Abstract: Designing biomass-derived activated carbons (ACs) is challenged by heterogeneous synthesis routes and multiobjective trade-offs among specific surface area (S_BET), total pore volume (V_T), and mass yield (M_yield). This study presents a mechanism-enhanced multitask distillation framework (AC-ResKD) built on a shared residual deep neural network (ResDNN) and dual priors from prediction-level and teacher-aware knowledge distillation (PD-KD and TA-KD). Process factors (agent, activation temperature/time, impregnation, heating rate, precursor composition) are modeled jointly with teacher predictions to learn an interpretable mapping across combined and separated one-step/two-step datasets. All results are reported as the mean and 95% confidence intervals over 20 repeated random 80/20 splits. On separated routes, TA-KD achieves robust accuracy for S_BET (one-step: R²=0.806 [0.787, 0.824]; two-step: R²=0.801 [0.773, 0.830]) and V_T (one-step: R²=0.787 [0.764, 0.810]; two-step: R²=0.817 [0.791, 0.843]). On the combined set, TA-KD yields the strongest gains for M_yield (R²=0.816 [0.802, 0.831]; root mean squared error E_RMS=5.047 [4.85353, 5.242]), while improving S_BET and V_T as well. Overall, M_yield is the most consistent beneficiary of distillation, and route-separated training exhibits improved monotonicity and reduced bias relative to combined training; the two-step route shows stronger V_T predictability consistent with a pre-carbonization priming effect. Explainability based on permutation feature importance (PFI) and Shapley additive explanations (SHAP) identifies teacher outputs, agent type, and thermal severity as dominant drivers. Impregnation governs V_T in one-step activation, while pyrolysis variables rise in importance in two-step activation. Robust Pareto screening with quantile-window extraction delivers agent- and route-specific operating envelopes (temperature–dose–time), enabling simultaneous S_BET/V_T improvement under bounded M_yield penalties. AC-ResKD thus provides accurate, interpretable, and actionable guidance for artificial intelligence (AI)-assisted AC design in heterogeneous, data-scarce settings.

基于机制增强的多任务蒸馏技术用于生物质活性炭的预测性、可解释性设计

[1]Belluati M, Tabasso S, Calcio Gaudino E, et al., 2024. Biomass-derived carbon-based catalysts for lignocellulosic biomass and waste valorisation: a circular approach. Green Chemistry, 26(15):8642-8668.

[2]Caruana R, 1997. Multitask learning. Machine Learning, 28(1):41-75.

[3]Chairunnisa, Yu H, Saren S, et al., 2024. A review of recent advances in sustainable preparation of high-performing activated carbon for dehumidification technology. Journal of Materials Science, 59(43):20121-20156.

[4]Devi R, Kumar V, Kumar S, et al., 2023. Recent advancement in biomass-derived activated carbon for waste water treatment, energy storage, and gas purification: a review. Journal of Materials Science, 58(30):12119-12142.

[5]Evans MJB, Halliop E, MacDonald JAF, 1999. The production of chemically-activated carbon. Carbon, 37(2):269-274.

[6]Fu J, Kang QH, Ao WY, et al., 2023. Comparison and analysis of one- and two-step activation for preparation of activated carbon from furfural residues. Biomass Conversion and Biorefinery, 13(6):4681-4694.

[7]Giudicianni P, Gargiulo V, Grottola CM, et al., 2021. Inherent metal elements in biomass pyrolysis: a review. Energy & Fuels, 35(7):5407-5478.

[8]Gorishniy Y, Rubachev I, Khrulkov V, et al., 2021. Revisiting deep learning models for tabular data. Proceedings of the 35th International Conference on Neural Information Processing Systems, article 1447.

[9]Heidarinejad Z, Dehghani MH, Heidari M, et al., 2020. Methods for preparation and activation of activated carbon: a review. Environmental Chemistry Letters, 18(2):393-415.

[10]Ibrahim AF, Hussein MA, 2025. Leveraging machine learning for prediction and optimization of texture properties of sustainable activated carbon derived from waste materials. Scientific Reports, 15(1):11313.

[11]Kadra A, Lindauer M, Hutter F, et al., 2021. Well-tuned simple nets excel on tabular datasets. Proceedings of the 35th International Conference on Neural Information Processing Systems, article 1832.

[12]Kaur M, Mittal N, Charak A, et al., 2023. Rice husk derived activated carbon for the adsorption of scarlet RR an anionic disperse dye. Evergreen, 10(1):438-443.

[13]Kolbadinejad S, Mashhadimoslem H, Ghaemi A, et al., 2022. Deep learning analysis of Ar, Xe, Kr, and O₂ adsorption on activated carbon and zeolites using ANN approach. Chemical Engineering and Processing - Process Intensification, 170:108662.

[14]Kundu S, Khandaker T, Anik MAAM, et al., 2024. A comprehensive review of enhanced CO₂ capture using activated carbon derived from biomass feedstock. RSC Advances, 14(40):29693-29736.

