Prediction carbon dioxide solubility in presence of various ionic liquids using computational intelligence approaches

Abstract

Ionic liquids (ILs) are highly promising for industrial applications such as design and development of gas sweetening processes. For a safe and economical design, prediction of carbon dioxide solubility by a trustworthy model is really essential. In this research, based on the pressure and temperature of system and the critical properties such as critical temperature (Tc) and critical pressure (Pc) and also acentric factor (ω) and molecular weight (Mw) of pure ionic liquids, a multi-layer perceptron artificial neural network (MLP-ANN) and an adaptive neuro-fuzzy interference system (ANFIS) have been developed to estimate carbon dioxide solubility in presence of various ILs over wide ranges of pressure, temperature and concentration. To this end, 728 experimental data points collected from the literature have been used for implementation of these models. To verify the proposed models, regression analysis has been conducted on the experimental and predicted solubility of carbon dioxide in ILs. Moreover, in this study, a comparison between experimental carbon dioxide solubilities and predicted values of carbon dioxide solubility by thermodynamic models based on Peng–Robinson (PR) and Soave–Redlich–Kwong (SRK) equation of states has been performed. For MLP-ANN, coefficient of determination (R2) between experimental and predicted values is 0.9972 and mean squared errors (MSEs) is 0.000133 and the values of R2=0.9336 and MSE=0.002942 were obtained for ANFIS model while, the values of R2 and MSE for PR-EOS were 0.7323 and 0.002702 respectively, and also, R2=0.6989 and MSE=0.005578 were obtained for SRK-EOS model. Therefore, in current study, ability and better performance of MLP-ANN as an accurate correlation for estimating carbon dioxide solubility in ILs was showed against other alternative models.