Measurement system evaluation

Measurement systems evaluation

The old saying garbage in, garbage out?is also applicable to quality-related data.?The data used must be valid.?That is why it is important?to evaluate the measurement system?to ensure you're using data that's indeed valid.?

The measurement system is the entire system used?to measure, collect and record data.?This includes sensors, instruments,?gauges, methods, procedures, human appraisers?and the operational definitions of the measurement?and the assessment criteria.?

When measurements are made,?the variations observed consist of two major components:?

• the actual part-to-part variation,?
• and the variation due to the measurement system?or measurement error.?

Ideally, the measurement system variation?should be zero?so that the observed equals the actuals.?But in practice, there's always some variation or error.?

Hopefully, the measurement system variation?or error is just a small percentage compared?to the total variation observed.?The rule of thumb is it should be less than 10%.?

Now, let's look at the components of variation.?Measurement system variation consists of variation due?to gauges, the measurement equipment,?and variation due to operators?or appraisers, the people doing the measuring.?

Measures in variation or errors?can be location-related or spread-related.?Location-related variations are bias,?stability and linearity?while spread-related variations?are repeatability, and reproducibility.?

Let's explore these ideas a bit more.?First, starting?with location-related measurement variations.?Here are the definitions.?

• Bias is the difference between the observed average?of measurements and the reference value?or correct master value.?Bias is often termed as accuracy?but to avoid confusion with the common meanings?of accuracy, it should not be used.?Bias is easily corrected through calibration.?
• Stability or drift is the total variation?in the measurements obtained with a measurement system?on the same master or parts?when measuring a single characteristic?over an extended time period.?Stability can be determined by using a control chart?in a gauge study.?Out-of-control conditions?in a control chart indicate instability.?
• Linearity is the difference in the bias values?throughout the expected operating range of the gauge.?

As for the spread-related variation,?here are the definitions.?

• Repeatability is the variation in measurements obtained?when one measurement instrument?when used several times by one appraiser?while measuring the identical characteristic?on the same part.??
• Reproducibility is the variation in the average?of the measurements made by different appraisers?using the same measuring instrument?when measuring the identical characteristic?on the same part.?

All of these components of variation?can be evaluated in an MSA study.?But remember, how you evaluate a measurement system?depends on the type of data.?

For categorical or attribute data,?conduct an attribute agreement analysis.?This is evaluating the validity?of pass-fall or go-no go inspections?or the classification of defect types.?For example, reason codes or classifications?for customer complaints and returns.?

For continuous or variable data,?conduct a gauge study,?such as a gauge R&R (repeatability and reproducibility) study.?

Such evaluations are called?Measurement System Analysis, or MSA.?So the next time you want to use existing data,?make sure that a measurement system has been validated.?If not, conduct a MSA study first.?If the measurement system is not valid,?improve it and then confirm it?with a follow-up MSA.?

The importance of having valid measurements?cannot be emphasized enough.?Managing quality requires?that the data used is valid.?Otherwise it's garbage in, garbage out.