Evolving machine learning models to predict hydrogen sulfide solubility in the presence of various ionic liquids

Substituting conventional solvents for gas sweetening with ionic liquids (ILs) is an interesting way to specify superior design from energy consumption in regeneration and reduction solvent loss. In this study, based on the critical temperature (T,), critical pressure (Pc), and molecular weight (Mw) of pure ionic liquids, a feed forward Multi-Layer Perceptron Artificial Neural Network (MLP-ANN), an Adaptive Neuro-Fuzzy Inference System (ANFIS) and a Radial Basis Function Artificial Neural Network (RBF-ANN) were developed to predict solubility of Hydrogen Sulfide in the presence of various ILs over wide ranges of temperature, pressure and concentration. To develop the aforementioned methods, 664 experimental data points collected from the literatures were employed. Moreover, to investigate the Hydrogen Sulfide solubility in ternary mixture containing Carbon Dioxide, Hydrogen Sulfide and ILs, MLP-ANN model was proposed. To propose MLP-ANN method for estimating H2S solubility in ternary mixture, 89 experimental data points collected from the previous published works were employed. To examine the ability of the methods suggested in this study different statistical criteria including R-Squared (R-2), Mean Squared Error (MSE), Standard Deviation (STD) and Mean Absolute Relative Error (MARE) were used. The values of R-2 and MSE achieved for the MLP-ANN model are 0.9951 and 0.000117 respectively. Furthermore, the values of R-2 and MSE for both ANFIS and RBF-ANN methods obtained 0.901, 0.002268 and 0.9679, 0.000787 respectively. In addition, R-2 and MSE of the MLP-ANN model for ternary mixtures are 0.9955 and 0.000082 correspondingly. Therefore, the ability and acceptable performance of using the MLP-ANN as an accurate model for estimating Hydrogen Sulfide solubility in ILs was showed versus other computational intelligence models. (C) 2016 Elsevier B.V. All rights reserved.