A hierarchical approach for solvent selection based on successive model refinement

Abstract

Liquid-liquid extraction has widespread use in industry, e.g., for separation of highly diluted components from fermentation broths. Feasibility and economic operation of the extraction processes depend critically on the selection of a suitable solvent. While the choice for an optimal solvent inherently depends on the overall process performance, common methods for solvent selection focus on much simpler performance indicators which can lead to suboptimal solutions. In order to improve solvent selection, we present a hierarchical approach with successive model refinement. The approach builds on the prediction of thermodynamic properties by COSMO-RS, avoiding the need for experimental data in early conceptual design phase. In the approach, advanced pinch-based shortcut models are combined with rigorous superstructure optimization to determine promising solvent candidates. The approach allows for an evaluation of several thousand potential solvents and identifies highly promising solvents based on the evaluation of process economics for an optimized process. The approach is illustrated for the extraction of γ-valerolactone from an aqueous feed stream.