This is an unedited manuscript accepted for publication and provided as an Article in Press for early access at the author’s request. The article will undergo copyediting, typesetting, and galley proof review before final publication. Please be aware that errors may be identified during production that could affect the content. All legal disclaimers of the journal apply.
Manish Kumar Jha,
- Ph.D. Scholar, Department of Computer Science Patliputra University, Patna, Bihar, India
Abstract
Thermoset polymer nanocomposites exhibit properties that are highly sensitive to molecular scale formulation decisions, yet the vast design space remains largely unexplored because of the high cost of experimental characterisation and fully atomistic simulation. This paper presents TNC GNN, a dual mode graph neural network framework developed for the computational design of thermoset nanocomposite formulations. The forward module employs an attention augmented Message Passing Neural Network with 3D geometric encoding to predict five thermos mechanical properties: glass transition temperature (Tg), tensile modulus, fracture toughness, thermal conductivity, and coefficient of thermal expansion. The inverse design module couples the trained predictor with a conditional variational auto encoder to generate candidate molecular structures that satisfy user specified property targets. Trained on a curated set of 4,076 thermoset formulations (approximately 48% containing nano particle fillers), the model achieves a Tg mean absolute error of 4.2K and recovers 76.4% of held out target profiles within 10% relative error. Ablation studies show that 3D geometric encoding reduces Tg MAE by 42.5% relative to a 2D topology only baseline (from 7.3K to 4.2K). Forward inference requires less than 12 ms per formulation, making high throughput computational screening feasible. The work is presented strictly as a methodological and computational contribution at a simulation and literature grounded conceptual stage. No experimental synthesis or characterisation of any proposed candidate has been performed; experimental validation remains an essential next step.
Keywords: machine learning; graph neural networks; polymer nanocomposites; inverse design; computational materials science; thermal and mechanical properties
References
[1] Njuguna J, Pielichowski K. Polymer nanocomposites for aerospace applications: properties. Adv Eng Mater. 2003;5(11):769-778.
[2] Paul DR, Robeson LM. Polymer nanotechnology: nanocomposites. Polymer. 2008;49(15):3187-3204.
[3] Grand View Research. Polymer nanocomposites market size, share & trends analysis report by type, by end-use, by region. San Francisco (CA): Grand View Research; 2024.
[4] May CA. Epoxy resins: chemistry and technology. 2nd ed. New York (NY): Marcel Dekker; 1988.
[5] Pascault JP, Williams RJJ. Epoxy polymers: new materials and innovations. Weinheim: Wiley-VCH; 2010.
[6] Doherty CM, Knapp CG, Clancy AJ. Machine learning approaches to accelerate thermoset composite formulation. Compos Sci Technol. 2022;218:109185.
[7] Zhao G, Xu T, Fu X, Zhao W, Wang L, Lin J, et al. Machine-learning-assisted multiscale modeling strategy for predicting mechanical properties of carbon fiber reinforced polymers. Compos Sci Technol. 2024;248:110455.
[8] Chen L, Pilania G, Batra R, Huan TD, Kim C, Kuenneth C, et al. Polymer informatics: current status and critical next steps. Mater Sci Eng R Rep. 2021;144:100595.
[9] Mukherji D, Stuehn T, Kremer K. Molecular dynamics simulation of epoxy nanocomposites: network topology and mechanical performance. Macromolecules. 2022;55(4):1412-1428.
[10] Bandyopadhyay A, Odegard GM. Molecular modeling of crosslink distribution in epoxy polymers. Model Simul Mater Sci Eng. 2012;20(4):045018.
[11] Zhou J, Cui G, Hu S, Zhang Z, Yang C, Liu Z, et al. Graph neural networks: a review of methods and applications. AI Open. 2020;1:57-81.
[12] Wu Z, Pan S, Chen F, Long G, Zhang C, Yu PS. A comprehensive survey on graph neural networks. IEEE Trans Neural Netw Learn Syst. 2021;32(1):4-24.
[13] Reiser P, Neubert M, Eberhard A, Torresi L, Zhou C, Shao C, et al. Graph neural networks for materials science and chemistry. Commun Mater. 2022;3:93.
[14] Schütt KT, Kindermans PJ, Sauceda HE, Chmiela S, Tkatchenko A, Müller KR. SchNet: a continuous-filter convolutional neural network for modeling quantum interactions. Adv Neural Inf Process Syst. 2017;30:991-1001.
[15] Gasteiger J, Giri S, Margraf JT, Günnemann S. Fast and uncertainty-aware directional message passing for non-equilibrium molecules. arXiv [Preprint]. 2020:arXiv:2011.14115.
