In this study, solubility and tie-line data of ternary system water + glycerol + 1-butanol were determined at 293. 2, 298. 2, and 303. 2 K and atmospheric pressure. This thermodynamic system is relevant for the production and purification of biofuels. Phase equilibrium data have been determined by the cloud-point titration method and the tie-lines were obtained by correlating the refractive index of the binodal curves as a function of mixture composition. All measured LLE data were modeled by UNIQUAC and NRTL activity coefficient equations obtaining a satisfactory accuracy with modeling errors lower than 0. 4%. Binary interaction parameters of tested thermodynamic models were estimated to predict the value of tie lines using a hybrid bio-inspired swarm intelligence optimization algorithm, that is, MAKHA without and with closure equation. This hybrid method was reliable to solve the global optimization problem for the binary interaction parameter identification of this ternary system. The results of this paper provide useful information for the design and modeling of industrial units for glycerol recovery, which is a relevant industrial feedstock.

This study aimed to measure the liquid−liquid equilibrium (LLE) data for a ternary (water–acetone– ethyl acetate) system to extract acetone. For this purpose, the experiments were conducted with a feed containing 80 wt. % water and 20 wt. % acetone in a T-shaped micromixer and then in a 0. 8 mm microtube. The operational variables, e. g., solvent-to-feed ratio (𝑆/𝐹, 0. 33-3. 00 v/v) and operating temperature (303. 15, 308. 15, 313. 15, and 318. 15 K), were calculated at 0. 1 MPa. According to the experimental results, the best selectivity (𝑆) and distribution coefficient (𝐷2) was achieved at 0. 50 v/v and 313. 15 K. Besides,the interaction parameters were obtained for the components by fitting UNIQUAC and NRTL models to experimental data. Both models exhibited high levels of accuracy, with standard deviations in the 0. 0002-0. 0346 range from experimental data. The results obtained from the UNIQUAC and NRTL models were consistent with the experimental data (root mean square deviations were 0. 0138 and 0. 0116 for the UNIQUAC and NRTL models, respectively).