Search

This paper describes the creation and description of a nuclear fuel isotopics dataset for irradiated fuel salt from a Molten Salt Reactor (MSR). The dataset has been created using the Monte-Carlo particle transport code, Serpent 2.1.32 (released February 24, 2021) and the calculation code SOURCES 4C (released October 09, 2002). The dataset comprises isotopic mass densities of 1362 isotopes (including fission products and major and minor actinides) and their corresponding contributions to decay heat, gamma activity, and spontaneous fission rates computed by Serpent 2.1.32 as well as overall neutron emission rates from spontaneous fission and (ɑ, n) reactions computed by SOURCES 4C. These quantities are computed for a model MSR core utilizing a full-core 3D model of the Seaborg Compact Molten Salt Reactor (CMSR) . The dataset spans a wide range of values of burnup (BU), initial enrichment (IE) and cooling time (CT) over which the above-mentioned quantities are reported. The structure of the dataset includes isotopic mass densities (in g/cm3), followed by isotope-wise contributions to decay heat (denoted by suffix ‘DH’ and reported in Watts), gamma photon emission rates (denoted by suffix ‘GS’ and reported photos per second), and spontaneous fission rates (denoted by suffix ‘SF’ and reported in fissions per second). In addition to these columns, the data also includes total neutron emission rates from 1) spontaneous fission (denoted by ‘SF’ and reported in neutrons per second per cm3), and 2) (ɑ, n) reactions (denoted by ‘AN’ and reported in neutrons per second per cm3). In total, the dataset has 310,575 rows of different combinations of fuel burnup, initial enrichment, and cooling time (BIC) values spanning the realistic possible range of these parameters. The dataset is made available for public use in a comma-separated value file that can be easily read using one of the numerous popular data analysis tools such as NumPy or Pandas.

This paper describes the methodology used to create a fuel data library comprising safeguards-relevant quantities that may be useful for verification of spent nuclear fuel (SNF) produced by simulating a concept Molten Salt Reactor (MSR). The Monte-Carlo particle transport code, Serpent2 and the calculation code SOURCES 4C were used to compile this fuel data library. The data library is based on the Compact Molten Salt Reactor (CMSR) concept being developed by Seaborg Technologies (based in Copenhagen, Denmark). The library includes data such as nuclide mass densities for a total of 1398 nuclides (in g/cm3), as well as total decay heat production (denoted by suffix the ‘TOT_DH’) in Watts, total gamma photon emission rates (denoted by the suffix ‘TOT_GS’) in photos per second, and the total activity (denoted by suffix ‘TOT_A’) in Becquerel. Lastly, the data also includes total neutron emission rates from 1) spontaneous fission (denoted by ‘SF’ and reported in neutrons per second per cm3), and 2) (ɑ, n) reactions (denoted by ‘AN’ and reported in neutrons per second per cm3) for the fuel salt. These quantities are reported for a range of burnup-initial enrichment-cooling time (or collectively known as, BIC) parameters. The resulting fuel data library is an extension of a previously published data library for the same reactor concept but with one significant change. The current library is based on a more realistic model of the CMSR involving movement of gaseous and volatile fission products (GFP and VFP) from the core via an Off-Gas System (OGS). The dataset is made available for public use in a compressed binary format as an HDF5 (or Hierarchical Data Format) file that can be parsed using data analysis tools such as Pandas., Order of authors in the list of papers of Vaibhav Mishra's thesis: V Mishra, E Branger, S Grape, Zs Elter, S Mirmiran

When analyzing and assessing properties of spent nuclear fuel (SNF) such as radionuclide inventories, the power history of the fuel during its time spent inside the reactor core plays an important role. This information can be very useful in the field of nuclear safeguards wherein a safeguards inspector can use it to verify the fuel properties such as burnup, initial enrichment and cooling time (or collectively termed as the “BIC” set of variables). However, such information may often be unavailable to the safeguards inspector or the level of detail in the available information may be lacking. Therefore, when analyzing SNF for various purposes (such as for safety, safeguards and back-end purposes), the power history of the fuel is most often disregarded altogether and the inspectors only look at the fuel BIC. If the power history-level information is considered, it is not uncommon to make simplifying assumptions about how the fuel is burned in the reactor. In this work, we perform an exploratory analysis of fuel cycle data from two PWR units of the Ringhals nuclear power plant in Sweden. The said analysis describes the variation in the number of cycles, cycle lengths, downtimes et cetera in order to develop a simplified yet representative model of irradiation that may be used to construct synthetic data libraries. Furthermore, we look into impact of changes in the power history on the nuclide inventories of key gamma emitters and isotopes responsible for decay heat in the SNF of the fuel in three different irradiation scenarios. Our results show that in most cases with fuels that are considerably long-cooled, it is acceptable and even preferred to use a simplified power history over an idealized or a representative irradiation model obtained from exploratory analysis of the fuel cycle data. However, for short-cooled fuels, using a simplified or even an idealized history is less preferable over modeling the detailed power history of the fuel due to the presence of sh

