Intelligent Ultrasonic Characterization of Bonding Conditions in Aircraft Honeycomb Sandwich Composite Panels

Abstract Aircraft horizontal-stabilizer honeycomb sandwich composite panels have complex structures, large inspection areas, and diverse defect types. Ultrasonic C-scan imaging can directly visualize the bonding condition between the face sheet and the honeycomb core; however, the large number of inspection images is conventionally interpreted by experienced inspectors, resulting in low evaluation efficiency and strong subjectivity. This study therefore proposes an intelligent classification method for ultrasonic C-scan images of the bonding interface in honeycomb sandwich composite panels. First, ultrasonic C-scan images of the face-sheet–honeycomb interface are acquired by interface-echo tracking, and a strategy is developed to expand the image dataset. Second, the images are divided into three target-region classes according to their amplitude distributions, representing three bonding-condition categories. Finally, a ResNet50 network is constructed, trained, and evaluated for bonding-condition classification. The results demonstrate that interface-echo tracking can generate C-scan images of the honeycomb bonding interface and that ResNet50 can identify the three target-region classes with good stability and accuracy.