Reliability Prediction, To include or not to include

[samuel80]

When performing a reliability prediction what is your thought on including things like heat sinks, fans, and other support type components that don't actually contribute to the "failure" of the system if it fails. Ex. A computer won't immediately fail if the fan fails, so would you count the failure rate of the fan into the computer's reliability prediction? Now eventually if the fan isn't replaced the computer will overheat causing temperature stresses for an eventual circuit component failure. Thanks for any experiences or input/opinions.

[penguin]

I think that the fans should be considered because of overheating. The probability of overheating should be counted in. If the fans can be replaced in time, the probability of overheating is zero or very small. On other words, if the fans cann't be replaced in time, the probability of overheating is equal to failure rate of the fans.

Expect other viewpoints. Thank you.

[Arai.M]

The question seems to be how to handle components that are not in simple series or parallel configurations.
Ignoring supporting components will generate an optimistic prediction, while considering them in series will result in pessimistic results.
Depending on the goal of the analysis, either approach might be acceptable.
One of the extremes is not always necessary. Components can be modeled in a case by case basis. For the fan example, penguin is right that the event that could be modeled is overheating. For that event to occur, the fan needs to fail and no replacement occuring within some time. That time needs to be determined, wether it is obtained via in-house testing, from field data, etc. This situation could then be modeled as a standby configuration. The active distribution is that one of the fan and the standby distribution is that of the time to overheating failure once there is a fan failure.
If the effect of a fan failure in other components is well understood more complex load sharing configurations can be used. A clear understanding of how the stresses affects the life of other components is then needed. Things can get very complex, very quickly. It is a good idea to look for the simplest approach that does the job.

Arai