Texas A&M researchers advance AI-driven nuclear reactor simulations

Texas A&M Engineering researchers, funded by a 2026 U.S. Department of Energy grant, are developing AI methods to accelerate nuclear reactor safety simulations without sacrificing accuracy.

Texas A&M Engineering researchers led by Dr. Yang Liu, with collaborators from Purdue University, Idaho National Laboratory, Argonne National Laboratory, and Oklo Inc., are developing AI-enhanced nuclear reactor simulations. The project, awarded a U.S. Department of Energy Nuclear Energy University Program (NEUP) grant in 2026, integrates machine learning with the Multiphysics Object-Oriented Simulation Environment (MOOSE) framework. This approach uses differentiable physics to enable continuous model updates during real-time simulation of reactor transients—addressing a key limitation where traditional machine learning models become fixed after training. The work aims to reduce computational time for reactor modeling while maintaining simulation accuracy.

The research represents a potential pathway to faster, more responsive nuclear reactor safety analysis. By enabling continuous refinement during simulation rather than static training, this method could eventually lower the computational barriers to real-time reactor monitoring. For energy abundance, this may help future-proof nuclear power as a scalable, low-carbon solution by making reactor safety assessments more efficient and responsive.

This remains early-stage research, not yet deployed technology. The project’s future-dated publication (August 2026) and NEUP grant focus on innovation development—not operational systems—mean the approach has not yet reached practical application. What matters now is whether this research can overcome current computational constraints to enable safer, more responsive nuclear energy systems later this decade.

Source: news.engineering.tamu.edu