Hygrothermal Aging and Strength Prediction of CFRP Single-Lap Joints: Experiments and Physics-Constrained Neural Networks

Abstract

Adhesively bonded carbon-fiber-reinforced polymer (CFRP) joints offer efficient load transfer and weight-saving potential in aerospace structures. The simple geometry of single-lap joints (SLJs) makes them useful for examining bond durability, although the effects of hygrothermal aging on their performance have not been sufficiently investigated. This research systematically integrates moisture absorption experiments, mechanical testing, finite element modeling (FEM), and deep neural network (DNN) modeling to provide a comprehensive analysis of moisture absorption behavior, strength degradation, and the evolution of failure modes within the adhesive layer. Initially, three types of adhesives were aged under various hygrothermal aging conditions to investigate their moisture absorption behavior and mechanical performance after aging. The tensile strength and modulus of the adhesives decreased linearly with increasing moisture uptake. In addition, hygrothermal aging tests were conducted on SLJs. Results demonstrated a significant reduction in joint shear strength during aging, accompanied by a transition in failure mode from cohesive to interfacial. Subsequently, a hygrothermally coupled FE model was developed using the experimental data and combined with a cohesive zone model to simulate joint strength evolution. The FE model achieved a calculation error within 5%. To address the high computational cost associated with finite element analysis of hygrothermal aging in SLJ structures and to enhance the physical interpretability of data-driven models, a novel prediction strategy was introduced. A three-stage serial neural network model was constructed using experimental and simulation data, with the quadratic stress criterion and power-law criterion embedded as physical constraints. Compared to unconstrained models, this physics-constrained neural network exhibited superior prediction accuracy and stability. The proposed method offers a valuable reference for the design and evaluation of the durability of CFRP bonded structures.