Calibration compares an instrument with a standard. Measurement system analysis asks a different question: is this measurement, made this way by these people on these parts, good enough for the decision it supports? A calibrated gauge can fail that test through coarse resolution, poor repeatability, or operators who measure differently.
Which measurement systems need a study
- Gauges that decide acceptance of a key characteristic or critical item.
- Gauges behind a capability number or a control chart, because the gauge's spread is part of what the number measures (E05, E06).
- Measurement systems a customer requires studies for, including those under AS13003 for the aero engine supply chain.
- New, repaired or relocated gauges on those uses, and any gauge after an out-of-tolerance result on them.
- Visual and go/no-go inspection that decides acceptance, through an attribute agreement study.
The MSA Planning Worksheet (E03) lists each of these as a line, with the study type, the reason, an owner and a date, and turns each result into a verdict and a re-study date. Lines that need no study say why. The Gage R&R Study Data Sheet (E04) runs the most common study, a crossed variable study by the average and range method, with the published constants, and records the decision.
Reading a gauge study in plain terms
| Result | What it tells you | Common guideline |
|---|---|---|
| %GRR of tolerance | How much of the tolerance the gauge's own spread uses up | Under 10% acceptable; 10% to 30% a decision for the use; over 30% not acceptable |
| %GRR of total variation | How much of what the study saw was the gauge rather than the parts | The same bands, for gauges used for SPC or capability |
| Number of distinct categories | How many groups of parts the gauge can tell apart | 5 or more |
| EV against AV | Whether the gauge (EV) or the people and method (AV) contribute more | No guideline: it tells you where to look first |
| Resolution share of tolerance | Whether the gauge's steps are fine enough | A tenth or finer |
Running a study well
- Plan it first on the study plan (E03): the parts, the people, the trials, the conditions and the acceptance criteria, before anyone measures.
- Choose parts that span the process, not ten in a row. Check their spread on a better instrument and keep the set for the next study.
- Use the people who do the job, at the place they do it, with the setting master and method they use.
- Code the parts and randomize the order each trial; hide earlier readings; leave time between trials.
- Watch the first trial. Differences in method show up at once and are cheaper to fix before the data is in.
- Check the ranges first. A range above its limit is a misread or a moved part; re-measure and record why, never delete.
- Read EV against AV before the percentages: it tells you whether to work on the gauge or the method.
- Record the decision, its conditions and the next study date, and put the study reference on the register and the MSA plan.
Attribute agreement in brief
For go/no-go gauges and visual inspection there are no readings, only decisions. An attribute agreement study gives two or three inspectors a set of parts (often 30 or more, with a good share close to each limit and a reference decision for each), two or three times in random order. It shows how often each inspector agrees with themselves, with the others and with the reference, and above all how often a bad part was passed. Boundary samples, agreed with the customer where the requirement is cosmetic, make the reference decisions defensible.
These figures are common guidelines from the Measurement Systems Analysis reference manual (AIAG), not requirements; your customer may set its own. A result in the conditional band is a decision, and the decision and its conditions go on the study record and the MSA plan.
Need more than one resource, or not sure where to start? Book a complimentary 30-minute Aegis Quality Guidance Call. We'll find the tools that fit your situation and a sensible place to start. No sales pressure, no obligation.
A management aid, not certification or a registrar audit. Using it does not by itself show conformity to AS9100D or any other standard, and your customers' requirements may add to it.
Clause numbers refer to AS9100D (SAE AS9100 Rev D) unless another document is named. The standards' text is not reproduced here: use your own licensed copy.
Aegis Insights are our own guidance, labeled wherever they appear: how we approach this work, not a requirement of any standard or customer.