Modeling surface tension of ten binary cryogenic mixtures with a thermodynamic method and artificial neural network

Abstract

The phase equilibrium calculations between the liquid and surface phase are conducted to predict the surface tension and interfacial mole fractions of the components for ten binary cryogenic systems. This thermodynamic model is combined with the perturbed chain statistical association fluid theory equation of state to determine the fugacity coefficients and molar volumes of the components. Based on the application of molar or partial molar volumes, 4 different strategies are applied to the molar surface area of this model. The results of the thermodynamic model indicate that the first strategy has the best predictions for most cases. Then an artificial neural network has been applied to the surface tension of these ten mixtures. This model contains four input parameters and 9 neurons with a single layer. The overall good predictive capability of the artificial neural network model is proved with an R² of 0.999 and an AAD_γ% of 0.94 for the entire dataset.