Multi-criteria optimization for parametrizing excess Gibbs energy models

Thermodynamic models contain parameters which are adjusted to experimental data. Usually, optimal descriptions of different data sets require different parameters. Multi-criteria optimization (MCO) is an appropriate way to obtain a compromise. This is demonstrated here for Gibbs excess energy (GE) models. As an example, the NRTL model is applied to the three binary systems (containing water, 2-propanol, and 1-pentanol). For each system, different objectives are considered (description of vapor-liquid equilibrium, liquid-liquid equilibrium, and excess enthalpies). The resulting MCO problems are solved using an adaptive numerical algorithm. It yields the Pareto front, which gives a comprehensive overview of how well the given model can describe the given conflicting data. From the Pareto front, a solution that is particularly favorable for a given application can be selected in an instructed way. The examples from the present work demonstrate the benefits of the MCO approach for parametrizing GE-models.