VDA-5 standard shows the way for future of quality engineering

There are two different methods of determining the accuracy or capability of measurements: uncertainty evaluation and measurement systems analysis.

Uncertainty evaluation is the ‘gold standard’ used to determine the uncertainty of measurements in physics experiments and calibration laboratories. 

When carried out properly, uncertainty evaluation fully considers all influences on a measurement and gives the most reliable indication of its uncertainty. 

The problem with uncertainty evaluation is that it’s based on mathematical models and can require highly skilled metrologists. Measurement systems analysis (MSA) is used for production measurements. It uses standardised tests such as gauge R&R (repeatability and reproducibility) studies and provides statistics that are based entirely on the observed variation in measurement results. MSA is easier to implement in industry but can sometimes miss systematic sources of uncertainty.

In their constant quest to improve quality, many manufacturers are starting to use uncertainty evaluation within product verification. In practice, this means implementing established MSA methods within an uncertainty framework. VDA-5, published by the German Association of the Automotive Industry, is the first standard covering this type of hybrid quality engineering.

What is an uncertainty framework?

The uncertainty framework used by calibration labs and physics experiments is the GUM uncertainty framework, set out in the Guide to the Uncertainty of Measurement. This requires all potential influences on the measurement result to be considered and a mathematical model constructed. The model quantifies the sensitivity of the measurement result to changes in the influence quantities. This enables two ways of evaluating uncertainty to be combined. 

Type-A methods involve repeated measurements of a reference and statistical analysis of the results, for example a gauge study. These can be used to evaluate the effect of random variation and systematic effects that can be easily varied in a gauge study. 

Type-B methods use other sources of information such as calibration certificates, material property data sheets, and environmental monitoring data logs. Since the mathematical model gives the sensitivity of the measurement result to a change in each of these influences, the uncertainties can be combined mathematically.

The simplest and most common way of combining uncertainties is using an uncertainty budget. This is a table listing each influence with its uncertainty and sensitivity coefficient. It provides an easy way to calculate the effect of each influence and the combined uncertainty for the measurement as a whole. 

What’s wrong with gauge studies?

It’s widely believed that uncertainty evaluation is more rigorous and scientific, but exactly what’s missing from a gauge study is less well known. The simplicity of gauge studies may at first suggest that they are infallible. Calibrated reference parts are measured many times, allowing the random variation and any bias in the measurement system to be calculated. 

This approach assumes that the uncertainty of the calibration is negligible, so the calibrated values can be assumed to be the true values of the reference parts. 

This can itself be problematic since it may not be possible to obtain such accurately calibrated references. Assuming it is possible to obtain suitable reference parts, there remains an issue that it’s probably not possible to vary all of the influences on the measurement so that the calculated variation and bias fully represent the actual uncertainty. 

Most gauge studies assume that the only influences on the measurement are the part and the operator, ignoring other factors. This can lead to gauge studies indicating that a measurement’s variation is many times less than its actual uncertainty.

How does VDA-5 implement MSA methods within an uncertainty framework?

VDA-5 specifies that an uncertainty budget should be created, considering all influences on the measurement, including systematic effects. 

It also requires that a gauge R&R study is used to quantify the random uncertainty within this budget. The resulting uncertainty can then be used in various ways. 

Tolerances can be tightened by the uncertainty to give conformance limits. Measurement results within these limits prove, with statistical confidence, that a part conforms to the tolerance. The ratio between the tolerance and the uncertainty can also be used to determine measurement capability.

VDA-5 also provides methods of evaluating the uncertainty of attribute gauges. These are tests that give a binary pass/fail result rather than a variable dimension. Examples are go/no-go plug gauges and visual inspection processes.

The hybrid approach is undoubtedly the future of quality engineering. VDA-5 provides relatively simple methods to implement it right now. Some technical errors remain, but for most purposes these will have a negligible effect. In any case, it represents a major step forward over simple MSA methods.