Most gage R&R studies recommend using at least 80% of the process’ variation. But why does this recommendation exist? We want to assure ourselves of the integrity of our assessment. That assessment includes an assumption that the gauge performs across all the measurement ranges of interest for the product or process being measured.
Getting parts across 80+% of a typical process variation is a simple statement, but impediments can impact these recommendations. We will exclude the cases in which a team does not want to spend the time and effort to make this happen. Of course, using leadership and interpersonal skills along with project management skills can help with a team that needs a little more persuasion.
The first option to consider, depending upon the cost and time of creating parts, can be a quick and easy solution. Gather items from the process right away from your process. If you have many lines or machines producing the same product, gather samples from each line. This approach is much easier for circumstances with shorter manufacturing or process cycle times. Stamping pieces of metal creates more parts from which to select than chemical batches that may take hours or days to produce the product. One other suggestion to gathering samples quickly, independent of cycle times, is to see if quality assurance or another group has samples saved from past production runs. Remember that the parts do not even have to be in the specification, just typical of the variation of the process.
The second thought for part selection is if process variation is impacted by the supplier or from seasonal effects? It may be difficult to see a full range of variation, and teams cannot wait months or quarters to see such variation. If you are blessed to be part of a larger organization that is worldwide, reach out to other regions that may be seeing the seasonal effects that you cannot see presently in the product. Do not forget that old samples that have not spoiled or deteriorated can be a good source. Another option to get variation is to search for specific examples of parts that may increase the range of variation than you can find quickly. This often takes specific process knowledge but can be useful. An example: many laminators have variations along the edges of the product that show typically thicker than the usual application of the material. If you could gather some of that material, it could be an example of process variation you will not find typically in a day but want to measure and confirm your gauge’s ability to measure. These options are not ideal, but it is better than not getting parts for a solid Gage Repeatability and Reproducibility (R&R.)
Third, let us remember what we mean by process variation. Let’s say a scale is used to measure weight of grades of paper. We all recognize that newsprint paper weighs differently than laser printer pages and construction paper used for school children’s projects. However, all three may be made on the same machinery or at least all grades may be measured in the same lab with the same devices or gauges. If one did a gage R&R with all three grades of paper it would create the process variation. At that point, we could then look solely upon the %study variation will almost certainly erroneously help us conclude the gage R&R is acceptable. Any variation by operators will be overwhelmed by the variation measured by the three grades improperly used in the same gage R&R. However, if the gage study mistakenly used all three grades of paper, the % tolerance result of the gage R&R will be an appropriate measure. We must remember the following:
% study variation = gage variation / process variation
% tolerance = gage variation / tolerance width
Tolerance width is the span of the upper specification limit MINUS the lower specification limit for ONE paper grade. Below, you can see the impact of using samples from three different grades in one gage study.
From the above figures, we learn that choosing improperly gathered samples can mislead a gauge study if wholly different products are selected to just create a range—especially if it creates a larger range than typical. Figure 4 shows how poor selection of dissimilar products can artificially lower the % study variation. This means it would not make sense to measure the diameter of oranges and grapes and watermelons and expect to get a good idea of the gage variation of the gage measuring the diameter. We want a gauge to perform for a product so we can detect variation of the product or process. This is important to have a good measurement system if we want to get more statistical confidence when using other tools in the Analyze and Improve phases of a typical DMAIC approach of Six Sigma.