The Differential Die-away Self Interrogation (DDSI) instrument was researched for many years under the Next Generation Safeguards Initiative Spent Fuel effort. Later a prototype instrument was manufactured and used to make non-destructive measurements of spent nuclear fuel in the Swedish Central Interim Storage Facility for Spent Nuclear Fuel (Clab) in Sweden in 2018. Results of DDSI research, based on either simulations or measurement time, have indicated that the instrument could successfully be used to draw safeguards-relevant conclusions about spent nuclear fuel.In this work we investigate how well the modelled response of the DDSI instrument, based on Serpent and MCNP simulations, corresponds to measured data of 17x17 pressurised reactor fuel. We also studied repeatability, i.e. to what extent repeated measurements on the same fuel assembly gave consistent results. We also investigated the dependence of tau on the selected time window. The results show that tau values determined from measurement data are consistently higher than tau values determined from simulations, and that the magnitude of tau is dependent on the choice of time window. We also note that tau is relatively insensitive to positioning in the DDSI instrument.

Master Thesis Project

Material test reactors have an extended use in irradiation testing of novel nuclear fuel materials and the fuel behavior in off-normal conditions. The performance of the nuclear fuel is examined in in-pile and out-of-pile post-irradiation examinations (PIEs), e.g., using Gamma Emission Tomography (GET). GET is a nondestructive assay that images the internal spatial distribution of gamma-emitting nuclides built up in the fuel due to irradiation. Since GET can be performed close to the reactor and without intrusion in the fuel object, it can potentially speed up the data generation from PIE in irradiation testing. The performance metrics of GET devices can be identified regarding time requirements, noise in the reconstructed image, signal-to-background ratio, and spatial resolution. However, these are complicated to determine, partly due to inherent trade-offs between the metrics themselves, partly because they depend on the fuel activity and its spectrum (i.e., object dependent), and, finally, on the GET setup and its configuration. This work proposes a structured methodology for optimizing the collimator design for a new generation of GET tomography setups, intending to improve spatial resolution by one order of magnitude: from the millimeter scale to the hundred-micron scale. The conflicting performance metrics are determined based on the controllable parameters of the GET setup and the uncontrollable parameters of an anticipated fuel object, able to provide a signal-to-background ratio above 100. The trade-off between the performance remaining metrics is then visualized by a Pareto approach, where dominated solutions are rejected. Finally, constraints on noise level and measurement time are used to find the optimal spatial resolution. Two GET setups are presented using the outlined method. Firstly, to upgrade the tomography test bench BETTAN at Uppsala University, a new segmented HPGe detector was planned to be tested using low-activity fuel rod mock-ups. Second

The Digital Cherenkov Viewing Device (DCVD) is one of the instruments used by safeguards inspectors to verify spent nuclear fuel in wet storage. The DCVD can be used for partial defect verification, where the inspectors verify that 50% or more of an assembly has not been diverted. The methodology is based on comparing the measured Cherenkov light intensity with a predicted intensity, calculated with operator information. Recently, IAEA inspectors have encountered fuel assemblies for which systematic deviations between predictions and measurements could be observed, indicating that the prediction model did not take into account all sources of Cherenkov light production. One contribution to the Cherenkov light intensity that is frequently omitted is the contribution from beta decays, where energetic electrons exit the fuel material and enter the water with sufficient energy to directly produce Cherenkov light. The objective with this work was hence to study beta contributions and evaluate whether that could be the cause of discrepancy between predictions and experimental data. By simulating the beta contribution for fuel assemblies where the discrepancy was experimentally observed, it was determined that beta decays were the cause. The fuel assemblies had fuel rods with relatively small radii, thin cladding, a short cooling time and an irradiation history that resulted in a relatively large beta contribution for assemblies that had a comparatively low burnup. Therefore, the beta contribution was significant, and caused 10-40% of the total Cherenkov light intensity. By including the beta contributions in the predictions, the RMSE of the deviation between prediction and measurement could be reduced from 20.7% to 11.6% for the available measurement data. The results highlight that the beta contribution can be significant and should be taken into account for accurate predictions.

