TaiwanManufacturing2017 – ongoingUnited Microelectronics Corporation (UMC)
Smart-fab AI at Fab 12A, a WEF Lighthouse factory
UMC's Fab 12A became a Lighthouse factory and reported 97% yield after a smart-fab programme with AI at its core
Problem
A mature-node wafer fab runs thousands of tools and many possible wafer routes. Defect causes are hard to trace, bottleneck tools cause waiting, and design kits for customers take time to deliver.
Approach
UMC's Fab 12A in the Southern Taiwan Science Park used image recognition with automated anomaly search to find defect root causes. An AI algorithm tied to the automated material handling system planned the transport path for each wafer. Machine-learning prediction and smart gas switching targeted bottleneck tools, and machine-learning tools shortened design-kit delivery. UMC said about 95% of its 48-plus solutions were developed in-house.
Results
Takeaway
The results belong to the whole smart-fab programme, which also includes non-AI work. The source does not say how much came from AI.
Categories
- Size
- Large enterprise
- Industry
- Manufacturing
- Business function
- Production and qualityIT and R&D
- Innovation type
- Process innovation
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Report on Fab 12A being named a World Economic Forum Lighthouse factory. It lists the AI systems used and gives yield of 97%, a 47% rise in work efficiency and a 57% cut in design-kit delivery time. The figures are UMC's and are not split by technology.
Full details are in the original sources.