Measurement uncertainty in practice: General observations
2. July 2012
AI-generated illustration of the target model.
Measurement uncertainty describes the accuracy of a measurement or test. Testing is an activity that establishes the extent to which a test object fulfils a particular requirement. It may be qualitative or quantitative. Measurement, by contrast, produces quantitative statements about a measurand, usually by giving a numerical value and a unit.
The approach to determining measurement uncertainty is very similar for measurement and test results. We therefore use the term “testing” in the following text to include measurements as well.
Why tests are never completely exact
Tests can never be one hundred per cent accurate, because even small influences cause test results to vary. Examples include natural variations in material properties within the sample volume, slight changes in the testing process or changes in the testing equipment over time.
Such effects inevitably cause results to scatter. This raises a central question: What effect do these variations have on the test result?
The “true value” and reality in the testing laboratory
For economic reasons, the true value of a material or component property can practically never be determined exactly. This would require a very large number of measurements. In practice, small samples are therefore used, for example a single tensile test or the evaluation of three spectral analyses.
This immediately raises another important question: How well do the results of these small samples represent the actual characteristic value of the material or component?
Measurement uncertainty and the confidence interval
Both questions can be answered using statistical methods. Measurement uncertainty u plays a central role. It describes an interval ±u around the sample mean x, known as the confidence interval. The true value µ lies within this range with a certain probability, known as the confidence level.
We can calculate the mean, but it is the true value that interests us. Measurement uncertainty indicates how closely the calculated mean matches the actual value.
A confidence level of 95% is typically used. This means that the true value µ lies within the range x ± u with a probability of 95%. If a higher confidence level is required, the confidence interval becomes correspondingly wider.
Why is measurement uncertainty so important? Three practical examples illustrate its significance.
- Comparing test results
If two laboratories want to compare their test results, their testing processes must also be comparable. Suppose two laboratories determine a mean carbon content of 0.40% and 0.43%, respectively, in the same steel sample. Is the difference of 0.03% a significant deviation, or does it fall within the range of normal variation?
This question cannot be answered without knowing the measurement uncertainty.
- Conformity assessments
Measurement uncertainty also plays a central role in conformity assessments, which determine whether a product meets a requirement. For example, a component must have a minimum hardness of 300 HV, but the test certificate states a mean hardness of 305 HV ± 15 HV.
Does the component meet the requirement or not? The assessment depends directly on the measurement uncertainty.
- Component design
Measurement uncertainty is also relevant when designing components. For example, a designer must design a bridge to accommodate a certain number of people. It is crucial whether the mean mass of a person must be assumed to be 80 ± 5 kg or 80 ± 15 kg. The uncertainty of the underlying data therefore directly influences the design of the structure.
These examples show that a test result must always be accompanied by its associated measurement uncertainty. Only then can tests be evaluated correctly, results compared and sound technical decisions made.
Measurement uncertainty in everyday life
Measurement uncertainty matters not only in the laboratory, but also in everyday life. A familiar example is radar speed measurement. How accurately do these devices actually measure? And when is speeding considered reliably established? Sometimes this uncertainty determines whether penalty points are imposed or not.
W.S. Werkstoff Service – Expertise in materials engineering
We offer practical seminars on measurement uncertainty in materials testing. Further information is available at: Measurement uncertainty online.
If you would like to deepen your knowledge of material testing or materials engineering (metals), take a look at the other seminars and courses offered by our W.S. TrainingCenter.
