Sheffield Endeavor R+R Question

[jimbob]

We have just completed an ID R+R study on our Sheffield Endeavor, running PC-DMIS 4.1. The ID spec is 63.487 +/- .013. I took 35 hits with a TP-20 using a 3mm x 40mm probe. The R+R results were ~18%. My question is, does this seem typical? The R+R was completed using 5 pieces, taking 3 different readings, and calibrating in-between each user. Any advice, remarks are appreciated.

Thanks!

[Matthew D. Hoedeman]

We have just completed an ID R+R study on our Sheffield Endeavor, running PC-DMIS 4.1. The ID spec is 63.487 +/- .013. I took 35 hits with a TP-20 using a 3mm x 40mm probe. The R+R results were ~18%. My question is, does this seem typical? The R+R was completed using 5 pieces, taking 3 different readings, and calibrating in-between each user. Any advice, remarks are appreciated.

Thanks!

Well, when doing any kind of an R&R study, alignments AND calibrations should stay the same, not be re-done for a part or an operator. Everything that you CAN keep the same should be kept the same. The only variables need to be the parts, the holding fixture and the operators. No alignments, calibrations, don't move the fixture, things like that.

[cmmguy]

Plus and Minus 13 microns??? that is pretty tight.
that would require a repeatability of around .002 total to get it under 10%

Dont recal tips at anytime during R&R.

[BKulpa]

My Leitz Reference can repeat better than that!

[cmmguy]

My Leitz Reference can repeat better than that!

Ok, do a 5-3-3 study, post your methods and your results or dont make comments like that.

[John Kingston]

My Leitz Reference can repeat better than that!

That's Incredible!

Ok, do a 5-3-3 study, post your methods and your results or dont make comments like that.

[bw_bob]

That's Incredible!

HAHAHAHAHAHAHAHAHAHAHAHAHAHAHA...........GOOD ONE

[BKulpa]

OK!
0.014mm diameter tolerance.
10x3x3 study, ANOVA method.
9.29% OR 0.0013MM

[cmmguy]

Not that easy... we need data and for you to describe the method that you used. Many incredible people(not implying you) post just as incredible results but dont post the true methods or data. Also when I say method I dont mean your calculations but what conditions were the tests done under. ie, fixtures, physically remove the part, 3 real different operators, etc...


btw .0013 range of a .014 tolerance would not yield a 9.295 GR&R. For a GR&R of .014 tolerance your number ranges should be in the .0005xx realm. So maybe this is where the methods helps with this.

[Wes Cisco]

Not that easy... we need data and for you to describe the method that you used. Many incredible people(not implying you) post just as incredible results but dont post the true methods or data. Also when I say method I dont mean your calculations but what conditions were the tests done under. ie, fixtures, physically remove the part, 3 real different operators, etc...


btw .0013 range of a .014 tolerance would not yield a 9.295 GR&R. For a GR&R of .014 tolerance your number ranges should be in the .0005xx realm. So maybe this is where the methods helps with this.

Bubble Buster ! ! !

[BKulpa]

I used a fixture, 3 independant operators. (It took approx. 20hrs.)
I measured a cylinder at 8 levels 17 points per level.

The 0.0013 is not the range I observed; It's the total repeatability and reproducability I expect to see when using the program in a production environment.

[cmmguy]

I used a fixture, 3 independant operators. (It took approx. 20hrs.)
I measured a cylinder at 8 levels 17 points per level.

The 0.0013 is not the range I observed; It's the total repeatability and reproducability I expect to see when using the program in a production environment.

Those are pretty good numbers. But that looks like a 10x3x1 study not a 10x3x3.

A 10x3x3 study would yield 90 different measurements. 10 parts measured by 3 operators, 3 separate times. Removing the part from the fixture on each cycle.

That is why most CMM studies lean toward the 5x3x3 study.