The interaction model parameters are the key thermodynamic parameters in activity coefficient models such as Wilson and non-random two-liquid (NRTL) models. Their accuracy directly determines the reliability of the results of vapor–liquid phase equilibrium and separation process calculations. The core bottlenecks of traditional parameter acquisition methods relying on experimental data regression and universal functional activity coefficient (UNIFAC) model estimation include limited data coverage, unstable solution of nonlinear equations and limited scope of application. To address these issues, we constructed a dual-branch graph neural network (GNN) model based on molecular structure characterization enabling efficient and direct prediction of the parameters of the Wilson interaction model. This model takes molecular diagrams as input, separately extracts the molecular features of solutes and solvents, and, through the message-passing mechanism, deeply learns the potential structural correlations between molecules. At the same time, we introduce a parallel residual strategy to reduce the prediction error. Using 4 068 binary system interaction model parameter data for training and testing, our proposed parallel residual GNN model significantly outperforms the UNIFAC method across multiple error indicators, and compared with the single GNN model, the prediction accuracy is further improved. The predicted parameters Aij, Aji, Bij, Bji of the Wilson interaction model were substituted into the Wilson equation to calculate the infinite dilution activity coefficient γ. The results were highly consistent with the calculated value in the literature, verifying the thermodynamic rationality and engineering applicability of our model.
LIZ, CONSTANTINOUL, BAURR, et al. Review of group contribution methods for prediction of thermodynamic properties of long-chain hydrocarbons[J]. Molecular Physics, 2025, 123(21-22): e2563020.
[2]
GMEHLINGJ, CONSTANTINESCUD, SCHMIDB. Group contribution methods for phase equilibrium calculations[J]. Annual Review of Chemical and Biomolecular Engineering, 2015, 6: 267-292.
[3]
HAYERN, WENDELT, MANDTS, et al. Advancing thermodynamic group-contribution methods by machine learning: UNIFAC 2.0[J]. Chemical Engineering Journal, 2025, 504: 158667.
[4]
CLETOP B, DURAUV G, KANDAL R S, et al. Phase equilibrium measurements and thermodynamic modeling of CO2+guaiacol, CO2+guaiacol and ethanol, and CO2+guaiacol, ethanol, and water[J]. Fluid Phase Equilibria, 2025, 591: 114319.
[5]
MARCILLAA, REYES-LABARTAJ A, OLAYAM M. Should we trust all the published LLE correlation parameters in phase equilibria? Necessity of their assessment prior to publication[J]. Fluid Phase Equilibria, 2017, 433: 243-252.
[6]
CRUZ SILVA KDA, ABREUC S, VIEIRAA W, et al. Aqueous two-phase systems formed by different phase-forming components: equilibrium diagrams and dye partitioning study[J]. Fluid Phase Equilibria, 2020, 520: 112664.
[7]
WINTERB, WINTERC, ESPERT, et al. SPT-NRTL: a physics-guided machine learning model to predict thermodynamically consistent activity coefficients[J]. Fluid Phase Equilibria, 2023, 568: 113731.
[8]
MEDINAE I S, LINKES, STOLLM, et al. Graph neural networks for the prediction of infinite dilution activity coefficients[J]. Digital Discovery, 2022, 1(3): 216-225.
RITTIGJ G, GAOQ, DAHMENM, et al. Graph neural networks for the prediction of molecular structure-property relationships[PP/OL]. V1. arXiv(2022-07-25)[2026-03-04].
WANGY, LIZ, FARIMANIA B. Graph neural networks for molecules[PP/OL]. V2. arXiv(2023-02-06)[2026-03-04].
[13]
QUC, SCHNEIDERB I, KEARSLEYA J, et al. Applying graph neural network models to molecular property prediction using high-quality experimental data[J]. Artificial Intelligence Chemistry, 2024, 2(1): 100050.
[14]
MURGODT R, REDDYP S, GADDAMS, et al. A survey on graph neural networks and its applications in various domains[J]. SN Computer Science, 2025, 6: 26.
[15]
ZHANGS, TONGH, XUJ, et al. Graph convolutional networks: a comprehensive review[J]. Computational Social Networks, 2019, 6: 11.
