Physics-Informed Machine Learning and Multiscale Modeling for Structure–Property Quantification of Polymer Composites

Notice

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.

Year : 2026 | Volume : 14 | 04 | Page :
By

Sandip R. Thorat,

Bhushan S. Chaudhari,

Sharad Ninu Kolte,

Amit V. Mohod,

Dipesh B. Pardeshi,

Atul A. Barhate,

P. William,

  1. Assistant Professor, Department of Mechanical Engineering, Sanjivani College of Engineering, Kopargaon, Maharashtra, India
  2. Professor, Department of Computer Science and Engineering, School of Engineering and Technology, Pimpri Chinchwad University, Pune, Maharashtra, India
  3. Lecturer, Department of Electrical Engineering, Prof. Ram Meghe college of Engineering and Management, Badnera, Maharashtra, India
  4. Assistant Professor, Department of Electrical Engineering, Prof. Ram Meghe college of Engineering and Management, Badnera, Maharashtra, India
  5. Professor, Department of Electrical Engineering, Sanjivani College of Engineering, Kopargaon, Maharashtra, India
  6. Associate Professor, Department of Electrical Engineering, Godavari College of Engineering, Jalgaon, Dr. Babasaheb Ambedkar Technological University, Lonere, Maharashtra, India
  7. Professor (Research), School of Computer Science and Technology, Karunya Institute of Technology and Sciences, Coimbatore, Tamil Nadu, India

Abstract

The growing need for light-weight, high strength, and sustainable polymer composites has led to the development of smart methods that enable accurate structural-property quantification and material design. However, conventional methods have been predominantly data-based, thus ignoring physical constraints as well as multi-scale interactions involving fiber, matrix, interface, and process parameters, leading to lower accuracy and poor robustness and interpretability of the models. In this study, a Cat Swarm Optimization-Tuned Physics-Informed Multi-Scale Intelligent Random Forest (CSO-PIM-IntRF) model is introduced to predict the mechanical properties of natural fiber-reinforced polymer composites. The approach involves a dataset of natural fiber-reinforced polymers with 15,000 samples, where each sample has different composition, reinforcement features, surface treatment, process parameters, interfacial property, and mechanical property. Bootstrap sampling has been used to increase data variability to boost model robustness and Min-Max normalization to standardize feature distribution to ensure consistent learning. Multi-scale features are derived through Principal Component Analysis (PCA), which is then combined with physics-informed features including fiber volume fraction, density, rule-of-mixtures, and fiber-matrix interface. CSO optimizes the parameters of the IntRF to enhance the learning process of non-linear structure-property relationship. It has been shown experimentally that the presented CSO-PIM-IntRF framework successfully predicts the properties of polymer composites, giving the following root mean square error (RMSE): 1.1937 MPa for tensile strength and 3.4267 MPa for flexural strength with corresponding coefficients of determination (R2): 0.9826 for flexural strength and 0.9874 for tensile strength.

Keywords: Physics-informed machine learning; Multiscale modeling; Polymer composites; Structure–property quantification; Natural fiber-reinforced composites.

How to cite this article: Sandip R. Thorat, Bhushan S. Chaudhari, Sharad Ninu Kolte, Amit V. Mohod, Dipesh B. Pardeshi, Atul A. Barhate, P. William. Physics-Informed Machine Learning and Multiscale Modeling for Structure–Property Quantification of Polymer Composites. Journal of Polymer & Composites. 2026; 14(04):-.
How to cite this URL: Sandip R. Thorat, Bhushan S. Chaudhari, Sharad Ninu Kolte, Amit V. Mohod, Dipesh B. Pardeshi, Atul A. Barhate, P. William. Physics-Informed Machine Learning and Multiscale Modeling for Structure–Property Quantification of Polymer Composites. Journal of Polymer & Composites. 2026; 14(04):-. Available from: https://journals.stmjournals.com/jopc/article=2026/view=255190

