Modeling of the Effect of Nanoparticles on CO2/CH4 Selectivity of Poly (4-methyl-1-pentene)-Based Mixed Matrix Membranes

Membrane-based technologies have a broad range of industrial applications for separating specific substances from either liquid or gas mixtures. Removing carbon dioxide (CO2) from natural gas is likely the most important application of membrane technology in the gas processing field. Although it is possible to fabricate many polymer-based membranes to separate CO2 from a gas mixture mainly consisting of methane (CH4), their CO2 permeability, CO2/CH4 selectivity, stability, and price are not good enough. So, researchers enhance the achievable CO2/CH4 selectivity by incorporating nanomaterials in the polymer structure, i.e., mixed matrix membranes (MMMs). Those MMMs constitute Poly (4-methyl-1-pentene) (i.e., PMP) and nanoparticles have recently found high popularity in this regard. Despite conducting many experiments to monitor the CO2-CH4 separation ability of these MMMs, there is no mathematical model to compute CO2/CH4 selectivity as a function of membrane chemistry and pressure. This study applies support vector machines (SVM) to build a model to precisely compute the CO2/CH4 selectivity of PMP-containing MMMs from the membrane chemistry (i.e., nanofiller type/dose and polymer dose) and pressure. The relevancy test proves that CO2/CH4 selectivity correlates well with all involved design features, so the nanofiller type and polymer dose are the strongest. Sensitivity analyses confirm that the SVM provides the lowest uncertainty once it is equipped with the Gaussian kernel function. This model predicts the collected literature data with an absolute average relative error of 3.23%. The leverage method identified four outliers among the experimentally measured CO2/CH4 databank of the considered PMP-containing mixed matrix membranes.