Elsmar Cove Forums Index - Tolerance vs Total Variation in Gage R&R report

Pages: 1

Tolerance vs Total Variation in Gage R&R report - MSA 3rd edition pg. 116
(View the original discussion thread with full colors, attachments and images)

Posted by: Geese

In the GRR report, TV (total variation) is calculated using the calculated GRR and PV. However, in the analysis section of the MSA 3rd edition pg. 116, it states that the TV may also be calculated using process variation / 6.00 if the process variation is known.

The reason I'm presenting this is that we have 2 separate report forms that calculate TV using these 2 different methods, only they call TV Tol in the 2nd report. And in the 2nd report, Tol is calculated using upperlimit - lowerlimit / 6.00 instead of process variation / 6.00. Is process variation = upperlimit - lowerlimit?

To verify the authenticity of this method or our report rather, I need to know what the upperlimit and lowerlimit numbers refer to. Do they refer to the tolerance of the actual part being measured, or are they a calculated number. Our report thinks they are the actual tolerances of the part and puts them in as such, but I want to make sure that is what they actually are, but process variation does not sound like it would be the actual tolerance on a part.

Posted by: Marc

Can anyone shed any light on this?

Posted by: Atul Khandekar

The "standard" way of representing %GRR is (GRR/TV) where TV is the Total Variation (stDev) observed during the study. PV in the GRR form is the observed Part Variation. These numbers depend on the way you select parts for the study - whether they truly represent the process or tolerance spread etc.

The two other methos are:
1. GRR as % of Tol: Where Tol= Upperlimit-LowerLimt. You can substitute (Tol/6) in place of TV above to get GRR as % of Tol.
2. GRR as %of Process Variation: This process variation (sigma) comes from your routine SPC studies on the characteristic under study.

Hope this helps.

Posted by: canhchi

Hello, nice to join this forum.

%GRR can be calculated by (GRR/TV) or (GRR/(Tol/6). But I wonder that in what case TV (Total variation) is used and what case use (Tol/6) to calculate as TV is not equivalent to (Tol/6).

Please explain to me.

Thanks and best regards,

Chi

Posted by: Howard Atkins

Welcome to the cove

Please see this post MSA using tolerances - Page 116 of the AIAG MSA manual for more information

Posted by: canhchi

Thanks for your reply!

I think I should ask more detail.

The page 74 of MSA book issue 3, refer to selection samples. If the sample parts DO NOT represent the production process, TV must be ignored and use Tolerance in assessment.
How are samples taken to represent the production process? And in what case, samples DO NOT represent? Please teach me.

I separated measuring devices into two groups: 1st group includes devices used in production line, and 2nd group includes devices used for final inspecting before shipment in Quality Inspection Division. With the 1st group, I taking one sample from process for several days and then calculate %GRR by 100*(GRR/TV). With the 2nd group, I don't take samples from process but taking only from a production lot (any lot) and calculate by 100*(GRR/(Tol/6). My thought is wrong?
Give me advice.

Thanks and best regards,
Chi

Posted by: boobab

hi,

i just joined this forum. im not sure if you're still interested but i would just like to share something. please disregard my input if somebody has already mentioned this in the previous threads.

i also take care of MSA studies in our company and this is how we consider tolerance and total variation.

tolerance is considered when the gauge being studied is used to measure product characteristics.

total variation is used when the gauge being studied is used to measure process characteristics.

Posted by: gszekely

Hello Chi !
I don't know if I1m right, but please have my interpretation:
1) Usually MSA is done in the early phase of product creation. At this stage you don't know your real process variation, even if you already have done a preliminary capability analyses. (ex. raw material, lot to lot variation)
2) In my opinion there is no process, which will not produce out of product tolerance parts.
So if at the stage, you conduct the MSA study, you can not include samples out of tolerance, then you should report to product tolerance.
If you can have parts out of tolerance, then you should report to the real process tolerance, which is larger, than product tolerance. Of course within a certain range.
There is not stated in the standard that you have to pick up your parts randomly from process. I usually try to collect samples, which covers the full production range:
EX: if my parts is 10mm +/- 0,1, then I would search for a parts something like this:
1) 9,88 out of tolerance
2) 9,93
3) 9,95
4) 9,97
5) 10.00
6) 10,03
7) 10,05
8) 10,08
9) 10.10
10) 10,12
Just my approach, how I interpret. Am I wrong ?
György

Posted by: Miner

Quote:

Originally Posted by canhchi

The page 74 of MSA book issue 3, refer to selection samples. If the sample parts DO NOT represent the production process, TV must be ignored and use Tolerance in assessment.
How are samples taken to represent the production process? And in what case, samples DO NOT represent? Please teach me.

