Decoding Gage R&R Output
What’s the punch line? Removing measurement system error from total observed error can be what I call a “quick win”.
Who couldn’t use a quick win these days. This guide assumes that your are a trained Lean Six Sigma Black Belt or have practical knowledge in conducting or interpreting the output of a gage R&R study (measurement system repeatability and reproducibility study).
Let’s get started. If you look at the normal distribution in the inset photo below, we ask ourselves, what percent of this variation is assigned to measurement system variation? The formula below (in the inset photo) also suggests that if we were able to significantly reduce measurement system variation, we would reduce overall observed variation. Not a bad deal, right?!
But before we can remove measurement system variation we have to be able to measure it and do something about it. Yes, you CAN do something about it and your process owner and champion will love you for it.
Study purpose and options to consider when selecting the study output and analysis include the following;
Variation of your product samples. In other words, the range of variation that these sample parts represent or should represent.
Product Tolerance for sample product.
Historical product standard deviation or variance.
That being said, there are four measures of study variation that one can choose. Question is, which one? See Minitab graphic above. Each color bar represents;
% Contribution
% Study Variation
% Process
% Tolerance
If you’re not an aficionado of gage R&R studies, the study options could make you scratch your head. In fact, you may have been exposed to all of the output options below during training and now you're wondering how to turn all of these off. Let’s keep it simple and get focused on one output that will always work,
.....% Contribution.
The only caveat is that you really need to know the purpose of your study and the appropriate width of your product sample variation. If you don’t have a handle on that, sign up for my next Lean Sigma Black Belt class that addresses both variable and attribute studies. :-)
Now, let’s clean up the output with just % Contribution and % Study Variation.
Notice the graphical output above has only two bars, % Contribution and % Study Variation. The computation for % Study Variation is not as clean as the % Contribution (shown inside red box).
In fact the % Study Variation data in the bottom table of the output below uses the VarComp data in the top table of the output. So what’s the purpose of the bottom table? Not so much in my opinion. Notice the %Contribution variance components add nicely together (i.e. total Gage R&R of 10.67 = Repeatability 3.10 + Reproducibility 7.56, etc.) If you try this with % Study Variation, it won’t work because standard deviation is a non linear value. In other words you can't add standard deviations together.
Notice that Reproducibility (7.56%) is further subdivided into Operator (2.19) and Operator*Part (5.37). Both sum up to 7.56%.
Ok, what should good measurement system variation look like? Here’s a quick table for an easy reference.
Total Gage R&R should be 5% or less. The 10.67 percent shown in this case study is marginal. This number indicates that 10.67% of the observed variation for the sample parts is assigned to measurement system variation. Consider using this measurement system only if the process capability is very good.
As a rule, when Repeatability is high, this usually indicates a problem with the measurement device. When Reproducibility is high, the measurement method is inconsistent between appraisers.
If you enjoyed this article, click the like button and share with your network. If you would more information about this topic or related initiatives contact us at succeed@cikata.com. Don’t forget to click on our social links for the latest.