Computer-aided multi-scale simulation of MOF-based membrane separation for CO₂ removal via integration of molecular and process simulation through machine learning

MOF membranes have shown high productivity in CO₂ removal from gas mixtures. In this work, we present a new integrated framework for reliable prediction of the process performance of MOF-based membranes (e.g. IRMOF-1). Firstly, molecular simulations are conducted to investigate the adsorption isotherms, self-diffusivity, activation energy, permeability and selectivity of IRMOF-1 for CO₂/CH₄ separation under different operating conditions. The simulated isotherms at 298 K are in good agreement with the experimental results. As the operating conditions vary, the permeability of CO₂ can scale from 40898 to 381776 barrer. A similar observation is made regarding selectivity, highlighting the dramatic effect of operating conditions on membrane performance. Further, predictive models are developed with machine learning, allowing the calculation of permeability and selectivity. As the mean squared error is 0.0086 and the R-Square is 0.9822, the predictive models are shown high reliability. The predictive models integrated with the tanks-in-series model of a hollow fiber membrane separation process are simulated by the finite-volume method. The feasibility and competence of the proposed framework are illustrated in three case studies.