机器学习驱动分子筛基CO吸附剂设计与优化

doi:10.11949/0438-1157.20250680

摘要/Abstract

摘要：

分子筛因其规整的孔道结构和可调控的化学组成在CO吸附领域展现出巨大的应用潜力，但众多影响因素致使通过传统实验方法和理论计算进行吸附剂设计与优化的效率低下。本研究提出分子筛吸附剂的结构特征片段化与实验数据结合的机器学习策略，整合文献数据与沸石结构信息，构建了表达分子筛结构与CO吸附性能构效关系的综合数据集。采用梯度增强决策树、随机森林、极端随机树和极端梯度提升（XGB）四种集成学习算法预测CO吸附，并运用嵌套交叉验证确保评估模型的预测准确性。结果表明，XGB模型表现最优，显示出优异的预测精度。通过分子筛结构特征片段化分析发现，三角晶系、 $F d 3 ¯ m$ 空间群、近圆形孔道和相互连接的笼状结构有利于CO吸附。分子筛骨架中的Ca²⁺和Ba²⁺等碱土金属离子有利于CO吸附。在金属离子改性分子筛中，引入的Cu(Ⅰ)负载量对CO吸附性能的影响最为显著，最优负载量为吸附剂总质量的13%~15%。吸附剂的比表面积在230~400 m²/g、总孔体积在0.10~0.15 cm³/g时表现最佳，其限制性孔径在4.5 Å（1 Å=0.1 nm）附近、最大孔径在5.0~5.5 Å内表现出较好的CO吸附性能。本研究为分子筛基CO吸附剂的合理设计和高效筛选提供了科学指导。

关键词: 分子筛, 吸附, 机器学习, 一氧化碳, 吸附剂设计

Abstract:

Zeolite exhibits great potential for CO adsorption due to its well-organized pore structure and tunable chemical composition. However, numerous influencing factors hinder the design and optimization of adsorbents using traditional experimental methods and theoretical calculations. In this paper, a machine learning strategy of combining the structural characteristics of molecular sieve adsorbents with experimental data was proposed, and a comprehensive data set was constructed to express the structure-activity relationship between molecular sieve structure and CO adsorption performance by integrating literature data and zeolite structure information. Four integrated learning algorithms, gradient enhanced decision tree, random forest, extreme random tree and extreme gradient lifting (XGB), were used to predict CO adsorption, and nested cross validation was used to ensure the prediction accuracy of the evaluation model. The results show that XGB model performs best and shows excellent prediction accuracy. The fragmentation analysis of molecular sieve structure characteristics showed that triangular crystal system, $F d 3 ¯ m$ space group, near circular pores and interconnected cage structure were conducive to CO adsorption. The alkaline earth metals such as Ca²⁺ and Ba²⁺ in the framework of molecular sieve are conducive to the adsorption of CO. In the metal ion modified molecular sieve, the Cu(Ⅰ) loading has the most significant effect on the CO adsorption performance, and the optimal loading is 13%—15% by weight. When the specific surface area of the adsorbent is 230—400 m²/g and the total pore volume is 0.10—0.15 cm³/g, the adsorbent has the best performance. Its limited pore size is around 4.5 Å and the maximum pore size is in the range of 5.0—5.5 Å, showing relatively good CO adsorption performance. This study provides scientific guidance for the rational design and efficient screening of molecular sieve based co adsorbents.

Key words: molecular sieve, adsorbent, machine learning, carbon monoxide, adsorbent design

中图分类号:

TQ 028.8

陶玟媛, 赵文凯, 沈海阔, 郭强, 肖永厚. 机器学习驱动分子筛基CO吸附剂设计与优化[J]. 化工学报, 2025, 76(12): 6453-6464.

Wenyuan TAO, Wenkai ZHAO, Haikuo SHEN, Qiang GUO, Yonghou XIAO. Machine learning-driven design and optimization of molecular sieve-based efficient CO adsorbents[J]. CIESC Journal, 2025, 76(12): 6453-6464.

