's actual operation; one such test on the digital twin demonstrated an algorithm that could predict power fluctuations with
