Correlation of Gage R&R Value with the Measured Value of Parts

[nicholas_yeoh]

Hi

I have done a gauge study based on MSA 3rd edition book.

However, is there a way that I can establish the uncertainty of measurement from a gauge R&R study and apply it to the measured value of parts?

If I need to be certain that all parts are within specification when measured using my equipment, can I use the results from my gauge study to calculate an uncertainty of measurement that can be applied to the drawing tolerance,

for example:
If the part specification is 10.0 ±0.5 mm
and
If the gauge uncertainty is 0.1 mm (based on GRR value),
then is it true that I would have to ensure that all parts measure between 9.6 and 10.4 mm to guarantee that they are within specification?

I have been involved in GRR in the last three years but I have never
understood the correlation of GRR with the measured value of parts. I have
carried out some research into the interpretation of GRR and all I can find
is the same old biblical guidelines as below:

% GRR under 10% of TV: Measurement system is acceptable
% GRR from 10% to 30% of TV: Measurement system may be acceptable based on
the application
% GRR over 30% of TV: Measurement system needs improvement

Please share your thoughts.


NY

 

[jerry_Malaysia]

Re: Correlation of GRR Value with the Measured Value of Parts

Hi

I have done a gauge study based on MSA 3rd edition book.

However, is there a way that I can establish the uncertainty of measurement from a gauge R&R study and apply it to the measured value of parts?

If I need to be certain that all parts are within specification when measured using my equipment, can I use the results from my gauge study to calculate an uncertainty of measurement that can be applied to the drawing tolerance,

for example:
If the part specification is 10.0 ±0.5 mm
and
If the gauge uncertainty is 0.1 mm (based on GRR value),
then is it true that I would have to ensure that all parts measure between 9.6 and 10.4 mm to guarantee that they are within specification?

I have been involved in GRR in the last three years but I have never
understood the correlation of GRR with the measured value of parts. I have
carried out some research into the interpretation of GRR and all I can find
is the same old biblical guidelines as below:

% GRR under 10% of TV: Measurement system is acceptable
% GRR from 10% to 30% of TV: Measurement system may be acceptable based on
the application
% GRR over 30% of TV: Measurement system needs improvement

Please share your thoughts.


NY

could you let me know how you get the gauge uncertainty to be 0.1 mm based on GRR value?

 

[nicholas_yeoh]

Re: Correlation of GRR Value with the Measured Value of Parts

could you let me know how you get the gauge uncertainty to be 0.1 mm based on GRR value?

Hi jerry_Malaysia

That was just an example and has not been formulated using the proper GRR method.

Ignoring the example as posted previously, I have attached an excel document showing the actual 50 sample measurement data taken on one tab and a GRR spreadsheet on the next tab. I am trying to establish how I can relate my GRR results with the 50 sample measurement data.

1.) Is it correct for me to assume the R&R value of 0.00082 as the gauge uncertainty?
2.) If this is so, is it correct for me to apply a plus/minus R&R value of 0.00082 to the 50 sample measurement data in order to provide me with a true value of measurement?

Thanks, NY

 

[Bev D]

Re: Correlation of GRR Value with the Measured Value of Parts

2.) If this is so, is it correct for me to apply a plus/minus R&R value of 0.00082 to the 50 sample measurement data in order to provide me with a true value of measurement?

no that would not be correct. The 50 units you measured already have measurement error incorporated into them. And at 50 units the probability that the true value of any of the parts being substantially greater or less than the 50 part min/max is miniscule...remember that standard deviations form different sources (measurement error and true part variation) do not combine linearly (you cannot perform straight forward addition on them).

When I get some time I will look at your data in more detail...unless of course Miner beats me to it! : )

By the way remember that the value 0.00082 is a single standard deviation of the measurement error.

 

[Miner]

I attached two files that will provide two different perspectives.

Open the MSA 3rd Ed. file and review the Gage Performance Curve tab. This illustrates the distribution of measurement variation, part variation and total (part + measurement) variation. In addition, it shows the probability of accepting a part depending on its true value. You can use this to establish guard band tolerances if desired. This file is based on AIAG methods.

Open the Wheeler file and review the Tolerance Analysis box of the report. This is based on Donald Wheeler's approach and shows various guard band levels. The PE value (+/- .000582) is the effective resolution of a single measurement. Follow the hyperlinks to read more on his approach.

BevD is correct. You are basing some key decisions on the standard deviation of 10 parts.

The confidence interval for SD=.00082, n=10 is:
90% CI of the SD 0.0005981 to 0.0013491
95% CI of the SD 0.0005640 to 0.0014970
99% CI of the SD 0.0005065 to 0.0018676

From GraphPad QuickCalcs
.

 

[Richard Pike]

Hi

I have done a gauge study based on MSA 3rd edition book.

However, is there a way that I can establish the uncertainty of measurement from a gauge R&R study and apply it to the measured value of parts?

If I need to be certain that all parts are within specification when measured using my equipment, can I use the results from my gauge study to calculate an uncertainty of measurement that can be applied to the drawing tolerance,

for example:
If the part specification is 10.0 ±0.5 mm
and
If the gauge uncertainty is 0.1 mm (based on GRR value),
then is it true that I would have to ensure that all parts measure between 9.6 and 10.4 mm to guarantee that they are within specification?

I have been involved in GRR in the last three years but I have never
understood the correlation of GRR with the measured value of parts. I have
carried out some research into the interpretation of GRR and all I can find
is the same old biblical guidelines as below:

% GRR under 10% of TV: Measurement system is acceptable
% GRR from 10% to 30% of TV: Measurement system may be acceptable based on
the application
% GRR over 30% of TV: Measurement system needs improvement

Please share your thoughts.


NY

If you could share your historical Process SD and Process Tolerance I would be pleased to give you a (non-biblical) alternative view point. PS I already have your R&R Data etc.

 

[nicholas_yeoh]

If you could share your historical Process SD and Process Tolerance I would be pleased to give you a (non-biblical) alternative view point. PS I already have your R&R Data etc.

I have attached a calculation spreadsheet that determines the process SD based on 20 subgroups (5 samples per subgroup). The process SD (or known as sigma in the spreadsheet) is 0.00305 and is calculated using Hartley's constant.

I am a bit puzzled with the term 'process tolerance'. I would assume the process tolerance is 0.02 mm (as the USL = 6.91 and LSL = 6.89).

And at 50 units the probability that the true value of any of the parts being substantially greater or less than the 50 part min/max is miniscule...remember that standard deviations form different sources (measurement error and true part variation) do not combine linearly (you cannot perform straight forward addition on them).

I appreciate the above advice.

One question: My gauge is calibrated externally and upon calibration, a certificate is issued. If the certificate of calibration states that the gauge uncertainty is (say 'x'), how do I incorporate the gauge uncertainty into my 50 sample measurement data and also how does this have an effect to the GRR study?