GR&R for Hardness Tester and Arc Spark Spectrometer
- Thread starter sinned
- Start date
If you have set up the study correctly and your results are poor that can be because your measurement system is actually not very repeatable given the tolerances you are dealing with.
Can you give details regarding the study design and the results?
Can you give details regarding the study design and the results?
We tried measure Pb, Ag, and Cu % content with the arc spark spectrometer using the same set of sample ingot from our process. The ANOVA method for destructive GR&R were used with 10 samples and 3 operators. (I doubt if this kind of measurement has the component of operator error indeed, because there is no fixture required in the measurement.) Nevertheless, the outcome %P/T is around 26 with spec. range 0.2%.
I think it could still be considered acceptable and my colleague is sending the results to customer for comment. I believe the uniformity of the sample content is important because the same spot can not be measurement repeatedly.
At the same time, Ppk has been done with result >1.66. I think it can also be supporting to the less-than-best GR&R results.
Btw, if the spec. range for Pb, Ag, and Cu are different, should I just do GR&R on the one with the smallest range ?
I think it could still be considered acceptable and my colleague is sending the results to customer for comment. I believe the uniformity of the sample content is important because the same spot can not be measurement repeatedly.
At the same time, Ppk has been done with result >1.66. I think it can also be supporting to the less-than-best GR&R results.
Btw, if the spec. range for Pb, Ag, and Cu are different, should I just do GR&R on the one with the smallest range ?
Arc Spark Spectometry has historically been considered a qualitative test. While newer equipment may be touted as more quantitative, it is still a pretty rough test when faced with a spec range of 0.2%.
It may simply be that the equipment is not capable of passing GRR at that level of need.
It may simply be that the equipment is not capable of passing GRR at that level of need.
What type of hardness tester (e.g., Rockwell, Brinell, Knoop, etc.)?
These will definitely be non-replicable studies. In most cases, there is too much within part variation in hardness, especially with the micro-hardness tests. I recommend using hardness test blocks as parts. Randomly assign the operators and repeat measurements to a grid location on each block and run the study.
% Tolerance is straight forward, but if you need % Study variation, I recommend having one person run a capability study of hardness on actual parts, then using this variation to calculate % Process Variation, which is a better metric than % Study Variation.
JeantheBigone
Quite Involved in Discussions
I haven't done Gage R&R for Arc Spark spectroscopy, but I've measured hundreds of compositions using one. This is an extremely sensitive measurement that checks a tiny portion of material.
If you're using castings and are sparking the surface, that would be my starting point. Castings have huge elemental segregation, and successive measurements on the same sample will reflect this. Heck, I've used that fact on cast stainless steels to show the elevated C and Cr content on the surface.
If that is the case, then I'd recommend using a wrought material for your gage R&R sample.
If you're using castings and are sparking the surface, that would be my starting point. Castings have huge elemental segregation, and successive measurements on the same sample will reflect this. Heck, I've used that fact on cast stainless steels to show the elevated C and Cr content on the surface.
If that is the case, then I'd recommend using a wrought material for your gage R&R sample.
I know very little about arc spark testing, so I don't know if the following is feasible.
Can you sub-divide the ingot into 10 pieces and perform the test? It is possible that there is significant within ingot variation in the metal content, and portions of the ingot in close proximity would presumably have less variation than portions further apart. Also, does the test itself change the product in any way that would prohibit tests in near proximity? If so, how much distance is required between test areas?
Also, as with the hardness test response previously given, are there standard test blocks that could be used instead of ingots?
JeantheBigone
Quite Involved in Discussions
If it is a casting, then portions of the ingot taken in close proximity may or may not have smaller variation than samples taken further apart. I would avoid using a casting as a sample for Gage R&R because of the inherent variability.
Brammer offers setup samples that are standardized and guaranteed to have a uniform and specified composition:
Brammer Standard Home Page
as does NIST (type in aluminum)
Standard Reference Materials | NIST
And yes, standard hardness blocks are available. Whoever does the calibration of the hardness tester should be able to offer them.
Last edited:
IF the capability study was performed correctly (using samples that span the full range variability of the process; different raw material lots, different processing events, etc.) then you R&R is certainly sufficient for measuring conformance to specification. Remind your customer that the observed variation in a capability study INCLUDES both part variation AND measurement variation. Measurement variance adds and subtracts from the true value; there is no true value out beyond the observed variation that was 'hidden' by measurement error...if there is undetected true results it will be because of an improper study design, not measurement error.
if you have a Ppk > 1.66 and you are reporting a '26%' measurement error, I'd say that most of the variation is measurement error, which is OK (and proves the weakness of the traditional R&R method...)
JeantheBigone
Quite Involved in Discussions
Again, I've never done Gage R&R for an Arc Spark Spectrometer, but my Spectro was one of the key instruments in my lab. 26 % measurement error sounds huge.
Are the operators performing calibrations using standard references prior to each measurement, or when the instrument is turned on? Or is this done at specified intervals?
Is the instrument left on continuously?
How often is the soot removed from the chamber?
Are the first three "burns" discarded?
What is the sample surface preparation?
Are the operators performing calibrations using standard references prior to each measurement, or when the instrument is turned on? Or is this done at specified intervals?
Is the instrument left on continuously?
How often is the soot removed from the chamber?
Are the first three "burns" discarded?
What is the sample surface preparation?
Similar threads
