JapanManufacturing2025 – ongoingKIOXIA Corporation
AI classification of semiconductor inspection images
KIOXIA used AI to classify inspection images and combined data from 8 fabs for analysis
Problem
KIOXIA's semiconductor factories generated more inspection and equipment data than employees could analyze manually. Engineers needed to identify defects and investigate process problems.
Approach
KIOXIA used supervised AI to classify inspection images using predefined labels. It also used unsupervised classification where frequent process changes made relabeling burdensome. Its shared data infrastructure connected factories at Yokkaichi and Kitakami for AI analysis.
Results
Takeaway
KIOXIA selected classification methods according to how often the manufacturing process changed and how much labeling work was practical.
Categories
- Size
- Large enterprise
- Industry
- Manufacturing
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Computer visionPredictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
KIOXIA described operating AI systems for inspection-image classification and equipment analysis. Its infrastructure combined data from 8 fabs, but it did not quantify the resulting yield improvement.
Full details are in the original sources.