How to predict with no failures?

Six Sigma – iSixSigma Forums Old Forums General How to predict with no failures?

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
  • #41427


    I have a test where we run the product under working conditions for 100 hours and we monitor two parameters and record their values every 10 min. The spec requires no more that 1% change in their value. In the last 6 months no units have failed. My goal is to reduce the test hours to 40 hrs or less but I am not sure what kind of analysis I should use to statistically back-up that change. I was thinking of degradation analysis but I have no failures. I would appreciate any suggestion. Thanks.



    Dear Sir,
    1.First you should study & analyse field failures.The phenomenon, which you have mentioned, is very common, when you test as per standards such as JIS or DIN.In spite of no failures in Lab. testing,there are failures in the field.The main reason is that the field stresses are different than the Designed stresses such as Vibrations, Thermal, Humidity,use & missuse by the customers. Good organisations treat use & missuse by customers as a noise factor during design stage.
    2.You should study the stresses operating on your product in the field & plan a Multi Environmental Overstress Testing.A good MEOST produces failure mostly within 32 hours.The failure may not be identical as in field, but if  failure from MEOST is similar to the one in field, you may treat testing as successful.By this you can improve the product & subject the improvement to MEOST for another 32 hours, may be 40 hours due to improvement. The purpose testing is to improve Reliability & Durability.
    3.You can get practicle knowledge on MEOST in the book ‘The World Class Reliability’ by Mr. Keki R. Bhote [ Publishers-American Management Association ].
    4.The methodology I have suggested will not only save your enormous testing time, sometimes 50:1,also improve reliabilty.You have to invest in new chambers.
    5. Since you will be giving a huge saving to the company,expect a big PROMOTION.Pl. don’t forget me for a treat.
    6.Pl feel free to ask any queries.
    With warm regards,



    You could try accelerated testing where by the product to be tested could be subjected to more severe operating conditions.You can read some testing procedures for enclosures of solenoid housings as good examples of accelerated testing.


    Michael Schlueter

    Hello Angelina,
    You can use Taguchi’s approach to Online Quality Engineering, when you are more concerned with economics than with statistics. His formulas puts into balance

    desired quality level (in monetary terms)
    frequency of inspections
    frequency of adjustments (to return to the on-target situation)
    adjustment limits
    by taking into account

    cost of measurement/inspection
    cost of adjustment
    current frequencies
    current adjustment limits.
    This approach improved both improves quality and cost of many processes. Hope this helps. Please contact me by posting to this reply when you need more.
    Kind regards, Michael Schlueter


    Mike Walmsley

    What are the parameters? You may be able to do a degradation analysis. Provide your e-mail an I’ll forward an example.
    If electronic, there are certainly many models available to select from.
    Ref Electronic Systems Reliability (2 volume set by Haklim & Pollino, Arkade House).
    Further questions would be;”What is the test and user environment?
    Doyou have a pre-defined level of sensitivity for the test ?
    I may be able to use ALTA 6 Pro to do the analysis.

Viewing 5 posts - 1 through 5 (of 5 total)

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