Just like the title says. How do you handle SPC of unilateral tolerances. Is there a way to apply a weibull distribution to data analysis
within DataPage?
You can have it calculate a CORRECT Pp/Ppk Cp/Cpk with a unilateral tolerance WERE it is possible for the readings to go 'beyond' zero, ie Surface with a tolerance of +1/-0, you CAN get negative values. It will NOT work for TP (which IS a unilater tolerance) because you can never get a negative reading.
OPTIONS
SPC CALCULATIONS
CAPABILITY THRESHOLD
There is a box to mark for Cp/Pp for unilateral tolerance, mark it.
Yeah I have that option checked. I'm measuring Maximum Pitch Error on a sprocket and when I chart it the 4sigma limit just happens to be far from zero but quite capable. I just wasn't sure how it might go if by the end of the cap study the mean skewed towards zero.
Yeah I have that option checked. I'm measuring Maximum Pitch Error on a sprocket and when I chart it the 4sigma limit just happens to be far from zero but quite capable. I just wasn't sure how it might go if by the end of the cap study the mean skewed towards zero.
As long as it is possible for a reading to go beyond EITHER limit, it SHOULD calculate it correctly. It is only in the case of an impossiblity of a 'negative' value that the numbers become screwed. Flatness, straightness, circularity, TP, those things where a negative reading can never happen.
HOWEVER, you CAN 'cheat' it to give you a correct Cpk/Ppk value in those cases by editing the tolerances for that feature (AFTER YOU HAVE RUN THEM ALL!!!!!) and giving it a negative tolerance equal to the MEAN (at minimum) or to the high (positive) tolerance. For example, if you are doing flatness of a plane to 0.5mm, if you give it a -0.5 tolerance instead of the zero it will have, then the calculations will be correct, at least as far as Cpk/Ppk will be. The Cp/Pp will not be correct in this case. Kind of a catch-22 thing going on there. The reason you have to wait to change the tolerance is because as soon as you send another sample to the database, it will update to the tolerance sent.
So, leave the tolerances alone for correct Pp/Cp values, give it a negative tolerance equal to the positive to get correct Cpk/Ppk if it is NOT possible for a negative value to appear, if you CAN get negative values, leave it all alone.
