Virginia Commonwealth University Richmond, Virginia, United States
With the increase of interest in the use and commercialization of High-Assay Low Enriched Uranium (HALEU) new methods for maintaining compliance with the Nuclear Regulatory Commission (NRC) will likely be needed. More methods are necessary to confirm enrichment levels to help in the commercialization and regulation of HALEU nuclear-fuel forms. This paper introduces the data generation meant to be fed into a machine learning model that will learn to perform a nondestructive assay for the differentiation between HALEU and highly enriched uranium (HEU). The spectra needed will be modeled using Monte Carlo N-Particle Transport Code (MCNP). Three detector types and various fuel forms will be represented in the model. Since civilian power fuels' enrichment for purposes must remain below 20% of the isotope of uranium-235. There are two critical U235 enrichment levels of interest 19.75% which is consider HALEU and 21% which is considered HEU. The objective is to bolster nuclear non-proliferation efforts by providing a cheaper, and more effective means of monitoring the boundary between these two materials. Traditional methods face challenges in distinguishing between HALEU and HEU due to subtle differences in their gamma radiation signatures. Machine learning, artificial intelligence and neural networks offer a potential solution by leveraging data-driven algorithms to analyze complex radiation spectra and improve classification accuracy.
