I have ID dimension that for some reason when I measure with caliper, I cant get a good R&R, (its a plastic part, maybe in the process the neck get deformed) they want us a use a method with less "noise" although we are in spec, we have (lots of variation according to the costumer), i will check with a CMM ( dimension are very similar hard to get differences between parts, can i suggest in the end a go not go, so we can finish this discussion?
Gage R&R and right way to measure
- Thread starter mcpe72
- Start date
Its a shame that it is an ID. Mitutoyo Low force caliper works well on outside dimensions. I assume that the CMM is not an option for ongoing process control? Another option of last resort is to use the average of multiple measurements as the measurement. This reduces the variation by the square root of the sample size. For example, the average of 4 measurements will have half the variation of individual measurements.
Column 7 is the standard deviation of column 5 which is the mean of the 4 measurements in columns 1-4.
Column 8 is the mean of column 6, which is the standard deviations of columns 1-4.
Note that column 7 is roughly half of column 8.
Column 7 is the standard deviation of column 5 which is the mean of the 4 measurements in columns 1-4.
Column 8 is the mean of column 6, which is the standard deviations of columns 1-4.
Note that column 7 is roughly half of column 8.
Last edited:
thank you, I'm using a bore gage too, I'm going to do 3 R&R, one with a caliper one with the CMM, to compare, but the problem is that they want a dimension in the parting line and one to 90 degrees so we can get the ovality too.
in production would take way too long to do all that wit the CMM
in production would take way too long to do all that wit the CMM
I've always found CMM to be faster than caliper or micrometers, personally.
What CMM do you have? How many samples are you testing?
Using a MicroVu Vertex 120 (old one from ~2005), we measured 5 samples of each lot for length and width in 3 places (20 data points total) in about a minute.
For circles...we were measuring center to center distance...that went much faster.
For circularity of a single circle, including diameter/radius...that would be about a 2 seconds (plus part mounting time) per part including data report...call it 10sec total.
What CMM do you have? How many samples are you testing?
Using a MicroVu Vertex 120 (old one from ~2005), we measured 5 samples of each lot for length and width in 3 places (20 data points total) in about a minute.
For circles...we were measuring center to center distance...that went much faster.
For circularity of a single circle, including diameter/radius...that would be about a 2 seconds (plus part mounting time) per part including data report...call it 10sec total.
Yes, I also found the OGP hard to use...thus I purchased the MicroVu.
Even with the OGP...if you take your time in proper fixturing (easy to mount the part in the right place)...and take your time in effectively arranging the measurement program, the per piece turnaround time should be measured in seconds even when including part load/unload.
(I am not trying to derail from your original question...but it seems that Miner has already answered that fully).
With a part loading/positioning/holding jig, and an automated CMM program...I would expect your GRR results to be on the order of 1.0-3.0 for most tolerances. (unless your tolerance is in microns)
Even with the OGP...if you take your time in proper fixturing (easy to mount the part in the right place)...and take your time in effectively arranging the measurement program, the per piece turnaround time should be measured in seconds even when including part load/unload.
(I am not trying to derail from your original question...but it seems that Miner has already answered that fully).
With a part loading/positioning/holding jig, and an automated CMM program...I would expect your GRR results to be on the order of 1.0-3.0 for most tolerances. (unless your tolerance is in microns)
Ummm.....Looking just at part #10, you have > .05mm variation on an ID only 26mm? On a CMM?
I would look at (in order):
- Is your program taking multiple readings as Miner describes, or just one ID? It's an automated CMM...you should be measuring literally hundreds of datum per second. Check your program. If you measure in different places each time, and your part varies that much (circularity), no wonder you're failing GRR.
- Is the part stable, or is it jiggling during measurement?
The demo I would run to show customers was to take off my wedding ring, put it on the stage, and hit "go".
In a couple of seconds, I got hundreds of ID measurements, hundreds of OD measurements, best fit circles for ID and OD and circularity for both.
Are you just using measurement #1 for your GRR?
Then I would put a penny on the stage, and we would measure the width of Abe Lincoln's knee (did you know he's in there?)...that's how a CMM should be giving you the data...not 0.05mm error.
(Note...his knee is less than 0.05mm wide...)
I would look at (in order):
- Is your program taking multiple readings as Miner describes, or just one ID? It's an automated CMM...you should be measuring literally hundreds of datum per second. Check your program. If you measure in different places each time, and your part varies that much (circularity), no wonder you're failing GRR.
- Is the part stable, or is it jiggling during measurement?
The demo I would run to show customers was to take off my wedding ring, put it on the stage, and hit "go".
In a couple of seconds, I got hundreds of ID measurements, hundreds of OD measurements, best fit circles for ID and OD and circularity for both.
Are you just using measurement #1 for your GRR?
Then I would put a penny on the stage, and we would measure the width of Abe Lincoln's knee (did you know he's in there?)...that's how a CMM should be giving you the data...not 0.05mm error.
(Note...his knee is less than 0.05mm wide...)
Last edited: