Scholarlink

<p>Carbon Fiber Reinforced Plastic (CFRP) is widely used in aerospace and other fields due to its high specific strength and high specific modulus. Hole-making is a critical process in the assembly of CFRP components. However, the anisotropy of CFRP often results in burr defects during this process, which compromise assembly precision and reduce service life. Therefore, accurate detection of burrs in CFRP hole-making is crucial for ensuring the safety of components. In response to the challenges of detecting burr defects in CFRP, such as severe background texture interference, blurred defect edges, a wide range of target scales, and difficulties in identifying small target defects, this paper proposes the Enhanced High-Extraction YOLO (EHE-YOLO) defect detection method. It introduces the Efficient Aggregation Enhanced Module (EAEM) to boost feature extraction against background and edge blurring, designs the High Resolution Boundary Enhanced Neck (HRBE-Neck) for cross-scale co-optimization, and constructs the Efficient Detail Capture Detection Head (EDCDH) for micro-burr recognition. Validated on the self-constructed CFRP-Burr dataset through ablation studies, comparative analyses, and visualization experiments, and validated through generalization experiments on the Severstal Steel Defect and VisDrone2019 datasets. The results demonstrate that EHE-YOLO achieves a detection accuracy of 91.51% for CFRP burrs, enabling high-precision detection of multi-scale and small targets. Finally, this paper discusses the advantages and limitations of the EHE-YOLO model, proposes targeted directions for further improvement, and summarizes the model’s value in industrial applications as well as its future development prospects.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

EHE-YOLO: An Enhanced Detection Approach for CFRP Hole-Making Burr Defects

  • 期刊论文

  • 发表日期：

    2026年4月29日

查看全文
  • Rongrong Wang,
  • Jigang Wu,
  • Youzhi Jiang,
  • Tai An,
  • Li Sun,
  • Jiaming Chen

摘要

Carbon Fiber Reinforced Plastic (CFRP) is widely used in aerospace and other fields due to its high specific strength and high specific modulus. Hole-making is a critical process in the assembly of CFRP components. However, the anisotropy of CFRP often results in burr defects during this process, which compromise assembly precision and reduce service life. Therefore, accurate detection of burrs in CFRP hole-making is crucial for ensuring the safety of components. In response to the challenges of detecting burr defects in CFRP, such as severe background texture interference, blurred defect edges, a wide range of target scales, and difficulties in identifying small target defects, this paper proposes the Enhanced High-Extraction YOLO (EHE-YOLO) defect detection method. It introduces the Efficient Aggregation Enhanced Module (EAEM) to boost feature extraction against background and edge blurring, designs the High Resolution Boundary Enhanced Neck (HRBE-Neck) for cross-scale co-optimization, and constructs the Efficient Detail Capture Detection Head (EDCDH) for micro-burr recognition. Validated on the self-constructed CFRP-Burr dataset through ablation studies, comparative analyses, and visualization experiments, and validated through generalization experiments on the Severstal Steel Defect and VisDrone2019 datasets. The results demonstrate that EHE-YOLO achieves a detection accuracy of 91.51% for CFRP burrs, enabling high-precision detection of multi-scale and small targets. Finally, this paper discusses the advantages and limitations of the EHE-YOLO model, proposes targeted directions for further improvement, and summarizes the model’s value in industrial applications as well as its future development prospects.