Early-Stage Degradation of Composite Materials Based on PPR Cohesive Force Constrained PINN

Abstract

Predicting the progressive early-stage degradation of composite laminates, which are characterized by complex interactions between intralaminar matrix cracking and interlaminar delamination, remains significant challenges in computational mechanics. Traditional mesh-based methods suffer from severe mesh dependency, complex contact definitions, and convergence difficulties when dealing with evolving 3D fracture topologies. In the paper, we propose a novel mesh-free framework based on physics-informed neural networks (PINNs) to simulate the coupled intralaminar and interlaminar damage evolution in 3D composite laminates. Driven by the principle of minimum total potential energy, the framework unifies the Phase Field Method (PFM) for bulk fracture and the Cohesive Zone Model (CZM) for interface delamination into a single optimization problem. Furthermore, a rigorous Finite Element Method (FEM) benchmark using Abaqus Explicit is conducted, utilizing the Hashin damage criterion and surface-based cohesive behavior. The PINN predictions exhibit excellent agreement with FEM results in macroscopic load-displacement responses and microscopic damage morphologies, successfully capturing punch-induced transverse shear delamination and bending-induced tensile cracking. The proposed method demonstrates vast potential for the efficient and mesh-free progressive failure analysis of advanced composite structures.