A Multi-Indicator Weighted Screening Strategy of Materials: Integrating Process Evaluation into Machine Learning-Assisted High-Throughput Screening for SF ₆ Recycling

The rapid development of metal–organic frameworks (MOFs) has created opportunities for identifying efficient adsorbents for SF6 recycling. However, existing studies have primarily relied on single-performance criteria in high-throughput screening and have overlooked a systematic assessment of material behavior under the temperature swing adsorption (TSA) cycle. To address this gap, this paper first proposes a multi-indicator weighted screening strategy of materials that integrates process evaluation into machine learning (ML)-assisted high-throughput screening for SF6 recycling. By integrating molecular simulations, this strategy is applied to identify MOFs exhibiting superior SF6/N2 separation potential during the TSA process. A novel decision index (ZTOPSIS) combining regenerability and the trade-off between selectivity and working capacity (TSN) is first introduced, enabling the successful selection of the top 10 MOFs from prescreened candidates. Six ML models are developed to predict material performance and identify key influencing features. The TSA cycle simulations are conducted for the top 10 materials, providing a comprehensive evaluation and multiobjective optimization of cycle performance indicators. The results indicate that 10 MOFs have ZTOPSIS values exceeding 0.8. The XGB model performs optimally in predicting TSN and ZTOPSIS, with R2 of 0.835 and 0.816, respectively. SF6 adsorption heat, PLD, C%, VF, and N2 adsorption heat are identified as the primary factors affecting the material performance. Among the top 10 MOFs ranked by ZTOPSIS, AZIVAI performs best in recovery and exergy efficiency, reaching 99.36 and 23.66%, respectively, along with a high purity of 99%. This study contributes to selecting ideal SF6 recycle materials that achieve both effective separation and low energy consumption.