(373t) Surrogate-Based Optimization for the Recovery of Critical Minerals and Rare Earth Elements

Critical minerals (CMs) and Rare Earth Elements (REEs) have received significant attention in recent years because they are necessary materials for several critical infrastructure technologies. Their recovery from industrial waste and recycling streams has the potential to minimize environmental issues and reduce costs in the supply chains where they are utilized. This motivates the rigorous and systematic optimization of CM and REE recovery process flowsheets.

In this work, we investigate various surrogate modelling techniques that can be used to simplify complex CM and REE flowsheets. We consider many surrogate modelling techniques, including deep learning and regularization. Through extensive computational testing, we determine which of the models are easier to construct and optimize, as well as which factors create computational bottlenecks in their solution. The results of our analysis inform the selection of the most appropriate algorithms for surrogate-based optimization of CM and REE process flowsheets. Our main conclusion is that interpretable surrogate models substantially reduce the computational cost while ensuring high quality of solutions.