Gamma-ray spectrometry using collimated detectors is a well-established examination method for irradiated nuclear fuel. However, the feasibility of examining a particular nuclide of interest is subject to constraints; the peak must be statistically determinable with the desired precision and the total spectrum count rate in the detector should not cause throughput issues. Methods were assembled for gamma spectrum prediction to optimize instruments for gamma emission tomography and to enable a priori feasibility evaluation of determination of single peaks of irradiated nuclear fuel. The aim was to find reliable results (~10% accuracy) regarding total spectrum and peak count rates with faster computation time than a full-Monte Carlo approach. For this purpose, the method is based on depletion calculations with SERPENT2, a point-source kernel method for the collimator response, a rig response matrix and a detector response matrix, both computed with MCNP6. The computational methodology uses as input the fuel properties (dimensions, materials, power history, and cooling time), and the instrumental setup (collimator and detector dimensions and materials). The prediction method was validated using measured data from a high-burnup, short-cooled test fuel rodlet from the Halden reactor. Absolute count rates and ratios of characteristic peaks were compared between predicted and measured spectra, showing a total count rate overestimation of 7% and discrepancies between 2-20% for the single peaks (same order of magnitude of the uncertainty). This level of agreement is deemed sufficient for measurement campaigns planning, and the optimization of spectroscopic instruments for use in gamma scanning and tomography of nuclear fuel.

The interest of the scientific community in alternate nuclear fuel cycles such as that of the Molten Salt Reactors (MSRs) has grown and waned in the past. In the recent decades however, there has been a resurgence of interest in them owing to the ever-rising demand for sustainable energy and growing concerns of climate change. However, these reactors are yet to enter the stage of commercial operation. The MSR family form a broad spectrum of reactor concepts with unique fuel cycles, designs, online reprocessing options et cetera and some of these have progressed from the phase of conceptualization to deployment as experimental reactors. Since MSRs differ significantly from the traditional Light Water Reactors (LWRs), both conceptually and from an operational point of view they come with new and unique challenges. On the issue of safeguards, the main challenges arise from the lack of experience in the industry on handling bulk fuel material such as highly radioactive and corrosive molten fuel salts containing nuclear material in place of traditional fuel items like spent nuclear fuel (SNF) assemblies. There is also a marked lack of measurement techniques and instruments that could be brought into use for enforcing safeguards on molten salt systems. There are novel ideas that have been proposed to increase the proliferation resistance of the fuel cycle in MSRs and deter the misuse of fuel material.Among the key MSR designs that have been developed at the Oak Ridge National Laboratory (ORNL) in the past was the Molten Salt Demonstration Reactor (MSDR). The MSDR concept was a 750 MW successor to the much smaller Molten Salt Reactor Experiment (MSRE) that was once operated at ORNL. While the MSDR was never built, its design, operation, safety, security and safeguards-related issues were studied extensively at several national labs in the United States. In the current study, we aim to model the irradiation of fuel salt used in the MSDR. The reactor specifications are used

The Differential Die-Away Self-Interrogation (DDSI) instrument detects neutrons in coincidence and is sensitive to the content of fissile material and neutron absorbing material in a fuel assembly. Its response to spent nuclear fuel with varying properties was investigated under the Next Generation Safeguards Initiative (NGSI) and the efforts included simulations as well as the manufacturing of a prototype instrument that was successfully tested in the field. Due to the interest of applying machine-learning techniques to support safeguards verification of spent nuclear fuel, we have investigated how to speed up the predictions of the early die-away time τ from the DDSI instrument for a large spent nuclear fuel inventory. One way to do this has been to develop a parametrisation function for τ as a function of the fuel parameters initial enrichment, burnup and cooling time. To assess the validity of the parameterisation function, the sensitivity of the DDSI response to various modelling parameters such as e.g. boron concentration, fuel pin geometry and operational history has been investigated. In this work, we summarise the recent efforts made to resolve these questions.

Under the Next Generation Safeguards Initiative, several different nuclear safeguards measurement techniques were studied. One of them was the Differential Die-AwaySelf-Interrogation technique, and the research showedthat its early die-away time τ was proportional to the fuel assembly multiplication and thus sensitive to the fissilecontent of the fuel assembly under assay. A prototype instrument was later built and tested in the field, and the measurements showed that the instrument could be usedsuccessfully in the field. This work builds on previous efforts, and systematically studies the effects of assumptions about the fuelproperties (such as its dimensions) and its irradiationconditions in the reactor, on the Rossi-Alpha Distribution(RAD) and τ. The motivation is twofold, firstly to betterunderstand if and what impacts such assumptions haveon the RAD and τ, and secondly to investigate how wellthe simulation model used to estimate the RAD and τ isable to generalize to other fuel types and irradiation conditions than those modelled. 20 spent nuclear fuel assemblies currently residing in theSwedish interim storage for spent nuclear fuel weremeasured by the prototype DDSI instrument. Theassemblies were modelled using Serpent2 and MCNP6 inthis work. Fuel depletion calculations were performed assuming both a standard irradiation cycle and the actualirradiation history as provided by the operator. Fuel properties and irradiation conditions were also modified and their effect studied. Based on the simulated DDSI instrument response inMCNP6, the RADs were created and τ determined. The analysis shows that each modelling assumption on itsown affects both the RAD and the τ value. However, some of the individual effects work in opposite direction and cancel out when considered at the same time. For this reason, the default model is considered to be a good and valid approximation of the more complex one and results are expected to generalize well.

In the current Digital Cherenkov Viewing Device (DCVD) measurement methodology, the DCVD is aligned over the centre of a fuel assembly when measuring emitted Cherenkov light. Due to the collimation of light, and due to the lifting handle of PWR fuel assemblies covering the fuel periphery, the DCVD is more sensitive to partial defects near the fuel assembly centre than near the periphery. Here, we investigate the sensitivity of the DCVD for detecting partial defects for different instrument alignments. By performing measurements at both the centre and near the assembly periphery, more accurate measurements near the periphery can be obtained. DCVD images were simulated for different partial defect scenarios with 30% of the fuel rods removed or replaced with low, medium or high-density rods. Simulations were run with different DCVD alignments, and the Cherenkov light distribution in the images were quantitatively analysed and compared to simulated images for a fuel assembly without defects. The simulation results were also compared with measurements of intact spent fuel assemblies. The simulations show that the local Cherenkov light intensity deviation due to a partial defect is not sensitive to the alignment. Hence, the current methodology is robust, and will not benefit from measuring at different alignments. Regarding the signal-to-noise ratio, combining measurements at different alignments can improve the measurements. However, the improvement is modest, and for the DCVD it may be preferred to simply use the current methodology and make longer measurements. For future autonomous Cherenkov measuring systems, combining images can be a way of improving the quality of the measurements.

Nuclear safeguards verification of spent nuclear fuel (SNF) is imperative to ensure peaceful use of nuclear material. To verify the correctness and completeness of operator declarations, non-destructive assay (NDA) measurements of gamma and neutron radiation from the SNF play a central role. Verification of fuel based on such measurements is done routinely by safeguards inspectors and is also expected to be conducted prior to preparation of SNF for final disposal. Traditionally, SNF verification has been carried out by analyzing data from a single NDA instrument at a time. In this study, we compare the performances of different machine learning algorithms in their ability to make predictions of the SNF parameters such as fuel burnup (BU), initial enrichment (IE), and cooling time (CT). Predictions were made based on simulated signatures such as gamma-ray intensities from individual radionuclides, the total Cherenkov light intensity, and the parameterized differential die-away time (tau). In this work, multiple machine learning algorithms have been trained and tested on a set of simulated data containing 596,181 fuel samples providing a broad range of these three parameters to encompass the majority of the spent nuclear fuel worldwide. Additionally, the resilience of the machine learning algorithms on noisy data was evaluated. The results show that the non-linear methods can provide highly reliable predictions of SNF parameters. In nearly all situations assessed in this work, we have found that shallow learning methods have a clear advantage over deep learning models investigated in the present study. We also found that shallow learning methods such as k-Nearest Neighbors outperform other tree-based methods as well as neural networks at noise levels above 5%.

The paper describes a data library containing material composition of spent nuclear fuel. The data is extracted from burnup and depletion calculations with the Serpent2 code. The simulations were done with a PWR fuel pin cell geometry, for both initial UO2 and MOX fuel load for a wide range of initial enrichments (IE) or initial plutonium content (IPC), discharge burnup (BU) and cooling time (CT). The fuel library contains the atomic density of 279 nuclides (fission products and actinides), the total spontaneous fission rate, total photon emission rate, activity and decay heat at 789,406 different BU, CT, IE configurations for UO2 fuel and at 531,991 different BU, CT, IPC configurations for MOX fuel. The fuel library is organized in a publicly available comma separated value file, thus its further analysis is possible and simple.

