DOE-NNSA

UQ and ML for Nuclear Forensics and Non-proliferation

We are part of the United States Department of Energy (US DOE) National Nuclear Security Administration (NNSA) Consortium for Nuclear Forensics (CNF). The CNF is a $25 million consortium over 5 years (2023-2028) led by the University of Florida. The CNF comprises 16 Universities and 7 National Laboratories that contribute to important research fields within nuclear forensics. This will be achieved by focusing on five main research areas: Rapid Turnaround Forensics, Advanced Analytical Methods, Ultrasensitive Measurement, Signature Discovery and Prompt Effects which will be achieved by using the combined expertise in radiochemistry, geochemistry, analytical chemistry, nuclear material science, shock physics, quantum-enabling sensing, high performance computing (HPC)/data science and training and education. Our goal is to develop inverse UQ methods for improving the quality of nuclear data for radiation transport simulations in nuclear forensics applications, as well as to combine scientific machine learning and UQ to improve the predictive modeling of fallout formation.

We are also part of the US DOE NNSA Office of Defense Nuclear Nonproliferation (DNN) Enabling Capabilities in Technology (TECH) consortium. TECH is also a $25 million consortium over 5 years (2025-2030) led by the The University of Tennessee, Knoxville. The TECH consortium has been formed to create new scientific knowledge in areas critical to DNN’s nuclear nonproliferation mission, to apply new knowledge toward the creation of new nuclear nonproliferation capabilities, and to generate the human capital needed for the DOE/NNSA laboratories to address emerging challenges in nuclear nonproliferation. TECH’s primary goals are to educate a talented group of professionals for roles in DOE National Laboratories and related sectors, and to promote innovation through collaboration with national lab partners. This initiative includes a comprehensive approach involving joint mentorship by faculty and lab experts, a broad research portfolio covering both fundamental and applied sciences, and an extensive educational and training program. Our goal is to apply inverse modeling and uncertainty quantification (UQ) to enhance the analysis of diagnostic measurements used in nuclear incident response. Data assimilation, machine learning, and inverse UQ methods will be integrated to develop approaches that can accurately infer the materials’ isotopic and chemical composition, geometric arrangement, shielding and reflection, neutron multiplication, and other properties, as well as the associated uncertainties.

Relevant publications:

1. Brady, C. and Wu, X. (2026). Nuclear Data Adjustment for Nonlinear Applications in the OECD/NEA WPNCS SG14 Benchmark – A Bayesian Inverse UQ-based Approach for Data Assimilation. Nuclear Science and Engineering.
   https://doi.org/10.1080/00295639.2025.2592175
2. Brady, C., Asadchykh, S., Xie, Z., and Wu, X. (2025). Preliminary Results on using Bayesian Inverse Uncertainty Quantification for OECD/NEA WPNCS Subgroup 14 Benchmark Exercise for Error Recovery and Experimental Coverage. In Proceedings of the 2025 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering (M&C 2025). Denver, Colorado, USA, April 27-30, 2025

Top of Page