SO there is a bit to unpack here…can you provide us the exact wording of your Customer’s request? Also can you tell us what your sampling scheme is? Ho many parts, how many measurements per part and how often do you sample parts. “Control Limit” has a very specific statistical meaning and from what you have posted I suspect that either your Customer or you do not understand what a control limit is and how it is used. This is fairly common and you are wise to come here to ask.

The proper use of the term Control Limit is when you are using SPC (Statistical Process Control) charting to monitor your process. Are you using SPC? Ho familiar are you with SPC and eh establishment of subgroups, frequency, charting and establishing statistical control?

What do you mean by process capability? Are you calculating one of the process capability indexes like Cpk or Ppk? What formula are you using? Within subgroup standard deviation or total standard deviation?

In many cases the control limits are calculated using subgroups of data that are between 2-5 sequentially produced parts with 1 measurement per part. The Chart is an X-bar, R Chart, In other words you are plotting the subgroup averages and the within subgroup range for each subgroup ( set of samples). The lower control limit will be the lower limit of expected variation of the subgroup average and as such this limit cannot be used for process capability calculations. In this case you will have to calculate the process capability index from all of the individual values after you have established that the process is in control.

In some cases - such as with destruct testing like tensile strength - a I, MR chart is a better choice (I, MR is an individuals, moving range chart). If this is yoru SPC charting method ten the Lower Control limit can be used for process capability index calculations...HOWEVER, since tensile strength is destruct an can be weirdly distributed it is essential that we understand teh questions above adn perhaps see some of your data…

First of all thank you for the answer!

Usually for other characteristics the control spec is defined by the mean and the standard deviation by 3 sigma.

But in this characteristics have tolerances of LSL and USL,

In the drawing it mentions exactly,

THE MINIMUM PULL STRENGTH SHALL BE HIGHER THAN 100N.

And the customer wants to perform 30 Pull Test with random sampling, and define a control limit, which I suppose that should be higher than 100N. Internally we have been discussing aprox 300 to 400N, because the Pull Strength that we get is above 800N to 2000N. The variation is large so i guess the process cabaility is not stable. But this part i checked with the customer and seems can be ommited.

So I don't know how can I get a accurate control limit statiscally, but not being too high to prevent issues of being as a NG part even though it is over the specification limit. For example control limit set is to 700N but I get 650N, even though it is 6.5 times higher than specification it is out of control spec....