New requirement for MSA

Hey all. We just received a new requirement from one of our customers for MSA studies where it says we have to do the following:

“To help assess the gauge, the organization shall report the value of +/- 2 Total Gauge R&R
Standard Deviations to understand the 95% prediction interval (uncertainty) of any one
measurement. This value can be used in conjunction with engineering judgment to help assess the distance between the edge(s) of the process distribution and the specification limit(s). The organization shall report gauge R&R as both a percent of study variation and a percent of
tolerance.”

To start, we used to just run a 3x3x10 ANOVA study for this customer and try to land under 10%. Now the change is they added the “+/- 2 Total Gauge R&R Standard Deviations” line as the change and I am not 100% sure what that entails.

Also for things like runout with one sided tolerance, it says to calculate by doing 6 total R&R standard deviations divided by the USL - lower boundary of zero. Does anyone have any insight on this?

Bump

@Miner can answer this one.

You might find this article useful.

Jayfaas said:

“To help assess the gauge, the organization shall report the value of +/- 2 Total Gauge R&R
Standard Deviations to understand the 95% prediction interval (uncertainty) of any one
measurement. This value can be used in conjunction with engineering judgment to help assess the distance between the edge(s) of the process distribution and the specification limit(s). The organization shall report gauge R&R as both a percent of study variation and a percent of
tolerance.”

To start, we used to just run a 3x3x10 ANOVA study for this customer and try to land under 10%. Now the change is they added the “+/- 2 Total Gauge R&R Standard Deviations” line as the change and I am not 100% sure what that entails.

Also for things like runout with one sided tolerance, it says to calculate by doing 6 total R&R standard deviations divided by the USL - lower boundary of zero. Does anyone have any insight on this?

Click to expand...

I pasted a Gage Performance Curve below to help illustrate your customer's request.

• The green distribution is the parts themselves.
• The yellow distribution is the Total Gage R&R.
• The red distribution is the combination of the parts plus the gage variation.
• The blue line is the probability of accepting a product of a given dimension. Ideally this line should be 0% when outside the specification and 100% within the specification. However, the Total Gage R&R allows some out-of-spec parts to be accepted and some in-spec parts to be rejected. Using +/- 2s of the total gage R&R would roughly approximate the interval outside the specification where the probability of accepting an out-of-spec part is greater than zero.

If your process is centered and highly capable, this would never be an issue, but if your process is marginally capable or is not centered, this would be an additional safeguard. The concept is similar to guard band tolerances.

Regarding the last comment about lower boundaries, see my blog post on one-sided tolerances. This should give you some insight on why and an argument against if warranted.

Ford has made this their requirement.

Johnnymo62 said:

Ford has made this their requirement.

Click to expand...

Which? The OP's customer's requirement, or the gage performance curve?

The OP's 2nd sentence. It's mentioned in their CSRs at the IATF website.


They also have a Powerpoint that explains it a bit (25 pages). This is confidential and shouldn't be shared here. It's avaliable through Ford's Covisint site.

Have they established requirements for it beyond reporting it and saying it may be used with engineering judgment?

So is this something I have to change physically with the population of parts, or a setting I need to change when running the study? Also I did not mention a name of our customer to protect the customer. I am more interested in how to run the study to make sure we are doing this correctly.