Normal vs. Weibull

[Robbie]

When is safe to use the normal distribution instead of a weibull for wear out cases? I'm having philisophical discussion with some six sigma people

[Arai.M]

The issue of using the normal distribution to describe life data is that the normal distribution is defined in the negative domain. Negative times/cycles/etc to failure are not very meaningful. That and the symmetrical shape of the normal distribution make it a distribution that is not often used in life data analysis. The Weibull distribution is very flexible in terms of the data it can represent and is therefore more commonly used. However, if the failure rate behavior of your product is reasonably described by a normal distribution and the probability of observing negative times/cycles/etc. is negligible, there is nothing wrong with using the normal distribution. The choice has to be based on understanding the failure rate behavior of your product rather than philosophy :)

[Robbie]

Thank you.

[Hilaire Perera]

The Weibull distribution is one of the most commonly used distributions in Reliability Engineering because of the many shapes it attains for various values of b (Shape Parameter). Weibull analysis continues to gain in popularity for reliability work, particularly in the area of mechanical component reliability, due to its inherent versatility. The useful metrics (Reliability, Failure Rate, Mean Time To Failure) can be easily calculated from Weibull probability density function variables. We can say that Weibull Analysis is good to show how a product behaves in its life




b < 1


Implies infant mortality. If product survives infant mortality, its resistance to failure improves with age



b = 1


Implies failures are random. An old part is just as good (or bad) as a new part



b > 1 & < 4


Implies early wearout


b > 4


Implies old age (rapid) wearout