Understanding measurement uncertainty: precision in conformity assessment

Measurement uncertainty indicates the accuracy of a measurement or test. A test is used to determine the extent to which a test object meets a specified requirement. Tests may be qualitative or quantitative. Measurements are a special case of testing and provide quantitative statements about a measurand, usually in the form of a numerical value with a unit.

Variations in test results

Materials testing can never be completely exact. Even minimal changes in the testing process or the testing equipment can lead to different test results. For practical reasons, the true value of a material property can never be determined exactly, as this would require a very large number of measurements.

Measurement uncertainty accounts for variations in test results and the influence of small samples. It defines a confidence interval around the sample mean within which the “true” material characteristics lie with a specified probability. Laboratories usually use a confidence level of 95%.

Since only samples are tested in practice, the results must be interpreted statistically so that they come as close as possible to the true material characteristics. DIN EN ISO/IEC 17025 requires accredited laboratories to take measurement uncertainty into account when making statements of conformity or carrying out internal calibrations. The same applies by analogy to accredited inspection bodies that rely on test results.

Illustrating measurement uncertainty

Assuming a normally distributed measurand and a 95% confidence level:

• The position of the mean marks the maximum of the distribution,
• the width of the confidence interval at the base of the distribution describes the measurement uncertainty,
• the vertical line indicates the conformity limit, for example the minimum value for a material property.

Six possible cases for assessing a product:

• a. The mean and measurement uncertainty indicate a conforming product → no risk in acceptance.
• b. The mean and measurement uncertainty indicate conformity → residual risk < 5%, acceptable in practice.
• c. The mean suggests conformity, but the measurement uncertainty does not permit a clear statement → the risk in acceptance is not insignificant.
• d. The mean and measurement uncertainty indicate a non-conforming product → no risk in rejection.
• e. The mean and measurement uncertainty indicate non-conformity → residual risk < 5%, acceptable in practice.
• f. The mean suggests non-conformity, but the measurement uncertainty does not permit a clear statement → the risk of rejecting a conforming product is not insignificant.

Practical recommendation

Measurement and test results should lie within the conforming range when measurement uncertainty is taken into account (case b). Complaints are justified when the results, including measurement uncertainty, lie within the non-conforming range (case e). Measurement uncertainty must not be used to declare results with possible non-conformity as conforming (case f).

Getting to the bottom of measurement uncertainty

We support you with the following services relating to measurement uncertainty in material testing and materials engineering:

• Article series “Measurement uncertainty in practice”
• Training on measurement uncertainty
• Assessment of testing processes and measurement uncertainty by our inspection body
• Materials testing with meaningful measurement uncertainty statements by our testing laboratory

Further information: https://www.werkstoff-service.de/en/blog/measurement-uncertainty-in-materials-testing/(opens in new tab).

***

Further information: Part 9 of our article series “Conformity assessments in materials testing” covers “Competence and trust: key factors in conformity assessment”.(opens in new tab)