New method maps neutron flux inside reactors using external measurements

Knowing the spatial distribution of neutron flux—a quantity that represents the number of neutrons moving through different regions of a reactor—is essential for safety, control and experiments, but installing sensors inside the reactor core can be challenging. High temperatures, intense radiation fields and limited space can shorten sensor life or make installation impractical, especially in advanced reactors and microreactors.

Researchers at the U.S. Department of Energy’s Argonne National Laboratory and Purdue University developed a new method to reconstruct the neutron flux distribution inside a reactor using measurements collected outside the fuel region. They call the approach Green’s-kernel Reconstruction from ACcessible Ex-core measurements, or GRACE. The work appeared in Nuclear Science and Engineering.

The idea is to treat the reactor a bit like a dark room that cannot be entered directly. Instead of placing sensors everywhere inside, GRACE uses signals measured around the edges to infer what is happening within. It is similar to estimating the shape of a campfire by feeling the heat around it rather than reaching into the flames.

The method combines reactor physics with machine learning. It uses a mathematical tool called a Green’s function, which helps describe how a change at one location in the reactor affects the neutron flux at another. The researchers trained a model to learn a reusable map that connects boundary measurements to the reactor interior. That means the system does not need a massive new training library for every operating case.

To test the method, the researchers used a detailed computational model of the Purdue University Reactor, known as PUR-1. After training GRACE on one control-rod setting, they tested its ability to infer neutron flux distributions for previously unseen settings. Using only simulated boundary data, the method reconstructed the three-dimensional neutron field with mean fractional errors below 5%.

The team also tackled a practical problem. Their reconstruction method needs not only neutron intensity at the boundary, but also how that intensity changes from place to place. Conventional ex-core instruments usually do not provide that additional information. So, the researchers designed a compact four-element sensor, called a quadrupole sensor, to measure neutron-flux gradients.

They built and tested an activation-based prototype at PUR-1 using gold wires. The sensor produced repeatable gradient patterns that were consistent with OpenMC simulation results. These results demonstrate that the neutron flux gradient measurements required by GRACE can be obtained experimentally, representing an important step toward its implementation in operating reactors.

Together, the results show a possible path toward reactor monitoring that relies entirely on external measurements. With further development, this strategy could help reactors that cannot support dense in-core instrumentation and may improve monitoring for compact or hard-to-access systems.