@

초록

Identifying suitable materials for industrial gas separation from a vast pool of candidates and performing preliminary process-level assessment of their separation potential remain challenging and time-consuming tasks. This study proposes a new end-to-end ultrafast screening framework for identifying optimal metal-organic frameworks (MOFs) for CO2/CH4 separation processes, using deep learning-based MOF property prediction. The proposed framework integrates molecular-level and process-level MOF screening evaluations. In the molecular-level, prescreening of the MOFs was conducted considering the physical descriptors of the MOF structures. Then, a lattice constant-hybridized graph neural network (GNN) was developed to predict CO2 and CH4 adsorption uptakes by combining global structural information with local atomic representations, overcoming the limited receptive field of standard GNN. The predicted uptakes enable pressure swing adsorption (PSA) process simulations to determine process-level metrics, including CH4 recovery and CO2 working capacity. The framework was applied to experimentally reported MOFs across three industrial scenarios with varying methane concentrations: landfill gas, biogas and coalbed methane. The proposed framework identified promising optimal MOF solutions for each application to enhance separation efficiency, including 290 for landfill gas, 42 for biogas, and 33 for coalbed methane. Structural analysis of top-performing MOFs provides recommended property ranges including pore limiting diameter of 5.12 ± 1.50 Å and volumetric surface area of 1183 ± 551 m2/cm3 that can guide adsorbent design. The techno-economic analysis demonstrates that the identified MOF-based PSA reduces separation cost by up to 51% and improves energy efficiency by up to 28% compared to conventional separation processes. This proposed screening framework provides a fast tool for narrowing large MOF databases to a manageable set of promising candidates, linking molecular-level prediction with process-level and techno-economic screening indicators that can guide subsequent rigorous evaluation.

키워드