Position：Home >> Abstract

Machine learning-based prediction and mechanistic analysis of water flux in thin-film nanocomposite membranes
Authors： LIU Jinyun1, YANG Haonan1, JI Zhanlin2, DONG Yingchao3
Units： 1. College of Artificial Intelligence, North China University of Science and Technology, 
Tangshan 063210, China; 2. College of Mathematics and Computer Science, Zhejiang A&F University, 
Hangzhou 311300, China; 3. School of Science and Engineering, The Chinese University 
of Hong Kong,  Shenzhen, Shenzhen 518172, China
KeyWords： thin-film nanocomposite membranes; seawater desalination; water flux; machine learning; artificial intelligence
ClassificationCode：TQ028.8
year,volume(issue):pagination： 2026, 46(4):164-171

Abstract：

To break through the performance bottleneck of TFN membranes and achieve efficient desalination and water treatment, this study constructed a data-driven framework that integrated machine learning with explainable artificial intelligence (XAI) to predict water flux and analyze structure-performance relationships. Through Grid Search and cross-validation, seven mainstream algorithms were systematically tuned and compared; the optimal model achieved an R2 of 0.860 0 and an MAE of 5.290 8 on the test set. XAI-based analysis revealed that the geometric characteristics of nanoparticles significantly affected mass transfer behavior, while a clear interaction existed between loading and surface wettability. Moreover, factors such as the structural evolution of the polyamide selective layer, interfacial hydration, and particle agglomeration under high loading were closely linked to nonlinear mass transfer behavior. Particle swarm optimization (PSO) further validated the rationality of the identified high-flux parameter window. The results demonstrate that this data-driven approach can accurately predict TFN membrane performance, uncover key structure-performance regularities, and provide effective guidance for the design of high-performance membrane materials.

Funds：

国家自然科学基金委员会与香港研究资助局联合项目（52261160381）； 深圳市基础研究计划一般项目（JCYJ20240813113500002）； 香港中文大学（深圳）大学发展基金（UDF01003350）

AuthorIntro：
刘劲芸（1987-），女，河北唐山人，主要研究方向为机器学习、图像处理.*通讯作者，纪占林，E-mail：jizhanlin@zafu.edu.cn;董应超，E-mail：ycdong@cuhk.edu.cn

Reference：

[1]郭艳艳, 宋程燕, 苏海建, 等. 碳基太阳能海水淡化复合膜的设计及其性能研究 ［J］. 膜科学与技术, 2025, 45(4): 133-142.
［2］Gao L, Lin J, Wang L, et al. Machine learning-assisted design of advanced polymeric materials ［J］. Acc Mater Res, 2024, 5(5): 571-584.
［3］Ji Z, Wang M, Guan H, et al. Artificial intelligence-powered MOF-incorporated thin-film nanocomposite desalination membranes for sustainable water purification ［J］. Desalination, 2026, 621: 119702.
［4］Laqueur H S, Shev A B, Kagawa R M C. SuperMICE: An ensemble machine learning approach to multiple imputation by chained equations ［J］. Am J Epidemiol, 2022, 191(3): 516-525.
［5］Abbaszadeh M, Soltani-Mohammadi S, Ahmed A N. Optimization of support vector machine parameters in modeling of Lju deposit mineralization and alteration zones using particle swarm optimization algorithm and grid search method ［J］. Comput Geosci, 2022, 165: 105140.
［6］Li X, Chen S, Song Z, et al. Genetic algorithms and Pearson correlation analysis synergy optimizes thermodynamic performance in multi-dephlegmation auto-cascade refrigeration cycle ［J］. Appl Therm Eng, 2026, 289: 129944.
［7］Chen H, Covert I C, Lundberg S M, et al. Algorithms to estimate Shapley value feature attributions ［J］. Nat Mach Intell, 2023, 5(6): 590-601.
［8］Zhao X, Yang H, Yao Y, et al. Factors affecting traffic risks on bridge sections of freeways based on partial dependence plots ［J］. Phys A: Statist Mechan Appl, 2022, 598: 127343.
［9］Wang F, Zheng T, Xiong R, et al. Strong improvement of reverse osmosis polyamide membrane performance by addition of ZIF-8 nanoparticles: Effect of particle size and dispersion in selective layer ［J］. Chemosphere, 2019, 233: 524-531.
［10］Wang Q, Liu L, Liu C, et al. Size effect in determining the water diffusion rate in carbon nanotubes ［J］. J Mol Liq, 2021, 334: 116034.
［11］Chen L, Ren X, Li Y, et al. Enhancing interface compatibility of UiO-66-NH2 and polyamide by incorporating dopamine into thin film nanocomposite membranes ［J］. J Membr Sci, 2022, 654: 120565.
［12］Lee T H, Oh J Y, Hong S P, et al. ZIF-8 particle size effects on reverse osmosis performance of polyamide thin-film nanocomposite membranes: Importance of particle deposition ［J］. J Membr Sci, 2019, 570-571: 23-33.
［13］Abdolhosseini Qomi M J, Miller Q R S, Zare S, et al. Molecular-scale mechanisms of CO2 mineralization in nanoscale interfacial water films ［J］. Nat Rev Chem, 2022, 6(9): 598-613.

Service：
【Download】【Collect】

《膜科学与技术》编辑部 Address: Bluestar building, 19 east beisanhuan road, chaoyang district, Beijing; 100029 Postal code; Telephone：010-80492417/010-80485372; Fax：010-80485372 ; Email：mkxyjs@163.com

京公网安备11011302000819号