Glossary · Manufacturing systems
Predictive quality
Also known as: Quality prediction
German: Predictive Quality
In manufacturing, predictive quality is the use of process, machine and material data, often with machine learning models, to predict the quality of products before or instead of measuring it, so processes can be adjusted or parts checked selectively.
- Manufacturing systems
- AI
In one sentence
Predictive quality uses process and machine data, often with machine learning, to predict product quality before it is measured.
Example
A model predicts weld seam quality from current, voltage and wire feed data, so only seams with a high defect risk go to X-ray inspection.
How it applies
- Engineering: Predictive quality needs linked data: process data, genealogy and quality results for the same units. Data preparation is typically the largest part of a project.
- Quality: Predictions can reduce inspection effort, but replacing measurements with predictions requires Validation of the model for the intended use and monitoring of its performance over time. Process changes can invalidate a model; see AI drift.
- Regulation: In regulated industries, a model that decides on product release is part of the quality system and may be subject to software validation and, depending on the product, AI regulation.
- Documentation: Document the model's purpose, input data, performance, limits and the fallback when the model is unavailable. Operators need to know how to read a prediction and what to do with it.
Predictive quality vs. predictive maintenance
Predictive maintenance predicts equipment failures. Predictive quality predicts product defects. Both use similar data and methods.