Tolerance definition based on expected Cp/cpk

[drhouse77]

Hello all,
I have a question. We are in technical review prior the offer, our customer is requesting a min Cp/Cpk of 2 on some characteristics.
Just as pure example, assuming that the dimension is 40 +/-0,3, that the Cp/Cpk in serial production has to be >2, how can I calculate in advance which should be UCL/LCL I have to keep in serial life or how should much I should close the tollerance range during setting my process in order to achieve the Cp/Cpk?
Many thanks for your help!!!

 

[Miner]

This is a misunderstanding of process capability. The process is what it is and creates its own control limits. You can reduce variation in three ways: 1) eliminate special causes; 2) reduce common cause variation by changing the process design, or 3) reduce common cause variation by using Taguchi parameter design methods. You cannot reduce variation by tightening the specifications unless you setup 100% automated inspection and sorted out everything, which would truncate the distribution and prevent you from calculating Cp/Cpk, which assume normality.

 

[drhouse77]

Hi Miner, yes I know you are totally right. Maybe I miss something in my description.
It's a plastic part, injection moulding process. Let's Imagine for a while that the only variability of the process Is the the mold. If i use the whole tolleranze range maybe I have cp/cpk 1.33 but if i want cp/cpk 2 or higher then i Need to reduce the portion of tollerance. How much i tight the tollerance Is equal to the effort to center the dimension. I know it's pure thoery but There Is a calculation whoch could support?

 

[Miner]

So you definitely have cavity to cavity variation, possibly mold to mold variation if you have more than one mold. If you run this mold on more than one injection press, you would have press to press variation. You will have variation from lot to lot of plastic resin, variation from the amount of regrind used, variation in moisture content of the resin, variation from the process parameters, etc., so the mold is not the only source of variation. Tightening the tolerance has no impact on these sources of variation. The process variation stays the same.

 

[optomist1]

...put another way if the "common cause variation is unacceptable", (special cause variation has been addressed/eliminated), the process should be looked at in detail....re-designed

 

[drhouse77]

I know all the variances you mentioned. I am talking from a purely hipothetical situation (totally out of real world), where i can only work in reducing the tollerance, Is there any way to calculate how much i have to reduce It?

 

[Jim Wynne]

I know all the variances you mentioned. I am talking from a purely hipothetical situation (totally out of real world), where i can only work in reducing the tollerance, Is there any way to calculate how much i have to reduce It?

In order to increase Cpk, you would need to expand the tolerance, not reduce it.

 

[drhouse77]

right, but in this case you should change customer design. To go in this direction we want to say like this:
1) you ask 40 +/-0,3 with cp/cpk 2, this means that we must tune our mould to 40 +/- xx (xx value is what I want to calculate which is less the +/-0,3 required in the dwg) and this has a huge developement cost
2) then considering the huge effert (anc cost) in centering the new tolelrance and taking into condideration that injetion process has several variants, proposal might be to increase orginal tollerances in dwg (i.e. 40 +/-0,5) or reduce cp/cpk to 1,33

 

[Mikey324]

Do you mean, for example, you are currently using up 80% of the tolerance and would like to know if improving to using only 5% of the tolerance would give you > 2?

 

[drhouse77]

Right, exactly what i mean