Genetically Optimized Bottleneck Index for Wafer Fabrication
A study presents the Dynamic Multi-Criteria Bottleneck Severity Index (DMBSI), a data-centric approach designed to pinpoint the most critical bottleneck in semiconductor wafer production in real time. This method evaluates various diagnostic signals, including cycle time, alterations in process parameters, and the effects of rework, to generate a comprehensive assessment of bottleneck severity. The experimental validation utilized Manufacturing Execution System logs from 22 wafer lots at a 200 mm fabrication facility operated by Seagate Technology. Through 5-fold cross-validation and optimization via genetic algorithms, DMBSI achieved a Pearson correlation of r = 0.80 with the observed contributions to cycle time.
Key facts
- DMBSI is a new data-driven methodology for bottleneck severity in wafer fabrication.
- It analyzes multiple diagnostic signals of cycle time and rework impacts.
- Validation used MES logs from 22 wafer lots at Seagate Technology's 200 mm line.
- GA-optimized DMBSI achieved Pearson correlation r = 0.80 with observed cycle-time contributions.
- 5-fold cross-validation was employed.
- The methodology addresses reentrant process flows and variable bottlenecks.
- The research is published on arXiv with ID 2607.24819.
- The paper focuses on semiconductor wafer manufacturing.
Entities
Institutions
- Seagate Technology