Uncertainty of measurement and accuracy are often confused with each other. The following discusses how these may affect the inspection report and the results contained therein. Ultrasonic wall thickness measurement has been selected as one example to illustrate the difference in meaning between uncertainty and accuracy.

Uncertainty of Measurement
Uncertainty of measurement is a quantitative indication of the variability of a result. It demonstrates how well the result represents the value of the quantity or reading being taken.

Uncertainty is an important element in any measurement because it must be recognised that the attainment of the ‘true value’ is an ideal concept which cannot be achieved with practice, and there is always some, although small, range of uncertainty associated with every measurement.

This uncertainty results from the presence of errors from many sources, e.g. in wall thickness measurement couplant thickness, surface condition, coating thickness, angle at which the probe is held, pits and uneven surface in the material etc, these errors can be considered in two categories, i.e. systematic errors and random errors.

Systematic errors are, as the name implies, constant when measurements are made under the same conditions or varying according to a definite requirement when conditions change. In general, systematic errors can be determined by calibration and these may be eliminated by applying a correction factor to the results of the measurement, or they may be incorporated in the statement of uncertainty on the report e.g. the wall thickness testing report.

Random errors vary in an unpredictable manner, both in magnitude and type, when a number of measurements are made and are made under effectively identical conditions. It is not possible to take account of random errors by the application of a correction factor. It is only possible to fix limits within which, with a stated probability, the wall thickness measurement errors, together with any uncorrected systematic errors associated with the result of a measurement are taken into account. The uncertainty of measurement should always be included in the report.

Uncertainty of a measurement is the range about the value recorded, within which the ‘true wall thickness value’ is believed to be, within a stated degree of confidence. The level of confidence may create a misleading impression that the measurement has been improved, but in fact there has been no improvement in accuracy. Therefore, it is not meaningful to report an uncertainty without also reporting the level of confidence to be associated with that uncertainty.

Small ranges of uncertainty together with high levels of confidence can be obtained from small measurements of high quality, but no statistical process will enable both these conditions to be satisfied from poor quality measurements.

General Sources of Uncertainty in Non-destructive Testing
These include:
• effects of environmental conditions on the measurement process
• personal bias in reading instruments
• instrument resolution
• errors in the graduation scale, e.g. time base on a UT instrument, moving iron meter etc
• a change in the calibration setting during readings
• calibration factor not being applied
• technical competence of individual performing the test

Accuracy
When a manufacturer refers to accuracy of the instrument it means that ‘whatever’ reading is shown on the instrument that this reading is accurate to the stated accuracy of the manufacturer. This must not be confused with the uncertainty of measurement. For example if the manufacturer states that the instrument is accurate to ±0.01mm then that is its accuracy and not its uncertainty of measurement.

General
Many important engineering decisions are based on the results reported from quantitative testing. It should be considered important that an indication of confidence of such results is available. Results of testing cannot be perfect for the reasons given above and therefore in order to have confidence in the results it is important that an uncertainty is given.

It must be remembered that measurement uncertainty is a vital part of the test result. The HERA Training Centre, in addition to other courses, offers training in NDT. For information on the courses offered visit: www.hera.org.nz or contact Peter Hayward on 09 262 4847 or email peter.hayward@hera.org.nz for any inspection and quality control matters or assistance.