机器学习辅助高通量筛选共价有机框架材料吸附铼酸

doi:10.19965/j.cnki.iwt.2025-0499

摘要/Abstract

摘要：

高效去除核废水中高浓度锝酸盐（⁹⁹TcO₄⁻）至关重要。现有吸附剂容量低，再生频繁，核素再释放风险高。共价有机框架（COFs）在吸附铼酸（锝酸替代物）方面潜力显著，但传统实验方式筛选高性能吸附剂效率较低。为此，提出基于机器学习的COFs筛选策略。通过收集文献中关于COFs特性和铼酸吸附实验参数的多维数据，构建了一个用于预测铼酸吸附性能的数据集。随机森林算法训练的模型表现最佳，可用来模拟吸附过程并从公开数据库中筛选出高容量吸附剂。关键结构特性分析表明，调控吸附剂孔直径（2~5 nm）和原子加权电负性（0.8~1.0）能显著提升水溶液中铼酸吸附性能。此策略能有效识别高性能水处理吸附剂，加速COFs材料在污染物去除中的应用，为发现优势吸附剂提供了新方法。

关键词: 机器学习, 共价有机框架, 铼酸吸附, 高通量筛选, 模型预测

Abstract:

Efficient removal of high concentrations of pertechnetate (⁹⁹TcO₄⁻) from nuclear wastewater is critical. Current adsorbents have limited capacity and require frequent regeneration, thereby increasing the risk of nuclide re-release. Covalent organic frameworks (COFs) show significant potential for adsorbing rhenate acid (ReO₄⁻) (a surrogate for ⁹⁹TcO₄⁻), but traditional experimental screening for effective COFs is inefficient. Consequently, we proposed using machine learning to screen COFs. A comprehensive dataset was created from literature data on COFs properties and ReO₄⁻ adsorption parameters. A random forest model, trained on this data, performed well and was used to simulate adsorption and identify high-performance adsorbents from a public database. Key structural characteristics indicated that adjusting the pore limited diameter (2-5 nm) and weighted atomic electronegativity (0.8-1.0) of adsorbents notably improved ReO₄⁻ adsorption. This approach efficiently pinpointed top adsorbents, accelerated COFs applications in pollutant removal, and provided a novel method for discovering optimized adsorbents.

Key words: machine learning, covalent organic frameworks, rhenate acid adsorption, high-throughput screening, predictive modeling

中图分类号:

X703

袁岭, 张涵, 陈晨, 赵昕, 吕路, 张炜铭. 机器学习辅助高通量筛选共价有机框架材料吸附铼酸[J]. 工业水处理, 2025, 45(12): 65-73.

Ling YUAN, Han ZHANG, Chen CHEN, Xin ZHAO, Lu LÜ, Weiming ZHANG. Machine-learning-guided high-throughput screening of covalent organic framework materials for rhenate adsorption[J]. Industrial Water Treatment, 2025, 45(12): 65-73.

https://www.iwt.cn/CN/Y2025/V45/I12/65

图/表 7

图1 本研究工作流程

Fig.1 Workflow of this study

图2 ML模型中数值变量的统计分布注：方框中的星形代表平均值，曲线代表数据分布的相对密度。

Fig.2 Statistical distributions of numeric variables for ML models

图3 所有描述符之间的皮尔逊相关系数矩阵

Fig.3 Pearson correlation coefficient matrix between all descriptors

图4 RF模型中输入变量的特征重要性（a）和基于递归特征消除的最佳描述符组合（b）

Fig.4 Feature importance of input variables in RF model（a） and optimal descriptor combination based on recursive feature elimination（b）

图5 用于预测ReO₄ ^-吸附到COFs上的ML算法的构建和性能注：（a）和（b）RF算法的不同测试集比例下的R ²和RMSE结果；（c）和（e）RF算法的超参数调优；（f）XGBoost算法的超参数调优；（d）和（g）RF算法和XGBoost算法的预测性能；（h）ML算法的R ²和RMSE；（i）以新发表数据作为验证集，RF算法的模型验证。

Fig.5 Construction and performance of ML algorithms for predicting ReO₄ ^- adsorption onto COFs

图6 pH和初始浓度的交互特征分析（a）；初始浓度的部分依赖性分析（b）；pH的部分依赖性分析（c）；吸附剂量的部分依赖性分析（d）注：实验参数的PDP图中，蓝色线表示模型对特征变化的响应，每个图表下方的黑色条形图用于展示数据值分布的密度。

Fig.6 Interaction characteristic analysis of pH and initial concentration （a）， partial dependence analysis of initial concentration（b）， partial dependence analysis of pH（c）， partial dependence analysis of adsorbent dose （d）

