United StatesTechnology and telecom2021 – ongoingIntel
Auto Crash wafer-failure pattern detection
Intel's Auto Crash models tagged about 2,500 wafers daily by late 2021
Problem
Yield engineers could not manually inspect every wafer for failure patterns. Delayed detection let manufacturing faults affect more material.
Approach
Intel built Auto Crash with machine learning and image processing to identify known failure patterns and flag unfamiliar ones. Results flowed into existing yield-analysis tools for engineers to investigate. Intel deployed the system across factories making its leading products.
Results
Takeaway
Automated detection expanded inspection coverage while engineers retained responsibility for root-cause investigation. Detection accuracy did not measure the resulting increase in manufacturing yield.
Categories
- Size
- Large enterprise
- Industry
- Technology and telecom
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and MLComputer vision
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Intel described automated wafer-failure detection integrated with engineering workflows. It reported about 2,500 wafers tagged daily and accuracy above 90% for baseline patterns.
Full details are in the original sources.
