Type 1 Gage R&R or something else?

[Steve Clark]

We have purchased a robotic gauging machine to test insertion and retention of cylindrical parts. There is no operator input with the exception of pushing the start button.
The robot picks up a cylinder that has tabs on one end
It then forces cylinder over a known gage pin and measures force required, it also at second station does the same but at this station it measures force needed to remove from gage pin. My question is how best to run a MSA. I do not have variation from different operators. I can not really run a type one gage R&R because every time the cylinder is measure and forced on a gage pin it affects the strength of the material, introduces wear and in turn the results.

 

[Ninja]

Before the GRR....
How are you reconciling the fact that your manufacturing In-Process testing changes the product?
Assuming GRR is awesome and done...how do you know the properties of what you are shipping, since they are no longer what you tested?

 

[Steve Clark]

Before the GRR....
How are you reconciling the fact that your manufacturing In-Process testing changes the product?
Assuming GRR is awesome and done...how do you know the properties of what you are shipping, since they are no longer what you tested?

We understand there is a change in the product after testing one time that is very minimal, and can be quantified by retesting to see exactly what the change is. My concern with a MSA is that running the same part 50 times will be a much greater change between run 1 and 50 than say between run 1 and 2

 

[Bev D]

you don't need a type 1 assessment.
you also don't need to test multiple operators.
my recommendation is to test 30 parts twice and plot it on a Youden. a Youden plot is a square 'scatter diagram' with a 45 degree 1:1 line. if there were no measurement error the two repeat measuremetns would fall on the 1:1 line. measurement error is the scatter perpendicular to the 1:1 line.
You can find the information on how to do it in the resources section. I published 2 items that will help: MSA tools will help you graph the output and MSA (not your father's gauge R&R) will describe how and why to do the analysis.
If there is slight damage that changes the part it will show up as bias to the 1:1 line. you can post your results here again for more interpretation help...

 

[Ninja]

Thanks for your answer.

My two cents...
1. Have you read @Miner 's blog on MSA? He covers MSA on destructive testing pretty well (read "awesomely").
2. You are concerned...is that concern valid by data?
Consider doing a 1x2x10 GRR (one operator, two samples, 10x each) and see if your part changes only once and plateaus or continues to change...

Added in edit: ...or what Bev said...

 

[John Predmore]

Another approach to assess gage repeatability in a process operation where one or both parts irreversibly change is to take 2 readings of every single insertion force event. The way to do that is Newton's third law, for every action there is an equal and opposite reaction. Put a force sensor on the insertion tool and a second force sensor on the fixture which holds the part. This 2-sensor method was championed by Dorian Shainin, and the paired sensor data points can be plotted on a Youden plot which @Bev D described. To reduce the risk that parallel measurement bias on both sensors in your experiment prevent you from seeing the bias, employ different technology sensors if possible - a piezoelectric load cell and a spring loaded LVDT, for example - or one sensor in compression and the other in tension.

 

[Welshwizard]

Hi Steve,

I would take at least 10 parts or more and measure each part 3 times.

Plot the data on an X Bar R Chart

Any signals on the Average Chart show that the measurement system can detect product variation

Any signals on the Range chart indicate problems with the measurement process.

Please post the data on here if possible for characterisation please see the link for more detail The Short EMP Study

Hope this helps you.