Scholarlink

<p>The article investigates the impact of varying illuminance levels and low-amplitude vibrations on the accurate detection of surface defects in rolled metal products using a deep learning approach. Automated surface inspection is crucial for quality control in metal manufacturing, and convolutional neural networks (CNNs) have shown promising results in this domain. However, real-world industrial environments often present challenges such as inconsistent lighting conditions and vibrations from machinery, which can negatively affect the performance of defect detection systems. This study explores the robustness of a CNN-based defect detection model under different luminance levels, simulating non-uniform illumination and, in the presence of low-amplitude vibrations, mimicking typical factory floor conditions. To achieve this, a software-hardware method for evaluating the efficiency of recognising metal defects is used. The research analyses the network’s performance across a range of illumination and vibration parameters, evaluating its ability to accurately identify and classify various surface defects. The results provide insights into the resilience of CNN-based inspection systems to these common industrial challenges and contribute to the development of more reliable and practical automated quality control solutions for rolled metal products.</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

Surface defects of rolled metal products recognised by a deep neural network under different illuminance levels and low-amplitude vibration

  • 期刊论文

  • 发表日期：

    2025年6月16日

查看全文
  • Pavlo Maruschak,
  • Ihor Konovalenko,
  • Yaroslav Osadtsa,
  • Volodymyr Medvid,
  • Oleksandr Shovkun,
  • Denys Baran

摘要

The article investigates the impact of varying illuminance levels and low-amplitude vibrations on the accurate detection of surface defects in rolled metal products using a deep learning approach. Automated surface inspection is crucial for quality control in metal manufacturing, and convolutional neural networks (CNNs) have shown promising results in this domain. However, real-world industrial environments often present challenges such as inconsistent lighting conditions and vibrations from machinery, which can negatively affect the performance of defect detection systems. This study explores the robustness of a CNN-based defect detection model under different luminance levels, simulating non-uniform illumination and, in the presence of low-amplitude vibrations, mimicking typical factory floor conditions. To achieve this, a software-hardware method for evaluating the efficiency of recognising metal defects is used. The research analyses the network’s performance across a range of illumination and vibration parameters, evaluating its ability to accurately identify and classify various surface defects. The results provide insights into the resilience of CNN-based inspection systems to these common industrial challenges and contribute to the development of more reliable and practical automated quality control solutions for rolled metal products.