PGET Monte Carlo simulations using Serpent

Since 2017, over 100 spent nuclear fuel assemblies at the Finnish nuclear power plants have been imaged with the Passive Gamma Emission Tomography (PGET) device in preparation of the implementation of PGET in the safeguards infrastructure of the Finnish geological repository. In order to increase understanding of the PGET method and guide its further development, we have recently implemented PGET in Serpent, a widely-used neutron and photon transport Monte Carlo simulation code. We will discuss the major aspects of this implementation and illustrate the usefulness of the simulations with a few examples. The PGET device as used in the measurements (which was developed under the guidance of IAEA and is approved for safeguards inspections) was implemented in a very realistic way based on its technical drawings. The simulation produces sinograms in user-defined energy windows as well as the uncertainty on these sinograms. Tomographic images are then reconstructed using the exact same algorithm as used for the measured data. A dedicated variance reduction scheme was implemented, increasing the computational efficiency by about a factor of 30. The simulation of the PGET response at one angular measurement position for 1 billion primary photons takes a few hours on a single 40-core node. The 1-sigma uncertainty in the highest intensity sinogram pixels is about a few percent. Aiming at improving the imaging of VVER-440 assemblies, we have simulated assemblies containing one or a few missing fuel rods or having only one emitting rod (the other rods being present but not emitting) in various well-chosen places, configurations that are not accessible in practice. The single-emitting rod results show in great detail those parts of the sinogram that contain most of the information for the particular rod position. How this information might be used for obtaining better images, especially of the central region of a fuel assembly, will be discussed.