Superstructure-based optimization of membrane network systems for multicomponent liquid mixture separation

The design of membrane networks to recover chemical components from liquid mixtures is intricate due to the large number of potential network configurations. Additionally, nonidealities present in liquid mixtures pose difficulties in describing membrane permeators. In this work, we present a framework for membrane network synthesis to recover valuable components from liquid mixtures using mixed-integer nonlinear programming (MINLP). First, to model membrane permeation, we develop a physics-based nonlinear surrogate model for crossflow membranes. Second, we construct a richly connected superstructure to account for all potential membrane network configurations. Third, we demonstrate how the two aforementioned elements can be integrated into a MINLP model to obtain the optimal membrane network configuration. Finally, through various applications, we illustrate how the proposed approach can achieve globally optimal solutions.