A new approach to parameterizing the NRTL model for binary liquid–liquid equilibria (LLE) is presented. It allows for the incorporation of a priori knowledge into the parameterization procedure rather than using the “black box” approach often employed. The method first converts the compositional Txx data to a set of unique binary interaction parameters, Tττ data. Thereafter, the second step improves the parameterization of frequently used temperature-dependent parameters (TDPs) by reducing the traditional nonlinear regression problem to simpler linear regression. This method is less susceptible to poor initial guesses, finding local minima, and significantly reduces computational requirements, with comparable/improved performance to published parameters. A total of 29 binary systems, including upper and/or lower critical solution temperatures, were evaluated using the approach to provide generalized recommendations and understanding of the TDP requirements for each system type. Inclusion of such information and use of the approach will significantly simplify multicomponent LLE model parameterization.

An improved understanding of the Non-Random Two Liquid model can improve the correlation of vapor liquid equilibrium (VLE) phase behavior beyond the trial-and-error approach often employed. This work provides a rigorous, systematic evaluation of the model’s capabilities to contribute unqualified, component-independent findings and recommendationsa limitation of previous studies. First, evaluations of the model’s isobaric and isothermal VLE prediction capabilities provide insight into multiple factors, including the influence of the model’s non-randomness parameter (0.20 and 0.47). Second, a new approach is presented to improve VLE correlation of common pressure maximum azeotropic behavior. It includes a method to unequivocally disqualify non-randomness parameter values based on a single azeotropic measurement. The approach further significantly simplifies the parameterization procedure by incorporating a priori knowledge, including parameter temperature dependence requirements and initial guesses therefor. The benefits of the presented approach are demonstrated in a case study on 17 (alcohol + hydrocarbon) binary systems.

No abstract