A framework for capturing and representing the process to classify nuclear waste and informing where processes can be automated

Decommissioning and dismantling of nuclear facilities are complex processes, where an accurate triage of visual and radiological characterisation is an important driver of how this process is executed. In-situ measurements before dismantling are essential for effective, optimized waste management solutions to ensure the safe and secure decommissioning of nuclear installations. Characterising nuclear structures includes a large amount of human involvement in decision making, physical inspections and even lifting and relocating radioactive waste items. The current process accounts for risks like close human contact with radioactive material for extended periods, and errors based on operator knowledge rather than automated detection systems. In this paper, we present a framework to explicitly outline the steps required to classify nuclear waste remotely, in-situ and non-destructively, and the subsequent evaluation of these steps to determine where they can be automated. This framework uses the CommonKADS methodology, a well-established approach for knowledge modelling systems, to identify the main decisions in the process of characterising a nuclear reprocessing cell in a nuclear facility. We capture the sources of knowledge required to support and justify decisions made, and the resulting models are reviewed to assess where decisions can be automated, or supported using AI tools, to ensure robust, reliable, and rapid decisions. This framework aims to provide the first step and help to support innovation, toward a system able to produce tangible benefits for enhancing the safety, economy and reliability of nuclear cell waste classification and decommissioning management. We illustrate the use of the framework with a case study application which demonstrates how a semi-automated decision support system could be built based on the framework.