- SPC
- Statistical Process Control: Tracking process metric stability against statistical control limits
- FDC
- Fault Detection and Classification: Monitoring equipment sensor traces to detect and classify tool anomalies
- APC
- Advanced Process Control: Adjusting tool recipe parameters run-to-run based on post-process metrology
- YMS
- Yield Management System: Integrating defect, metrology, and test data for comprehensive yield analysis
Focus on Core Engineering Questions Rather Than System Names
| System Category | Primary Operational Mission | Role in Root-Cause Investigations |
|---|---|---|
| SPC | Monitor Process Stability and Identify Anomalous Trends | Confirm Metrics, Control Rules, and Time Windows |
| FDC | Collect Equipment States and Detect and Classify Process Anomalies | Inspect Associated Waveforms, Features, and Alarms |
| APC / R2R | Adjust the Process Using Models and Feedback | Understand Actual Compensation, Model Versions, and Applicable Windows |
| YMS / Analytics Platform | Combine Inspection, History, and Test Data for Analysis | Reuse Grouped, Commonality, Spatial, and Statistical Analyses |
- 01MonitorExisting Systems Detect Deviations
SPC and FDC systems generate metric alarms, trend shifts, and tool event notifications.
- 02AggregateIngest Existing Analytics Results
Leverage existing pre-computed metrics and reports, avoiding redundant recalculations.
- 03InvestigateExecute Method-Driven Verification Checks
Within authorized interfaces and permissions, execute standardized checks or formulate supplementary analyses.
- 04ReviewDeliver Verifiable Findings and Next Steps
Engineers review attributed data sources, candidate hypotheses, and prioritized verification actions.
Common Misconceptions Regarding Fab Systems
- Statistical control limits are not product specification limits; SPC is not restricted to univariate normal distributions.
- FDC is not limited strictly to tool hardware; modern platforms correlate sensor traces with wafer-level metrology outcomes.
- APC does not only compensate prior step errors; run-to-run controllers can adjust subsequent processing on the same tool depending on model architecture.
- YMS is not just offline reporting; existing platforms may include spatial analysis, AI/ML, guided analytics, and online rules.
What Burdens Can Remain in Collaborative Investigation?
Even with mature enterprise software, engineers must still frame the investigation objective, verify comparable baseline conditions, handle non-standard split exceptions, and synthesize competing supporting and contradictory evidence. Investigation bottlenecks should be evaluated against actual fab workflows rather than assuming teams rely solely on manual spreadsheets.
What Are the Respective Roles of Stable Workflows and the Agent?
Prioritize established standard checks and existing pre-computed reports. An intelligent agent operates within verified engineering constraints to interpret observations, propose next steps, and structure deliverables; approved underlying tools execute the actual computations. Investigations advance, hold for evidence, or conclude based on accumulated findings.
WAFERPILOT / APPLICATION TO THIS TOPIC
Enhancing Existing Fab Workflows with Autonomous Investigation Capabilities
WaferPilot coordinates investigation workflows across existing connected systems, providing engineers with structured reports, candidate mechanisms, and transparent evidence traces. Software value is measured against existing automation baselines; the agent does not unilaterally release held equipment or authorize recipe changes.
Deliverables for Engineers:A unified investigation dossier synthesizing existing system findings with targeted diagnostic checks, stating which tools were queried, which questions were answered, and which hypotheses require verification.
View the Current Demo and Capability Scope →Related Reading & References
Below are verifiable industry references; specific processes, permissions, and acceptance criteria are governed by fab site specifications.