GlossaryKind of term: Manufacturing termMeasurement uncertainty
What is measurement uncertainty?
Measurement uncertainty states the margin around a measured value within which the true value lies with a stated probability, usually as U with k = 2.
No measurement is exact. Measurement uncertainty states how far the true value of a feature can reasonably lie from the measured value. In calibration certificates, and in inspection reports that state it, it usually appears as expanded uncertainty U with coverage factor k = 2; for a normal distribution that corresponds to a probability of about 95 percent. A measurement result is complete only with its uncertainty, for example 12.004 mm ± 0.002 mm (k = 2).
What does measurement uncertainty depend on?
Uncertainty comes from several sources: calibration and resolution of the instrument, the temperature of part, instrument and surroundings, the form and roughness of the feature itself, how the part is held and probed, the measuring strategy and the person measuring. Together these contributions make up the total uncertainty. A finer scale only reduces the resolution contribution. A form deviation can weigh more: an oval bore has no single diameter, and a two-point measurement gives a different value from one with many points.
What does the standard say about measurement uncertainty?
The method is set out in the Guide to the expression of uncertainty in measurement (GUM), published as JCGM 100 and as ISO/IEC Guide 98-3. ISO 1 sets the reference temperature for geometrical product specifications at 20 °C. ISO 14253-1 gives decision rules for proving conformity: unless something else is agreed, the uncertainty counts against whoever is proving. The supplier proves conformity within the tolerance reduced by U, the customer proves nonconformity outside the tolerance enlarged by U. Customer and supplier can agree a different decision rule.
How does measurement uncertainty work in practice?
An example: with a tolerance width of 0.01 mm and an uncertainty U of 0.003 mm, 0.004 mm is left for proving conformity, and the process has to hold that. A tight tolerance can therefore call for a more stable process, a better instrument or measuring at controlled temperature. An uncertainty of at most a tenth to a quarter of the tolerance is common practice, as a rule of thumb rather than a standard. If you cannot prove a tolerance with your own equipment, raise it at the quoting stage.
What is measurement uncertainty often confused with?
Measurement uncertainty is not resolution: a display with three decimals says nothing about how reliable those decimals are. Nor is it measurement error: a known systematic error is corrected, the uncertainty remains. And it is not process variation: variation lives in the parts, uncertainty in the measurement, although the spread of a series of readings contains both, and they combine in quadrature, not by simple addition.
Frequently asked questions
How accurate does my instrument need to be for a tolerance?
As a rule of thumb, at most a tenth to a quarter of the tolerance width. At 0.02 mm a calliper is too coarse for that; a micrometer or a comparative measurement against a gauge block comes closer.
Who carries the measurement uncertainty, me or the customer?
Under the default rule of ISO 14253-1, whoever is proving something. To prove a dimension is good, the supplier measures within the tolerance minus their uncertainty. To reject it, the customer measures outside the tolerance plus theirs. A different agreement is allowed.