The limits of Weibull Parameters

[raed kareem]

Hello:

I test power modules of reliability and i analyze the results staticly with Weibull. But i get a larg beta (about 11). Is this o.k? And how can i interpret the test results according to the calculated weibull paramaters

thx for every help

raed

[RS Support]

There is a general discussion on the effects of the parameters at: http://www.weibull.com/LifeDataWeb/characteristics_of_the_weibull_distribution.htm
You may also find an article in the Edge “A High Value of Beta is Not Necessarily Cause for Concern” useful. href: http://www.reliasoft.com/newsletter/1q2001/beta.htm

Now as to your specific question, the higher the beta the closer the failures times (or the values of whatever random variable you are measuring) are to each other. A beta of infinity would imply that everything failed at exactly the same instant. (Just as a side note -- with fairly high betas (20+) investigation of other candidate distribution, such as the lognormal, may be warranted.)

Now with a beta of 11 that would imply that once you reach a specific point in time (or a specific value of the random variable) failures start and they continue to occur very close to each other. If your random variable is some particular stress (instead of time) then that would imply that as soon as you reach that stress failures happen fairly quickly.

Now that could be good or bad. If the scale parameter (eta) is higher than the expected life (or stress value) then that is good – if it is lower then you do have a problem!

Of course the above statement assumes that there are no issues with your reliability test itself. Other factors can influence this (i.e. failures happening very close to each other) such as an improperly designed experiment including stresses applied and data recording/collection/interpretation.

Hope that helps.