Modeling and Experimental Study of Liquid–liquid Extraction of Water + Formic Acid + 1-Octanol with NaCl and KCl Using Non-random Two-liquid and Artificial Neural Network Models

Abstract

This study investigates the liquid–liquid extraction behavior of a ternary system composed of water, formic acid, and 1-octanol in the presence of inorganic salts (NaCl and KCl) at varying concentrations of 0%, 5%, 10%, and 15%. Each salt was examined individually to assess its impact on the extraction efficiency. Experimental solubility data and tie-line compositions were obtained. The results demonstrate that the addition of salt significantly improves the efficiency of extraction. NaCl was found to induce a stronger salting-out effect than KCl, especially at 10% concentration, where the highest selectivity and distribution coefficient were observed. To model the phase behavior, both the Non-Random Two-Liquid (NRTL) thermodynamic model and an Artificial Neural Network (ANN) were employed based on the experimental results. A Neural Architecture Search approach was implemented to optimize ANN structure. Both models exhibited strong predictive capability; however, the ANN model demonstrated superior performance, achieving higher accuracy and lower prediction errors than the NRTL model, particularly at high salt concentrations.