Tracking of individual TRISO-fueled pebbles through the application of X-ray imaging with deep metric learning

Modern pebble-bed reactor concepts utilizing TRISO-fueled pebbles use on-line continuous refueling, where fuel are continuously circulated through the core. Presently, no method exists for tagging, identification, and tracking of individual as they enter exit This leaves room improvement in nuclear material accountability safeguards pebbles. work presents a methodology to identify by exploiting unique distribution TRISO-coated particles, which is imprinted during manufacturing process, within By combining X-ray imaging deep learning, our learns mapping from radiographic images compact Euclidean space distances provide direct measurement similarity pebble radiographs. A convolutional neural network trained optimize image triplet loss implemented enforce greater distance between mappings that different dataset consisting 1,250 Monte Carlo N-Particle (MCNP) Transport simulations generated training testing learning algorithm, achieves an accuracy 93.49% ± 9.35% first transforming with Gaussian blur transformations it 98.70% 2.60%.