Destructive Testing GR&R (Gage R&R) - Track tester - JMP Program

[Bookworm]

First and foremost, please bear with me since I am new to Gage R&R testing

I'm trying to run a GR&R test for a track tester. Basically the machine is programmed to push a device in a pre-fabricated track and record its force measurements over a programmed distance. Once a test is run once, the testing material will get deformed, so running the same material through the tester again would give different force measurements.
Since this is a destructive test, do I calculate values like the % Gage R&R in the same manor as I would a non-destructive test? Do I have to actually find a way to change my testing procedure to make it non-destructive? Any ideas?
Also, I'm using the JMP program which asks me to enter a tolerance value (USL-LSL), and I'm not really sure what value I should enter into here, but I know what I enter here has a dramatic effect on my % Gage R&R value.We have no set tolerances or standard devices to use as reference since this is a first round series of tests. Is it then up to the engineers best educated guess on this tolerance value?

[Miner]

The MSA manual 3rd edition has a section (Chapter IV - Section A & C) that deals with destructive testing.

You can also perform a nested gage R&R.

There are two basic approaches to obtaining the necessary repeat measurements. One is to collect the test specimens, and sub-divide them into smaller samples if possible. This would work for chemical testing and similar needs. The second approach is to use consecutive homogeneous parts from a single lot for the repeats, and parts from different lots for the "parts" themselves.

I am not familiar with JMP, but MSA software is normally looking for the tolerance range (USL minus LSL). Do you have a unilateral tolerance (i.e., a max or min spec)?

Unless your customer requires the P/T Ratio (uses tolerance), you may be better off using the %GRR, which is GRR as a percent of the process variation. This does not use the tolerance, and is better suited to evaluate the gage's suitability for process control.

Also, search this forum. There are a lot of threads dealing with GRR and destructive testing.

[Bookworm]

[The second approach is to use consecutive homogeneous parts from a single lot for the repeats}

So then basically I could measure one design of a device from one lot and then measure the same design of a device from the same lot and compare those results to get my gauge value? And this process is called Nested gage R&R?

I mentioned that the device will permanantely deform after one run. Will I be able to see if there is uniformity in force measurements after deformation between devices if I ran each device two times and compared all force values from the first and second run to get a %GRR? Or should I compare them seperately? ie: find a %GRR for Run 1 of two devices from a single lot, and another seperate %GRR for Run 2 of those same devices


{and parts from different lots for the "parts" themselves}

Finally I wasnt sure what was meant by this statement. Would you mind clarifying? Thanks again!

[Miner]

In Reply to Parent Post by Bookworm

[The second approach is to use consecutive homogeneous parts from a single lot for the repeats}

So then basically I could measure one design of a device from one lot and then measure the same design of a device from the same lot and compare those results to get my gauge value? And this process is called Nested gage R&R?

For the sake of discussion, lets use the "standard" 10 parts x 3 repeats x 3 operators MSA layout. In your situation, the 10 parts would be 10 different lots of the product. These should be the same design, but different lots. Different designs will inflate your process variation (PV) and give you erroneous results. The three repeats and three operators would be 9 consecutive parts. Operator 1 would test parts 1, 4 & 7. Operator 2 would test parts 2, 5 & 8, and Operator 3 the remainder. Analyze the results using the ANOVA method for a nested study. A crossed study is the normal method used when parts can be measured repeatedly. A nested study is used when the can only be measured once.

In Reply to Parent Post by Bookworm

I mentioned that the device will permanantely deform after one run. Will I be able to see if there is uniformity in force measurements after deformation between devices if I ran each device two times and compared all force values from the first and second run to get a %GRR? Or should I compare them seperately? ie: find a %GRR for Run 1 of two devices from a single lot, and another seperate %GRR for Run 2 of those same devices

An MSA is not really the correct method to use for this. You could perform an MSA on the 1st time results, and a second MSA on the 2nd time results to see whether a pretest reduces test variation (your second option). To quantify the change between tests, you have a number of options, including but not limited to regression analysis. Your first option would inflate the process variation and give erroneous results.

In Reply to Parent Post by Bookworm

{and parts from different lots for the "parts" themselves}

Finally I wasnt sure what was meant by this statement. Would you mind clarifying? Thanks again!

See the first section response.