drift detection

**Drift detection** is the **monitoring and analytics process that identifies gradual parameter shifts indicating equipment or process degradation before limit violations occur** - it turns slow failure signatures into early maintenance and process interventions. **What Is Drift detection?** - **Definition**: Detection of non-random trend movement in sensor, metrology, or performance signals over time. - **Signal Types**: Pressure creep, temperature offsets, power changes, cycle-time elongation, and defect trend rise. - **Methods**: SPC trend rules, model-based anomaly scoring, and slope-threshold analytics. - **Action Output**: Early alerts tied to inspection, maintenance, or recipe adjustment workflows. **Why Drift detection Matters** - **Preventive Response**: Finds degradation before sudden failures or yield excursions occur. - **Downtime Reduction**: Planned intervention replaces emergency outage when drift is caught early. - **Quality Stability**: Limits subtle process shifts that can accumulate into major defect events. - **Asset Longevity**: Controlled correction avoids prolonged operation in damaging conditions. - **Data-Driven Operations**: Enables objective trigger points instead of reactive judgment. **How It Is Used in Practice** - **Baseline Integration**: Compare live signals against golden trajectories and allowed drift bands. - **Alert Prioritization**: Rank drift events by criticality and expected time-to-threshold. - **Verification Loop**: Confirm root cause after intervention and adjust detection sensitivity as needed. Drift detection is **a high-value early-warning capability in semiconductor manufacturing** - catching slow degradation early protects yield, uptime, and maintenance efficiency.