PSI - Issue 75

ScienceDirect Available online at www.sciencedirect.com ScienceDirect Available online at www.sciencedirect.com ScienceDirect Structural Integrity Procedia (2025) 000 – 000 Available online at www.sciencedirect.com Procedia Structural Integrity 75 (2025) 129–139 Structural Integrity Procedia (2025) 000 – 000

www.elsevier.com/locate/procedia

www.elsevier.com/locate/procedia

2452-3216 © 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under the responsibility of Dr Fabien Lefebvre with at least 2 reviewers per paper 10.1016/j.prostr.2025.11.015 2452-3216 © 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the Fatigue Design 2025 organizers 2452-3216 © 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the Fatigue Design 2025 organizers The primary objective of this study is to simulate the fatigue properties of a high-strength unidirectional carbon/epoxy ply, a material widely used in aerospace, automotive, and other demanding sectors requiring an optimal combination of lightness and strength. This work focuses on the characterization and modeling of the fatigue behavior of high-performance carbon/epoxy composite laminates using Physics-Informed Neural Networks (PINNs). The laminates are manufactured from unidirectional carbon fibers (TR50) embedded in an epoxy resin system (R367-2). The main objective is to reconstruct the fatigue limit at the elemental ply with only a single quasi isotropic laminate. This knowledge is essential for modelling the fatigue behaviour of composite structures. © 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under the responsibility of Dr Fabien Lefebvre with at least 2 reviewers per paper Abstract This study investigates the fatigue behavior of high-strength carbon/epoxy laminates using Physics-Informed Neural Networks (PINNs). Laminates composed of TR50 unidirectional fibers and R367-2 epoxy were tested under cyclic tension and fatigue at room temperature. Self-heating tests were also conducted to efficiently estimate the fatigue limits. A strong agreement was found between classical fatigue tests and the self-heating method. A PINN-based approach was employed to reconstruct S – N curves by embedding a fatigue model into the learning process. The model, implemented in Python using Keras/TensorFlow, accurately predicts fatigue performance, and a fully commented code example is provided Keywords: S – N ; Wöhler ; Self-heating, PINNs ; carbon ; epoxy ; Neural Networks ; PINNs 1. Introduction The primary objective of this study is to simulate the fatigue properties of a high-strength unidirectional carbon/epoxy ply, a material widely used in aerospace, automotive, and other demanding sectors requiring an optimal combination of lightness and strength. This work focuses on the characterization and modeling of the fatigue behavior of high-performance carbon/epoxy composite laminates using Physics-Informed Neural Networks (PINNs). The laminates are manufactured from unidirectional carbon fibers (TR50) embedded in an epoxy resin system (R367-2). The main objective is to reconstruct the fatigue limit at the elemental ply with only a single quasi isotropic laminate. This knowledge is essential for modelling the fatigue behaviour of composite structures. Abstract This study investigates the fatigue behavior of high-strength carbon/epoxy laminates using Physics-Informed Neural Networks (PINNs). Laminates composed of TR50 unidirectional fibers and R367-2 epoxy were tested under cyclic tension and fatigue at room temperature. Self-heating tests were also conducted to efficiently estimate the fatigue limits. A strong agreement was found between classical fatigue tests and the self-heating method. A PINN-based approach was employed to reconstruct S – N curves by embedding a fatigue model into the learning process. The model, implemented in Python using Keras/TensorFlow, accurately predicts fatigue performance, and a fully commented code example is provided Keywords: S – N ; Wöhler ; Self-heating, PINNs ; carbon ; epoxy ; Neural Networks ; PINNs 1. Introduction Fatigue Design 2025 (FatDes 2025) Deep learning based predictions of Wöhler curve using physics informed neural networks: self-heating determination of fatigue limits for laminated composite materials Fatigue Design 2025 (FatDes 2025) Deep learning based predictions of Wöhler curve using physics informed neural networks: self-heating determination of fatigue limits for laminated composite materials Laurent Gornet GeM, UMR CNRS 6183, Ecole Centrale de Nantes , Rue de la Noë, 44321 Nantes Laurent Gornet GeM, UMR CNRS 6183, Ecole Centrale de Nantes , Rue de la Noë, 44321 Nantes