Real-time damage diagnosis using digital image correlation method and deep learning

Real-time crack detection is vital for structural health monitoring in high-safety systems. This study integrates a Digital Image Correlation (DIC) system with a lightweight Convolutional Neural Network (CNN) for automated damage diagnosis. The DIC system enables high-resolution full-field strain measurement under dynamic loading, and over 1500 annotated DIC images were used to train the CNN. The model achieved 81.9% precision, 83.4% recall, and 82.1% F1-score in detecting crack initiation and propagation. Compared to traditional DIC post-processing, the proposed method provides automated, accurate, and early detection while maintaining real-time performance. The system runs efficiently on standard computing hardware without GPU acceleration, demonstrating its robustness and feasibility for industrial applications. Experimental validation on metallic specimens confirmed the method’s reliability and practical applicability in structural health monitoring.