| Abstract Scope |
The US Department of Energy manages Hanford Site's nuclear waste, processing it into low-activity waste (LAW) and high-level waste fractions through vitrification methods. Waste vitrification has been practiced worldwide as the preferred method of treating highly radioactive nuclear wastes. Previously, algorithms for LAW glass formulation were developed using traditional methods such as partial quadratic mixture models. Since machine learning (ML) has been successfully used to model glass properties, this work reviews the integration of ML in nuclear waste immobilization, including efforts on database management, model development, uncertainty qualification, and glass formulation. Compared to the previous framework, ML-based optimization methods, as demonstrated in proof-of-principle studies, offer improved LAW glass designs and a streamlined approach to generation of optimally designed data and near real-time updates. Moreover, the review highlights advancements in the glass optimization framework and available tools with examples. Finally, it discusses prospects for glass property model development and formulation strategies. |