ADVANCES IN MACHINE LEARNING FOR SAFEGUARDING A PUREX REPROCESSING FACILITY

摘要

The average IAEA inspector spends approximately 100 days in the field per year verifying activities at various nuclear facilities. Detecting diversion of nuclear material from bulk handling facilities, such as a reprocessing plant, is a goal that can prove to be both challenging and resource intensive as it often relies on destructive analysis of numerous samples taken across the facility. Emerging technologies in the fields of machine learning and data science hold promise in reducing the in-person days to safeguard facilities while enhancing the effectiveness and efficiency of safeguards. This work builds on previous efforts to develop a machine learning framework built on unattended Non-Destructive Assay to safeguard a PUREX facility. This work shows that under certain conditions machine learning based approaches can exceed the performance of traditional safeguards approaches.