A dropped caliper
At a routine calibration, a caliper turns out to read 0.05 mm too high over part of its measuring range. It has been used every day since the last calibration, six months ago. The quality manager now has to answer three questions: which measurements were made with it, which of them were so close to a limit that the error may have changed the decision, and have such parts been delivered?
The questions can only be answered if there is a link between the gauge and the measurements. The calibration itself only shows that something was wrong. It is the rest of the routine that decides whether you can find out what it meant.
What calibration is, and is not
The word calibration is often used for several different things. The international vocabulary of metrology (VIM, JCGM 200:2012) tells them apart:
- Calibration is comparing the gauge’s reading with a reference of known value and known uncertainty, and documenting the difference. Calibration changes nothing on the gauge.
- Adjustment is setting the gauge so that it reads correctly, or closer to correctly. After an adjustment, a new calibration is needed to show the result.
- Verification is showing that the gauge meets a specified requirement, for example a maximum permissible error. This is often what you really want to know: is the gauge approved for its use?
A calibration certificate therefore normally states the deviation and the measurement uncertainty at a number of points, and sometimes a decision on whether the gauge is within its requirement. If a gauge was adjusted, it helps if the certificate shows the values both before and after, because it is the value before that says something about the measurements already made.
Metrological traceability
A gauge is calibrated against a reference, for example a set of gauge blocks. The gauge blocks are in turn calibrated against more accurate references at an accredited laboratory, which has been calibrated against national standards. This unbroken chain is called metrological traceability. It should not be confused with product traceability, which is about being able to follow a part through manufacturing.
Every step in the chain has a measurement uncertainty, and the uncertainty grows the further down the chain you go. So it is not enough that the reference is calibrated: its uncertainty must also be known and small enough for the gauge calibrated against it. ISO 9001:2015 (clause 7.1.5.2) requires measuring equipment used to show that products meet requirements to be calibrated or verified, at specified intervals, against standards traceable to international or national standards. If no such standard exists, the basis used for calibration shall be documented.
Intervals
There is no general answer to how often a gauge should be calibrated. The interval depends on how much the gauge is used, in what environment, how sensitive it is and how close to the limit the measurements usually are. A first interval is often set from the manufacturer’s recommendation and experience with similar gauges.
Then the interval should be followed up. ISO 10012:2003 describes that intervals should be reviewed and adjusted based on data from earlier calibrations. If a gauge has been well within its requirement at several calibrations in a row, the interval can be extended. If it has drifted close to the limit, or been outside it, the interval should be shortened. A gauge that has been dropped, damaged or started showing strange values should be checked at once, regardless of the interval.
Between calibrations, a simple check in production, for example against a check piece at the start of a shift, can catch an error early. It does not replace calibration, but it shortens the time before you discover a problem.
If a gauge turns out to be wrong
This is where the calibration routine shows its value. ISO 9001:2015 requires the organisation to determine whether earlier measurement results have been affected when a gauge is found to be unfit for its purpose, and to take action. IATF 16949:2016 (clause 7.1.5.2.1) goes further and requires, among other things, an assessment of the risk, action for parts that may be affected and that the customer is informed if suspect parts have been shipped.
In practice this means you need to be able to:
- Find all the measurements made with the gauge since the last approved calibration.
- See which of them were so close to a tolerance limit that the deviation may have changed the decision.
- Link these measurements to order, batch or serial number, so that you know which parts are concerned and where they are.
- Document what was assessed and decided.
Without a link between gauge and measurement, the answer is often that everything made during the period has to be treated as suspect.
Common mistakes
- Believing that a calibrated gauge is automatically good enough. Calibration shows deviation and uncertainty. If they are too large in relation to the tolerance, the gauge is unsuitable, even though it is calibrated.
- Saving only the value after adjustment. Then the measurements already made cannot be assessed.
- Leaving the intervals untouched for years, even though the history shows they should be changed.
Summary
A calibration answers the question of whether a gauge measures correctly, at one point in time. A register, intervals, traceability to standards and a link to the measurements make the answer hold over time, and let you sort out the consequences when at some point it does not. The next step is to understand how large the uncertainty in a measured value is, and when it matters for the decision.