[16] Gilmer J, Schroeder SC, Riley PF, Vinyals O, Dahl GE. Neural message passing for quantum chemistry. In: Proceedings of the 34th International Conference on Machine Learning; 2017. p. 1263-1272.
[17] Queen O, McCarver GA, Thatigotla S, Bhansali SP, Bera C, Pilania G, et al. Polymer graph neural networks for multitask property learning. NPJ Comput Mater. 2023;9:90.
[18] Kuenneth C, Ramprasad R. Polymer informatics at scale with multitask graph neural networks. Chem Mater. 2023;35(4):1560-1567.
[19] Sanchez-Lengeling B, Aspuru-Guzik A. Inverse molecular design using machine learning: generative models for matter engineering. Science. 2018;361(6400):360-365.
[20] Zheng Z, Ravikumar N, Sutton C, Bligaard T, Reuter K. AI-guided inverse design and discovery of recyclable vitrimeric polymers. Adv Sci (Weinh). 2024;11(15):e2411385.
[21] Kingma DP, Welling M. Auto-encoding variational Bayes. arXiv [Preprint]. 2013:arXiv:1312.6114.
[22] Goodfellow I, Pouget-Abadie J, Mirza M, Xu B, Warde-Farley D, Ozair S, et al. Generative adversarial nets. Adv Neural Inf Process Syst. 2014;27:2672-2680.
[23] Ho J, Jain A, Abbeel P. Denoising diffusion probabilistic models. Adv Neural Inf Process Syst. 2020;33:6840-6851.
[24] Yan C, Feng X, Li G. From drug molecules to thermoset shape memory polymers: a machine learning approach. ACS Appl Mater Interfaces. 2021;13(52):60508-60521.
[25] Xu X, Ke Z, Yang X, Hu Z, Li F, Kang Z, et al. Discovery of thermosetting polymers with low hygroscopicity, low thermal expansivity, and high modulus by machine learning. J Mater Chem A. 2023;11(24):12918-12927.
[26] Rogers D, Hahn M. Extended-connectivity fingerprints. J Chem Inf Model. 2010;50(5):742-754.
[27] Bicerano J. Prediction of polymer properties. 3rd ed. Boca Raton (FL): CRC Press; 2002.
[28] Zhao Y, Mulder RJ, Houshyar S, Le TC. A review on the application of molecular descriptors and machine learning in polymer design. Polym Chem. 2023;14(29):3325-3346.
[29] Kuenneth C, Schertzer W, Ramprasad R. Copolymer informatics with multitask deep neural networks. Macromolecules. 2021;54(13):5957-5961.
[30] Gao Q, Dukker T, Schweidtmann AM, Weber JM. Self-supervised graph neural networks for polymer property prediction. Mol Syst Des Eng. 2024;9:1130-1143.
[31] Cui Y, Zhu J, Zhou W, Zang H, Ren Y, Xu J, et al. SA-GNN: prediction of material properties using graph neural network based on multi-head self-attention optimization. AIP Adv. 2024;14(5):055033.
[32] Chen C, Ye W, Zuo Y, Zheng C, Ong SP. Graph networks as a universal machine learning framework for molecules and crystals. Chem Mater. 2019;31(9):3564-3572.
[33] Lou Q, Guo K, Li B, Bhatt M, Smidt T, Krishnamurthy A, et al. Scalable deeper graph neural networks for high-performance materials property prediction. Patterns (N Y). 2022;3(5):100491.
[34] Li D, Ru Y, Liu J. GATBoost: mining graph attention networks-based important substructures of polymers for a better property prediction. Mater Today Commun. 2024;38:107577.
[35] Shim H, Sakamoto S, Kanemura T, Tsuji Y, Ishitani A, Yoshida K. Knowledge-embedded message-passing neural networks: improving molecular property prediction with human knowledge. ACS Omega. 2021;6(48):32569-32577.
[36] Shahriari B, Swersky K, Wang Z, Adams RP, de Freitas N. Taking the human out of the loop: a review of Bayesian optimization. Proc IEEE. 2016;104(1):148-175.
[37] Coley CW, Barzilay R, Green WH, Jaakkola TS, Jensen KF. Convolutional embedding of attributed molecular graphs for physical property prediction. J Chem Inf Model. 2017;57(8):1757-1772.
[38] Kingma DP, Welling M. An introduction to variational autoencoders. Found Trends Mach Learn. 2019;12(4):307-392.
[39] Sohn K, Lee H, Yan X. Learning structured output representation using deep conditional generative models. Adv Neural Inf Process Syst. 2015;28:3483-3491.
[40] Yue T, Tao L, Varshney V, Li Y. Benchmarking study of deep generative models for inverse polymer design. Digit Discov. 2025;4:910.
