Every gauge certificate carries a measurement result and, if it is worth the paper, an uncertainty. In cap inspection the uncertainty is often larger than the tolerance it is being compared with, and the reason is almost always temperature rather than the instrument.
Building the budget takes twenty minutes and tells you which of five contributors to attack.
| Contributor | Typical magnitude | How to reduce it |
| Reference standard uncertainty | 1 µm to 2 µm | A better setting rod or master |
| Repeatability of the measurement | 1 µm on a good micrometer | Same operator, same technique, three readings |
| Temperature difference from 20 °C | 11 µm per metre per K for steel | Soak the part, measure fast |
| Thermal expansion mismatch | Depends on materials | Measure steel against steel |
| Geometry and elastic deformation | 1 µm to 3 µm on a curved surface | Use the right anvil, control the force |
Steel expands about 11 micrometres per metre per kelvin, which sounds small until it is applied to a 100 mm setting rod: 5 K away from 20 °C is 5.5 µm, which is the entire gauge tolerance on many cap dimensions. The part and the gauge both move, but not necessarily by the same amount, because the part may be brass and the gauge steel.
Brass expands at about 19 micrometres per metre per kelvin, nearly twice steel. A brass cap measured at 26 °C is bigger than the same cap at 20 °C by more than the go gauge allowance on a nominal 26 mm diameter. In summer, in an inspection room without temperature control, a batch can fail in the morning and pass in the afternoon without anything changing in the process.
The practical rule is to soak the parts in the inspection room for long enough that a probe on the surface reads the room temperature, not to measure them as they arrive from the moulding shop.
Three routes, in order of cost. Control the temperature, which is a room and a soak period. Reduce the number of transfers by using a comparator against a master of the same material. Or change the decision rule: gauge to a guard band so that parts within the uncertainty of the limit are treated as conditional rather than accepted.
None of these requires a better micrometer. Most cap gauging problems are thermal, and the instrument is usually the smallest term in the budget. A cap gauge set with a certificate stating both the uncertainty and the reference temperature is the starting point, and the site index lists the setting masters by cap family.
Standards referenced: ISO/IEC Guide 98-3; ISO 14253-1.