Alternatives to GR&R (Gage R&R) - Only 4 sample parts and one operator

[Miner]

Re: Alternatives to GR&R (Gage R&R)

Actually there are ways to apply MSA methodology to both destruct testing and small sample sizes that can provide information that is usefull - especially given the situtation that the OP describes: early in the design process.

FYI, the OP said this was non-destructive testing. I agree that there is a lot that can be learned early in the design process.

To the OP: the coefficient of variation can be a useful alternative to the standard R&R approach. However, you do NOT have sufficient sample size to use it at this time. There is also a very special experimental approach to using the CV.

Can you elaborate on this approach in a separate thread? I am not familiar with it, but I do not want to derail this thread.

 

[Bev D]

Re: Alternatives to GR&R (Gage R&R)

FYI, the OP said this was non-destructive testing. I agree that there is a lot that can be learned early in the design process.

OOPS mis-read that...since that's the case teh OP should simply perform a standard R&R test. In this case he might beenfit from many multiple reads of each piece...say 10-20 or so. this will provide the best estimate of repeatability given the small sample size

Can you elaborate on this approach in a separate thread? I am not familiar with it, but I do not want to derail this thread.

Yes I will it's a common situation in biological assays.

 

[Richard Pike]

In my experience we sometimes make things more complicated than necessary. What is the measurement error you can live with? (if its large then do you really need to quantify it) Is the current method used the "only" method available to you? (can comparative tests be conducted) Is the characteristic generically similar to other characteristics. (is history available). In my view these questions must be asked and answered before we get into the theory and semantics of deciding which (if any) statistical or graphical technique would best address your concern. Hope this helps.

 

[Kalpol92]

Firstly thanks for the reponses you have certainly got me thinking, Miner i will check out your blog.

I thought this might raise some good discussion.

Bev D,
The type of testing i would be carrying out are things like height, length, surface roughness etc on replacement hips like your gran and grandad has.

This is not for any one individual test but i am preparing a procedure on Test method validation, simple for production but R&D a little more difficult. I think in many cases brahmaiah's idea of experience along with some sort of risk assessment would be the only way forward.

 

[Bev D]

Bev D,
The type of testing i would be carrying out are things like height, length, surface roughness etc on replacement hips like your gran and grandad has.

This is not for any one individual test but i am preparing a procedure on Test method validation, simple for production but R&D a little more difficult. I think in many cases brahmaiah's idea of experience along with some sort of risk assessment would be the only way forward.

I guess I am confused then. These are fairly simple test methods and even with small sample sizes a standard repeatability test - using 5-15 repeats - is not only simple and easy it will provide you directly with the best information you might need on the suitability of the method in the early phases.

I truly don't understand why people prefer the more complicated approaches when a simple approach will do. Remember that in R&D the verification and early validation (Design optimization) testing don't have to be the full blown approach advocated by the automotive industry. Form follows function.

A risk assessment is all well and good but too often in design I've seen such seemingly simply methods not work well at all in practice because the system - not just the gage - is not properly assessed and the reality of manufacturing is not comprehended in the assessment. I too am in the medical device industry and my practical experience says do the repeat testing. Risk assessment without testing is theoretical and theory never matches practice. My company could have saved themselves a lot of heartache and delayed launches and regulatory scrutiny if someone had just done a simple test in the design phase before it was too late. But, hey that's just my experience...

 

[Richard Pike]

Unless I am misunderstanding the situation there is no problem. For the characteristics you mention it is not necessary to conduct the tests on the same product. i.e for surface finish the error related one one product would be the same on a similar (generic) product. Even if the test required a time span of several weeks (in order to collect sufficient data) you can still use standard MSA techniques. i.e the small sample (4) is not applicable. Yes i understand stability may enter into the equation, but in my metrology experiance this is not going to be significant as the useage of the equipment is minimal. hope that this helps rather than confuses !

 

[Miner]

you can still use standard MSA techniques. i.e the small sample (4) is not applicable.

