Global optimization for accurate and efficient parameter estimation in nanofiltration

One of the most well-established frameworks for modeling multicomponent transport in nanofiltration (NF) is Donnan-Steric Pore Model with Dielectric Exclusion (DSPM-DE). Conventional DSPM-DE characterizes across NF membranes through four governing membrane parameters: (1) pore radius; (2) effective thickness; (3) charge density; and (4) dielectric constant inside pores. The process quantifying these parameters typically sequential. First, neutral solute experiments are performed to determine radius thickness. Next, charged species conducted, data used regress out remaining parameters. resulting regressions often using local search algorithms that can struggle provide low residuals robust fits. In addition, this two-step approach tends to: require a substantial number uncharged experiments; introduce assumed relationships between size water flux, such as Hagen-Poiseuille equation, which may not be representative complex networks. To address issues, we propose use metaheuristic global optimization techniques supplemented gradient-free maximum likelihood estimation simultaneously all directly from experiments. We validate our against eight independent datasets diverse input salinities, compositions, membranes.