Enhanced detection of fine damage in composite materials using integrated X-ray computed tomography, digital volume correlation, and U-Net

Accurate detection of fine damage in composite materials is essential for advancing research and engineering practices in various industries. This study introduces a novel method for detecting fine damage by integrating Digital Volume Correlation (DVC) and U-Net. X-ray computed tomography (XCT) images are analyzed to correlate microstructural features with damage progression. DVC residuals highlight new damage, while U-Net enhances segmentation precision. This approach addresses challenges in differentiating inherent flaws from new cracks, improving detection accuracy. Experimental validation and synthetic data generated by the discrete element method (DEM) demonstrate the method's robustness. The results show that the method is susceptible to newly formed microcracks and can accurately identify fine cracks, with over 60 % being nearly invisible in XCT images. Cyclic loading experiments on two mortar samples reveal that cracks are the predominant form of damage, exhibiting different evolution characteristics during the loading process. The method significantly advances the non-destructive damage evaluation of composite materials. •Innovative method integrates XCT, DVC, and U-Net for enhanced damage detection.•DVC residuals help distinguish new damage from inherent flaws effectively.•Precisely identifies fine cracks, with over 60 % being nearly invisible in the XCT images.•Fatigue damage in mortar predominantly occurs in cracks, with minimal porosity change.•Verify accuracy using synthetic data from the discrete element method.