MSA, Precision to Total/Tolerance and Gage RR?
Six Sigma – iSixSigma › Forums › Old Forums › General › MSA, Precision to Total/Tolerance and Gage RR?
 This topic has 25 replies, 9 voices, and was last updated 17 years, 6 months ago by Swaggerty.

AuthorPosts

February 3, 2005 at 3:28 pm #38284
Can anyone clarify the following ?
I have carried out a gauge study using 3 operators, 10 parts and one piece of measuring apparatus. Each operator measured the same part twice. Results where recorded separately.
I entered the data into Minitab and carried out a Gage RnR for the data.
However in parallel i though i would work through the manual calculations provided by my BB training, for familiarity reasons. SO………….
I calculated “sigma repeat”, using [RBar Bar / d2 for a subgroup of 3], where R is the Range and d2 is a Shewhart constant.
I calculated “sigmas repro”, using XmaxXmin / d2 for a subgroup of 3.
Squaring these, summing them together and rooting them gives me “sigma measure”.
I then carried on calculating the sigma level of all of the results taken, using sigma level = 3 * Cpk, where Cpk is the min value of (USLmean/3sigma) or (meanLSL/3sigma).
From this I could determine the sigma of the dimension,
sigma tot = sigma dimension + sigma measure, and I continued to find PTot andPTol.
However I can’t quite link this to what Mintab feedsback, in fact Minitab does not require USL and LSL to complete the gage study, and does not mention Ptot and Plot
Can anyone explain all of this ?
Regards, Simon.0February 3, 2005 at 3:36 pm #114418You may want to question your training. You should know the answer to this. Who trained you?
0February 3, 2005 at 3:46 pm #114419Stan,
If this is the best advise you can give a neawly trained Black Belt, (who payed for all of his training himself and took leave of absence for 4 weeks in doing so), and who is bravely attempting to turn the culture around in the factory where he works who isn’t interested in six sigma, then I say to you may wish to question your training…… ass hole !
Simon.0February 3, 2005 at 3:57 pm #114420I dont want to read more into your response than you intended, but were you appreciative of Stans input or not? It was difficult to tell. Could you please elaborate? Thanks.
Vinny0February 3, 2005 at 4:11 pm #114422Vinny,
No. If someone you did not know, was to ask you a question regarding a topic that they were attempting to grasp, (and provide some detail of there attempts), this topic being one that you that you had a good appreciation of, would you tell them to go away and read your notes again or would you offer a some constructive help.
While it may be possible that I did not absorb every aspect of the equations, calculations and concepts in the mere 4 weeks of my training, (considering Mintab sheltered me from almost all of the equations), I did learn that no question is too trivial to warrant such a pompous response, considering Stan does not know who I am, what I do, how old I am, what my current educational background is, who trained me, what my course note content was, etc.etc.
Are you about to give me a similar reply or are you mature enough to know that we are all human at the end of the day !
Simon.
0February 3, 2005 at 4:24 pm #114423
Don’t seriousParticipant@Don'tserious Include @Don'tserious in your post and this person will
be notified via email.Simon
Don’t serious with Stan’ Standard. Many people had asked him to leave this WEB. But there is no effect.
The easy way is Don’t read Stan answer. And the way you’re being is good.0February 3, 2005 at 4:26 pm #114424Did you click on ANOVA method or xbar/R method in Minitab?
This is how Minitab calculates its values.
http://www.minitab.com/support/docs/CalculatingVarCompsbyHand.pdf0February 3, 2005 at 4:28 pm #114425I, unlike many both on this forum and life in general, am exceedingly intelligent, mature, handsome, and worldly. So asking me about how I might or might not respond in a specific circumstance might not give you the best sample with which to base forum contributory determinants.
But, I assure you, if I were to respond in a pompous and rude manner, not that I would you understand, I would only do it in an anonymous manner on a technical forum its just my way.
I really just wanted to see whether or not youd call Stan some more ugly names if provoked. And I really think hes got it in for you. I dont know why he does. Because you seem like a really nice fellow.
Vinny0February 3, 2005 at 4:44 pm #114426It looks like you are trying to do to much, and combining your apples and oranges. Things knid of fall apart after you calculate sigma measure.
The gage study does not require tolerances, unless you want to report % of Tolerance. The gage study is supposed to use parts that represent the total variation within the process. You use the sigma variation in the parts to obtain the process variation, not a Cpk calculation. The Sqrt (process variation**2 + measure variation**2) gives you the total variation.0February 3, 2005 at 5:09 pm #114427
Thomas C. TribleMember@ThomasC.Trible Include @ThomasC.Trible in your post and this person will
be notified via email.Simon:
I have a few solutions to Gauge R&R problems that I solved long hand, including an Excel spreadsheet for completing the calculation fairly easily. If you post an email address, I will send what I have to you.
TC Trible0February 3, 2005 at 5:19 pm #114428
AnonymousParticipant@Anonymous Include @Anonymous in your post and this person will
be notified via email.Remember that in a Gage R&R, you want to see how much variation is due to the measurement system compared to other sources of variation. For this purpose, I’d suggest using control charts. These are easy to understand and create, you can look graphically at lots of different “issues”, and calcuate you sighat^2 (Rbar/d2), an estimate of the variation. Using the control chart, you can quickly compare those to the USL & LSL and tell what kind of shape you’re in. It’s not as accurate as the ANOVA method, but I think it’s a lot easier to interpret and digest rather than MiniTab telling you what it thinks is right.
