Title: MACHINE LEARNING DRIVEN PREDICTION OF MECHANICAL AND FATIGUE PROPERTIES OF FIBRE-REINFORCED COMPOSITES

Authors: Vishwas Kesarwani, Anuranjana

DOI: https://doi.org/10.33599/GL.2026.INCOMAT.TP26-0013

Abstract: Fibre-Reinforced Polymer (FRP) composites are indispensable in aerospace, defence, automotive, and energy sectors owing to their high specific stiffness, tailorable anisotropy, and multifunctional performance. However, the mechanical response and fatigue life of FRP laminates are governed by complex interactions among fibre architecture, matrix behaviour, stacking sequence, and manufacturing parameters. Traditional approaches for estimating elastic properties, damage progression, and fatigue degradation rely heavily on extensive experimental techniques, resulting in long development cycles and high cost. This study presents a machine-learning-driven predictive framework to estimate key mechanical and fatigue-related properties of FRP laminates using material composition, laminate architecture, and manufacturing process parameters as inputs. The dataset integrates ASTM-compliant experimental data, curated public composite databases, and laboratory-generated datasets from advanced composite testing facilities. Multiple regression and learning architectures, including Random Forest, Gradient Boosting, Support Vector Machines, and Feed-Forward Artificial Neural Networks are implemented using Python and MATLAB environments. The results demonstrate that these models and deep learning architectures effectively capture nonlinear interactions between laminate configuration, material properties, and damage indicators, achieving high predictive accuracy even under data-limited conditions. The proposed framework enables rapid virtual screening of composite designs, reduction in experimental iterations, and early-stage optimization of manufacturing routes. By integrating data-driven intelligence with composite damage mechanics

References: [1] Saxena, A., and Goebel, K. “NASA Prognostics Data Repository: Composite Specimens Fatigue Datasets.” NASA Ames Research Center, 2024. Hou, X., and Hu, X. “Machine learning boosts the design and discovery of nanomaterials.” ACS Sustainable Chemical Engineering 9 (2021): 6130–6147. [3] Sanman, S., and Sunil, S. K. “Experimental study on natural fiber reinforced hybrid polymer composites.” Materials Today: Proceedings 72 (2023): 2021–2026. [4] Komninos, P., Kontogiannis, A., Eleftheroglou, N., and Zarouchas, D. “Fatigue Monitoring of Composites (F-MOC) dataset.” Mendeley Data, V1 (2024). DOI: 10.17632/4zm6jh8jkd.1. [5] Pathan, M. “Rapid surrogate modeling for composite structures using FFT-based models.” Virtual Physical Prototyping 18(1) (2024). [6] Alsheghri, A., and Alhammadi, A. “Predicting mechanical properties of CFRP composites using data-driven models.” PMC Materials 11 (2025): 1126. [7] Michaloglou, A., and Papadimitriou, I. “Advancements in Physics-Informed Neural Networks for Laminated Composites.” Mathematics 13(1) (2025): 31. [8] Wu, W. “Machine learning for mechanical design of composite materials and structures.” MechanoEngineering 1(1) (2026): 014303. [9] Wang, H., and Cao, L. “Machine learning-driven multidomain nanomaterial design.” ACS Applied Nano Materials 7(23) (2024): 26579–26600. [10] Ge, W., and Stenzel, M. H. “Machine learning in polymer research.” Advanced Materials 37(11) (2025): 2413695.

Conference: INCOMAT 2026

Publication Date: 2026/03/13

SKU: INCOMAT.TP26-0013

Pages: 14

Price: $28.00