图/表 11

参考文献 30

[1]	Fujimori S, Inoue S. Carbon monoxide in main-group chemistry[J]. Journal of the American Chemical Society, 2022, 144(5): 2034-2050.
[2]	Kapetanaki S M, Burton M J, Basran J, et al. A mechanism for CO regulation of ion channels[J]. Nature Communications, 2018, 9(1): 907.
[3]	唐磊, 王振菲, 李聪利, 等. Co-MOF-74和Mg-MOF-74的CO工作吸附容量及操作条件[J]. 化工学报, 2025, 76(5): 2279-2293.
	Tang L, Wang Z F, Li C L, et al. CO working capacity and operating conditions of Co-MOF-74 and Mg-MOF-74[J]. CIESC Journal, 2025, 76(5): 2279-2293.
[4]	Yin Y, Wen Z H, Shi L, et al. Cuprous/vanadium sites on MIL-101 for selective CO adsorption from gas mixtures with superior stability[J]. ACS Sustainable Chemistry & Engineering, 2019, 7(13): 11284-11292.
[5]	Li Y, Mei Y R, Zhang T, et al. Paths to carbon neutrality in China's chemical industry[J]. Frontiers in Environmental Science, 2022, 10: 999152.
[6]	Cheng X, Liao Y, Lei Z, et al. Multi-scale design of MOF-based membrane separation for CO₂/CH₄ mixture via integration of molecular simulation, machine learning and process modeling and simulation[J]. Journal of Membrane Science, 2023, 672: 121430.
[7]	Ma X Z, Albertsma J, Gabriels D, et al. Carbon monoxide separation: past, present and future[J]. Chemical Society Reviews, 2023, 52(11): 3741-3777.
[8]	郭强, 肇启东, 肖永厚. 双回流变压吸附高效分离CO/H₂制备高纯H₂和CO[J]. 化工学报, 2024, 75(11): 4298-4308.
	Guo Q, Zhao Q D, Xiao Y H. Preparation of high-purity H₂ and CO by efficient separation of CO/H₂ using dual-reflux pressure swing adsorption process[J]. CIESC Journal, 2024, 75(11): 4298-4308.
[9]	Du S J, Huang J W, Ryder M R, et al. Probing sub-5 Ångstrom micropores in carbon for precise light olefin/paraffin separation[J]. Nature Communications, 2023, 14(1): 1197.
[10]	熊波, 陈健, 李克兵, 等. 工业排放气二氧化碳捕集与利用技术进展[J]. 低碳化学与化工, 2023, 48(1): 9-18.
	Xiong B, Chen J, Li K B, et al. Technical progress in carbon dioxide capture and utilization of industrial vent gas[J]. Low-Carbon Chemistry and Chemical Engineering, 2023, 48(1): 9-18.
[11]	Yang S H, Xiao Y H, Zhang W Y, et al. Facile preparation of C u ( Ⅰ ) / 5 A via one-step impregnation with highly dispersed CuCl in ethanol single solvent toward selective adsorption of CO from H₂ stream[J]. ACS Sustainable Chemistry & Engineering, 2022, 10(48): 15958-15967.
[12]	Yue X, Wang S, Gao J X, et al. Effects of mesopore size on ethyl acetate adsorption-desorption behaviors over hierarchical ZSM-5/MCM-41 molecular sieves[J]. Separation and Purification Technology, 2024, 336: 126228.
[13]	Xue C L, Hao W M, Cheng W P, et al. Effects of pore size distribution of activated carbon (AC) on CuCl dispersion and CO adsorption for CuCl/AC adsorbent[J]. Chemical Engineering Journal, 2019, 375: 122049.
[14]	Oo W, Park J H, Zaw Win M, et al. Dual preservative effects of SnO₂-chitosan on Cu¹⁺-doped boehmite composites for stable CO adsorption properties[J]. Separation and Purification Technology, 2024, 348: 127631.
[15]	Li Y X, Zhong W, Zhou J J, et al. Reversible light-controlled CO adsorption via tuning π-complexation of Cu⁺ sites in azobenzene-decorated metal-organic frameworks[J]. Angewandte Chemie International Edition, 2022, 61(46): e202212732.
