A fuel assembly operated in a nuclear power plant typically contains 100–300 fuel rods, depending on fuel type, which become strongly radioactive during irradiation in the reactor core. For operational and security reasons, it is of interest to experimentally deduce rod-wise information from the fuel, preferably by means of non-destructive measurements. The tomographic SPECT technique offers such possibilities through its two-step application; (1) recording the gamma-ray flux distribution around the fuel assembly, and (2) reconstructing the assembly׳s internal source distribution, based on the recorded radiation field. In this paper, algorithms for performing the latter step and extracting quantitative relative rod-by-rod data are accounted for.
As compared to application of SPECT in nuclear medicine, nuclear fuel assemblies present a much more heterogeneous distribution of internal attenuation to gamma radiation than the human body, typically with rods containing pellets of heavy uranium dioxide surrounded by cladding of a zirconium alloy placed in water or air. This inhomogeneity severely complicates the tomographic quantification of the rod-wise relative source content, and the deduction of conclusive data requires detailed modelling of the attenuation to be introduced in the reconstructions. However, as shown in this paper, simplified models may still produce valuable information about the fuel.
Here, a set of reconstruction algorithms for SPECT on nuclear fuel assemblies are described and discussed in terms of their quantitative performance for two applications; verification of fuel assemblies׳ completeness in nuclear safeguards, and rod-wise fuel characterization. It is argued that a request not to base the former assessment on any a priori information brings constraints to which reconstruction methods that may be used in that case, whereas the use of a priori information on geometry and material content enables highly accurate quantitative assessment, which may be particularly useful in the latter application.
Two main classes of algorithms are covered; (1) analytic filtered back-projection algorithms, and (2) a group of model-based or algebraic algorithms. For the former class, a basic algorithm has been implemented, which does not take attenuation in the materials of the fuel assemblies into account and which assumes an idealized imaging geometry. In addition, a novel methodology has been presented for introducing a first-order correction to the obtained images for these deficits; in particular, the effects of attenuation are taken into account by modelling the response for an object with a homogeneous mix of fuel materials in the image area. Neither the basic algorithm, nor the correction method requires prior knowledge of the fuel geometry, but they result in images of the assembly׳s internal activity distribution. Image analysis is then applied to deduce quantitative information.
Two algebraic algorithms are also presented, which model attenuation in the fuel assemblies to different degrees; either assuming a homogenous mix of materials in the image area without a priori information or utilizing known information of the assembly geometry and of its position in the measuring setup for modelling the gamma-ray attenuation in detail. Both algorithms model the detection system in detail. The former algorithm returns an image of the cross-section of the object, from which quantitative information is extracted, whereas the latter returns conclusive relative rod-by-rod data.
Here, all reconstruction methods are demonstrated on simulated data of a 96-rod fuel assembly in a tomographic measurement setup. The assembly was simulated with the same activity content in all rods for evaluation purposes. Based on the results, it is argued that the choice of algorithm to a large degree depends on application, and also that a combination of reconstruction methods may be useful. A discussion on alternative analysis methods is also included.
Elsevier, 2015. Vol. 783, 128-141 p.
Tomography; Nuclear fuel; Reconstruction; Radon transform; Analytic algorithm; Algebraic algorithm