This project attempts to automate the differentiation of High Assay Low-Enriched Uranium (HALEU) and Highly Enriched Uranium (HEU) through low-resolution gamma spectroscopy by utilizing machine learning techniques to aid in the identification of 238U and 235U spectra. This process currently requires a trained spectroscopist to identify the differences of each material’s characteristics, which making it costly, but with new advances in artificial intelligence this process can become more automated. With the increase in the United States’ interest in advanced nuclear power reactors, the Inflation Reduction Act invests $700 million to support the development of a domestic supply chain for high assay low-enriched uranium (HALEU)[1], and with it requires the United States to increase its safeguards capabilities in order to efficiently and cost effectively regulate these materials. The purpose of this project is to create a way for new safeguards and inexpensive measurement techniques in the regulations of HALEU materials, which border on the line of highly enriched uranium (HEU), according to the nuclear regulatory commission and the International Atomic Energy Agency standards. By utilizing gamma spectra, a machine learning model can be trained to differentiate nuclear materials that lie closely on the border between HEU and low enriched uranium (LEU), such as HALEU. Comparisons of machine learning methods for automated gamma-ray spectroscopy have been done and have shown great promise as a possible solution to be applied in this situation, and there are even existing open-source packages available that can differentiate between isotopes using artificial neural networks (ANNs).
Year
2024
Abstract