Physics-informed neural networks for predicting deformation and stress fields in asymmetric composite adhesive joints

Gu, Zewen and Liu, Yalong and Ding, Xiaoxuan and Zhao, Jianwei and Liu, Jianlin and Ye, Jianqiao and Hou, Xiaonan (2026) Physics-informed neural networks for predicting deformation and stress fields in asymmetric composite adhesive joints. Composite Structures, 391: 120568. ISSN 0263-8223

Text (COST Revised manuscript (unmarked version))
COST_Revised_manuscript_unmarked_version_.pdf - Accepted Version
Available under License Creative Commons Attribution.
Download (2MB)

Abstract

The widespread use of composite adhesive joints in aerospace, automotive, and advanced technological applications is attributed to their lightweight nature and specific strength relative to metals and mechanical fasteners. However, the nonlinear loading conditions within the adhesive layer pose significant challenges for accurately predicting joint strength and stress distribution. This study integrates experiments, finite element analysis (FEA), theoretical modeling, and physics-informed neural networks (PINNs) to predict deformation and stress fields in bonded joints under tensile loading. Tests and simulations on CFRP–CFRP and Al–CFRP joints were conducted to validate the FEA model. Based on these results, PINNs using Timoshenko beam theory and two-dimensional elasticity were developed to predict deformation and stress fields, respectively. Both frameworks demonstrated high accuracy, particularly in asymmetric and mixed-material joints where nonlinearities dominate. R2 remained above 0.85, with displacement errors below 0.1 mm and stress field errors under 5%. The proposed method offers a fast, reliable approach for predicting joint strength by overcoming the difficulty of solving partial differential equations under complex nonlinear loads. This work advances the understanding of adhesive joint mechanics and highlights the potential of PINNs for solving complex governing equations, providing a powerful framework for future research on intricate joint structures.