1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.
Dismiss Notice
You must be a registered member in order to post messages and view/download attached files in this forum.
Click here to register.

evaluation of the process produced on the another machine

Discussion in 'Capability - Process, Machine, Gage …' started by essegn, Jul 30, 2019.

  1. essegn

    essegn Member

    Feb 5, 2016
    Likes Received:
    Trophy Points:
    The process X is being done at the machine 1. The process is stable and capable and they exist more than 100 measurements. The most important is Std.Dev. The mean could be easily adjusted.

    The process X should be introduced to the machine 2 (as backup).

    Every run is expensive and time consuming. This is why there is no possible to collect data easily.


    How could be evaluated, that the process X on the machine "works well"? (Cpk>0.8)
    What sample size should be used in order to have the Power at at least by 80%?


    The way, how i usually manage it is as following:
    1. Simply run the process X at the machine 2 3-5 times (to get a sense about mean and Std.Dev.)
    2. Define (simulation in Minitab - Calc - Random Data - Normal) the worst possible process, that is still in tolerance (for instance St.Abw. 0,9 , Cpk=0,66-0,74, 0-3,3% out of spec)
    3. Power and Sample Size - 2 Variances

    After that is obvious, which sample size correspond to 80% Power.

    Is there any other way?


    I do not want to compare the process X on the machine 1 and the machine 2. The process X could be on the machine2 could be worse than the machine 1, but it needs to be stable.
    I would like to find out (in the cheapest way), if we can rely on the outputs. Stable process have Cpk>1, but more than 0,8 should be sufficient.

    The stability of the process has to be judged with a SPC Chart- in this example I-MR. If you have a sample size of 10, the control chart will give you an idea about the stability of the process.

    In order to calculate the process capability there are at least 25 measurements required.

    When it is not necessary, also 10 runs are too much.

    There could be also compared variances between the process X produced at the machine 1 and machine 2 - to get sense, whether the variation is comparable.


    How could be the sample size in the most economical way determined, in order to get at least 80% Power?