iSixSigma

Mixed Failure Modes in Process Validation

Six Sigma – iSixSigma Forums General Forums Tools & Templates Mixed Failure Modes in Process Validation

Viewing 4 posts - 1 through 4 (of 4 total)
  • Author
    Posts
  • #247641

    Brian Walsh
    Participant

    Hi All,

    Sorry if this is already covered elsewhere, but i did a search and didn’t see anything.

    How do you typically handle mixed failure modes in a process validation.  A specific example where I have run into would be bonding validations.  Lets say the joint has 25 Newton spec.  Sometimes the adhesive joint fails and sometimes the substrate fails.    Alternatively, the substrate always fails and the joint never does.  However, the peak load is always above 25 Newtons.

    Our company tends rely more heavily on confidence/reliability as acceptance criteria and usually have an observational Ppk (e.g. we will calculate Ppk but its not acceptance criteria), which is probably not the most robust way from a process validation perspective, but that is a whole different discussion.

    If we are using C/R as acceptance criteria and the failure modes are mixed, I think the best way to assess the results would be to analyze each failure mode separately, however  that could hurt our stats by decreasing the sample size and also isolating a failure mode with a lower peak load (assuming the adhesive bond fails at a lower value than the substrate).  If all samples have the substrate fail i would assume you treat it as attribute (the joint passes or fails), but that can drive up your sample size and you may not wind up having enough samples to satisfy the requirements.

    I have no idea on where to start on how to assess Cpk or Ppk with mixed failure modes/all substrate failures.

    In either case (Cpk/Ppk or C/R) in which it is acceptable to group the data even though substrate failure is not technically the load at which the adhesive bond fails?

    Any help is greatly appreciated.  Thanks.

    0
    #247675

    Strayer
    Participant

    I would do a formal FMEA exercise.  Failure Mode Effects Analysis will help you to understand and prioritize the various failure modes so that you can address them in order beginning with the highest RPN (Risk Priority Number).  If you aren’t familiar with FMEA you can find instruction here or by web search, but the definitive handbook is from the AIAG https://www.aiag.org/quality/automotive-core-tools/fmea   As for your statistics with mixed failure modes, when you mix apples and oranges it won’t tell you much about either apples or oranges, but only about apple-orange.

    0
    #247701

    Brian Walsh
    Participant

    @Straydog, thanks for the input.  A PFMEA and DFMEA have already been drafted and that is what is driving the specified C/R level.

     

    0
    #247962

    Roger Hill
    Participant

    Hello, as I understand your description, the test “passes” in that the failure (bond or substrate) is always greater than the requirement, correct?

    since you are testing to destruction something will always fail.  Does the failure mode matter if you are well above the requirement. Given you use C/R, and it seems you have lots of data, can you not calculate that the low end of distribution of strength at failure for a given C/R, is above the requirement?

    0
Viewing 4 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic.