Introduction
Nuclear power plants (NPPs) have been reported to be one of the most effective ways for both reducing greenhouse gas emissions and satisfying electricity demand [1–3]. A pressurized water reactor (PWR), a common reactor type deployed at NPPs, typically uses nuclear fuel enriched with 3% to 5% 235U to generate electricity through nuclear fission. During this process, spent nuclear fuel (SNF) is generated as high-level radioactive waste [4, 5]. SNF contains various fission fragments such as 235U and 239Pu, which can be diverted for the purposes of legitimate military application or nuclear terrorism. Since SNF emits a large amount of heat and radiation even after being withdrawn from the nuclear reactor vessel, it is generally stored and managed in the following order: wet storage, dry storage, interim storage, and deep geological repository [6]. The loss or theft of SNF during transportation between storage sites could lead to harmful radiation exposure to the public and biosphere. Certainly, it is necessary to develop accurate inspection techniques for non-proliferation of nuclear materials.
The International Atomic Energy Agency (IAEA) has developed and proposed inspection techniques such as gamma emission tomography (GET) for detection of partial-defects within SNF. The IAEA and associated European research groups have developed a semiconductor-detector-based GET instrument named passive GET. It was officially authorized for use in SNF inspection in 2017, and was routinely applied to inspection of SNF during transportation to dry storage in 2021 [7–10]. Further, in a previous study, our research team developed the gadolinium aluminum gallium garnet (GAGG) scintillation-crystal-based GET instrument named Yonsei single-photon emission computed tomography (YSECT), and its performance was experimentally validated using a mock-up of unirradiated nuclear fuel in air [11–19]. However, the GET instruments have a crucial limitation, specifically low inspection accuracy in the central region, due to the high-density of nuclear fuel rods and the reduced detection efficiency due to self-absorption. To solve this problem, the energy window for obtainment of the tomographic image was optimized, and thereby, utilizing gamma rays above 662 keV, the tomographic image quality could be improved over the entire SNF region. However, for conjugation of such high-energy gamma rays in image reconstruction, an advanced technique is required, due specifically to the low linear energy transfer. The present study developed, and this paper proposes, a multisensor-based YSECT instrument for enhancement of partial-defect detection accuracy in the central region of SNF. The image quality in the SNF’s central region was assessed according to the crystal material in Monte Carlo (MC) simulations.
Materials and Methods
1. Optimization of Scintillation Crystal Using Monte Carlo Simulation
In this study, the GEometry ANd Tracking version 4 (GEANT4) toolkit was employed for detection efficiency optimization [20]. The scintillation crystal was modeled in water so as to define the appropriate material for image quality improvement in the central region of SNF (Fig. 1). The shape of the scintillation crystal was defined as trapezoidal, similarly to a previous study using the conventional YSECT instrument. Also, the length, thickness, and height (short axis) were determined to be 45, 3, and 40 (3) mm, respectively.
The Oak Ridge Isotope GENeration-Automatic Rapid Processing (ORIGEN-ARP) program was utilized to implement the source term of 10 years cooled Westinghouse (WH) 14×14-type SNF [21]. To acquire the corresponding gamma-ray energy spectrum, the burn-up was fixed at 45,000 megawatt-days per metric ton of uranium (MWd/MTU), in agreement with the average burn-up in the Republic of Korea (40,000–50,000 MWd/MTU). The uranium concentration and total uranium mass were determined as 4.1% and 430.8 kg, respectively [22, 23]. The combustion cycles were set at three cycles. The burn-ups for each cycle were determined as 15,105 (431 days), 15,202 (422 days), and 14,693 (433 days) MWd/MTU, respectively [24]. The shape of the source term was implemented as a plane and placed 29.92 cm from the scintillation crystal.
As explained above, a high-density scintillation crystal is required to detect high-energy ( > 662 keV) gamma rays. Since properties such as density, energy resolution, and light yield vary for each type of scintillation crystal material, it is necessary to select a crystal that is suitable for the application conditions and gamma-ray energy. Due to the high-level background radiation in the SNF inspection context, a reasonable energy resolution is required in order to reduce the effect of both the background radiation and gamma-ray scatter. Consequently, the representative scintillation crystal materials listed in Table 1 were selected. The gamma-ray energy spectrum was obtained at 662 keV, and the net counts within the ± 2.5% energy window were compared to select the most appropriate scintillation crystal for the proposed YSECT instrument.