Before encapsulation of spent nuclear fuel in a geological repository, the fuels need to be verified for safeguards purposes. This requirement applies to all spent fuel assemblies, including those with properties or designs that are especially challenging to verify. One such example are quivers, a new type of containers used to hold damaged spent fuel rods. After placing damaged rods inside the quivers, they are sealed with a thick lid and the water is removed. The lid is thick enough to significantly reduce the amount of the gamma radiation penetrating through it, which can make safeguards verification from the top using gamma techniques difficult. Considering that the number of quivers at storage facilities is foreseen to increase in near future, studying the feasibility of verification is timely. In this paper we make a feasibility study related to safeguards verification of quivers, aimed at investigating the gamma and neutron radiation field around a quiver designed by Westinghouse AB and filled with PWR fuel rods irradiated at the Swedish Ringhals site. A simplified geometry of the quiver and the detailed operational history of each rod are provided by Westinghouse and the reactor operator, respectively. The nuclide inventory of the rods placed in the quiver and the emission source terms are calculated with ORIGEN-ARP. The radiation transport is modeled with the Serpent2 Monte Carlo code. The first objective is to assess the capability of the spent fuel attribute tester (SFAT) to verify the content for nuclear safeguards purposes. The results show that the thick quiver lid attenuates the gamma radiation, thereby making gamma radiation based verification from above the quiver difficult.

This paper describes a methodology to identify partial defects in modelled spent nuclear fuel using passive gamma spectroscopy data. A fuel library, developed with Serpent2, was used to calculate the material composition of spent nuclear fuel. Two fuel configurations were investigated in this work; one where the fuel assembly configuration was intact and one where 30% of the fuelrods were substituted with stainless steel rods in a random configuration. Emission and detection of gamma radiation from 134Cs, 137Cs and 154Eu was simulated using a model of a passive gamma spectroscopy measurement station mimicking the Clab measurement station in Sweden. A simple HPGe detector model was implemented, and its detector efficiency was assessed using a range of different source energies. Realistic total gamma attenuation coefficients were calculated using the XCOM database.The modelled estimates of detected full-energy peak counts were then used in a Principal Component Analysis in order to investigate whether it was possible to distinguish between intact and partial defect fuel assemblies or not. The results showed that partial defects could be identified using the simultaneous analysis of all three peak intensities, and that the ability to do so increased when only gamma emission energies from 154Eu were considered.

Before encapsulation of spent nuclear fuel in a geological repository, the fuels need to be verified fors afeguards purposes. This requirement applies to all spent fuel assemblies, including those with properties or designs that are especially challenging to verify. One such example are quivers, a new type of containers used to hold damaged spent fuel rods. After placing damaged rods inside the quivers, they are sealed with a thick lid and the water is removed. The lid is thick enough to significantly reduce the amount of the gamma radiation penetrating through it, which can make safeguards verification from the top using gamma techniques difficult. In this paper we make a first feasibility study related to safeguards verification of quivers, aimed at investigating the gamma and neutron radiation field around a quiver using a simplified quiver geometry. The nuclide inventory of the rods placed in the quiver is calculated with Serpent and Origen-Arp, and the radiation transport is modeled with Serpent. The objective is to assess the capability of existing non-destructive assay instruments, measuring the gamma and/or neutron radiation from the object, to verify the content for nuclear safeguards purposes. The results show that the thick quiver lid attenuates the gamma radiation, thereby making gamma-radiation based verification from above the quiver difficult. Verification using neutron instruments above the quiver, or gamma and/or neutron instruments on the side may be possible. These results are in agreement with measurements of a BWR quiver using a DCVD, performed by the authors.