I separated measuring devices into two groups: 1st group includes devices used in production line, and 2nd group includes devices used for final inspecting before shipment in Quality Inspection Division. With the 1st group, I taking one sample from process for several days and then calculate %GRR by 100*(GRR/TV). With the 2nd group, I don't take samples from process but taking only from a production lot (any lot) and calculate by 100*(GRR/(Tol/6).

The second group is easy. They are used for product acceptance and should be evaluated as P/T Ratio of measurement variation as a percent of tolerance.

The first group depends on how you are using the gages. If you are using them for inspection, the answer is the same as group 2. If you are using them for process control such as SPC, you need to assess the gage using the PV process variation. Then the selection of samples need to reflect the actual variation of the process. The samples selected need to form a subgroup with a standard deviation that is representative of the short term variation (standard deviation) of the process itself.

Posted by: gszekely

Miner ! Thanks for the answer. In our case we use one measurement, in production phase, as the part is shipped directly to the customer. we do use the data for SPC, as well. (ex. injection molding) I met the situation, when my part tolerance was + /- 0,1, and my actual process variation + / - 0,02. Reporting to process variation, I could not get acceptable GR&R result, below 10%, while I was quite good, reported to part tolerance. can we say, that if process wariation, is less than product tolerance, we can use the last one ?
György

Posted by: Miner

Quote:

Originally Posted by gszekely

Miner ! Thanks for the answer. In our case we use one measurement, in production phase, as the part is shipped directly to the customer. we do use the data for SPC, as well. (ex. injection molding) I met the situation, when my part tolerance was + /- 0,1, and my actual process variation + / - 0,02. Reporting to process variation, I could not get acceptable GR&R result, below 10%, while I was quite good, reported to part tolerance. can we say, that if process wariation, is less than product tolerance, we can use the last one ?
György

This means that the gages are acceptable for use to inspect product to specification, but may not be acceptable for use as a process control (SPC) gage.

I say "may not" because there are several items to investigate.

First, were the samples actually representative of the full short-term process variation? If they were not, the analysis will give you worse results. Second, if your results were greater than 10% but less than 30%, the gage may still be acceptable for use.

I would also check to see if the gage has enough resolution to detect the process variation. The Range chart from the MSA study should have at least 5 distinct levels (including 0) under the UCL-R.

Posted by: gszekely

Thanks again Miner ! The answer is clear. The measuring device, is a video CMM, resolution up to 4 or 5 digits. Howewer, I usually don't care about the 3rd one allready. I was talking about milimeters, and I foud quite difficult to measure microns, having the certainity that you have measured what is there. On the other hand, I would not spend time, improving the process, with the above mentioned results. I set the control limits for the process, to 80% of the tolerance limits, and as far as my process is between, I don't issue corrective actions. A lot of start ups, batch changes, and interventions, due to cosmetical problems alraedy. I didn't have any customer claim, during thze last 5 years, due to dimensions.
Thanks
György

Posted by: mahender

Hi There,
My interpretatin was that for a new peice of equipment that has no spc we can use the tolerance.
once the equipment has become established and the spc should decrease its limits to reflect actual process variation (UCL and LCL) as oppose to up and lower FFU.
in practice you should get parts that represent the process spread as mentioned above but in some cases for us we would have to manufacturer these specifically. but the value of the part is not the goal it is the R+R of the gauge.
so when you do the study. if you use tolerance your value will be better than using 6 sigma from spc.
and using 6 sigma your part variation value will be large.

Posted by: metrocat - 2009

Hi,

In our company, the tolerance approach for %GRR calculations is used when the measurement system has a larger tolerance range, greater than 5%. And the Total variation approach is used when the measurement system has a tight tolerance range, lower thar 5%.

I hope this can be helpful.

Posted by: bobdoering

In a simplified manner, the way I like to describe the difference is that % of Total Variation is the participation of the gage system error to the total variation (gage system variation and part variation). % of Tolerance is the amount of tolerance used up by the gage system error - so no longer available for the part to use.