[Paper Review] Fuel rod classification from Passive Gamma Emission Tomography (PGET) of spent nuclear fuel assemblies

Riina Virta, Rasmus Backholm+6 more|arXiv (Cornell University)|Sep 24, 2020

Advanced X-ray and CT Imaging11 references8 citations

TL;DR

This paper presents a robust, non-destructive method for fuel rod classification in spent nuclear fuel using Passive Gamma Emission Tomography (PGET) with iterative image reconstruction. By simultaneously reconstructing activity and attenuation maps while incorporating geometric and material priors, the method achieves high-accuracy classification of fuel rods—including missing, partial, water channels, and burnable absorber rods—across diverse fuel types and parameters, with only rare false negatives for present rods.

ABSTRACT

Safeguarding the disposal of spent nuclear fuel in a geological repository needs an effective, efficient, reliable and robust non-destructive assay (NDA) system to ensure the integrity of the fuel prior to disposal. In the context of the Finnish geological repository, Passive Gamma Emission Tomography (PGET) will be a part of such an NDA system. We report here on the results of PGET measurements at the Finnish nuclear power plants during the years 2017-2020. Gamma activity profiles are recorded from all angles by rotating the detector arrays around the fuel assembly that has been inserted into the center of the torus. Image reconstruction from the resulting tomographic data is defined as a constrained minimization problem with a data fidelity term and regularization terms. The activity and attenuation maps, as well as detector sensitivity corrections, are the variables in the minimization process. The regularization terms ensure that prior information on the (possible) locations of fuel rods and their diameter are taken into account. Fuel rod classification, the main purpose of the PGET method, is based on the difference of the activity of a fuel rod from its immediate neighbors, taking into account its distance from the assembly center. The classification is carried out by a support vector machine. We report on the results for ten different fuel types with burnups between 5.72 and 55.0 GWd/tU, cooling times between 1.87 and 34.6 years and initial enrichments between 1.9 and 4.4%. For all fuel assemblies measured, missing fuel rods, partial fuel rods and water channels were correctly classified. Burnable absorber fuel rods were classified as fuel rods. On rare occasions, a fuel rod that is present was falsely classified as missing. We conclude that the combination of the PGET device and our image reconstruction method provides a reliable base for fuel rod classification.

Motivation & Objective

• To develop a reliable, non-destructive assay (NDA) method for verifying spent nuclear fuel integrity prior to geological disposal in Finland.
• To address limitations of existing NDA techniques that only detect gross material deviations, by enabling rod-level detection of anomalies.
• To improve image reconstruction and classification accuracy for diverse spent fuel assemblies with varying burnup, cooling time, and enrichment.
• To classify fuel rods into categories (present, missing, abnormal, or modified) to support safeguards verification and reduce false alarms.
• To integrate PGET with future safeguards workflows by automating data acquisition, image reconstruction, and classification.

Proposed method

• The PGET system uses two linear arrays of collimated CdZnTe (CZT) gamma detectors arranged in a torus to acquire tomographic projections from 360 angles around a spent fuel assembly.
• Image reconstruction is formulated as a constrained minimization problem minimizing a functional that includes data fidelity and regularization terms for activity, attenuation, and detector sensitivity.
• Regularization incorporates prior knowledge on fuel rod locations, diameters, and expected activity/attenuation bounds to stabilize the inverse problem.
• The Levenberg–Marquardt algorithm is used to solve the nonlinear minimization problem, enabling simultaneous reconstruction of activity and attenuation maps.
• A support vector machine (SVM) classifies rods based on their activity relative to neighbors, adjusted for radial distance from the assembly center.
• The method accounts for detector response and includes energy windowing (400–600 keV, 600–700 keV, 700–1500/2000 keV, >1500/3000 keV) to target key gamma-emitting fission products.

Experimental results

Research questions

• RQ1 Can PGET with iterative image reconstruction reliably detect single missing fuel rods across diverse spent fuel assembly types and parameters?
• RQ2 How accurate is the classification of burnable absorber rods and partial fuel rods when they have similar gamma activity or density to standard fuel rods?
• RQ3 What impact do geometric simplifications (e.g., idealized rod positions) have on image reconstruction and classification accuracy?
• RQ4 Can the classification system be enhanced to detect rods that are modified (e.g., replaced with low-activity material) without classifying them as missing?
• RQ5 How can the forward model be improved to include gamma ray scattering and more realistic material properties for better reconstruction fidelity?

Key findings

• All measured fuel assemblies—spanning 10 types, 5.72–55.0 GWd/tU burnup, 1.87–34.6 years cooling time, and 1.9–4.4% initial enrichment—were correctly classified.
• Missing fuel rods, partial rods, water channels, and burnable absorber rods were all correctly identified, with burnable absorber rods classified as fuel rods.
• Only rare false classifications occurred, where a present rod was misclassified as missing, indicating high sensitivity and specificity.
• Abnormal rods with lower-than-average activity (e.g., due to lower burnup) were detected as missing by the current SVM classifier, though their physical presence was visible in attenuation maps.
• The method demonstrated robustness across different measurement campaigns (2017–2020), with consistent performance using 360 projection angles and 800–924 ms integration time per angle.
• Future improvements include revising geometry assumptions for corner rods, refining activity-attenuation bounds, and incorporating scattering in the forward model to enhance realism.