Rule, calibration and model training are three different changes
One rule shows known specialist logic. Calibration adapts parameters within a defined model to reference values. Model training changes a statistical model based on a selected data set.
Customer data will not be regulated by uploading to training data alone. Purpose, release, data set, measurement method, version and rollback option must be separately regulated for each change.
Key points
Methodology is explained in such a way that inputs, procedures, effects and limits in the CNC calculation remain comprehensible.
- Rule: Explicit, verifiable logic with clear version and validity.
- Calibration: Controlled adjustment of parameters to tested reference values.
- Training: Separate learning process with released dataset and independent evaluation.


