Download as txt, pdf, or txt
Download as txt, pdf, or txt
You are on page 1of 1

Of course, the denominator will ordinarily be 1, because the device has a cumula tive probability of 1 of failing some time

from 0 to infinity. Thus it is a char acteristic of probability density functions that the integrals from 0 to infinit y are 1. As a result, the mean time to fail can usually be expressed as

If we substitute the exponential density f(t) = le-lt into this equation and eva luate the integral, we get MTTF = 1/l. Thus the mean time to fail for an expo nential system is the inverse of the rate.

Now let's try something a little more interesting. Suppose we manufacture a batc h of dual-redundant widgets, hoping to improve their reliability in service. A d ual-widget is said to be failed only when both sub-widgets have failed. What is the failure density for a dual-widget? This can be derived in several different ways, but one simple way is to realize that the probability of both sub-widgets being failed by time t is

so this is the cumulative failure distribution F(t) for dual-widgets. From this we can immediately infer the density distribution f(t), which is simply the deri vative of F(t) (recalling that F(t) is the integral of f(t)), so we have

Notice that this is not a pure exponential distribution anymore (unlike the dist ribution for failures of a single widget). A plot of this density is shown below :

Remember that the failure density for the simplex widgets is a maximum at t = 0, whereas it is zero for a dual-widget. It then rises to a maximum and falls off. What is the mean time for a dual-widget to fail? As always, we get that by eval uating equation (5) above, but now we use our new dual-widget density function. Evaluating the integral gives MTTF = (3/2)(1/l).

You might also like