References

  1. Recio-Colmenares, C.L., Recio-Colmenares, R.B. and García-García, C.A., 2026. Cross-study machine learning analysis of structure–property–performance relationships in macroporous PolyHIPE-based magnetic polymer composites. Journal of Composites Science, 10(3), p.128. https://doi.org/10.3390/jcs10030128
  2. Wang, W. and Thai, H.T., 2025. A physics-informed neural network framework for laminated composite plates under bending. Thin-Walled Structures, 210, p.113014. https://doi.org/10.1016/j.tws.2025.113014
  3. Zeng, B., Wu, S. and Yao, S., 2025. Machine Learning-Driven Paradigm for Polymer Aging Lifetime Prediction: Integrating Multi-Mechanism Coupling and Cross-Scale Modeling. Polymers, 17(22), p.2991. https://doi.org/10.3390/polym17222991
  4. Wei, H., Wu, C.T., Hu, W., Su, T.H., Oura, H., Nishi, M., Naito, T., Chung, S. and Shen, L., 2023. LS-DYNA machine learning-based multiscale method for nonlinear modeling of short fiber-reinforced composites. Journal of Engineering Mechanics, 149(3), p.04023003. https://doi.org/10.1061/JENMDT.EMENG-6945
  5. Nguyen, V.D. and Noels, L., 2022. Micromechanics-based material networks revisited from the interaction viewpoint; robust and efficient implementation for multi-phase composites. European Journal of Mechanics-A/Solids, 91, p.104384. https://doi.org/10.1016/j.euromechsol.2021.104384
  6. Nguyen, V.D. and Noels, L., 2022. Interaction-based material network: A general framework for (porous) microstructured materials. Computer Methods in Applied Mechanics and Engineering, 389, p.114300. https://doi.org/10.1016/j.cma.2021.114300
  7. Huang, T., Liu, Z., Wu, C.T. and Chen, W., 2022. Microstructure-guided deep material network for rapid nonlinear material modeling and uncertainty quantification. Computer Methods in Applied Mechanics and Engineering, 398, p.115197. https://doi.org/10.1016/j.cma.2022.115197
  8. Gajek, S., Schneider, M. and Böhlke, T., 2022. An FE-DMN method for the multiscale analysis of thermomechanical composites. Computational Mechanics, 69(5), pp.1087–1113. https://doi.org/10.1007/s00466-021-02131-0
  9. Dey, A.P., Welschinger, F., Schneider, M., Gajek, S. and Böhlke, T., 2022. Training deep material networks to reproduce creep loading of short fiber-reinforced thermoplastics with an inelastically informed strategy. Archive of Applied Mechanics, 92(9), pp.2733–2755. https://doi.org/10.1007/s00419-022-02213-2
  10. Li, T., 2024. Micromechanics-informed parametric deep material network for physics behavior prediction of heterogeneous materials with a varying morphology. Computer Methods in Applied Mechanics and Engineering, 419, p.116687. https://doi.org/10.1016/j.cma.2023.116687
  11. Wu, L. and Noels, L., 2025. Stochastic deep material networks as efficient surrogates for stochastic homogenisation of non-linear heterogeneous materials. Computer Methods in Applied Mechanics and Engineering, 441, p.117994. https://doi.org/10.1016/j.cma.2025.117994
  12. Jean, J.G., Su, T.H., Huang, S.J., Wu, C.T. and Chen, C.S., 2024. Graph-enhanced deep material network: multiscale materials modeling with microstructural informatics. Computational Mechanics, 75(1), pp.113–136. https://doi.org/10.1007/s00466-024-02493-1
  13. Wei, T.J. and Chen, C.S., 2025. Foundation model for composite microstructures: Reconstruction, stiffness, and nonlinear behavior prediction. Materials & Design, 257, p.114397. https://doi.org/10.1016/j.matdes.2025.114397
  14. Maia, M.A., Rocha, I.B.C.M., Kerfriden, P. and van der Meer, F., 2023. Physically recurrent neural networks for path-dependent heterogeneous materials: Embedding constitutive models in a data-driven surrogate. Computer Methods in Applied Mechanics and Engineering, 407, p.115934. https://doi.org/10.1016/j.cma.2023.115934
  15. Maia, M.A., Rocha, I.B.C.M., Kovačević, D. and van der Meer, F., 2024. Physically recurrent neural network for rate and path-dependent heterogeneous materials in a finite strain framework. Mechanics of Materials, 198, p.105145. https://doi.org/10.1016/j.mechmat.2024.105145
  16. Zhao, Y., Chen, Z. and Jian, X., 2023. A high-generalizability machine learning framework for analyzing the homogenized properties of short fiber-reinforced polymer composites. Polymers, 15(19), p.3962. https://doi.org/10.3390/polym15193962
  17. Kumar, P. William, S. Raghunathrao Deshmukh, A. Sharmila, K. Kaushik and J. Shreyas, “Optimizing Litter Detection in Images: An Integration of Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs),” in IEEE Access, vol. 14, pp. 3325-3344, 2026, doi: 10.1109/ACCESS.2025.3646075.
  18. Anber Abraheem Shlash Mohammad, N. Yogeesh, Suleiman Ibrahim Shelash Mohammad, N. Raja, Lingaraju, P. William, Asokan Vasudevan, Nawaf Alshdaifat, and Mohammad Faleh Ahmmad Hunitie. 2025. An Intuitionistic Fuzzy Graph Model: Matrix Representations and Applications. Cybern. Inf. Technol. 25, 4 (Dec 2025), 3–19. https://doi.org/10.2478/cait-2025-0030
  19. Sun, Y., Hanhan, I., Sangid, M.D. and Lin, G., 2024. Predicting mechanical properties from microstructure images in fiber-reinforced polymers using convolutional neural networks. Journal of Composites Science, 8(10), p.387. https://doi.org/10.3390/jcs8100387
