Using Artificial Neural Network and Ideal Adsorbed Solution Theory for Predicting the CO2/CH4 Selectivities of Metal-Organic Frameworks: A Comparative Study

Predicting the adsorptive separations of gaseous mixtures by metal-organic frameworks (MOFs) is a difficult task, because it is a complex function of the textural properties of the MOF, its surface chemistry, and the operating conditions. In this research, the separation behavior of multicomponent mixtures of CO2/CH4 in MOFs was predicted by means of artificial neural networks (ANNs). To validate the as-designed ANN model, a well-known MOF, CuBTC or HKUST-1, was synthesized in our laboratory, and the selectivity of CO2/CH4 mixtures was determined using ideal adsorbed solution theory (LAST) and the extended Langmuir model (ELM). On the basis of the obtained results, it was demonstrated that ANN modeling could be a good candidate for predicting the separation behavior of CO2/CH4 mixtures in MOFs in the absence of data from laboratory experiments.