[16]
THEISENM F, MEESTERSG M H, SCHWEIDTMANNA M. Graph neural networks for soft sensors: learning from process topology and operational data[J]. Computers and Chemical Engineering, 2026, 206: 109532.
[17]
BENTOA P, HERSEYA, FÉLIXE, et al. An open source chemical structure curation pipeline using RDKit[J]. Journal of Cheminformatics, 2020, 12: 51.
[18]
WUJ N, WANGT, CHENY, et al. t-SMILES: a fragment-based molecular representation framework for de novo ligand design[J]. Nature Communications, 2024, 15: 4993.
[19]
HASHEMIANS, KHANZ, KALHANP, et al. Learning SMILES semantics: Word2Vec and transformer embeddings for molecular property prediction[J]. Algorithms, 2025, 18: 547.
[20]
SCHWEIDTMANNA M, RITTIGJ G, KÖNIGA, et al. Graph neural networks for prediction of fuel ignition quality[J]. Energy & Fuels, 2020, 34(9): 11395-11407.
[21]
GILMERJ, SCHOENHOLZS S, RILEYP F, et al. Neural message passing for quantum chemistry[C]//Proceedings of the 34th International Conference on Machine Learning. Sydney, 2017: 1263-1272.
[22]
SHENC, WANGQ, PRIEBEC E. One-hot graph encoder embedding[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(6): 7933-7938.
[23]
FEY M, LENSSENJ E. Fast graph representation learning with PyTorch Geometric[PP/OL]. V3. arXiv(2019-04-25)[2026-03-04].
[24]
VON STOSCHM, OLIVEIRAR, PERESJ, et al. Hybrid semi-parametric modeling in process systems engineering: past, present and future[J]. Computers and Chemical Engineering, 2014, 60: 86-101.
[25]
GLASSEYJ, VON STOSCHM. Hybrid modeling in process industries[M].Boca Raton: CRC Press, 2018: 19-44.
[26]
MCBRIDEK, MEDINAE I S, SUNDMACHERK. Hybrid semi-parametric modeling in separation processes: a review[J]. Chemie Ingenieur Technik, 2020, 92(7): 842-855.
CONSTANTINESCUD, GMEHLINGJ. Further development of modified UNIFAC (Dortmund): revision and extension 6[J]. Journal of Chemical & Engineering Data, 2016, 61(8): 2738-2748.
[29]
JIRASEKF, BAMLERR, MANDTS. Hybridizing physical and data-driven prediction methods for physicochemical properties[J]. Chemical Communications, 2020, 56(82): 12407-12410.
[30]
LVJ, WANGL. Hybrid modeling of adsorption process using mass transfer and machine learning techniques for concentration prediction[J]. Journal of Saudi Chemical Society, 2025, 29: 12.
[31]
GARGALOC L, MALANCAA A, AOUICHAOUIA R N, et al. Navigating industry 4.0 and 5.0: the role of hybrid modelling in (bio)chemical engineering’s digital transition[J]. Frontiers in Chemical Engineering, 2024, 6: 1494244.
[32]
MIENYEI D, SUNY. A survey of ensemble learning: concepts, algorithms, applications, and prospects[J]. IEEE Access, 2022, 10: 99129-99149.
[33]
ZHAOH K, LIUW H, WANGY X, et al. Comparative analysis of algorithmic approaches in ensemble learning: bagging vs. boosting[J]. Scientific Reports, 2025, 15: 34218.
[34]
WOOLEYR J, PUTSCHEV. Development of an ASPEN PLUS physical property database for biofuels components[R]. Golden: National Renewable Energy Laboratory (NREL), 1996.
[35]
WILSONG M. Vapor-liquid equilibrium. XI. A new expression for the excess free energy of mixing[J]. Journal of the American Chemical Society, 1964, 86(2): 127-130.
[36]
DOMINGO-FERRERJ, TORRAV. Median-based aggregation operators for prototype construction in ordinal scales[J]. International Journal of Intelligent Systems, 2003, 18(6): 633-655.
[37]
BROUWERT, SCHUURB. Model performances evaluated for infinite dilution activity coefficients prediction at 298.15 K[J]. Industrial & Engineering Chemistry Research, 2019, 58(20): 8903-8914.