Measurement system analysis (MSA) and gage R&R
Every measurement contains error. Measurement system analysis (MSA) checks whether the combined variation of the gage, the method, the operator and the environment is small enough to make good decisions about the product or process. If measurement variation is large, process capability looks worse than it really is, and good parts get rejected or bad parts accepted.
Types of study
- Bias: the difference between the observed average and the reference value.
- Linearity: how bias changes across the operating range.
- Stability: how the measurement changes over time.
- Repeatability and reproducibility (gage R&R): variation when the same part is measured repeatedly (equipment variation, EV) and differences between appraisers (appraiser variation, AV).
- Attribute agreement: agreement of good/bad decisions in attribute inspection.
Running a gage R&R study
Typically 10 parts covering the full process spread are measured 2–3 times by 3 appraisers in random order, without seeing their earlier results. The data is evaluated with ANOVA or the average and range method. %GRR is measurement variation relative to total variation or to the tolerance; ndc (number of distinct categories) shows how many groups the measurement can reliably separate the parts into.
| Metric | Acceptable | Conditionally acceptable | Unacceptable |
|---|---|---|---|
| %GRR | < 10% | 10–30% (depending on application, cost and customer) | > 30% |
| ndc | ≥ 5 | – | < 5 |
Type 1 study (Cg, Cgk)
In the type 1 study, common in German practice (VDA 5 and OEM standards), one appraiser measures a master or reference part typically 25–50 times. Cg relates measurement variation, and Cgk additionally bias, to 20% of the tolerance; the usual requirement is Cg ≥ 1.33 and Cgk ≥ 1.33.
Acceptance limits may vary with customer requirements; customer-specific requirements always apply.