Vision-based sensor for early detection of periodical defects in web materials

Area of Science:

• Materials Science
• Industrial Engineering
• Computer Vision

Background:

• Web material production (plastics, textiles, metal) often involves rolls that can develop flaws.
• Flawed rolls can cause recurring surface defects, significantly degrading product quality.
• Early detection systems are crucial for maintaining high production standards.

Purpose of the Study:

• To present a novel vision-based sensor system for early detection of periodic defects in web materials.
• To demonstrate the sensor's adaptability to various materials and noisy data conditions.
• To validate the system's performance in a real-world industrial setting.

Main Methods:

• Development of a vision-based sensor system for defect detection.
• Adaptation of the sensor for inspecting web materials, including noisy data.
• Experimental validation using 36 hot steel strips from ArcelorMittal Avilés.
• Optimization using a full-factorial experimental design and verification on separate datasets.

Main Results:

• The vision-based sensor system successfully detected periodic defects in hot steel strips.
• Performance was evaluated and optimized using experimental design on 36 steel strips.
• The system demonstrated superior performance compared to a commercial, globally used inspection system.

Conclusions:

• The proposed vision-based sensor offers an effective solution for early detection of periodic defects in web materials.
• The system is robust and adaptable, even with noisy input data.
• This technology presents a significant improvement over existing commercial systems for quality control in web material production.