I ran this gage study on Thursday while our customer was here on a new piece of equipment. We felt pretty comfortable with the results, that is until I "plugged" them into the spreadsheet. Can anyone help us understand why this gage r&r fails? I thought it looked pretty good.

The NDC is at 4, you need 5 min to pass on that sheet.

Also I would have Hector remeasure parts 2 and 9.

I suppose operator C with part number 9 seems to high. in all the other case part range is aprox 0.01. In that case is 0.03 with no range inside operator C.

may be someone else can give You a statistical description

Did you pick randomly 10 parts or did you select a group of parts that would represent the process? If it was random 10 part selection, then the process may become artificially narrow with results as you have shown.

Ok... took all the suggestions.. Hector remeasured parts 2 and 9 and they were closer to the others... NDC now at 5 - R&R still fails.

I've actually forgotten most of the statistics behind a gage r&r (damn computers making things too simple for us)... possible problem with the calculations?

BTW - these were 10 parts that my customer pulled from the line while it was running.

The gage EV-TV is to high, have reddy recheck part #2. If I get time I will type it in the minitab.

http://elsmar.com/Forums/showthread.php?t=27782&highlight=MINER

Use the forms Miner has posted, I think that the sheet you used is bad.

Here is what the mighty minitab says...

QMMike,

Maybe some amplifying remarks on what others have said.

What will the gage be used for? If for product release to fairly liberal specifications, you may be OK. Key on the P/T ratio. Use the spreadsheets Miner made if you don't have MINITAB.

If for analytic studies, control charting etc, then you need to correct your flawed study. It is flawed because of the samples taken, UNLESS they do represent the normal variation of the process. From how you describe the sample selection, this would not be the case.

Get samples that span the normal range of the process and repeat. This will dramatically improve all the numbers, especially the NDC, which is the failure item driving the rest of the metrics.

In summary; if you will use the gage to release product only, it may well be OK as is. Calculate the P/T.

If you want it for analytic studies (process improvement, control charts) or if you are meeting some customer requirement, then you need to repeat a fairer evaluation.

If the gage will be used for these kinds of studies and the parts span in the original GRR IS reflective of the actual process spread, than it is not suitable. You will need to improve the gage or take repeat averaged measurements
.

Thanks again, all...

I'll be repeating this R&R again with a truer representation of the process.

Get samples that span the normal range of the process and repeat. This will dramatically improve all the numbers, especially the NDC, which is the failure item driving the rest of the metrics. .

Very true, unless the normal range of the process is incredibly small...then it will show really bad R&R results.

Kind of unfair, the better you get at making the part, the less variation there is, and the better the gage has to be to see this reduced variation.

Sooner or later in CI activities you get to a point where the best gage you can find is lacking the resolution needed to separate your excellent parts, and come up against the somewhat unhelpful advice to develop a better gage then the current best one...

But of course this is a good problem to have...

http://elsmar.com/Forums/showthread.php?t=27782&highlight=MINER

Use the forms Miner has posted, I think that the sheet you used is bad.

Here is what the mighty minitab says...
From the Range chart, there are only two possible measures below the UCL. There should be at least 5. This indicates that the gage resolution is inadequate. This is the primary cause for the high %SV Repeatability.

If however, the gage will be used for inspection, you can ignore this per Dave's suggestion as the %Tol was good.