iSixSigma

Process Yield Estimation

Six Sigma – iSixSigma Forums General Forums Tools & Templates Process Yield Estimation

This topic contains 3 replies, has 3 voices, and was last updated by  James Heidt 1 year, 1 month ago.

Viewing 4 posts - 1 through 4 (of 4 total)
  • Author
    Posts
  • #55859

    Marc Trumpi
    Participant

    Hello everyone,

    Quick question, which I can’t seem to find anywhere.

    Say I have a process which produces parts with several critical dimensions. In my case 4. From each of those dimensions I can determine a cpk value and a . How can I estimate a total process yield, combining the yields or cpk values from the individual dimensions? I know that there are influences between some dimensions, so I can’t just multiply all individual yields.

    Thanks.

    #201948

    Chris Seider
    Participant

    sigma squared total = sigma squared of first part plus sigma squared of second part ….if they are linearly connected and you’re looking at total length.

    One can predict total ppm defective of final product by adding up ppm defective individually BUT there’s some overlap probably that may be difficult to model.

    It’s not an exact method but…

    #201951

    Marc Trumpi
    Participant

    Hi Chris,

    Thanks, but unfortunately the measurements are not connected linearly. Translated to an XYZ axis I’m looking at 1 measurement in X, 2 in Y and 1 in Z. And as you say, there’s some overlap (not probably, but surely) in the defective values. Eg, if X becomes larger, there is a significant chance that one of the Y’s is smaller.
    In that case the summation of the ppm would be far too pessimistic, I’m afraid.

    #201978

    James Heidt
    Participant

    If your objective to take the 4 individual Cpk’s to find an overall process Cpk, the ‘overlap’ would seem to make this a problem. You could try a sample to estimate how many of the parts overlap in defects, and then develop a model of how often one dimension failing leads to other dimensions failing and then test the model for reliability against a sample. But unfortunately I don’t see an easy solution here.

Viewing 4 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic.