Browsing by Author "Zheng, Ruoxin"

Carbon Fiber-Reinforced Polymer (CFRP) composites are widely employed in industrial sectors including aerospace, automotive, maritime and sports due to their stiffness and lightweight properties. However, these materials also vulnerable from impacts or long-term use, which can result in various damages that may influence their expactation lifetime and safety. X-ray micro tomography provides three-dimentional reconstruction with high resolution images, making it an ideal method for damage detection among traditional non-destructive testing (NDT) methods. To ultilize X-ray micro-CT images, this work explores the potential of deep learning-based object detection methods, with a particular focus on the YOLOv8 model developed by Ultralytics with transfer learning. The use of a highly complex pre-trained model with a limited annotated dataset of 60 images presents high precision and recall rates. Furthermore, the evaluation on different dataset sizes proves that 5\% of annotated images from entire sample is sufficient for training a reliable model. The predicted damages on the entire sample with quantitative information reveals the potential for impact level evaluation with the location and distribution of damages. The results showed not only the successful implementation of cutting-edge deep learning methods on material science and X-ray microtomography, but also established a foundation for AI-based automation in industrial applications such as quality control and lifetime monitor.