[16]	Wang Q, Wang M, Chen H W, et al. Fluorination strategy in GME zeolitic imidazolate frameworks for enhanced ethane/ethylene separation[J]. Separation and Purification Technology, 2025, 364: 132423.
[17]	Li C L, Wang J, Wang Z F, et al. Understanding the vacuum autoreduction behavior of Cu species in CuCl/NaY adsorbent for CO/N₂ separation[J]. Microporous and Mesoporous Materials, 2024, 365: 112904.
[18]	Rao F, Liu M L, Liu C H, et al. Synthesis of binder-free pelletized Y zeolite for CO₂ capture[J]. Carbon Capture Science & Technology, 2024, 10: 100166.
[19]	文一如, 付佳, 刘大欢. 基于机器学习的MOFs材料研究进展: 能源气体吸附分离[J]. 化工学报, 2024, 75(4): 1370-1381.
	Wen Y R, Fu J, Liu D H. Advances in machine learning-based materials research for MOFs: energy gas adsorption separation[J]. CIESC Journal, 2024, 75(4): 1370-1381.
[20]	Pardakhti M, Moharreri E, Wanik D, et al. Machine learning using combined structural and chemical descriptors for prediction of methane adsorption performance of metal organic frameworks (MOFs)[J]. ACS Combinatorial Science, 2017, 19(10): 640-645.
[21]	Yuan X Z, Suvarna M, Low S, et al. Applied machine learning for prediction of CO₂ adsorption on biomass waste-derived porous carbons[J]. Environmental Science & Technology, 2021, 55(17): 11925-11936.
[22]	Tao W Y, Zhao W K, Zhao Q D, et al. Ensemble-learning-guided optimization design for metal-organic framework adsorbents toward CO adsorption[J]. Inorganic Chemistry, 2025, 64(18): 9237-9250.
[23]	Tao W Y, Cui Y J, Zhao Q D, et al. Prediction of the enhanced performance of C u ( Ⅰ ) - m o d i f i e d porous materials towards CO adsorption by using tree-based machine learning models[J]. Separation and Purification Technology, 2025, 359: 130850.
[24]	Li J, Liu T Y, Palansooriya K N, et al. Zeolite-catalytic pyrolysis of waste plastics: machine learning prediction, interpretation, and optimization[J]. Applied Energy, 2025, 382: 125258.
[25]	Al-Sakkari E G, Ragab A, So T M Y, et al. Machine learning-assisted selection of adsorption-based carbon dioxide capture materials[J]. Journal of Environmental Chemical Engineering, 2023, 11(5): 110732.
[26]	Zhang W J, Chen J F, Huang G H, et al. Machine learning-assisted prediction and exploration of the homogeneous oxidation of mercury in coal combustion flue gas[J]. Environmental Science & Technology, 2025, 59(22): 11073-11082.
[27]	Jayarathna R, Onsree T, Drummond S, et al. Experimental discovery of novel ammonia synthesis catalysts via active learning[J]. Journal of Materials Chemistry A, 2024, 12(5): 3046-3060.
[28]	Xu H, Mguni L L, Yao Y L, et al. Machine learning-assisted high-throughput screening of MOFs for efficient adsorption and separation of CF₄/N₂ [J]. Journal of Cleaner Production, 2024, 461: 142634.
[29]	Xiong T, Cui J W, Hou Z M, et al. Prediction of arsenic adsorption onto metal organic frameworks and adsorption mechanisms interpretation by machine learning[J]. Journal of Environmental Management, 2023, 347: 119065.
[30]	Zheng G T, Zhang S Y, Meng L H, et al. Machine learning-guided design and synthesis of eco-friendly poly(ethylene oxide) membranes for high-efficacy CO₂/N₂ separation[J]. Advanced Functional Materials, 2024, 34(51): 2410075.

