Advanced separation technologies are required in the context of critical materials and spent nuclear fuels, in order to increase the supply of key elements, minimize nuclear waste production, optimize storage conditions, and alleviate proliferation concerns associated with reprocessing. However, rationally designing advanced separations schemes has traditionally been a time-consuming and expensive process. This project implements a multidisciplinary effort that integrates modern data-centric approaches to develop new predictive capabilities for separations. In this presentation, we share our recent progress on these integrated tasks. Separation efficiency is driven by the difference between two chemical systems. To capture the essence of this problem, we developed a new method, pairwise difference regression (PADRE), that is a mathematical instantiation of the intuitive concept that it is often easier to deduce relative differences between two chemical systems than it is to grasp a single system alone. This simple and generic approach can be applied to machine learning models for improved performance and the generation of uncertainty quantification with improved performance compared to the state-of-the-art techniques.

Data-driven separation design for F-elements

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.