  20. Zhang, S., Mojumder, S., Liu, W.K., Chen, W. and Apley, D.W., 2023. Accelerated and interpretable prediction of local properties in composites. APL Machine Learning, 1(3), p.036112. https://doi.org/10.1063/5.0156517
  21. Gavallas, P., Stefanou, G., Savvas, D., Mattrand, C. and Bourinet, J.M., 2024. CNN-based prediction of microstructure-derived random property fields of composite materials. Computer Methods in Applied Mechanics and Engineering, 430, p.117207. https://doi.org/10.1016/j.cma.2024.117207
  22. V. Kulkarni, P. William, Pravin B Khatkale, Sandip R. Thorat, Santosh Kumar Sharma, Apurv Verma. An Innovative AI-Integrated Approach for Identifying the Tensile Robustness of Polymeric Materials. Journal of Polymer & Composites. 2024; 13(01):90-97.
  23. William, P., Mishra, V.P., Khalaf, O.I., Mukundan, A., N, Y. et al. (2025). Dynamic Multi-Objective Gannet Optimization (DMGO): An Adaptive Algorithm for Efficient Data Replication in Cloud Systems. Computers, Materials & Continua, 84(3), 5133–5156. https://doi.org/10.32604/cmc.2025.065840
  24. Heidenreich, J.N., Gorji, M.B. and Mohr, D., 2023. Modeling structure-property relationships with convolutional neural networks: Yield surface prediction based on microstructure images. International Journal of Plasticity, 163, p.103506. https://doi.org/10.1016/j.ijplas.2022.103506
  25. Rashid, M.M., Pittie, T., Chakraborty, S. and Krishnan, N.A., 2022. Learning the stress-strain fields in digital composites using Fourier neural operator. iScience, 25(11), p.105452. https://doi.org/10.1016/j.isci.2022.105452
  26. Park, D., Jung, J., Gu, G.X. and Ryu, S., 2022. A generalizable and interpretable deep learning model to improve the prediction accuracy of strain fields in grid composites. Materials & Design, 223, p.111192. https://doi.org/10.1016/j.matdes.2022.111192
  27. Su, H., Guan, T. and Liu, Y., 2023. A three-dimensional prediction method of stiffness properties of composites based on deep learning. Computational Mechanics, 71, pp.583–597. https://doi.org/10.1007/s00466-022-02253-z
  28. Yin, B., Huang, J. and Sun, W., 2024. Peridynamics-fueled convolutional neural network for predicting mechanical constitutive behaviors of fiber-reinforced composites. Computer Methods in Applied Mechanics and Engineering, 431, p.117309. https://doi.org/10.1016/j.cma.2024.117309
  29. Gholami, K., Ege, F. and Barzegar, R., 2023. Prediction of composite mechanical properties: Integration of deep neural network methods and finite element analysis. Journal of Composites Science, 7(2), p.54. https://doi.org/10.3390/jcs7020054
  30. Bhaduri, A., Gupta, A. and Graham-Brady, L., 2022. Stress field prediction in fiber-reinforced composite materials using a deep learning approach. Composites Part B: Engineering, 238, p.109879. https://doi.org/10.1016/j.compositesb.2022.109879
  31. Saha, I., Gupta, A. and Graham-Brady, L., 2024. Prediction of local elasto-plastic stress and strain fields in a two-phase composite microstructure using a deep convolutional neural network. Computer Methods in Applied Mechanics and Engineering, 421, p.116816. https://doi.org/10.1016/j.cma.2024.116816
  32. Li, M., Wang, B., Hu, J., Li, G., Ding, P. and Ji, C., 2024. Artificial neural network-based homogenization model for predicting multiscale thermo-mechanical properties of woven composites. International Journal of Solids and Structures, 301, p.112965. https://doi.org/10.1016/j.ijsolstr.2024.112965
  33. Bahtiri, B., Arash, B., Scheffler, S., Jux, M. and Rolfes, R., 2023. A machine learning-based viscoelastic–viscoplastic model for epoxy nanocomposites with moisture content. Computer Methods in Applied Mechanics and Engineering, 415, p.116293. https://doi.org/10.1016/j.cma.2023.116293
  34. V. Kulkarni, P. William, Pravin B Khatkale, Harshal P. Varade, Sandip R. Thorat. Developing an AI-Based Novel Forecasting Framework for Surface Irregularity in Metal Matrix Materials. Journal of Polymer & Composites. 2024; 12(05):48-56.
  35. Bahtiri, B., Arash, B., Scheffler, S., Jux, M. and Rolfes, R., 2024. A thermodynamically consistent physics-informed deep learning material model for short fiber/polymer nanocomposites. Computer Methods in Applied Mechanics and Engineering, 427, p.117038. https://doi.org/10.1016/j.cma.2024.117038
  36. Elaskalany, M. and Behdinan, K., 2024. Stochastic multiscale modeling of electrical conductivity of carbon nanotube polymer nanocomposites: An interpretable machine learning approach. Advanced Engineering Materials, 26, p.2401233. https://doi.org/10.1002/adem.202401233
  37. Joo, C., Park, H., Kwon, H., Lim, J., Shin, E., Cho, H. and Kim, J., 2022. Machine Learning Approach to Predict Physical Properties of Polypropylene Composites: Application of MLR, DNN, and Random Forest to Industrial Data. Polymers, 14(17), p.3500. https://doi.org/10.3390/polym14173500
Ahead of Print Subscription Original Research
Volume 14
04
Received 12/08/2026
Accepted 27/08/2026
Published 10/09/2026
Publication Time 29 Days

Login

My IP

PlumX Metrics