iSixSigma

Weibull/minitab

Six Sigma – iSixSigma Forums Old Forums General Weibull/minitab

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #38659

    Steven
    Member

    Field failure analysis…I have been researching MTBF and have come up with a generic MTBF for complete systems since I can come up with system data, but I need to derive MTBF for individual PCAs. I may have to push to get all the ship data for all the PCAs and come up with it that way. I have 1 year’s worth of data for PCAs and lifetime for whole systems. Field failure analysis…PCA Analysis……
    I am not familiar with Weibull and I use SPC XL and DOE KISS 97 with EXCEL. The biggest thing I will need is all the Parts Library that Relex has. The actual calculations may be easy compared to finding/inputting all the data for the individual parts. Steven

    0
    #116024

    Dayton
    Member

    Steven,
     
    Your MTBF calculations at systems level can be done two ways, retrospectively using your empirically derived data or you can use a predictive morphology and the process that I described earlier, that of determining the MTBF of each of the subassemblies separately and then determining the reciprocal of each MTBF and adding the reciprocals of all subassemblies together and taking the reciprocal of the summation – which gives you the compounded MTBF of the system. 
     
    If you have sufficient and appropriate past failure data it will be (generally) easier and quicker to use your past failure data to create your MTBF model.  The predictive approach is predicated upon understanding enough about your components and their interplay to build the model from the ground up – and for ongoing products or systems it’s a pretty good bet that you don’t have that immediate wherewithal or you’d already have done it.
     Vinny

    0
    #116026

    Dayton
    Member

    Also, one thing that can plague you when you are using your empirical data is that you have to be able to appropriately factor in a utilization factor – you can’t just say x-units in the field for y-duration against z-failures yields w-MTBF. 
     
    You need to be able to factor in how many hours per day your units are used, days per month/year, and number of power ups, etc., and know whether past application is typical of future application.  That’s what I meant by utilization factor.   
     
    I have had similar issues with various batteries used in medial devices, and utilization factors, trickle charges, ambulatory utilizations, etc., meaning you have to apply what you know about your customer’s current use and predicted future use of your equipment to your assessment.
     Vinny  

    0
    #116041

    Bob J
    Participant

    Steven,
    Sounds like predictive MTBF (Mil-HDBK-217) is the way to go…   As Vinny pointed out, it’s pretty easy to do…  Take the EBOM and map it against the failure rates (FITS=Failures in Time) in Relex…  For example, most resistors and caps are 5 FITS, diodes are 20 FITS etc…  The active silicon stuff (RAM, Arrays etc) are typically 100-300 FITS… 
    After you’ve mapped the EBOM, add up the total FITS…  This is the total predicted failure rate per 1000000 hours.  The model was originally developed for the Military so there are scaling factors that can be applied for especially rigorous environments…  Since I assume that you are working with commercial grade products, the scaling doesn’t apply. 
    Your predicted MTBF is one over your total FITS…
    Good luck….  If you have any other questions and want to take this offline you can reach me at [email protected].
    Best Regards,
    Bob J

    0
    #116043

    Dayton
    Member

    Steven,
    Bob J. is correct.  He and I should have added though that the predictive model described is applicable to systems with components in series versus in parallel, but that caveat usually only impacts some components in an otherwise dynamic and series concentrated system.   Should be of little impact to you but a quick schematic review will tell you the potential for calculation impact due to lack of MTBF compounding from a few potentially high importance but relatively low reliability components in parallel – my bet is low impact to your predictive MTBF calculation but it’s something for your back pocket anyway.
    Vinny

    0
    #116060

    Bob J
    Participant

    Vinny,
    No problem…  Our development guys use this as a way to show that a new (or revised) product design meets the predicted reliability targets… 
    Of course we also have a serious derating program so the actual field performance averages almost 4 times the predicted values….
    Good stuff…;-)
    Best Regards,
    Bob J

    0
Viewing 6 posts - 1 through 6 (of 6 total)

The forum ‘General’ is closed to new topics and replies.