2. Image Quality of Proposed Scintillation Crystal with YSECT
To obtain the tomographic image, the YSECT instrument and the WH 14 × 14-type SNF were implemented on the GEANT4 toolkit as depicted in Fig. 2. The YSECT comprises eight detection modules, each of which contains a tungsten parallel collimator and 46-channel GAGG scintillation crystals. A total of 179 nuclear fuel rods was arranged within the WH 14 × 14-type SNF, each nuclear fuel rod being composed of ZIRLO cladding, a helium gap, and a nuclear fuel pellet of UO 2. In consideration of simulation efficiency and the field of view of the YSECT, the length of the WH 14× 14-type SNF was determined to be 40 cm. An identical source term to that described in the previous section was used, and its length was fixed at 1 cm.
To validate the performance of the multisensor-based YSECT instrument, tomographic images were obtained, the quality of which was compared based on the signal-to-noise ratio (SNR) in the Equation (1):
where RMSSignal indicates the root mean square of the signal pixel intensity, and RMSNoise refers to the root mean square of the noise pixel intensity.
Results and Discussion
1. Evaluation of Scintillation Crystals for Energy Spectrum and Detection Efficiency
Fig. 3 plots the gamma-ray energy spectrum of each of the scintillation crystal materials. The shapes of the spectra of GAGG and CsI(Tl) exhibited a resemblance owing to their comparable properties of energy resolution and light yield. On the other hand, CdWO4 had a relatively low energy, which rendered distinguishing energy peaks below 500 keV difficult. However, it was found to have superior detection efficiency based on its high density. As for PbWO4, its major energy peaks, specifically 662, 873, 1,274, and 1,489 keV, could not be discriminated due to its poor energy resolution.
To quantitatively validate the detection efficiency for each scintillation crystal material, the net counts within the energy window, considering each energy resolution, were scrutinized and compared as depicted in Table 2 and Fig. 4. PbWO4 was excluded since its energy peaks were not distinguishable. CdWO4 was found to have a detection efficiency twice as high as that of GAGG across all of the major gamma-ray energy ranges on account of its high-density despite relatively poor light yield. As for CsI(Tl), it was assessed as having the lowest net counts due to its low density.
Subsequently, the deposited energy distributions were compared for GAGG, CsI(Tl), and CdWO4, as shown in Fig. 5. To that end, the maximum deposited energy was normalized from 0 to 1. The comparison indicated that, across all of the scintillation crystals, most of the gamma-ray energy had been delivered to the surface. Relative to CdWO4, GAGG’s and CsI(Tl)’s relative maximum deposited energies were calculated as 0.71 and 0.64, respectively. However, high-detection efficiency does not always guarantee improved tomographic image quality. Therefore, to enhance the tomographic image quality in the central region of SNF, energy window optimization studies were carried out in order to utilize gamma-ray energy higher than 662 keV. To enhance the detection efficiency of such high-energy gamma rays, it is necessary to employ a high-density scintillation crystal, and the optimal scintillation crystal material, in consideration of both detection efficiency and density, was determined to be CdWO4.
2. Comparison of Image Quality According to Scintillation Crystal
As shown in Fig. 6, tomographic images were obtained to validate the applicability of the multisensor-based YSECT. The source term was implemented at the innermost nuclear fuel rod, and the applied energy windows of GAGG and CdWO4 were ± 2.5% and ± 5.5%, respectively. In our previous study, an iterative reconstruction algorithm was developed to improve tomographic image quality. Since this algorithm extremely limits the image size, the inherent characteristics of the instrument could not be considered. Therefore, in the present study, a filtered back-projection algorithm with Ram-Lak filter was employed to reconstruct the sinogram into a tomographic image. Streak artifacts were noticeable in both images due to the high density of the nuclear fuel rods. The maximum pixel intensity of the CdWO4 image was higher than that of GAGG by a factor of 2. The SNR of the CdWO4 image was calculated as 7.85, which was higher than that of the GAGG image, which had an SNR of 6.18. Conversely, the standard deviation in the noise region for each tomographic image was determined to be 0.72 and 1.15, respectively. These results indicate that the poor energy resolution of the CdWO4 scintillation crystal increased the effects of gamma-ray attenuation and scatter within the energy window.
