Abstract Objective The mechanical response of adhesive layers with randomly distributed voids is complex, and conventional finite element analysis (FEA) is computationally intensive. This study aims to clarify how these defects influence stiffness and strength, and to provide an efficient predictive approach for analyzing tensile and fracture properties. Method A viscoelastic-fracture constitutive model was developed using material parameters of DP460 adhesive. Geometric models with random voids were generated via X-ray CT scanning and 3D reconstruction. A finite element model (FEM) simulated tensile and fracture behavior, validated against experiments. Using the validated FEM, a mechanical dataset for various void characteristics was created. A backpropagation (BP) neural network was trained to predict equivalent Young’s modulus and ultimate tensile strength from statistical void descriptors (including void ratio, averaged radius, radius standard deviation, and the mean value of the 3D weighted L-function), followed by sensitivity analysis. Result and conclusion The reduction in stiffness and strength of adhesive layers under tensile loading due to void defects was accurately captured, with a relative error of less than 4% against the experimental data. The trained BP neural network predicts equivalent Young’s modulus and ultimate tensile strength with relative errors within 5% for the test set. Sensitivity analysis indicates that the void ratio presents the most significant influence on the reduction of both stiffness and strength, followed by the spatial concentration of voids, while variations in void size have negligible effects.
Zhang et al. (Mon,) studied this question.