Dataset	Mean	Std. deviation	Min	25% percentile	50% percentile	75% percentile	Max	Missing data/%
MPD/Å	6.62	1.47	5.00	5.00	7.30	8.10	8.10	0.00
LPD/Å	5.54	1.66	3.80	4.10	4.10	7.40	7.40	0.00
cage size/Å	10.95	1.56	0.00	11.20	11.20	11.40	11.40	0.00
VS	1.46	0.69	0.00	1.00	1.00	2.00	3.00	0.00
Si/Al	5.05	30.13	0.20	1.00	1.00	2.35	450.00	0.00
SSA/(m²/g)	547.07	197.15	149.86	392.00	579.70	669.90	1005.00	0.00
TPV/(cm³/g)	0.28	0.13	0.07	0.18	0.27	0.31	0.64	1.85
Cu(Ⅰ) loading/%(mass)	5.77	10.71	0.00	0.00	0.00	10.39	46.10	0.00
T/K	288.09	37.69	150.00	293.00	298.00	303.00	338.00	0.00
P/kPa	294.90	291.43	1.25	80.00	103.15	500.00	1000.00	0.00
CO adsorption capacity/(mmol/g)	1.88	0.95	0.01	1.15	1.89	2.56	4.35	0.00

Dataset	Mean	Std. deviation	Min	25% percentile	50% percentile	75% percentile	Max	Missing data/%
MPD/Å	6.62	1.47	5.00	5.00	7.30	8.10	8.10	0.00
LPD/Å	5.54	1.66	3.80	4.10	4.10	7.40	7.40	0.00
cage size/Å	10.95	1.56	0.00	11.20	11.20	11.40	11.40	0.00
VS	1.46	0.69	0.00	1.00	1.00	2.00	3.00	0.00
Si/Al	5.05	30.13	0.20	1.00	1.00	2.35	450.00	0.00
SSA/(m²/g)	547.07	197.15	149.86	392.00	579.70	669.90	1005.00	0.00
TPV/(cm³/g)	0.28	0.13	0.07	0.18	0.27	0.31	0.64	1.85
Cu(Ⅰ) loading/%(mass)	5.77	10.71	0.00	0.00	0.00	10.39	46.10	0.00
T/K	288.09	37.69	150.00	293.00	298.00	303.00	338.00	0.00
P/kPa	294.90	291.43	1.25	80.00	103.15	500.00	1000.00	0.00
CO adsorption capacity/(mmol/g)	1.88	0.95	0.01	1.15	1.89	2.56	4.35	0.00

Model	Test R²	RMSE	MAE	Outer CV R²
GBDT	0.90	0.29	0.17	0.87
RF	0.85	0.35	0.26	0.84
ERT	0.88	0.31	0.23	0.86
XGB	0.91	0.29	0.16	0.87

Model	Test R²	RMSE	MAE	Outer CV R²
GBDT	0.90	0.29	0.17	0.87
RF	0.85	0.35	0.26	0.84
ERT	0.88	0.31	0.23	0.86
XGB	0.91	0.29	0.16	0.87

Model	Optimal hyperparameter
GBDT	{'learning_rate': 0.2, 'max_depth': 8, 'min_samples_leaf': 3, 'min_samples_split': 10, 'n_estimators': 200, 'subsample': 0.8}
RF	{'max_depth': 10, 'max_features': 'sqrt', 'min_samples_leaf': 1, 'min_samples_split': 2, 'n_estimators': 50}
ERT	{'max_depth': 7, 'min_samples_leaf': 1, 'min_samples_split': 2, 'n_estimators': 100}
XGB	{'colsample_bytree': 0.8, 'learning_rate': 0.1, 'max_depth': 7, 'n_estimators': 500, 'subsample': 0.5}