HOW DO YOU PERFORM A COMPLETE MSA WHEN YOUR TEST METHOD IS A DESTRUCTIVE TEST??

Well, for complete MSA info we would need more data.

There's 1 'general' method akin to an R&R in intention (consideration of repeatability and reproducability) described near the end of the MSA presentation (if I remember correctly) in the pdf_files directory here.

Anyone want to go into details?

The test I am referring to is bond pull, both destructive and non destructive. We use ultrasonic bonders with aluminum wire and periodically we do a non destructive pull test and then on separate units, a destructive bond pull. I do not know how to measure repeatability or linearity in these two cases.

Use a test coupon, its the only way you can do it as a continuous process, short of dedicated bonders for product. Have been doing this for years for mil standard defense/space programs, as well as AS/QS and has been acceptable. I hear the frustration..been there, done that. that process does NOT fit the mold for SPC stuff at all.

How can I do in case I cannot find/use dummy coupon? Is it possible to use within Lot material to explain Repeatability and between Lot to explain Reproducibility?
What about your idea? Thanks for your idea.

Originally posted by barb butrym:
Use a test coupon, its the only way you can do it as a continuous process, short of dedicated bonders for product. Have been doing this for years for mil standard defense/space programs, as well as AS/QS and has been acceptable. I hear the frustration..been there, done that. that process does NOT fit the mold for SPC stuff at all.

Destructive measurement system analysis can be accomplished via SPC. You should referene "Evaluating the Measurement Process When Testing is Destructive" by Donald J. Wheeler and "EMP-Evaluating the Measurement Process", by Donald J. Wheeler and Richard W. Lyday. Both publications can be obtained from SPC Press, Inc. 615-584-5005

The best method I found for MSA on destructive testing is through using the ISOPlot method developed by Dr. Dorian Shainin (see www.shainin.com (http://www.shainin.com) for information).

Basically, I would take 2 consecutive samples as a group and collect a total of 30 groups. Next I would use sample 1 of each group as Trial 1 and sample 2 of each group as Trial 2. Then I would plot Trial 1 versus Trial 2 and determine the discremination ratio. It is very similar to correlation, but the method of calculating discrimination ratio is the key to qualify the measurement system. There is a software to do this from Shainin (www.shainin.com/RedXTracker/).

Check it out!

MSA reference manual give some examples of approaches to MSA including single specimen. Let's see in Appendix D pp.107 and let's comment on that.

This may be of use. File attached on non replicable GRR Case Study.

The key is to use Nested ANOVA and have parts that you can at least assume (Pretend) to be homogenous and treat such as you are testing the same product.