The OP stated that he only had one operator available. This makes the standard R&R approach problematic.

Yes i understand stability may enter into the equation, but in my metrology experiance this is not going to be significant as the useage of the equipment is minimal. hope that this helps rather than confuses !

The point of the stability study in this case is not to determine the stability. It is to obtain an estimate of Repeatability from the Rbar of the stability test. Remember that the within-subgroup variation of a stability subgroup is entirely measurement variation since you are measuring and remeasuring the same part.

 

[Bev D]

WARNING: RANT AHEAD

I think the major confusion here is the parochial view that the method as described in the AIAG manual is the "standard method". Measurement Systems Analysis is an entire body of knowledge itself and it goes beyond the AIAG manual. It has many many valid applications and uses in all kinds of industry. The automotive industry isn't the only one in the universe - thank goodness given the state of the the american automotive industry!

The OP is in the medical device industry and V&V is entirely different there.
Understanding the measurement system in actual use is a fundamental element. Some rather mundane gages like those for height or surface finish may seem trivial but if they are measuring critical features of a medical device it is incumbent on the organization to understand and verify and validate these gages as a total system. The earlier this is done in the design process the better - especially if these gages will be used for optimizing and then validating the design itself.

we must also remember that design engineers tend to not acknowledge the existance of variation and so they might not even think about the fact that within piece variation can effect the measurement result looking like measurement error (when I was in aerospace this very thing happened and the egnineer compensated for the alleged measurement error by averaging the readings across the piece, wrote it into the QC specification and promptly got inflight engine shutdowns...and it was a 'simple' wall thickness measurement....); how you hold the gage can have an effect on length or thickness...


and the OP said: "However when we complete product design validation/verification any studies/investigation reports that the results are used in we have to have evidence that the level of precision we have within the method is acceptable." And this is true for medical devices. How will you assure that the gages you are using have the correct precision for DESIGN V&V which comes before mass production? you will naturally have small sample sizes. And objective evidence means doing the test not doing a risk assessment. Depending on the nature of the device, the FDA might accept studies performed on similar parts with the same gage types but since the test is relatively easy to perform and the consequences of a poor sustem detected after launch is so severe, why wouldn't the OP just measure each of 4 devices several times to get a first pass understanding of the actual measurement variation?

 

[Richard Pike]

The OP stated that he only had one operator available. This makes the standard R&R approach problematic.
The point of the stability study in this case is not to determine the stability. It is to obtain an estimate of Repeatability from the Rbar of the stability test. Remember that the within-subgroup variation of a stability subgroup is entirely measurement variation since you are measuring and remeasuring the same part.

Yep R&R is of course not possible with one appraiser. That,s why I said standard MSA Techniques. I referred to stability only to clarify that because the test samples and data could be collected over a longer period the effects of stability might have an outside chance of affecting the results.

I am going to back out of this discussion because working with only half the information of the process is dangerous. . I have written several Training Courses on MSA, have lectured on the subject at various levels for almost 20 years and have assisted in the implementation of QM Systems for many medical device Organisations. If the originator of this thread would care to contact me thro e-mail i,d be pleased to attempt to offer more detailed opinions or advice.

 

[Bev D]

Yep R&R is of course not possible with one appraiser. That,s why I said standard MSA Techniques.

yeah. Too often people say (or in this case type) gage R&R when what they really mean is MSA. Some have come to equate "Gage R&R" as a generic reference to MSA, like Kleenex = tissue paper meant for wiping one's runny nose (not to be confused with the 'tissue paper' one uses to wrap glass or clothing and the like with). I've been guilty of the same lazy reference. Others of course have become stuck on AIAG Gage R&R as the only MSA one can or should do...hence a reference to MSA can be interpreted by these individuals as Gage R&R. NOT to accuse Miner of this - he is one of our experts in Gage R&R and MSA here. We all mis read posts from time to time. AND I don't mean that statement pejoratively at all - seriously - I meant it only as my observation of behaviors.