Accurate prediction of miscibility of CO2 and supercritical CO2 in ionic liquids using machine learning

Abstract

In this study, the solubility of CO2 and supercritical (SC) CO2 in 20 ionic liquids (ILs) of different chemical families over a wide range of pressure (0.25–100.12 MPa) and temperature (278.15–450.49 K) were predicted, using a robust machine learning method of multi-layer perceptron neural network (MLP-NN). The developed model with the R2 of 0.9987, MSE of 0.6293 and AARD% of 1.8416 showed a great accuracy in predicting experimental values. In another approach for predicting the CO2 solubility, an empirical correlation with several constants was developed. With the R2 of 0.9922, MSE of 3.7874 and AARD% of 3.5078 the empirical correlation showed acceptable results; nevertheless weak compared to the ANN. The significance of this correlation is that it needs no physical property of the ILs or their mixture, and for its estimation, even a simple calculator is sufficient. A comprehensive statistical assessment conducted to assure the robustness and generality of the model. In addition, the applicability of the model and quality of experimental data was fully investigated by Leverage approach.