0February 3, 2005 at 5:28 pm #114429
AnonymousParticipant@Anonymous Include @Anonymous in your post and this person will
be notified via email.Don’t fret, Simon. Stan is just light years ahead of us and doesn’t have time to answer “small” questions (even though he is a common 1st responder with snide comments). With his busy six sigma plate, he comes out just to taunt us mere mortals!
There are more mortals on this forum than Gods like Stan… if you have a question, just ask.0February 3, 2005 at 6:20 pm #114436
SwaggertyParticipant@George Include @George in your post and this person will
be notified via email.Simonwar,
Minitab does not require spec limits however you can put them in in order to get a % tolerance value.
Click the options box and put this in under process tolerance. It is a range( USL – LSL).
I dont use the calculation that you have mentioned however here is my understanding of the signifigance of the main Minitab numbers.
As far as the Minitab output goes the Total Gage R&R % contribution (On the first table) is the measurement system variation with respect to the total variation.You use this number to evaluate you measurement system in comparison to the variance in the parts you are trying to measure.
If you look at the bottom table there will be a %tolerance column that will give you the measurement variation compared to the specification width if you entered a process tolerance.Keep in mind by default you are looking at 6 times the standard deviation divided by the spec width. Use this number to evaluate your measurement system with respect to your spec width.
The last value that MiniTab outputs is the number of distinct catagories. This is really putting the first number (% contribution) into a way that is easyier to understand at a glance.
If you have 3 distinct catagories your system can put the parts into three buckets, low, medium and high. With five catagories, very low, low, medium, high and very high. If you were evaluating a measurement tool for use in control charting you can see that 3 catagories is just not enough.
Hope this helps
0February 4, 2005 at 3:48 am #114450Okay, I’ll bite – who are you, what do you do, how old are you, what is your current educational background is, who trained you, what was your course note content was, etc.etc
Thanks for calling me an ass hole. What does that make someone who paid for poor training and took 4 weeks leave to get bad training. If you want to learn, contact me at [email protected] – I will answer your questions.0February 4, 2005 at 3:50 am #114451Could be I want you to question your crap training!
0February 4, 2005 at 3:53 am #114452Vinny, I think this guy is a metrosexual – careful, we’ll be spreading rumours your consort with this type.
0February 4, 2005 at 4:28 am #114455Nope, I just sensed a reticence on his part to fully express himself while in a state of apparent angst regarding your admittedly brusque retort. I was trying to draw him out a bit believing that he had a tad more to say about the sting from his faceless interface with you earlier.
While fundamentally a rugged heterosexioscientific kind of guy, Im also gifted with amazing empathetic abilities and powers much like Lieutenant Commander Deanna Troi on Star Trek Next Generation (except of course she was a hot alien chick and I, as I have already established, am a rugged heterosexioscientific kind of guy) and I sensed the pain of a noncommunicated and somewhat stifled series of expletives on his part which as we all know can danger an already fragile psyche if left pent up and nonexpressed. I was just trying to help, figuring that if he leads off with ah—, no telling where hell go once he loosens up.
Vinny0February 4, 2005 at 4:51 am #114456Maybe he will go learn something
0February 4, 2005 at 8:56 am #114464NS,
I used the default ANOVA. I am assuming that this accounts for the differences, as my manual methods use the “xbar and R”.
Thanks for the Minitab extract, I have not had time to digress this fully, (but I will), and appreciate the detail.
I will double check my data against the example data, as I have an almost identical scenario; 10 parts, 3 operators, 2 measurements each, but don’t get any Reproducability figure from Minitab ANOVA method ? I must have entered it incorrectly?
Thanks, Simon.
0February 4, 2005 at 10:42 am #114469Jim,
Thanks for the detail. I have returned to my data and used the std deviation of the measured results for my sigma total, and returned to Minitab and selected Gage RnR (Crossed) and changed to xbarR to see if my numbers tally. (However I have been told that the ANOVA method is better – why I don’t know?)
I have checked some basic stats with my data and Minitab. All looks exact and the Std Devations are exactly the same to 4 decimal places.
But before I go any further Minitab is returning zero for my Reproducability.
Have I entered my data wrong, i have checked it and in my opinion I have entered it correctly ?
Thanks, Simon.
0February 4, 2005 at 10:53 am #114471Thomas,
Thanks for the offer. At the moment I have been swamped with options. Let me get a better grasp of where I am and I will possible take you up on the offer.
Thanks, Simon.0February 4, 2005 at 11:00 am #114473George,
Something isn’t right? I calculated a sigma Repeat and sigma Repro using Xbar and Rbar, (because I had 10 parts, 3 operators and measured the same part twice).
However Minitab is returning zero for Reproducability ?
Thanks, Simon.0February 4, 2005 at 11:02 am #114474Anonymous,
Thanks for the detail. I have a lot to digest at the mo but hopefully a pennt will drop eventually.
Thanks, Simon.0February 4, 2005 at 11:40 am #114475For starters, sigma level isn’t 3* Cpk. Second, Your hand calculations will give slightly different results because using an estimate of std dev derived from Rbar will be less accurate than computing it directly. The method you used has it’s roots in the 1960’s when people did this by hand and they simplified the calculation to minimize the computational error.
0February 4, 2005 at 1:40 pm #114478The ANOVA method is “better” than the Xbar method because it gives you a “p” value for the “operator*part” interaction.
This interaction lets you know if 1 or more operators are having difficulties measuring 1 particular part.0February 4, 2005 at 5:25 pm #114493
SwaggertyParticipant@George Include @George in your post and this person will
be notified via email.Hard to say what the problem is. Minitab does some weird things if there is a data entry error. If there is a way you can post the data I can run it and see.
I would not give up on Minitab. I think it is the best way to analyze a gage R&R.
0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.