Evolutionary optimization strategies for Liquid-liquid interaction parameters

A multi-level program for regressing thermodynamic parameters was tested on a lately quantified salting-out equilibrium involving acetonitrile + water + ammonium bicarbonate (of pseudo-type-II), plus on other two type-I case studies (n-heptane + benzene + methanol and water + ethanol + toluene), both with the UNIQUAC and NRTL models. The presented algorithm is based on already validated ones, and mixes the isoactivity condition of coexisting phases with sequential tests on the Gibb's energy shapes. The thermodynamic parameters are refined with an evolutionary strategy. Evaluating the outcome of 14 regression cases, it was found that: a) the unmodified UNIQUAC equation slightly outperforms the NRTL model for the given systems, b) the consistency check of the excess energy function acts as a bottleneck, and c) the use of different phases optimizer has a great impact on the overall performance.