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.

cpk on non-normal distribution (limited intervals)

Discussion in 'Capability - Process, Machine, Gage …' started by Marco2mac, Sep 4, 2020.

  1. Marco2mac

    Marco2mac New Member

    Joined:
    Sep 4, 2020
    Messages:
    3
    Likes Received:
    0
    Trophy Points:
    1
    Hello!
    first of all, thanx in advance for your attention and your support
    I've a lack of knowledge than is blocking my analysis.

    We register the efficiency of a process, measured as percentage.
    Requirement is that minimum values has to be >95%, our data populations are distributed between 97% and 100%

    I was required to define a ppk of this variable.... but I can't really approach it because I don't know how to handle cases where there is a physical upper (or lower) limits.

    Could someone indicate me the correct approach?
    thank you in advance
    Marco
     
  2. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    474
    Likes Received:
    516
    Trophy Points:
    92
    Location:
    Maine
    Welcome Marco2mac! :)

    By efficiency do you mean yield, uptime, on-time delivery, productivity? Or something else? Knowing the specific measure will help us.

    That said, a Ppk index of these types of metrics is not warranted. Ppk is for continuous data characteristics that have engineered specifications.
     
    Marco2mac and Andy Nichols like this.
  3. Marco2mac

    Marco2mac New Member

    Joined:
    Sep 4, 2020
    Messages:
    3
    Likes Received:
    0
    Trophy Points:
    1
    thanx alot for your feedback
    And sorry, u are right, it was given just for myself.
    I'm talking about a "filtration efficiency", where targets can be ie 90% or 95% or 99,5%
    Let's say, I can't control each batch (test is quite expensive) and we do a revalidation each 3 months
    My customer asked us for this kind of reports, but clearly values aren't normally distributed (somehow they are at 90% but definitely not at 99%)
    thanks alot for any useful hint
     
    Last edited: Sep 27, 2020
  4. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    474
    Likes Received:
    516
    Trophy Points:
    92
    Location:
    Maine
    Can you provide a bit more Information on the filtration efficiency? This sounds like you are producing a filter or filtration system of some kind and that the ‘grading’ process is based on well it filters out target material. How do you calculate the efficiency? This is critical to our answer...

    Is the Customer asking for the Cpk/Ppk index or just a capability study?
     
    Marco2mac likes this.
  5. Marco2mac

    Marco2mac New Member

    Joined:
    Sep 4, 2020
    Messages:
    3
    Likes Received:
    0
    Trophy Points:
    1
    the test itself is somehow easy to descrive, but quite complex in the details. sample is challenged with a liquid where a certain quantity of contaminant is constantly added added during the test, with 2 particle counters counting the amount of particles upstream and downstream and value of efficiency is calculated each channel (that means, particle dimension) as a standard formula 1- (Ndown/Nup). actually, N is not constant because while injection is constant, contaminant in the circuit is increasing because what is not filtered out still circulating. So, very complex, almost not realistic, to use particle number. Output of the test is ie 99,6% at certain size etc etc, so we have populations of these percentage values. And yes, customer (automotive) is asking for a cpk calculated on the minimum value given by specifications.
     

Share This Page