Can anyone help me to determine the best way to calculate capability for a dimension that has a one-sided tolerance. I am currently calculating using 6 sigma. Would it not be better to calculate it using 3 sigma in order to better analize the data that falls in a very tight window? See attache example.
Calculating Capability on a One-Sided Tolerance Dimension
- Thread starter bwilliams
- Start date
I typically use the 6 sigma approach you identify. It may be specified as on sided, but it is not a unilateral tolerance, such as roundness. The process should be able to be centered within the tolerance zone. If you loaded your target in, you would get Cpm, but I would not personally buy into that, unless the process had a compelling reason to do so.
As usual, it would be nice to know what the process is for a more accurate opinion.
As usual, it would be nice to know what the process is for a more accurate opinion.
it would also be helpful to all if you search this forum first as we have addressed this question many times. Then you can ask more spcific questions about your situation...in the meantime Bob is correct there is nothing special about a one-sided tolerance unless of course it's naturally unilateral or physically bounded such that it is impossible to get values beyond some limit. In this case special care must be taken as usually, getting closer to the boundary is getting better capability.
Last edited:
Unilateral = one-sided.
D
Darius
One-Sided Tolerance? the worksheet show a 2-sided Tolerance...
Also something strange; is the presence of 2 populations, you should check for the cause in order to reduce the variation.
Also something strange; is the presence of 2 populations, you should check for the cause in order to reduce the variation.
I am suspicious that it is measurement (not gage) error. That is why it would be good to know the dimension and process. It could be a sampling error if the data was not actually taken in the order it was plotted.
Last edited:
You cannot just center a unilateral specification to make it bi-lateral. For example, a specification of a maximum of 5% is not the same as a nominal of 2.5% +/- 2.5%. If you have a unilateral or one-sided, there is no Cp and only a Cpk calculated. Not sure why we need to talk in 6S or 3S, use the capability indices as defined.
Because there are so many people/Customers out there blindly requiring "statistical" analyses such as Cp/Cpk, Gage R&R, etc. that too many individuals in overwhelmed companies are trying to 'check all of the boxes' without benefit of real education and understanding of the methods. It's not their fault and to their credit many are coming here to at least get some level of understanding.
The six sigma approach is the correct approach, with LSL 65.0 mm and USL 65.5. Looks like it could use some centering, and more data would be handy, too. Is this just one mold cavity, one lot, consecutive parts?
Last edited:
Similar threads
