Liquid−liquid equilibrium (LLE) data for the (water + valeric acid + methyl ethyl ketone or methyl isobutyl ketone) ternary systems were measured at T = (298.2, 308.2 and 318.2) K and atmospheric pressure. Solubility curves and tie-line end compositions were determined for the investigated systems. Both ternary systems show type-2 behavior of the LLE systems. The experimental tie-line data were correlated to assess consistency with the Othmer−Tobias and Hand equations. Distribution coefficients and selectivity factors were determined to evaluate the solvent extraction capacity. The maximum of separation factors for methyl ethyl ketone (MEK) and methyl isobutyl ketone (MIBK) as solvent were determined as 96.8 and 469.5 at 298.2 K, respectively, and the separation factors decreased with increasing temperature. As a result, MEK and MIBK have relatively high separation factors and are suitable separating agents for valeric from water. Binary interaction parameters of NRTL and UNIQUAC thermodynamic models were estimated for these systems by using genetic algorithm (GA). The average RMSD values between the experimental and the calculated mass fraction of the NRTL and UNIQUAC models are 0.0021 and 0.0222, respectively. The RMSD values demonstrate that in order to describe the phase behavior of the studied LLE systems the NRTL model is a more suitable model in comparison to the UNIQUAC model.

The experimental data of the solubility curve and the tie line for the ternary system containing compounds of water, acetic acid, and 1-octanol were studied over the temperature range (298 to 328) K and atmospheric pressure. This ternary system showed the behavior of type 1 liquid–liquid equilibrium (LLE) systems. Also, to test the tie line consistency, the obtained experimental data were correlated with Othmer-Tobias and Han equations. Distribution coefficients, selectivity factors, and percentage of extraction were determined to evaluate the extracting capacity of the solvent. Interaction parameters of the nonrandom two-liquid (NRTL) and universal quasichemical (UNIQUAC) thermodynamic models were estimated using a genetic algorithm (GA) and particle swarm optimization (PSO) which are evolutionary optimization methods. The mean RMSD values between the calculated and experimental mass fractions of the NRTL and UNIQUAC models using GA were 0.0047 and 0.0089, and using PSO were 0.0023 and 0.0042, respectively. The RMSD values indicate that the NRTL and UNIQUAC models are suitable for describing the phase behavior of the studied LLE systems and the accuracy of the PSO method is slightly more than that of the GA.