← Cases
ChinaManufacturing2025 – ongoing

Baosteel

AI recognition of hot-rolled steel surface defects

Baosteel's hot-rolled surface-defect model reached 96% accuracy and spread to multiple production bases

WorkingCompany's own claim

Problem

Baosteel needed to identify surface defects in hot-rolled steel during production.

Approach

Baosteel deployed an AI model for hot-rolled product surface-defect recognition. Its broader industrial AI program included collaboration with Huawei. The surface-defect model reached 96% accuracy within 6 months and was replicated on production lines at multiple bases.

Results

96%hot-rolled surface-defect recognition accuracy reached within 6 months, as reported for BaosteelCompany's own claim[1]

Takeaway

Baosteel expanded the model beyond an initial production setting. It did not disclose the test set or false-rejection rate behind the accuracy figure.

Categories

Business function
Production and quality
Innovation type
Process innovation
AI technology
Computer vision
Evidence
Company's own claim

Sources

  1. 大模型让高炉更“聪明”
    People's Daily14 Jan 2026News article

Source summaries

[1]People's Daily14 Jan 2026

People's Daily described Baosteel's industrial AI deployments and collaboration with Huawei. It reported 96% accuracy for hot-rolled surface-defect recognition and replication across production bases.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026
Open case page ↗