South KoreaManufacturing2022 – ongoingHyundai Steel
Semi-supervised classification of special-steel surface defects
Hyundai Steel put AI surface-defect classification into its Pohang production lines
Problem
Special-steel surface defects required highly trained specialists because different defects could look similar. A shortage of labeled production data hindered the use of deep learning.
Approach
Hyundai Steel's Pohang bar-production team worked with Handong Global University to develop automated defect classification. The researchers combined process-domain algorithms with semi-supervised learning. The system passed factory validation and entered production lines in January 2022, reporting defect types and their proportions within individual bars.
Results
No published figures
Takeaway
Semi-supervised learning allowed the project to proceed despite limited labeled data. The partners reported production deployment without publishing an accuracy rate.
Categories
- Size
- Large enterprise
- Industry
- Manufacturing
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Partner co-development
- Evidence
- Vendor case study
Sources
Source summaries
Electronic Times reported Handong Global University's account of a jointly developed classification system deployed at Hyundai Steel's Pohang works. It described the shortage of labeled data and confirmed that production introduction followed successful factory validation.
Full details are in the original sources.