From failure in time to FIT

[avovan]

This might have been asked before... I am currently calculating a time to failure of a product, which is a function of several parameters, including a failure probability F. Given an arbitrary F (let's say 1e-6) then a time to failure can be calculated.

From this data, how can I calculate the failure rate (given in FITs)? My problem lies in the fact that if I change the arbitrarily chosen F, then the time to failure will obviously also change (but not linearly), and I end up with a different FIT value.

It is possible at all to go from failure probability and time to failure rate data?

Thanks in advance.

[Pantelis]

Well it depends on the model. With a simple exponential model (constant failure rate) -- then yes one can do it with a single F. With a time varying failure rate model (i.e. Weibull) the failure rate is function of the two model parameters, thus one needs two values of F to compute the two parameters and so forth.

In other words if you have two values of F, and in the case of Weibull, you have two equations and two unknowns (the model parameters). Once you solve for the parameters you can use them in the failure rate equation.

Hope this helps.

[avovan]

Thank you for your answer... I'll restate my question...

If, for example, I choose to work with a failure probability F of 1e-6, I end up with a time to failure of 25.68 years.

Keeping all the other variables constant, if I use an F of 1e-5, I end up with a time to failure of 721.13 years.

Thus:
F->Time to failure
1e-6=25.68 yrs
1e-5=721.13 yrs
1e-4=20250.40 yrs

How do I convert these figures to FIT values?

Thanks again.

[Pantelis]

OK -
Based on what you provided you have a decreasing failure rate.

The parameters of a Weibull model fitting the data are:
Beta: 0.6904
Eta: 1.259E+010

Using the equation for the weibull failure rate (see http://www.weibull.com/LifeDataWeb/weibull_statistical_properties.htm)you can compile values of the failure rate at any time.

To convert to FIT multiply by the appropriate multiplier (I don't use FITs - but if I recall correctly FITs have a 10e9 multiplier)

[avovan]

OK,

Thank you. This was exactly what I was looking for...