Gage R&R Across Multiple Sites in a Company

G

GM_Quality

Good afternoon all,

I'm working on a Gage R&R for a set of air gaging that my company and 2 of our suppliers recently purchased. Each company bought the exact same air gaging units and heads.

Each site then conducted their own Gage R&R study. The results were pretty fantastic, since the precision and accuracy of air gaging far exceeds our tolerance bands for the 2 features of interest. Each site passed their own "long" gage R&R (10 pieces measure three times by three operators) well under the required 30%.

Now I want to compare all 3 Gage R&R's together. My first thought was to take each operator's three measurements and average them together (starting with 30 measurements for each operator, this would yield 10 measurements). Repeating across each site would turn 90 data points into 30 averages. I could then run a new study with the 30 average for each of the three sites representing one "operator's" measurements, effectively yielding a new Gage R&R study.

The results weren't nearly as fantastic. One feature passed at 19%, but the other feature (same tolerance, but different gaging head and size) came in around 44%. Looking at the Xbar and R charts, they have a much wider spread and overall measured parts several .0001" smaller than my company and the other supplier.

Is there another way to evaluate this multi-site data? Or is this approach my best bet?

Thank you for reading - I look forward to hearing everyone's feedback!
 

Miner

Forum Moderator
Leader
Admin
Do you have access to Minitab? If so, there is an option to do an Expanded Gage Study. You can take your existing data (provided the same 10 parts were used in all 3 studies) and add Location as an additional factor. This will add Location to Location variation as an additional element within the R&R results.

If you do not have Minitab, this analysis is still relatively easy to perform using ANOVA. You will just have to calculate the percentages manually as well as recreating the graphs manually.

Do not average any of the results. This results in a loss of information and will artificially reduce the variation by the square root of 3. If you would attach your data, I would be happy to analyze it for you if you do not have Minitab.
 

howste

Thaumaturge
Trusted Information Resource
...provided the same 10 parts were used in all 3 studies...
I think this is the key. If each site used different parts, then it will need to be redone with the same parts at each site to get what you want.
 
G

GM_Quality

Miner, thank you for this reply, it was very helpful. I do have access to Minitab 16, and I was able to run an Expanded Study with Location as an additional variable. When analyzing the results, I realized the study assumed that my 3 inspectors were constant across all 3 locations. Unfortunately, this was not the case in our study - 9 different inspectors were used (3 at each location).

Is there a way to account for this in the study?

Thank you again :thanx:
 
G

GM_Quality

I think this is the key. If each site used different parts, then it will need to be redone with the same parts at each site to get what you want.

Howste, good point. The same 3 parts were used at all 3 locations.
 

Miner

Forum Moderator
Leader
Admin
That would be ideal. If they are not, it could be analyzed as a nested study, but that is less than optimal.
 

Miner

Forum Moderator
Leader
Admin
Miner, thank you for this reply, it was very helpful. I do have access to Minitab 16, and I was able to run an Expanded Study with Location as an additional variable. When analyzing the results, I realized the study assumed that my 3 inspectors were constant across all 3 locations. Unfortunately, this was not the case in our study - 9 different inspectors were used (3 at each location).

Is there a way to account for this in the study?

Thank you again :thanx:

This is a more complicated scenario. It means that the inspectors are nested within location. It can still be analyzed, but must be analyzed as an ANOVA or GLM. The canned R&R analyses cannot do this.
 
G

GM_Quality

Thank you both for the feedback. Do I need special software to include nested inspectors, or is this something that can be done within Minitab?
 

Miner

Forum Moderator
Leader
Admin
You can do it within Minitab using the General Linear Model (GLM), but you have to know what you are doing, and many of the metrics and graphs must be reconstructed the hard way.
 
Top Bottom