Structural health monitoring (SHM) of composite structures using surface strain fields measured by digital image correlation (DIC) has been widely demonstrated; however, accurate damage quantification remains challenging. This study proposes a hybrid framework integrating finite element (FE) modeling, machine learning (ML), and peridynamics (PD). A CFRP specimen with a notch was subjected to cyclic loading, and damage evolution was monitored using DIC and validated by ultrasound measurements. A validated FE model generated synthetic strain-field datasets for ML training, enabling defect detection and quantitative characterization directly from surface strains. The trained models achieved high accuracy, including perfect notch detection and low prediction errors. A calibrated PD model captured internal damage evolution and fatigue behavior. The combined DIC–ML–PD approach enables accurate, non-contact damage identification and prognosis, supporting physics-informed digital twins for composite structures.
Vaitkūnas et al. (2026) studied this question.