§10.1 · APPLYING · 5 MIN READ · EDITION 1.0
Predictive quality on process and test data
Predictive-quality use cases forecast defect likelihood from process parameters; they support investigation priority, not automatic disposition.
A high-risk prediction should route a part or lot to a defined human review, not to automatic rejection or automatic release.
Model drift is a live concern here: process changes over time, and a model trained on last year's process may quietly lose relevance.
PUT THIS INTO PRACTICE
- 01Route predictions to review, not automatic disposition.
- 02Monitor for process drift on a defined cadence.
- 03Retrain or revalidate against the current process.
RELATED INSTRUMENTS
CITE THIS SECTION
§10.1 · Edition 1.0 · Predictive quality on process and test data