Year
2024
Abstract
| Tristructural-isotropic (TRISO) fuel is currently one of the most mature fuel types for pebble bed reactor (PBR) designs. Fuel units in PBRs are made in the form of a billiard-size ball, which contains thousands of randomly-distributed TRISO fuel particles of 500-1000 µm diameter surrounded by graphite or silicon carbide. In PBRs, TRISO-fueled pebbles can be re-introduced into the reactor core several times to thoroughly burn the fuel. Reliable techniques to individually tag and identify fuel pebbles are needed to maintain accountability of special nuclear material and gain a detailed understanding of pebble history, burnup, and structural integrity throughout the pebble’s lifespan. The identification technique should be non-destructive, fast, and robust against measurement noises to meet the PBR’s operational constraints.The 3D spatial distribution of TRISO fuel kernels is a unique feature associated with each individual fuel pebble and can be used for the purpose of identification. X-ray CT scans of a mockup fuel sample with an attenuation coefficient close to actual TRISO pebbles were acquired with an industrial CT scanner. 3D image reconstruction and segmentation algorithms were developed to accurately segment TRISO particles and extract their unique 3D distribution, forming a point cloud. The image processing took approximately 41 s for a single pebble. The uncertainty associated with the measured kernel position was below 0.5 mm and approximately 90% of the kernels were correctly identified.The subsequent task is to identify a randomly rotated point cloud with noises from a library of 100,000 point clouds, each representing a unique pebble. Rotation-invariant features were extracted from the 3D point clouds and by comparing the features, the number of candidates were reduced from 100,000 to 100. Afterwards, a point-cloud registration algorithm was employed to align the input point cloud with each candidate, and the candidate that yielded the minimal difference after alignment was returned. The identification algorithm was tested on randomly-rotated pebbles and an identification accuracy of 100% was achieved in 10,000 tests, with positional noises up to 1.5 mm and kernel outliers up to 30%. It took 7 seconds to identify a single rotated pebble from a dataset of 100,000 pebbles. Overall, the identification procedure was performed in 50 seconds, which is a time frame compatible with reactor operation. The developed identification method can be applied to TRISO-fueled units of any shape, such as the cylindrical TRISO fuel elements in HTGRs. |