Physics-Constrained Surrogate Modeling for Microscale Elastic Homogenization of Additively Manufactured Continuous Carbon Fiber Composites
Tongwha KIM, Kamran BEHDINAN
Abstract. In this paper, a physics-constrained surrogate model is presented for the microscale elastic homogenization of additively manufactured continuous carbon fiber reinforced polymer (CCFRP) composites. Stochastic representative volume elements (RVE) of the CCFRP filament cross-section are generated across four domain sizes, and homogenized stiffness tensors are extracted from periodic finite element analyses. The surrogate model maps scalar RVE descriptors to five effective elastic constants; Voigt-Reuss bounds and thermodynamic consistency conditions are enforced as analytical hinge penalties during training. The model achieves R2 ≥ 0.93 and MAPE < 1.3% across all outputs on the held-out test set, demonstrating both predictive accuracy and physical admissibility. The proposed framework provides a computationally efficient alternative to direct finite element homogenization, supporting future integration into multiscale process-structure-property frameworks.
Keywords
Surrogate Modeling, Physics-constrained, Machine Learning, Additive Manufacturing, Continuous Carbon Fiber Composites
Published online 11/8/2026, 5 pages
Copyright © 2026 by the author(s)
Published under license by Materials Research Forum LLC., Millersville PA, USA
Citation: Tongwha KIM, Kamran BEHDINAN, Physics-Constrained Surrogate Modeling for Microscale Elastic Homogenization of Additively Manufactured Continuous Carbon Fiber Composites, Materials Research Proceedings, Vol. 73, pp 95-99, 2026
DOI: https://doi.org/10.21741/9781644904459-19
The article was published as article 19 of the book Mechanics of Advanced Materials and Structures
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 license. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
References
[1] Y. Zhang, W. Zheng, Y. Wang, K. Ma, X. Feng, Q. Ji, W. Guo, and B. Lu, “A review of 3D printing continuous carbon fiber reinforced thermoplastic polymers: Materials, processes, performance enhancement, and failure analysis,” Polymer Composites, vol. 46, no. 14, pp. 12619-12649, 2025, doi: https://doi.org/10.1002/pc.29895
[2] R. Nakka, D. Harursampath, and S. A. Ponnusami, “A generalised deep learning-based surrogate model for homogenisation utilising material property encoding and physics-based bounds,” Scientific Reports, vol. 13, no. 1, p. 9079, 2023, doi: 10.1038/s41598-023-34823-3
[3] Y. Yamanaka, S. Matsubara, N. Hirayama, S. Moriguchi, and K. Terada, “Surrogate modeling for the homogenization of elastoplastic composites based on RBF interpolation,” Computer Methods in Applied Mechanics and Engineering, vol. 415, p. 116282, 2023, doi: https://doi.org/10.1016/j.cma.2023.116282
[4] J. N. Fuhg and N. Bouklas, “On physics-informed data-driven isotropic and anisotropic constitutive models through probabilistic machine learning and space-filling sampling,” Computer Methods in Applied Mechanics and Engineering, vol. 394, p. 114915, 2022, doi: https://doi.org/10.1016/j.cma.2022.114915
[5] J. Jiang, J. Wu, Q. Chen, G. Chatzigeorgiou, and F. Meraghni, “Physically informed deep homogenization neural network for unidirectional multiphase/multi-inclusion thermoconductive composites,” Computer Methods in Applied Mechanics and Engineering, vol. 409, p. 115972, 2023, doi: https://doi.org/10.1016/j.cma.2023.115972
[6] E. Pitz and K. Pochiraju, “A neural network transformer model for composite microstructure homogenization,” Engineering Applications of Artificial Intelligence, vol. 134, p. 108622, 2024, doi: https://doi.org/10.1016/j.engappai.2024.108622
[7] H. Maher Saad Hayder, H. Afrasiab, and M. Gholami, “Efficient generation of random fiber distribution by combining random sequential expansion and particle swarm optimization algorithms,” Composites Part A: Applied Science and Manufacturing, vol. 173, p. 107649, 2023, doi: https://doi.org/10.1016/j.compositesa.2023.107649
[8] COMSOL Multiphysics®. COMSOL AB, Stockholm, Sweden.
