Scholarlink

<p>Seamless Stainless Steel Pipes (SSSP) are crucial for high-end equipment manufacturing, energy transmission, and the chemical industry, as their structural integrity directly impacts industrial system safety and efficiency. However, SSSP production is complex, resulting in surface defects such as cracks and scratches. Undetected and unrepaired defects can cause leaks, explosions, and other accidents, leading to economic losses and environmental risks. Existing inspection methods face issues such as high labor intensity, low accuracy, and difficulty identifying subtle defects. To address these issues, this study proposes YOLO-DSCW, a surface defect detection model designed explicitly for SSSP. Via modules like DFBCSPELAN4, the model enhances the capture of irregular defect features, strengthens key features, optimizes feature fusion, and alleviates sample imbalance. Validated on a self-built SDSSP dataset with 1,200 images and four defect categories, YOLO-DSCW achieves 88.1% mean average precision, which surpasses the mainstream YOLOv9 model’s 84.8%. However, the limited dataset scale may weaken its generalization in cross-factory and multi-batch scenarios; its detection accuracy and inference speed also need improvement for high-standard production lines.</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

YOLO-DSCW: a seamless stainless steel pipe surface defect detection model

  • 期刊论文

  • 发表日期：

    2025年9月22日

查看全文
  • Chengbo Zhang,
  • Jiahao Zhang,
  • Yangyu Wang,
  • Hui Wang,
  • Yizhi Wang

摘要

Seamless Stainless Steel Pipes (SSSP) are crucial for high-end equipment manufacturing, energy transmission, and the chemical industry, as their structural integrity directly impacts industrial system safety and efficiency. However, SSSP production is complex, resulting in surface defects such as cracks and scratches. Undetected and unrepaired defects can cause leaks, explosions, and other accidents, leading to economic losses and environmental risks. Existing inspection methods face issues such as high labor intensity, low accuracy, and difficulty identifying subtle defects. To address these issues, this study proposes YOLO-DSCW, a surface defect detection model designed explicitly for SSSP. Via modules like DFBCSPELAN4, the model enhances the capture of irregular defect features, strengthens key features, optimizes feature fusion, and alleviates sample imbalance. Validated on a self-built SDSSP dataset with 1,200 images and four defect categories, YOLO-DSCW achieves 88.1% mean average precision, which surpasses the mainstream YOLOv9 model’s 84.8%. However, the limited dataset scale may weaken its generalization in cross-factory and multi-batch scenarios; its detection accuracy and inference speed also need improvement for high-standard production lines.