TY - JOUR
T1 - Capturing progressive interfacial decohesion and thermal residual stress effect in unidirectional elastoplastic composites
AU - Quan, Henglei
AU - Chen, Qiang
AU - Xiao, Ce
AU - Yang, Zhibo
N1 - Publisher Copyright:
© 2026 Elsevier Ltd.
PY - 2026/8/10
Y1 - 2026/8/10
N2 - An asymptotic homogenization framework is presented for predicting the effective response and local stress fields of unidirectional composites exhibiting progressive interfacial debonding and elastoplastic deformation. The interfacial traction-separation relation is modelled using a coupled cohesive zone model (CZM), where interface degradation is represented by a scalar damage variable that progressively reduces the interfacial stiffness. To accurately capture the initiation of interfacial damage, thermal residual stresses arising from the consolidation process are incorporated into the micromechanics simulations prior to the application of mechanical loading. In addition, a fully connected neural network, trained using a Bayesian regression approach, is employed to identify the parameters associated with interfacial damage and matrix plasticity from macroscopic transverse stress–strain data. The proposed framework is further validated against experimental results under complex oligocyclic and off-axis loading conditions. The numerical results demonstrate that the developed approach can reliably capture the progressive interfacial degradation and the resulting nonlinear mechanical response of composites.
AB - An asymptotic homogenization framework is presented for predicting the effective response and local stress fields of unidirectional composites exhibiting progressive interfacial debonding and elastoplastic deformation. The interfacial traction-separation relation is modelled using a coupled cohesive zone model (CZM), where interface degradation is represented by a scalar damage variable that progressively reduces the interfacial stiffness. To accurately capture the initiation of interfacial damage, thermal residual stresses arising from the consolidation process are incorporated into the micromechanics simulations prior to the application of mechanical loading. In addition, a fully connected neural network, trained using a Bayesian regression approach, is employed to identify the parameters associated with interfacial damage and matrix plasticity from macroscopic transverse stress–strain data. The proposed framework is further validated against experimental results under complex oligocyclic and off-axis loading conditions. The numerical results demonstrate that the developed approach can reliably capture the progressive interfacial degradation and the resulting nonlinear mechanical response of composites.
KW - Homogenization
KW - Interfacial debonding
KW - Micromechanics
KW - Parameter identification
KW - Thermal residual stresses
UR - https://www.scopus.com/pages/publications/105040741513
U2 - 10.1016/j.engfracmech.2026.112330
DO - 10.1016/j.engfracmech.2026.112330
M3 - 文章
AN - SCOPUS:105040741513
SN - 0013-7944
VL - 343
JO - Engineering Fracture Mechanics
JF - Engineering Fracture Mechanics
M1 - 112330
ER -