Abstract:Experimental vapor–liquid equilibrium (VLE) data, and the activity coefficient model parameters obtained by their correlation, are frequently used in separation processes design. The inspection of this information for ternary systems has revealed some problems that are not widely known and could lead to serious mistakes. A small sampling of ten ternary systems, with VLE data sets at constant pressure, arbitrarily selected has been considered to illustrate these problems. The experimental VLE data and the NRTL model parameters for these systems have been taken from different sources, one database and two commercial simulation packages. A MATLAB code has been developed to prepare 3D graphs where the T/x and T/y surfaces calculated by the activity coefficient model (NRTL) are represented along with the experimental VLE data. The results of the analysis of this small sample are very worrying. In fact, half of the selected systems present severe inconsistences between the experimental VLE data and the T/x,y surfaces calculated using the correlation results published or implemented in the simulation software. These inconsistences are of different natures, but the most notable is the existence of parameter sets that do not correlate at all the VLE behavior of the system that they supposedly represent. These irregularities may occur only in the ternary region, only in the binary subsystems, or, in some cases, in both. These problems go frequently unnoticed for different reasons. As discussed in this work, the typical 2D graphs and deviations in T or P and vapor composition are not enough in some cases so as to have an adequate idea of both the quality of the data and the suitability of the model to represent them. This is particularly true when considering, as is strongly advisible, the complete set of data, i.e. the ternary and binary VLE data simultaneously. Some recommendations are presented in this work to overcome the problems detected both when planning the experimental design to determine VLE data and to correlate them with a model.

Vapor–liquid phase equilibrium prediction for mixtures of binary systems using graph neural networks

MMGNN: A Molecular Merged Graph Neural Network for Explainable Solvation Free Energy Prediction

Predicting CO$_2$ Absorption in Ionic Liquids with Molecular Descriptors and Explainable Graph Neural Networks

Models for predicting multi-component phase equilibrium ratio based on artificial neural networks

Machine Learning Models for Vapor-Liquid Equilibrium of Binary Mixtures: State of the Art and Future Opportunities

Prediction of Vapor Liquid Equilibrium of Binary CO2-Contained Mixtures for Carbon Capture and Sequestration using Artificial Intelligence

The Vapor Liquid Equilibrium Mixing Rules based on Excess Free Energy Models

Evaluating to Nasrifar-Moshfeghian equation and applying it to vapor-liquid equilibrium

Prediction of critical temperature of binary refrigerant mixtures by neural network

A Data-Driven Phase Change Model for Injection Flow Modeling

Machine learning prediction of BLEVE loading with graph neural networks

Modeling of thermophysical properties and vapor-liquid equilibrium using Gaussian process regression

A Neural Network Approach to Predict Gibbs Free Energy of Ternary Solid Solutions

Prediction of ternary vapor-liquid equilibria for 33 systems by molecular simulation,

Artificial neural network aided vapor–liquid equilibrium model for multi-component high-pressure transcritical flows with phase change

Advancing Vapor Pressure Prediction: A Machine Learning Approach with Directed Message Passing Neural Networks

Analysis of Vapor–Liquid Equilibrium in Ternary Systems for an Adequate Planning of Their Experimental Determination and Correlation

Graph Neural Networks for Temperature-Dependent Activity Coefficient Prediction of Solutes in Ionic Liquids

Graph Neural Networks and Structural Information on Ionic Liquids: A Cheminformatics Study on Molecular Physicochemical Property Prediction

Machine Learning for Predicting Thermodynamic Properties of Pure Fluids and Their Mixtures

Finite Volume Graph Network(FVGN): Predicting unsteady incompressible fluid dynamics with finite volume informed neural network