PVC detection in Sellafield repackaging procedures via hyperspectral imaging

Sellafield Ltd (United Kingdom) has traditionally used multi-layered packages to store Special Nuclear Material (SNM), where these packages contain metallic cans and also an over-layer of Polyvinyl Chloride (PVC). However, it has been reported that this plasticized PVC can be affected by radiation and heat, which could lead to corrosion of the metallic cans. Therefore, repackaging procedures to remove PVC are under development at Sellafield for safe long-term storage. This work aims to evaluate the use of Hyperspectral Imaging (HSI) based instrumentation for the automated detection of PVC on metallic surfaces. The spectral response of PVC and 316L stainless steel (same material as is used to construct the metallic cans) is captured and evaluated in the Shortwave Infrared (SWIR) range, where data processing is applied to capture the differential features of each material. After this processing, all pixels in an image can be classified either into PVC or 316L via a machine learning model, the Support Vector Machine (SVM). Experiments included the generation of detection maps that highlight the presence of any PVC on the metallic surface for a given scenario, where small PVC samples of different sizes are randomly distributed across a 316L layout. Results showed that HSI, with appropriate data processing, is able to detect and highlight the presence of every PVC sample in the image, quantifying the percentage of the area covered by each material and demonstrating potential to become onsite instrumentation for decision support in this task. HSI technology, going beyond the visible range, could be applied for material detection and quantification in similar applications within nuclear power plants facilities.