ISO/IEC 17025 Consulting - Guidance on Measurement Uncertainty Evaluation for Testing and Calibration Laboratories

1. Overview of Measurement Uncertainty According to ISO/IEC 17025

Measurement Uncertainty (MU) evaluation is a core technical requirement of ISO/IEC 17025:2017, specified in Clause 7.6. It is one of the key elements used to demonstrate the technical competence of both Testing Laboratories and Calibration Laboratories.

According to JCGM 100:2008 – Guide to the Expression of Uncertainty in Measurement (GUM) and TCVN 9595:2013, measurement uncertainty is defined as:

“A parameter associated with the result of a measurement that characterizes the dispersion of the values that could reasonably be attributed to the measurand.”

In other words, measurement uncertainty reflects the level of confidence in a measurement result and indicates the range within which the true value of the measurand is expected to lie with a specified level of confidence.

2. ISO/IEC 17025 Requirements for Measurement Uncertainty

For Calibration Laboratories

Calibration laboratories shall:

• Identify all sources that may influence measurement results.
• Establish an appropriate measurement model.
• Evaluate and calculate measurement uncertainty for all calibrations within the scope of accreditation.
• Report the expanded measurement uncertainty on calibration certificates.

For Testing Laboratories

Testing laboratories shall:

• Identify factors contributing to measurement uncertainty.
• Perform measurement uncertainty evaluation whenever it is relevant to the validity of the test results.
• Utilize data obtained from:

For certain test methods where the nature of the method does not permit a rigorous quantitative evaluation based on a complete measurement model, the laboratory shall perform a reasonable estimation using available technical information and supporting data.

3. Measurement Uncertainty Evaluation Process According to GUM

Step 1: Define the Measurand

The measurand is the quantity intended to be measured.

Examples:

• Vernier caliper calibration: the measured dimension or length indicated by the caliper.
• Balance calibration: the mass value indicated by the balance.
• Temperature calibration: the temperature value measured by the thermometer.
• Chemical testing: concentration of a substance in a sample.
• Mechanical testing: tensile strength of a material.

A clear and precise definition of the measurand is essential because it forms the basis for identifying uncertainty sources and developing the measurement model.

Where:

• L: Final measured value
• Lr: Reading obtained from the measuring instrument
• C: Correction value obtained from the reference standard
• ΔT: Temperature correction

This relationship is referred to as the Measurement Model.

Step 2: Identify Sources of Measurement Uncertainty

Common sources influencing measurement results include:

Step 3: Classify Measurement Uncertainty Components

According to the GUM, uncertainty components are evaluated using two approaches:

3.1 Type A Evaluation

Type A evaluation is based on the statistical analysis of repeated observations or measurements.

Example:

A component is measured ten times, yielding results such as:

• 50.01 mm
• 50.02 mm
• 49.99 mm
• 50.00 mm
• ...

The following parameters are then calculated:

• Arithmetic mean
• Standard deviation
• Standard deviation of the mean
• Standard uncertainty

The standard uncertainty obtained from repeated observations is calculated statistically and represents the random variation associated with the measurement process.

This Type A uncertainty component is commonly expressed as:

Where:

• s = standard deviation
• n = number of measurements

Type A uncertainty reflects the variability observed from experimental measurement data.

3.2 Type B Evaluation

Type B evaluation is based on sources of information other than the statistical analysis of repeated measurements.

Examples include:

Note:

Type A and Type B are methods of evaluating uncertainty components and should not be confused with the classification of random errors and systematic errors. Both random and systematic effects may contribute to either Type A or Type B uncertainty evaluations depending on how the information is obtained.

4. Sensitivity Coefficients

When the measurand depends on multiple input quantities, each uncertainty component contributes differently to the final measurement uncertainty. Therefore, it is necessary to determine a Sensitivity Coefficient for each input quantity.

The sensitivity coefficient indicates how much the output quantity changes when an input quantity changes by one unit.

Mathematically, the sensitivity coefficient is defined as the partial derivative of the measurement model with respect to the corresponding input quantity:

Where:

• ci = sensitivity coefficient of input quantity xi
• f = measurement function (measurement model)
• xi = input quantity

The sensitivity coefficient is used to convert the standard uncertainty of each input quantity into its contribution to the uncertainty of the measurand.