[41] Chen X, Li Y. On-demand reverse design of polymers with PolyTAO. NPJ Comput Mater. 2024;10:264.
[42] Ding Z, Attar HR, Wang H, Liu H, Li N. Integrating convolutional neural network and constitutive model for rapid prediction of stress-strain curves in fibre reinforced polymers: a generalisable approach. Mater Des. 2024;241:112849.
[43] Yi Y, An HW, Wang H. Machine-learning-assisted design of highly tough thermosetting polymers. ACS Appl Mater Interfaces. 2022;14(38):43796-43805.
[44] Guo J, Kim JH, Lei Y. Thermal conductivity prediction of epoxy composites using machine learning. Mater Des. 2021;206:109808.
[45] Xu X, Chen J, Zhou J, Li B. Thermal conductivity of polymers and their nanocomposites. Adv Mater. 2018;30(17):1705544.
[46] Landrum G. RDKit: open-source cheminformatics software. Version 2023.09.1. Available from: https://www.rdkit.org
[47] Riniker S, Landrum GA. Better informed distance geometry: using what we know to improve conformation generation. J Chem Inf Model. 2015;55(12):2562-2574.
[48] Veličković P, Cucurull G, Casanova A, Romero A, Liò P, Bengio Y. Graph attention networks. arXiv [Preprint]. 2017:arXiv:1710.10903.
[49] Huang G, Sun Y, Liu Z, Sedra D, Weinberger KQ. Deep networks with stochastic depth. In: Proceedings of ECCV; 2016. p. 646-661.
[50] Higgins I, Matthey L, Pal A, Burgess C, Glorot X, Botvinick M, et al. Beta-VAE: learning basic visual concepts with a constrained variational framework. In: Proceedings of ICLR; 2017.
[51] Doan Tran H, Kim C, Chen L, Chandrasekaran A, Batra R, Venkatram S, et al. Machine-learning predictions of polymer properties with Polymer Genome. J Appl Phys. 2020;128(17):171104.
[52] Huo Z, Yao Y, Liu G, Zhang K, Fan H, Gu J, et al. Machine learning models for prediction of properties of epoxy/MWCNT nanocomposites. Compos Sci Technol. 2021;211:108823.
[53] Hu W, Fey M, Zitnik M, Dong Y, Ren H, Liu B, et al. Open graph benchmark: datasets for machine learning on graphs. Adv Neural Inf Process Syst. 2020;33:22118-22133.
[54] Barton JM. The glass transition temperature. Adv Polym Sci. 1985;57:111-162.
[55] Fox TG, Flory PJ. Second-order transition temperatures and related properties of polystyrene. J Appl Phys. 1950;21(6):581-591.
[56] MarketsandMarkets. Electronics packaging market. Pune (India): MarketsandMarkets Research Pvt. Ltd.; 2024.
[57] Settles B. Active learning literature survey. Madison (WI): University of Wisconsin-Madison; 2009. Computer Sciences Technical Report 1648.
[58] Martin TB, Audus DJ. Emerging trends in machine learning: a polymer perspective. ACS Polym Au. 2023;3(3):239-258.
[59] Doherty C, Wight M. Environmental degradation modeling of epoxy nanocomposites: from molecular dynamics to prediction. Polym Degrad Stab. 2023;214:110430.
[60] Odegard GM, Bandyopadhyay A. Physical aging of epoxy polymers and their composites. J Polym Sci B Polym Phys. 2011;49(24):1695-1716.
[61] Fuchs FB, Worrall D, Fischer V, Welling M. SE(3)-transformers: 3D roto-translation equivariant attention networks. Adv Neural Inf Process Syst. 2020;33:1970-1981.
[62] Schütt KT, Unke OT, Gastegger M. Equivariant message passing for the prediction of tensorial properties and molecular spectra. In: Proceedings of the 38th International Conference on Machine Learning; 2021.
[63] Batzner S, Musaelian A, Sun L, Geiger M, Mailoa JP, Kornbluth M, et al. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials. Nat Commun. 2022;13:2453.
[64] Patel RA, Borca CH, Webb MA. Featurization strategies for polymer sequence or composition design by machine learning. Mol Syst Des Eng. 2022;7(6):661-676.
[65] Mannodi-Kanakkithodi A, Pilania G, Huan TD, Lookman T, Ramprasad R. Machine learning strategy for accelerated design of polymer dielectrics. Sci Rep. 2016;6:20952.
Journal of Polymer & Composites
| Volume | 14 | |
| 04 | ||
| Received | 12/06/2026 | |
| Accepted | 12/09/2026 | |
| Published | 15/09/2026 | |
| Publication Time | 95 Days |