SMART Lab will lead or participate in three major research projects supported by the U.S. Department of Energy (DOE), spanning generative and agentic AI, scientific machine learning, advanced reactor modeling and simulation, uncertainty quantification, and AI-assisted nuclear licensing.
Together, these projects advance the SMART Lab’s mission of integrating state-of-the-art artificial intelligence and machine learning with nuclear engineering to accelerate the development and deployment of advanced nuclear energy technologies.
1. SHIELD: Secure Human-in-the-loop Intelligence for Engineering and Licensing Deployment of Advanced Nuclear Systems
Led by Dr. Yang Liu at Texas A&M University, SHIELD was selected as part of DOE’s inaugural Genesis Mission, a national initiative to use artificial intelligence, advanced computing, and scientific infrastructure to accelerate scientific discovery and technology development. Texas A&M is leading four university projects for the Genesis Mission.
In collaboration with Argonne National Laboratory, the SHIELD project will develop a human-in-the-loop, AI-enabled framework to automate key components of nuclear reactor engineering analysis and licensing workflows. The system will integrate AI agents with reactor modeling and simulation tools to conduct safety analyses and generate supporting engineering and regulatory documentation, while maintaining expert review and approval at critical stages. The project will demonstrate the framework for both advanced sodium-cooled reactor systems and large light-water reactors.
The project directly addresses a major challenge in advanced nuclear deployment: reducing the time and human effort required for engineering analysis and licensing while maintaining the rigor, traceability, and safety required for nuclear applications.
2. Integrating Graph Neural Operator with MOOSE Framework for Enhanced Transient Modeling and Real-Time Data Assimilation
SMART Lab also received a $1 million DOE Nuclear Energy University Program (NEUP) award, led by Dr. Liu, to develop a new scientific machine learning framework for advanced reactor simulation. The project brings together collaborators from Pennsylvania State University, Argonne National Laboratory, Idaho National Laboratory, and Oklo Inc.
The project will develop a GNO-MOOSE differentiable hybrid solver that integrates Graph Neural Operators directly into the MOOSE multiphysics framework. The approach is designed to accelerate computationally intensive, long-duration reactor transient simulations while preserving physics-based predictive capabilities. It will also couple the solver with an Ensemble Kalman Filter for real-time data assimilation, enabling reactor models to continuously incorporate observational data and update their predictions.
The framework will be demonstrated on challenging sodium fast reactor and molten salt reactor transient problems, with the broader goal of enabling efficient and real-time predictive simulation capabilities for next-generation nuclear systems.
3. Comprehensive Uncertainty Quantification for High-Fidelity Multiphysics Simulation of Sodium-Cooled Fast Reactors
SMART Lab will also participate in a second $1 million DOE NEUP project, led by Dr. Xu Wu at North Carolina State University, to establish a comprehensive uncertainty quantification methodology for high-fidelity multiphysics reactor simulations. Dr. Liu serves as a collaborator together with researchers from NC State, Argonne National Laboratory, and Idaho National Laboratory, with technical advisors from the U.S. Nuclear Regulatory Commission and TerraPower.
The project will integrate Bayesian inverse uncertainty quantification, quantitative model validation, scientific machine learning, and high-fidelity modeling and simulation into a unified framework. Using sodium-cooled fast reactor applications and the DOE NEAMS ecosystem as demonstration cases, the research aims to establish trustworthy uncertainty estimates for safety-relevant predictions, particularly in regimes where experimental validation data are limited.
These three projects highlight SMART Lab’s growing research portfolio at the intersection of AI, scientific machine learning, advanced reactor modeling, digital engineering, and nuclear safety and licensing. They also strengthen the lab’s collaborations with DOE national laboratories, universities, industry, and the nuclear regulatory community as we work toward trustworthy AI-enabled technologies for the next generation of nuclear energy systems.
