Passive Gamma Emission Tomography (PGET) is used for the non-destructive assay of spent nuclear fuel in cooling pools. Effective image reconstruction techniques are crucial for obtaining high-quality cross-sectional images of the inspected spent fuel bundle for fuel safety and material accountability. In this work, we have developed a linear forward model to obtain the PGET response to a fuel bundle with an arbitrary shape and number of fuel rods. The image reconstruction is then performed by solving a regularized linear inverse problem using the fast-iterative shrinkage-thresholding algorithm (FISTA). After reconstructing the image, we used a convolutional neural network (CNN) to identify automatically the bundle type, detect any missing fuel pins and extract the locations of the present pins from the images. In this work, we demonstrate that this computational method is able to identify successfully missing pins in mock-up fuel assemblies with 10% and 50% missing pins.
Year
2020
Abstract