A New Fast Estimation\nMethod for Critical Properties\nof Mixtures Based on the Modified Redlich–Kister Method-Part\n2: Prediction of Critical Properties for Binary and Ternary Mixtures

The\nvapor–liquid critical properties of mixtures,\nwhich\nrepresent the end points of vapor–liquid equilibrium curves,\nare crucial for the development of next-generation environmentally\nfriendly working fluids and advancements in supercritical fluid technology.\nExperimental measurement and theoretical prediction are the main means\nto obtain the critical properties. However, the existing theoretical\nprediction methods have the problems of low accuracy, especially when\npredicting critical volumes of binary mixtures and critical properties\nof ternary mixtures. Moreover, in most prediction methods, all the\ndata are used to fit the adjustable parameters. No prediction data\nset was added to test its prediction ability. In this work, new prediction\nmodels for critical properties, including critical temperature, critical\npressure, and critical volume of binary mixtures and ternary mixtures,\nwere proposed. New methods extend our previous work (Tang et al.’s\nmodel) to the prediction of critical volumes and critical properties\nof ternary mixtures and accurately evaluate the extrapolation ability.\nNew prediction models inherit many advantages of Tang et al.’s\nmodel, including considering the effect of molecular polarity to some\nextent, possessing four fewer adjustable parameters, a simpler form\nin use, and not requiring the critical volume data of pure substances\nwhen predicting critical temperatures and critical pressures of mixtures.\nMoreover, three new methods possess fast prediction abilities for\ncritical properties of mixtures, showing higher accuracy in predicting\ncritical properties of binary and ternary mixtures in the fitting\ndata set and prediction data set. When predicting critical temperatures\nand critical pressures, new models can be applied to binary and ternary\nmixtures consisting of methane-free alkanes, alkenes, alkynes, alicyclic\nhydrocarbons, benzene and its derivatives, NH<sub>3</sub>, CO<sub>2</sub>, halogenated hydrocarbon, N<sub>2</sub>O, Kr, Xe, sulfur-compounds,\nand oxygen-containing organic compounds. Notably, new methods are\napplicable only to class I (continuous curve of the critical point\nof two pure components) of the vapor–liquid critical locus,\nclassified by Van Konynenburg and Scott. Class I covers many systems\nand most of the available critical property experimental data. Almost\nall systems can be covered in predicting critical volumes. About 9000\ncritical data points including critical temperatures, critical pressures,\nand critical volumes for binary mixtures and ternary mixtures are\ncollected. About 80% of the data for binary mixtures were used to\nfit the empirical parameters. The remaining data were used to test\nthe prediction ability for binary and ternary mixtures for three new\nprediction methods. The prediction model 1 shows the highest correlation\nand prediction accuracy among the three prediction models. The average\nabsolute relative deviations are 1.13, 3.91, and 5.90 when correlating\ncritical temperatures, critical pressures, and critical volumes; 1.43,\n4.94, and 6.93 when predicting critical temperatures, critical pressures,\nand critical volumes of binary mixtures; and 1.12 and 3.99 when predicting\ncritical temperature and critical pressures of ternary mixtures.