The Design of Optimal Mixtures from Atom Groups Using Generalized Disjunctive Programming.

A comprehensive computer-aided mixture/blend design methodology for formulating a general mixture design problem where the number, identity and composition of mixture constituents are optimized simultaneously is presented in this work. Within this approach, Generalized Disjunctive Programming (GDP) is employed to model the discrete decisions (number and identities of mixture ingredients) in the problems. The identities of the components are determined by designing molecules from UNIFAC groups. The sequential design of pure compounds and blends, and the arbitrary pre-selection of possible mixture ingredients can thus be avoided, making it possible to consider large design spaces with a broad variety of molecules and mixtures. The proposed methodology is first applied to the design of solvents and solvent mixtures for maximising the solubility of ibuprofen, often sought in crystallization processes; next, antisolvents and antisolvent mixtures are generated for minimising the solubility of the drug in drowning out crystallization; and finally, solvent and solvent mixtures are designed for liquid-liquid extraction. The GDP problems are converted into mixed-integer form using the big-M approach. Integer cuts are included in the general models leading to lists of optimal solutions which often contain a combination of pure and mixed solvents. (C) 2018 The Authors. Published by Elsevier Ltd.