TY - JOUR
T1 - A novel concurrent multiscale damage analysis method enhanced by physics-informed neural network for composite joint
AU - Hu, Wenlong
AU - Cheng, Hui
AU - Zhang, Kaifu
AU - Li, Yapeng
AU - Yang, Haozhe
AU - Li, Yuan
AU - Bai, Renzi
AU - Liang, Biao
N1 - Publisher Copyright:
© 2025 Elsevier Ltd.
PY - 2026/3/1
Y1 - 2026/3/1
N2 - The concurrent multiscale damage analysis allows to capture the macro and micro damage information simultaneously for carbon fiber reinforced polymer composite (CFRP) joint, which is beneficial for elucidating its complex multiscale damage failure mechanisms. However, existing concurrent multiscale methods have failed to balance the efficiency and accuracy, posing challenges for the concurrent damage simulation of CFRP joint. To address this issue, this work proposed a novel concurrent multiscale method integrating finite element method (FEM) and physics-informed neural network (PINN) based self-consistent clustering analysis (SCA) method, aiming at efficiently and accurately predicting the multiscale damage behavior of CFRP joint. The PINN-based SCA method was employed to efficiently compute the stress and damage state of the unidirectional representative volume element (UD-RVE) at microscale, while the modified macro stress homogenization method and the energy-based macro damage state calculation method were adopted to accurately compute the stress and damage of corresponding macro material point. The effectiveness of the proposed method was validated through in-situ loading experiments and Digital Image Correlation (DIC) experiments, demonstrating its capability to effectively capture the multiscale damage behavior of CFRP joint. In the end, different joint forms (bolt forms and lapped forms) were analyzed with this method to investigate the influence of joint forms on the damage around the bolt-holes, providing a useful analysis tool for the design of CFRP joint structure.
AB - The concurrent multiscale damage analysis allows to capture the macro and micro damage information simultaneously for carbon fiber reinforced polymer composite (CFRP) joint, which is beneficial for elucidating its complex multiscale damage failure mechanisms. However, existing concurrent multiscale methods have failed to balance the efficiency and accuracy, posing challenges for the concurrent damage simulation of CFRP joint. To address this issue, this work proposed a novel concurrent multiscale method integrating finite element method (FEM) and physics-informed neural network (PINN) based self-consistent clustering analysis (SCA) method, aiming at efficiently and accurately predicting the multiscale damage behavior of CFRP joint. The PINN-based SCA method was employed to efficiently compute the stress and damage state of the unidirectional representative volume element (UD-RVE) at microscale, while the modified macro stress homogenization method and the energy-based macro damage state calculation method were adopted to accurately compute the stress and damage of corresponding macro material point. The effectiveness of the proposed method was validated through in-situ loading experiments and Digital Image Correlation (DIC) experiments, demonstrating its capability to effectively capture the multiscale damage behavior of CFRP joint. In the end, different joint forms (bolt forms and lapped forms) were analyzed with this method to investigate the influence of joint forms on the damage around the bolt-holes, providing a useful analysis tool for the design of CFRP joint structure.
KW - CFRP joint
KW - Concurrent multiscale
KW - Damage behavior
KW - Self-consistent clustering analysis
UR - https://www.scopus.com/pages/publications/105024481342
U2 - 10.1016/j.compscitech.2025.111483
DO - 10.1016/j.compscitech.2025.111483
M3 - 文章
AN - SCOPUS:105024481342
SN - 0266-3538
VL - 275
JO - Composites Science and Technology
JF - Composites Science and Technology
M1 - 111483
ER -