Fig. 7 shows the tomographic images with the 1,275 keV energy window for each scintillation crystal, as obtained in the same manner as Fig. 6. The SNR of the GAGG and CdWO4 images were calculated as 8.21 and 9.71, respectively. Despite low net counts, the SNR of the tomographic images with the 1,275 keV energy window was estimated to be 1.2 times higher than that of the images with the 662 keV energy window. Since high-energy ( > 662 keV) gamma rays were used to image the cross-section of the SNF, the extent of streak artifacts due to the high-density of the nuclear fuel rods was reduced in both images, as shown in Figs. 8 and 9, which depict the tomographic images for each scintillation crystal and energy window when the source was positioned at the outermost nuclear fuel rod. By the same token, the maximum pixel intensity of the CdWO4 image was higher than that of the GAGG image owing to its high density. On the other hand, salt and pepper noise, also called impulse noise, was shown in both images. These results imply that both scintillation crystals could not completely absorb the energy of the 1,275 keV gamma rays since the components of the YSECT instrument such as the collimator and scintillation crystal, had been optimally designed to detect 662 keV gamma rays.
3. Single vs. Multisensor–Based YSECT
To validate the applicability of the multisensor-based YSECT instrument, tomographic image quality was compared between it and the single-sensor–based YSECT, as illustrated in Fig. 10. The synthesis image of the multisensor-based YSECT was produced by summation of the tomographic images obtained with each of the scintillation crystals and energy window. In that image, the streak artifacts and salt and pepper noise were remarkably reduced. Furthermore, the SNR of the synthesis image was determined to be 9.85, which was higher than that of the tomographic image obtained with the single-sensor-based YSECT by a factor of 1.59. And note, in this regard, that in the case of on-site testing, which is marked by high attenuation and scatter conditions, it would be necessary to reduce the noise level caused by the instrument’s inherent characteristics. Based on these results, therefore, we believe that the proposed YSECT instrument can enhance inspection accuracy for partial defects arising in PWR-type SNF.
Conclusion
In this study, a multisensor-based GET instrument named YSECT was optimized to improve partial defect inspection accuracy, and its applicability was evaluated using MC simulation. For detection of high-energy ( > 662 keV) gamma rays, the proper scintillation crystal material was determined to be CdWO4 given, specifically, its density, energy resolution, and light yield properties. The applicability of the optimized YSECT instrument was validated using tomographic images for two patterns, where the source was positioned at the innermost or outermost nuclear fuel rod, respectively. The SNR of the tomographic images obtained with CdWO4 was determined to be 2 times and 1.2 times higher than that of the images using GAGG for the 662 keV and 1,275 keV energy windows, respectively. In comparisons between the conventional and proposed YSECT instruments, streak artifacts and salt and pepper noise potentially leading to partial-defect misreading were significantly reduced. Furthermore, the tomographic image quality was improved when the proposed YSECT instrument was employed. Based on the applicability of the validation results, it is expected that the proposed YSECT instrument could be applied to partial-defect inspection of PWR-type SNF for high accuracy and efficiency.
However, CdWO4 scintillation crystal has a high cross-section for neutrons. As this might cause induced noise or property changes in the crystal, use of CdWO4 for SNF inspection could result in deterioration of tomographic image quality. For this reason, other methods, such as pulse-shape discrimination and machine learning algorithms that can discern signals between gamma rays and neutrons, should be applied when utilizing CdWO4 scintillation crystal for partial-defect inspection. In further studies, the effect and influence of neutrons will be evaluated, and noise-reduction methods will be developed.
With respect to the main limitation of the current study and paper, their development and proposal, respectively, of the multisensor-based YSECT instrument were and are based only on simulation. Considering particularly the efficiency and duration of the applied MC simulation, actual experimental conditions were not additionally implemented. And in fact, both simulation and experimentation can cause over- and/or under-estimation of system performance. The fact that SNF emits various types of radiation, such as gamma rays, neutrons, and alpha particles, render obtainment of high-quality tomographic images problematic. In the future, therefore, the performance of proposed system will be experimentally evaluated with the prototype instrument under conditions that mimic those of actual inspections.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation 
Print 
