A Sound Evaluation Of Safeguards Verification Measurements Based On Statistical Analysis And Visualization Of Historical Nda Data

Year
2021
Author(s)
Andrea Cerasa - European Commission- Joint Research Centre
Marc Couland - European Commission, DG ENER
Domenico Perrotta - European Commission- Joint Research Centre
Alice Tomanin - European Commission, DG ENER
Cristina Versino - European Commission, Joint Research Centre
File Attachment
a136.pdf695.77 KB
Abstract
Euratom Safeguards performs sets of conformity assessment activities in the form of accountancy and physical verifications in installations using or owning nuclear material. The on-site inspection process makes extensive use of independent measurement systems to determine the flows, quantities or characteristics of nuclear material. The deployment of an increasing number of unattended systems associated with remote data transmission offers an opportunity to improve inspection effectiveness and efficiency by systematic data analysis.At present, dedicated IT applications are used to evaluate acquired measurement data. The results of their evaluations, being reported by individual inspection, are only representative of a situation over a limited period in time. However, Euratom Safeguards long-term strategy foresees an assessment of risk based on the results of past verification activities and the evaluation of confidence factors capturing limits and uncertainties encountered during the process.The evaluation of measurement data can directly support those two concepts by including, for instance, a deeper analysis of historical trends or an automated assessment of the measurement systems performances. To this end, the development of dedicated analysis packages associated to a centralized measurement data repository was initiated.The objective is to provide inspectors with data analysis tools that combine robust statistical techniques and time series analysis aimed to detect anomalous patterns in the measurements. The prompt detection of anomalies such as structural breaks or outliers together with a suitable visualization of these patterns may indeed support the identification of discrepancies in inspections’ findings. In this paper, we describe an exploratory study of possible analysis tools applied to Non-Destructive Assay data designed to provide a solid basis for the development of a structured and sound statistical framework for the analysis of inspections’ outcomes.