This automatic system tracks the movement of TRISO-fueled pebbles through the reactor core at nuclear power plants. Nuclear power plants provide 55 percent of America's clean energy, producing 809 billion kWh of electricity in 2019. Pebble bed reactors have gained increased interest over the past few years, using tri-structural isotropic (TRISO) pebbles, designed to withstand high temperatures. Modern pebble-bed reactor concepts use online refueling, where fuel pebbles are continuously circulated through the reactor core. Presently, pebble identities are not tracked because there is no viable method to tag individual pebbles without altering the manufacturing process.

Researchers at the University of Florida seek to address this by developing an automatic system to track and identify individual TRISO-fueled pebbles at the entry and exit of a reactor core through the combination of X-ray imaging with deep learning. Maintaining pebble identity enables measurements of pebble transit time through the core, validation of computational physics models, and improvements concerning nuclear fuel safety and material accountability.

Application

Automatic identification of individual TRISO-fueled pebbles using X-ray imaging and deep learning to track pebbles as they enter and exit the reactor core

Advantages

Provides a method to identify and tag individual TRISO-fueled pebbles
Requires no modification to the current manufacturing process
Enables a method to accurately measure pebble transit time through the reactor core to validate computational physics models and determine if any pebbles are retained in the core for unexpectedly long periods of time

Technology

This system takes advantage of an intrinsic feature of the manufacturing process for TRISO-fueled pebbles where the compaction of thousands of fuel particles into a hard sphere results in an arbitrary and distinct distribution of particles, or pebble fingerprint. This pebble fingerprint can be visualized and digitized through radiation imaging, and further extracted and compared to a collection of unique fingerprints using neural networks. This makes it possible to identify and tag individual fuel pebbles. Each pebble undergoes X-ray imaging before insertion into the reactor core. These images are stored in a database. When a pebble exits the core, it is imaged again. A deep learning-based classification algorithm then uses these output images to compare and match the pebble fingerprint to a pebble stored in this database.