Re: [Hampshire] [OT] MTBF

2009/7/19 Philip Stubbs <philip@???>:
> 2009/7/19 James Courtier-Dutton <james.dutton@???>:
>
>> To get a MTBF figure:
>> Approx MTBF = ( Number of Devices tested * Test time ) / Number of failures.
>
> As ever, that is not always the case. I used to work for an aerospace
> company, and they would quote MTBF before one sample had even been
> built, let alone tested to failure. In that case, the MTBF was
> calculated based on some rules. I don't know the details but I do
> remember that number of components were a key variable. Therefore the
> pumps they made were designed to have a few parts as possible. The
> MTBF number was therefore a result of some clever work by
> statisticians. I would not be surprised if that is the case here. No

It is my understanding that once one has the MTBF of each single component.
Once one knows how they are fitted together, one can calculate
availability figures for the sum of the parts.
So, if one has a network with redundant devices, one can take the MTBF
together with the mean time to restore (MTTR) of each device on the
network and come up with a total availability for the entire network.
The knowledge area is called RAMS analysis. (Reliability,
Availability, Maintainability and Safety).
The case is similar for components on a circuit board. One can have
the MTBF of each resistor, etc. on the circuit board and calculate
the the MTBF for the entire circuit board.
As a general rule, more devices in series makes the result worse, more
devices in parallel make the results better. This is why more
components in the pump made it worse because I expect few of them were
redundant components.

James