•Before placing spent nuclear fuel in in a geological repository, they will be characterized and their declared properties will be verified. •We have created large library of modelled spent nuclear fuel (SNF) assemblies and estimated their activity of gamma-ray emitting fission products, the early die-away time τ and the Cherenkov light intensity. •We have used Random Forest regression to evaluate the capability to determine the fuel parameters initial enrichment (IE), burnup (BU) and cooling time (CT) using data from non-destructive assay (NDA) techniques

The differential die-away self-interrogation (DDSI) instrument developed and built in Los Alamos National Laboratory (LANL) is being considered for verification before final disposal. One of the signals from this instrument, the early die-away time, has been shown to be proportional to the multiplication of the spent fuel assembly. Full-scale simulations of the instrument response using MCNP are time consuming. This may become a problem in cases when the instrument response to a large number of fuel assemblies is required, such as in the case of training machine learning models. In this paper, we propose a parametrization of the early die-away time as a function of initial enrichment (IE), burn-up (BU) and cooling time (CT), for intact PWR spent fuel assemblies. The parametrization is calculated from a dataset of 1040 simulated PWR spent fuel assemblies with fuel parameters in the range of IE=2-5%, BU=15-60 GWd/tU and CT=5-70 years. The simulations are done using Serpent2 for the depletion calculation and MCNP6 for the neutron transport and detection in the DDSI. It was found that the CT dependence can be decoupled from the BU and IE dependence, and that it follows an exponential decay. The BU and IE dependences have been fitted with several different functions, and the best fit was chosen based on the chi-square value. The determination of the die-away time using the parametrization has been tested on a separate dataset, resulting in a root mean square error (RMSE) of 0.6 µs (the early die-away time ranges from 28 µs to 84 µs). A description of this work is given in the paper together with details on the choice of parametrizing function, and qualitative arguments for that choice.

The operator declarations of spent nuclear fuel assemblies are routinely verified for nuclearsafeguards purposes to ensure their non-diversion and integrity. Many countries considerthe possibility of eventually placing the fuel in a geological repository and prior to this it isexpected that the fuel assemblies need to be carefully characterized and verified, for instanceusing gamma spectroscopy measurements. It can be expected that in connection to suchactivities, verifying parameters such as burnup (the energy outtake from the nuclear fuel),cooling time (the time the fuel spent outside the reactor after operation), initial enrichment(the amount of fissile material in the fuel before operation) and integrity (whether pins insidethe assembly have been manipulated) may become an important task of nuclear safeguardsinspectors, since discrepancies may indicate unauthorized activities at the facilities. Ideally,the verification should be done with non-destructive assay systems.Passive gamma spectroscopy provides a robust and relatively simple method to analyze spentfuel since the characteristics of spent nuclear fuel strongly affect the gamma radiation emittedfrom the fuel (Jansson, 2002). Lately, passive gamma tomography has also become a possiblemethod to characterize and analyze properties of spent nuclear fuels (Mayorov et al., 2017).In both cases, gamma radiation is measured around the fuel assembly from a distance usingone or more collimated detectors with spectroscopic capabilities. Of great interest is thedetector efficiency of these systems, i.e., the ratio between number of detected particles andnumber of particles emitted by the source. The detector efficiency is the product of thegeometric efficiency (probability that emitted particles reach the detector region) and theintrinsic efficiency of the detector (probability that the particles are detected)

Material test reactors have an extended use in irradiation testing of novel nuclear fuel materials and the fuel behavior in off-normal conditions. The performance of the nuclear fuel is examined in in-pile and out-of-pile post-irradiation examinations (PIEs), e.g., using Gamma Emission Tomography (GET). GET is a nondestructive assay that images the internal spatial distribution of gamma-emitting nuclides built up in the fuel due to irradiation. Since GET can be performed close to the reactor and without intrusion in the fuel object, it can potentially speed up the data generation from PIE in irradiation testing. The performance metrics of GET devices can be identified regarding time requirements, noise in the reconstructed image, signal-to-background ratio, and spatial resolution. However, these are complicated to determine, partly due to inherent trade-offs between the metrics themselves, partly because they depend on the fuel activity and its spectrum (i.e., object dependent), and, finally, on the GET setup and its configuration. This work proposes a structured methodology for optimizing the collimator design for a new generation of GET tomography setups, intending to improve spatial resolution by one order of magnitude: from the millimeter scale to the hundred-micron scale. The conflicting performance metrics are determined based on the controllable parameters of the GET setup and the uncontrollable parameters of an anticipated fuel object, able to provide a signal-to-background ratio above 100. The trade-off between the performance remaining metrics is then visualized by a Pareto approach, where dominated solutions are rejected. Finally, constraints on noise level and measurement time are used to find the optimal spatial resolution. Two GET setups are presented using the outlined method. Firstly, to upgrade the tomography test bench BETTAN at Uppsala University, a new segmented HPGe detector was planned to be tested using low-activity fuel rod mock-ups. Second