The ability of artificial neural network in prediction of the acid gases solubility in different ionic liquids

In this work, the solubility of carbon dioxide and hydrogen sulfide, in different ionic liquids (ILs) have been investigated by applying the artificial neural networks (ANNs). According to the economic benefits of CO2 as an inexpensive, non-toxic sources of carbon, many studies have done in capturing of CO2 from the main resources in ILs due to their specific properties such as negligible vapor pressure. Solubility is a key parameter in the phase equilibria calculations. According to the complexity of ILs structure, the phase behavior modeling for these systems is complicated. ANNs are the nonlinear mathematical models which can make a relation between the inputs and the outputs. In this paper 2930 and 664 solubility data of CO2 and H2S are used respectively. Network was trained, validated and tested by 70, 15 and 15 percent of total data with one hidden layer through hyperbolic tangent sigmoid transfer function. Optimum neurons are 23 and 14 for CO2 and H2S solubility respectively. AAD% and R2 are 3.58 percent and 0.9947 for CO2 and 2.07 and 0.9987 for H2S system. In addition, the Peng-Robinson EoS with and without optimized kij and an empirical correlation with different constants are used to compare their deviations with the ANN model. Results showed that the ANN model can correlate the solubility of acid gases in ILs with a high accuracy and its error is minimum among three approaches.