[15]Lamb WF, Grubb M, Diluiso F, et al., 2022. Countries with sustained greenhouse gas emissions reductions: an analysis of trends and progress by sector. Climate Policy, 22(1):1-17.

[16]Li GY, Xue B, Yu H, et al., 2025a. Performance improvement of waste heat upgrading adsorption heat pump by employing copper oxide-loaded composite zeolites for high-temperature steam generation. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 711:136330.

[17]Li GY, Yu H, Ji DX, et al., 2025b. Pine cone-based activated carbon via dual physical activation for efficient carbon dioxide capture: experimental and molecular simulation studies. Energy, 328:136506.

[18]Li XH, Huang ZH, Shao SS, et al., 2024. Machine learning prediction of physical properties and nitrogen content of porous carbon from agricultural wastes: effects of activation and doping process. Fuel, 356:129623.

[19]Liao MC, Kelley SS, Yao Y, 2019. Artificial neural network based modeling for the prediction of yield and surface area of activated carbon from biomass. Biofuels, Bioproducts and Biorefining, 13(4):1015-1027.

[20]Lu YH, Zhang SL, Yin JM, et al., 2017. Mesoporous activated carbon materials with ultrahigh mesopore volume and effective specific surface area for high performance supercapacitors. Carbon, 124:64-71.

[21]Merchant A, Batzner S, Schoenholz SS, et al., 2023. Scaling deep learning for materials discovery. Nature, 624(7990):80-85.

[22]Muhammed F, Lavaggi T, Advani S, et al., 2021. Influence of material and process parameters on microstructure evolution during the fabrication of carbon–carbon composites: a review. Journal of Materials Science, 56(32):17877-17914.

[23]Othman NAB, Kim T, Imamura A, et al., 2013. Investigation on H₂S removal factors of activated carbons derived from waste palm trunk. Journal of Novel Carbon Resource Sciences, 7:7-11.

[24]Przybek A, Setlak K, Piątkowski J, et al., 2025. Alkaline-activated materials for CO₂ capture–literature review, own observations, and future perspectives. Evergreen, 12(3):1664-1679.

[25]Reiser P, Neubert M, Eberhard A, et al., 2022. Graph neural networks for materials science and chemistry. Communications Materials, 3(1):93.

[26]Ruder S, 2017. An overview of multi-task learning in deep neural networks. arXiv:1706.05098.

[27]Sevilla M, Mokaya R, 2014. Energy storage applications of activated carbons: supercapacitors and hydrogen storage. Energy & Environmental Science, 7(4):1250-1280.

[28]Shukla AK, Alam J, Mallik S, et al., 2024. Optimization and prediction of dye adsorption utilising cross-linked chitosan-activated charcoal: response surface methodology and machine learning. Journal of Molecular Liquids, 411:125745.

[29]Song YH, Li JK, Chi DZ, et al., 2025. AI-driven advances in metal–organic frameworks: from data to design and applications. Chemical Communications, 61(82):15972-16001.

[30]Sosa JA, Laines JR, García DS, et al., 2023. Activated carbon: a review of residual precursors, synthesis processes, characterization techniques, and applications in the improvement of biogas. Environmental Engineering Research, 28(3):220100.

[31]Yu H, Seo SW, Mikšík F, et al., 2023. Effects of temperature and humidity ratio on the performance of desiccant dehumidification system under low-temperature regeneration. Journal of Thermal Analysis and Calorimetry, 148(8):3045-3058.

[32]Yu H, Mikšík F, Thu K, et al., 2024. Characterization and optimization of pore structure and water adsorption capacity in pinecone-derived activated carbon by steam activation. Powder Technology, 431:119084.

[33]Yuan XZ, Suvarna M, Low S, et al., 2021. Applied machine learning for prediction of CO₂ adsorption on biomass waste-derived porous carbons. Environmental Science & Technology, 55(17):11925-11936.

[34]Yudha CS, Hutama AP, Gustiana HSEA, et al., 2024. An efficient Nypa fiber waste conversion to activated carbon for Li-ion battery anode material. Evergreen, 11(3):2193-2201.

[35]Zhang K, Zhong SF, Zhang HC, 2020. Predicting aqueous adsorption of organic compounds onto biochars, carbon nanotubes, granular activated carbons, and resins with machine learning. Environmental Science & Technology, 54(11):7008-7018.

[36]Zhao P, Zhou JH, Li GY, et al., 2025. Prediction of physisorption on zeolites using graph integrated adsorption network and spline data augmentation. Thermal Science and Engineering Progress, 66:104047.

[37]Zhou ZY, Yu H, Li GY, et al., 2026. Material–scenario coupled techno-economics comparison of carbon capture adsorbents: feasibility and suitability across various contexts. Energy Conversion and Management, 348:120589.

[38]Zou RG, Yang ZB, Zhang JH, et al., 2024. Machine learning application for predicting key properties of activated carbon produced from lignocellulosic biomass waste with chemical activation. Bioresource Technology, 399:130624.