AUTOMATING SPENT FUEL DETECTION WITH FUSED DCVD AND GET DATA

摘要

Within the safeguards regime, spent fuel monitoringhas long been implemented to verify facility declarations.As the number and size of facilities under safeguardsincrease, so does time spent on highly repetitive spent fuelmonitoring, making it a natural fit for automation.However, due to international proliferation concerns,automation must be robust and provide estimates ofconfidence in the results. Cerenkov Viewing Devices(CVDs) are widely implemented for spent fuel monitoring,but their single defect detection ability is limited. In thisresearch, we augment CVD measurements with GammaEmission Tomographer (GET) data to increase defectdetection above the level of either detector methodindividually. Here we present the development andimplementation of a Bayesian data fusion algorithm forclassification of individual fuel rods as defects or nondefectsacross two burn-up and cooling time scenarios.Results show a nearly 75% single defect detectioncapability with a 10% false positive rate for 17x17 PWRfuel assemblies. The Bayesian framework also enablescalculation of uncertainties associated with eachclassification. These results also show a moderate abilityto generalize across both high burn-up and short coolingtimes as well as low burn-up and long cooling times.