Tracking Of Individual TRISO-fueled Pebbles Through The Application Of X-ray Imaging With Deep Metric Learning

Modern accident tolerant pebble-bed reactor concepts implementing TRISO-fueled pebbles utilize on-line refueling, where fuel pebbles are continuously circulated through the reactor core. Up to 400,000 pebbles may be contained in a reactor core at one time, and, when combined with continuous, online-refueling and recirculation of these pebbles, necessitates development of material control and accountability capabilities for this non-traditional nuclear fuel form. Presently, no method exists to identify, or track, individual TRISO-fueled pebbles as they enter and exit the reactor core. In response to this need, a methodology has been developed to track individual TRISO-fueled pebbles by exploiting the unique distribution of the TRISO particles that is imprinted within each fuel pebble during the manufacturing process. By combining X-ray imaging and artificial intelligence, our method trains a deep convolutional neural network to learn a mapping from radiographic images to a lower-dimensional Euclidean space where distances provide a direct measure of the similarity between radiographic images of TRISO-fueled pebbles. Neural networks are complex mathematical functions that aim to mimic the biological processes of the brain. These functions contain a large number of coefficients, or parameters, that must be fitted to experimental data in order to achieve a desired output. Convolutional neural networks use image processing techniques, such as convolutional filters, to extract and compare details between images. This work presents the first application of neural networks towards the unique identification of individual TRISO-fueled pebbles. The results of this work show that our algorithm can achieve high classification accuracies in excess of 98%, and future work is planned to refine this technique to achieve as close to 100% accuracy as possible.