OK i was asked to do a GR&R on some equipment we have. This equipment feeds vials onto a conveying system and uses a single camera to make a measurement of each vial. How the vial is presented to the camera is likely to be important in the measurement. To check if the measurements are repeatable and reproducable accross all vial holders on the conveying system I was asked to do a GR&R.

As there are 60 holders and getting the data from the equipment takes about 1 min for each sample. It was not reasonable to measure each vial on each holder repeated times. So I took 60 "identical" vials and labled them. I ran these vials through the equipment in order and noted which holder the first vial loaded onto. I then repeated 2 more times ensuring that the first vial always loaded onto the same holder. All the other vials followed in sequence.

I then repeated the test 3 more time ensuring that for this set of three tests the first vial loaded onto a different holder than was used for the first three runs.

I then pulled the 360 data points from the photos the camera took of the parts.

All this went into Minitab in hopes of runninga GR&R BUT it refuses to run saying the design is not balanced. Is this a requirement? If it is I can not run 60 vials on each holder. there are 60 holder that would be 3600 runs. Ineed this done by Monday. Is there another way to get the same indication of what the components of the measurement variabilty are?

Hrm I wish I could be of some use but I'm totally unprepared...

I therefore refer you to http://elsmar.com/Forums/showpost.php?p=98579 with an excel template for GRR that should be suitable to your variables measurement.

Someone else might assist you with your technique and issue of operators = 0.

Some questions... What is the resolution of the camera? Can it discriminate between 1cc or .01cc difference in fluid level, for example?

What is the acceptable tolerance on volume in the vials?

Is this your first GR&R analysis on the vial camera measurement system?

And curious myself, is this Banner Engineering equipment and software?

I guess the way I am looking at it the "operators" are the holders and I have 60 of them.

For each holder "operator" I have 6 data points, 3 one one vial and 3 on another. To me this is just a highly fractionated DOE. I can not get "Operator" part variability. But I should be able to obtain "operator", part and total variability correct? The total - "operator" - part should be the ("operator" by part and the repetability).

Does this make sense?

Can you post your data? The way you describe it, it should be possible to analyze the data set.

Excell file should be attached.

Part - 1-60 numbers parts
Chuck -1-60 number of holder on conveyer belt
edge_one - first response
edge_two - second response
fail - not relavent ot this question.

Can anyone helpme get some data from this?

When I plot EDGE_one by vial I get a great deal of scatter, by chuck I get scatter as well but many of the ranges from one chuck to another do not overlap. This makes me suspect that which chuck is used makes a difference in the result. ANOVA of the data supports this conclusion. BUT how do I determine what fraction of total variation in the data is due to chuck?

The reason Minitab choked on your data was because you have a nested design. The standard GRR format is a fully crossed design. You must use the nested study format within Mintab.

I have attached the results of this analysis below. I also included some graphical plots of the data. It appears that you had a major shift between your first response and your second response. This shows up in the Repeatability source of variation.

Miner,

Thanks for your help. Running the nested GR&R works. BUT..... I get differnt results than you show. Did you treat EDGE-one and EDGE-two as repeats of the same measuremnt? They are actually different measurements not repeats. If do a GR&R on EDGE_ONE I get:

%Contribution
Source VarComp (of VarComp)
Total Gage R&R 7.68795 99.75
Repeatability 0.36026 4.67
Reproducibility 7.32770 95.07
Part-To-Part 0.01937 0.25
Total Variation 7.70733 100.00

Which I am interpreting to mean that going from chuck to chuck give different results. Differences between operators is high.


When I do the same thing on EDGE_two I get:

%Contribution
Source VarComp (of VarComp)
Total Gage R&R 4.3291 18.71
Repeatability 4.1779 18.06
Reproducibility 0.1513 0.65
Part-To-Part 18.8076 81.29
Total Variation 23.1367 100.00

Which I am interpreting to mean that the part to part variation is high compared to the operator to operator variation.

Make sense?

Thanks again for all your help.

This is why as quality professionals we should be able to do this type of analysis by long hand if necessary. We are getting too dependent of software and are losing the ability to function without it.

Just my :2cents:

Miner,

Thanks for your help. Running the nested GR&R works. BUT..... I get differnt results than you show. Did you treat EDGE-one and EDGE-two as repeats of the same measuremnt? They are actually different measurements not repeats. If do a GR&R on EDGE_ONE I get:

%Contribution
Source VarComp (of VarComp)
Total Gage R&R 7.68795 99.75
Repeatability 0.36026 4.67
Reproducibility 7.32770 95.07
Part-To-Part 0.01937 0.25
Total Variation 7.70733 100.00

Which I am interpreting to mean that going from chuck to chuck give different results. Differences between operators is high.


When I do the same thing on EDGE_two I get:

%Contribution
Source VarComp (of VarComp)
Total Gage R&R 4.3291 18.71
Repeatability 4.1779 18.06
Reproducibility 0.1513 0.65
Part-To-Part 18.8076 81.29
Total Variation 23.1367 100.00

Which I am interpreting to mean that the part to part variation is high compared to the operator to operator variation.

Make sense?

Thanks again for all your help.

You are correct. I misunderstood your earlier post regarding edges 1&2 as first response and second response. That sounded like a repeat measurement.

I repeated the analysis and get the same results as your last post. Without a better understanding of your equipment, I cannot provide a definitive answer, but your interpretation seems appropriate.

When you plot edge 1 by chuck, you get one chuck with distinctly lower results than the rest plus a few chucks that are suspicious.

Regarding the edge 2 results, if this is a different measurement of the same vials, it appears to be less sensitive to the chuck variation and more sensitive to the part variation, which is good.

This is why as quality professionals we should be able to do this type of analysis by long hand if necessary. We are getting too dependent of software and are losing the ability to function without it.

Just my :2cents:
While I agree with the comment regarding the understanding of how an analysis works, and not blindly accepting what the software spits out, this was a different situation entirely.

Many posters do not provide all of the necessary information or word it in such a way that it can be misinterpreted. A responder has a choice between asking many questions before attempting to help, or providing the best help that they can with the information available. We do not always get it right.

Miner

Thanks for all the help. As you can tell i am very new at this. My company does not have a "quality professional." I am actually a chemical engineer by trainning. While I have had some statistics trainning I often know just enough to get myself into trouble. This site has been a valuable resource to me.

Regarding being able to do the calculations by hand. When faced with this issue I had a choice, spend several hours or days locating and reviewing the details on the calculation or look for some quick pointers on this site. Given the other 10,000 things I need to get doen in a day I took the easy way out.

Thanks again.

I understand your predicament completely, and you did the right thing. I did not originate the comment about hand calculations and do not mean any criticism about your questions. That is what this forum is for, to ask questions.

I also agree that based on the circumstances given you at the time, you did what needed to be done to get it done. Good job on that. No critiicism of you intended, I am sorry if it came across that way.:o

Thanks to you both for your help.