Neutron Tomography of Spent Fuel Casks

Abstract

Dry casks for spent nuclear fuel (SNF) ensure the safe storage of SNF and provide radiation shielding. However, the presence of the thick casks encompassing several layers of steel and concrete makes inspection of the SNF a challenging task. Fast neutron interrogation is a viable method for the nondestructive assay of dry storage casks. In this study, we performed a Monte Carlo simulation-based study associated with a machine-learning-based image reconstruction method to verify the content of SNF dry storage casks. We studied the use of neutron transmission and back-scattered measurements to assess the potential damage to fuel assemblies or fuel pin diversion during transportation of dry casks. We used Geant4 to model a realistic HI-STAR 100 cask, MPC-68 canister and basket, and GE-14 fuel assembly irradiated by a D-T neutron generator. Several bundle diversion scenarios were simulated. The angular distribution of the neutrons scattered by the cask was used to identify the diversions inside the fuel cask. A fuel bundle with at least 75% of its pins removed can be identified with a drop in the back-scattered signature larger than $$2\sigma$$ compared with a fully loaded scenario. We combined an iterative reconstruction algorithm with a convolutional neural network (CNN) to obtain a cross-sectional image of the fuel inside the cask. The proposed imaging approach allows locating the position of a missing fuel bundle with at least 75% of the pins removed when performing tomographic imaging of a canister with an overall scan time of less than two hours, when using a commercial neutron generator with a source strength of $$10^{10}$$ n/s in the 4 $$\pi$$ solid angle.