Rotation-Invariant Rapid TRISO-Fueled Pebble Identification Based on Feature Matching and Point Cloud Registration

Abstract

Pebble bed reactor (PBR) relying on TRISO-fueled pebbles is one of the most promising Gen-IV reactor designs because of intrinsic safety and thermal efficiency. Fuel pebbles flow through PBR's core and the identification of individual pebbles exiting the core will be beneficial to improve safeguards and fuel management. We propose a pebble identification method that is fast, accurate, robust, and applicable to PBRs containing hundreds of thousands of pebbles. The identification relies on the internal distribution of TRISO fuel particles, which is a unique feature of each pebble. We experimentally demonstrated that X-ray CT can extract the particle distribution with high accuracy. We then applied the algorithm to identify a single pebble in a data set of 100,000 pebbles achieving 100% identification accuracy in 90,000 tests with the presence of arbitrary rotations and measurement noises. The average time to identify one pebble is below 50 s, compatible with PBR operation.

36 pages, 18 figures

References in corpus (6)

• scikit-image: Image processing in Python
• Non-convex Optimization for Machine Learning
• A comprehensive survey on point cloud registration
• Quantitative Imaging and Automated Fuel Pin Identification for Passive Gamma Emission Tomography
• Feasibility of Neutron Coincidence Counting for Spent Fuel
• Boron Coated Straw-based Neutron Multiplicity Counter for Neutron Interrogation of TRISO Fueled Pebbles