125 PC Capability Study Sample.

We have a customer that now requires a 125 pc Capability Study. They used to just require a 30 piece capability study but have now increased it to 125 pieces. Does anyone have any examples that they would like to share with me? Thanks

 

Hi Bev, we currently use an excel spread sheet that we can enter the data into that calculates the Cpk. However, this sheet is only set up to collect 30 pieces of information. I am not sure if I change the format to the 125 pieces if it will calculate correctly. That is why I asked if someone had one so that I could compare the 2 forms to see if what I did actually worked.

Atul, I will check the AIAG SPC manual for that and see what I can come up with.

Thank you both for the help.

 

You should be able to modify your spreadsheet fairly easily (or input the data into another tab and only pull in the averages) to reflect the subgroups. You should group these into sub-groups of 5 readings only, and that is what you will do the capability on. There is no real difference between doing the sub-group averages and your normal 30 piece. The benefit of the sub-grouping is that it tends to blend your data, so hopefully it will improve your capability. You will need to confirm that this is what your customer wants though.

 

Could you share the excel sheet you are currently using? This will help fellow members help you.

 

Ahhh. The growing n problem. Here, then, is my question. If I am going from n = 30 to n = 125 then along with that, my confidence interval on my mean and standard deviation gets much smaller. Since both of these are inputs to Cpk, should I not allow my Cpk hurdle to be lower? I know much better where I am at ....

In other words, on n=30 with Cpk = 1.29 I would call that a fail. But if n=125 .... wouldn't that be OK given my "more certain" knowledge of where my mean was?

 

(The converse argument then being .... if I do NOT allow for a shorter hurdle, what is my motivation for checking more parts? That costs money. If I don't gain anything by n=125 over n=30, then why am I measuring the 95 more samples? Someone please explain ....)

 

Yes. That's what I mean. So if the hurdle I want to guarantee is 1 (where a +/- 3 sigma process would fit within specification if it were EXACTLY centered and we knew every measurement) then looking at your Monte Carlo of n=30, I see a CI width of 0.63. That means if I want to guarantee 1.0 then I must have my target + half my confidence. Or 1.0 + 0.63/2 = 1.32 (close enough to the hallowed hurdle of 1.33). Increasing my n to n=125 to thus guarantee a 1.0, then my hurdle needs to be 1.20, less because we all know more samples = more confidence. That's my point. The industry talked for years about samplings in the 30s with a hurdle of 1.33 If I keep all things the same, and increase my sampling, should I not then lower my hurdle? In other words, a capability study with n=30 and Cpk = 1.33 is equivalent to a capability study of n=125 and Cpk = 1.20