图7 COFs的结构-吸附容量关系

Fig.7 Structure-adsorption capacity relationships of COFs

参考文献 22

[1]	武莉娅，顾平，张光辉. 去除放射性废水中锶的研究进展［J］．工业水处理，2013，33（1）：6-9．
	WU Liya， GU Ping， ZHANG Guanghui. Research progress in the removal of strontium from radioactive wastewater［J］. Industrial Water Treatment，2013，33（1）：6-9．
[2]	ROSA E A， TULER S P， FISCHHOFF B，et al. Nuclear waste：Knowledge waste ［J］. Science，2010，329（5993）：762-763． doi:10.1126/science.1193205
[3]	WANG Yougan， HU Qinghua， HUANG Juan，et al. Synthesis of cationic polymer decorated with halogen for highly efficient trapping ⁹⁹TcO₄ ^-/ReO₄ ^- ［J］. Journal of Hazardous Materials，2023，443：130325． doi:10.1016/j.jhazmat.2022.130325
[4]	YANG Heng， DENG Hao， LIANG Pengliang，et al. Photocatalytic reduction of perrhenate and pertechnetate in a strongly acidic aqueous solution［J］. Environmental Science & Technology，2024，58（27）：12237-12248． doi:10.1021/acs.est.4c02511
[5]	ZHU Lin， SHENG Daopeng， XU Chao，et al. Identifying the recognition site for selective trapping of ⁹⁹TcO₄ ^- in a hydrolytically stable and radiation resistant cationic metal-organic framework［J］．Journal of the American Chemical Society，2017，139（42）：14873-14876．
[6]	LI Jie， DAI Xing， ZHU Lin，et al. ⁹⁹TcO₄ ^- remediation by a cationic polymeric network［J］. Nature Communications，2018，9：3007． doi:10.1038/s41467-018-05380-5
[7]	LOHSE M S， BEIN T. Covalent organic frameworks：Structures，synthesis，and applications［J］. Advanced Functional Materials，2018，28（33）：1705553． doi:10.1002/adfm.201870229
[8]	何边燕，王玉冰，李珊珊，等. 共价有机框架材料的合成策略及其在重金属离子吸附中的研究进展［J/OL］. 化工进展，2024.DOI：10.16085/j.issn.1000-6613.2024-1408．
	HE Bianyan， WANG Yubing， LI Shanshan，et al. Recent progress on synthesis strategies of covalent organic framework materials and their adsorption application for heavy metal ions［J/OL］. Chemical Industry and Engineering Progress，2024.DOI：10.16085/j.issn.1000-6613.2024-1408．
[9]	DAS S， FENG Jie， WANG Wei. Covalent organic frameworks in separation［J］. Annual Review of Chemical and Biomolecular Engineering，2020，11：131-153． doi:10.1146/annurev-chembioeng-112019-084830
[10]	OUYANG Zhengdong， LI Xiaofeng， ZHANG Jie，et al. Zwitterionic adsorbents derived from self-Exfoliated ionic covalent organic frameworks for efficient co-removal of anionic and cationic radionuclides［J］. Chemical Engineering Journal，2024，497：154882． doi:10.1016/j.cej.2024.154882
[11]	XIONG Xiaohong， YU Zhiwu， GONG Lele，et al. Ammoniating covalent organic framework（COF） for high-performance and selective extraction of toxic and radioactive uranium ions［J］. Advanced Science，2019，6（16）：1900547． doi:10.1002/advs.201900547
[12]	YANG Chenghao， CHANG J S， LEE D J. Covalent organic framework EB-COF：Br as adsorbent for phosphorus （Ⅴ） or arsenic （Ⅴ） removal from nearly neutral waters［J］. Chemosphere，2020，253：126736． doi:10.1016/j.chemosphere.2020.126736
[13]	冯丁，刘晶静，马文丹，等. 基于机器学习的生物炭吸附重金属建模研究进展［J］. 工业水处理，2024，44（12）：1-11．
	FENG Ding， LIU Jingjing， MA Wendan，et al. Research progress on modeling of heavy metal adsorption by biochar based on machine learning［J］. Industrial Water Treatment，2024，44（12）：1-11．
[14]	XU Hong， MGUNI L L， YAO Yali，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． doi:10.1016/j.jclepro.2024.142634
[15]	许慕舰，计成汉，袁岭，等. 基于混合模型的重金属吸附树脂穿透曲线预测［J］. 工业水处理，2025，45（5）：54-61．
	XU Mujian， JI Chenghan， YUAN Ling，et al. Breakthrough curve prediction of resins for heavy metal adsorption by hybrid model［J］. Industrial Water Treatment，2025，45（5）：54-61．
[16]	YUAN Ling， ZHANG Han， YU Hang，et al. Machine learning reveals key adsorption mechanisms for oxyanions based on combination of experimental and published literature data［J］. Environmental Science & Technology，2025，59（22）：11401-11413． doi:10.1021/acs.est.5c03992
[17]	QIU Ye， LI Zhejun， ZHANG Tong，et al. Predicting aqueous sorption of organic pollutants on microplastics with machine learning［J］. Water Research，2023，244：120503． doi:10.1016/j.watres.2023.120503
[18]	JIANG Mingxing， FU Weiwei， WANG Ying，et al. Machine-learning-driven discovery of metal-organic framework adsorbents for hexavalent chromium removal from aqueous environments［J］．Journal of Colloid and Interface Science，2024，662：836-845． doi:10.1016/j.jcis.2024.02.084
[19]	SHEN Tian， PENG Haoyi， YUAN Xingzhong，et al. Feature engineering for improved machine-learning-aided studying heavy metal adsorption on biochar［J］. Journal of Hazardous Materials，2024，466：133442． doi:10.1016/j.jhazmat.2024.133442
[20]	ZHANG Jing， FU Kaixing， WANG Dawei，et al. Refining hydrogel-based sorbent design for efficient toxic metal removal using machine learning-Bayesian optimization［J］. Journal of Hazardous Materials，2024，479：135688． doi:10.1016/j.jhazmat.2024.135688
[21]	ONGARI D， YAKUTOVICH A V， TALIRZ L，et al. Building a consistent and reproducible database for adsorption evaluation in covalent-organic frameworks［J］. ACS Central Science，2019，5（10）：1663-1675． doi:10.1021/acscentsci.9b00619
[22]	LAN Youshi， HAN Xianghao， TONG Minman，et al. Materials genomics methods for high-throughput construction of COFs and targeted synthesis［J］. Nature Communications，2018，9：5274． doi:10.1038/s41467-018-07720-x

选择文件类型/文献管理软件名称

选择包含的内容