Determining Optimal Factor Levels for EMP Studies: How Many Appraisers, How Many Samples, How Many Repeats? How Do I Know?
Jed Campbell, Quality Manager, US Synthetic
What is the rational basis for determining the correct number of appraisers, parts and repeats for a measurement system analysis? Is there any rational reason for the rules of thumb that exist, such as the three appraisers, 10 parts, three repeats guideline (AIAG 2002)? Building a model allows for the specification of repeatability and reproducibility variation components of a hypothetical process. The model also specifies the number of appraisers, parts and repeats that an experimenter would use to conduct an EMP study. The model builds a sample EMP study based on the factors entered and populates it with random-normal data based on the variation components chosen. The model then back-calculates and reports the variation components determined from the study. A Monte Carlo simulation is used to determine the accuracy of the model in determining the true variation components over a very large sample size. A full-response surface model was then created, including a number of appraisers, parts, repeats and hypothesized variation components. The resulting DOE led to a JMP application which includes a profile that allows users to interactively search for optimal factors' levels for their EMP studies, as well as a utility to build the EMP data table for data gathering and analysis.