Analysis of Unattended Monitoring System Data Using Knowledge Generation Software

摘要

Unattended monitoring systems can reduce the need for on-site human presencewhile still assuring the proper safeguards of nuclear material. However, such systems generate large quantities of raw sensor data that then have to be related to known or declared activities and material accountancy records. We previously described a concept and technical approach to analyzing this data, based on the use of finite-state machine process models. We have now applied this technique to the analysis of sensor data from unattended monitoring systems at two facilities: an integration laboratory used to simulate material handling facilities in the DOE complex and a bunker used to simulate semi-static storage of high-value assets. The analysis of the integration laboratory data focused on verifying the occurrence of declared activities, even in the presence of 'noise' due to people walking around the facility. The analysis of the bunker data considered questions of data integrity and system integrity including how to modify process analysis results based on the quality of the data. The paper will describe the models used to perform the analyses and the results obtained. We will also discuss how additional data could strengthen the conclusions and discuss the implications for monitoring system design.