Unattended Monitoring and Machine Learning for Safeguarding a PUREX Reprocessing Facility

Safeguarding facilities with high throughput of nuclear material is challenging and often requires many attended measurements. Emerging technologies in the fields of machine learning hold promise in reducing the amount of in-person days required to safeguard such facilities while enhancing the effectiveness and efficiency of safeguards. Based on previous efforts, this work demonstrates the applicability of a machine learning method, the one-class support vector machine, to a PUREX reprocessing facility. This algorithm can make use of unattended process monitoring data to draw conclusions that a human might miss. This method is uniquely suited for safeguards applications in that it requires no knowledge of potential off-normal or facility upset conditions. Comparisons between this new machine learning method and traditional safeguards methods will be made. Results of this approach will be shown in the full paper with recommendations for pursuing other approaches, if applicable.