Yes, a failure is a failure. If something keeps having early life failures, it's MTBF will be low, no point hiding the early life failures, as ultimately, they are a failure that has an impact on the plant / business.
MTBF and any other metric for that matter must be used in context of something. MTBF is usually used on the context of a reliability program to help prioritize where attention should be paid to solve problems, so a low MTBF = problem = do something to stop the startup failures.
If MTBF is being used on contractual / performance guarantee context, then you would need to refer to whatever the contract says, but it'd be surprised if it sadi to ignore commissioing / early life failures.
Andrew O'Neill
Specialist Mechanical Engineer
Australia
Seems to me that would depend on the cause of the failure. Was it just a bad setup? A one-off failure isn't something that you can really do remedial actions for. But, otherwise, yes, they would count. Note that typically, MTBF is a misnomer, since that only applies to the middle of life for lots of systems, there's infant mortality on the front end and there's wear mechanisms on the back end.
We once had a testing house mis-wire power to our system and fried 80% of the electronics; we didn't count that toward MTBF, but did have to explain that to the customer.
TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!
To answer your original question - any failure at any time during the OPERATING life of the machine is part of the MTBF calculation. That includes start-up failures, in-service failures, and - in some rare cases - decommissioning failures (think nuclear stuff).
The key thing for any MTBF-type calculation is to PRECISELY define what the term "failure" means. For example:
"Anything that causes my generator to be unable to produce the nameplate electric power" can include all of the following (which is not an exhaustive list, by any means!).
1) coolant failure (requires a derate)
2) installed above design altitude (requires a derate)
3) short circuit of stator winding
4) short circuit of rotor winding
5) open circuit of stator winding
6) open circuit of rotor winding
7) failure of excitation (assuming excitation is from separate source)
8) bearing failure on generator (can't turn the shaft)
4) coupling failure (can't transmit mechanical energy to rotate the generator shaft)
5) generator shaft failure (if it's broke, won't turn)
11) any failure of the prime mover (which opens up another whole can of worms! But not your problem, since you're the generator guy.)
12) etc.
Converting energy to motion for more than half a century
Here's fun. You are in default on the contract if a control wire is left loose when the power switch is turned on and nothing happens. MTBF is zero and the contract can be terminated for breech on failure to meet requirements.
To avoid this, be specific about the period of operation over which the MTBF will be evaluated.
We had a customer complaining about early failures.
The replacement costs were 50x the equipment cost, and what they really wanted was a predictable cost.
So we built special units for them. They cost us 2x.
Our customer paid noting for the first 100 days, and then began paying a daily rate.
At 200 days we had recouped all of our costs,
But the beefed-up equipment often ran 400-600 days, with them paying us for each one.
It really depends on where the money is as to how this is handled.
If you are talking about small replicable items that are used in numbers (disc drives in arrays) then a few early life failures have little impact on the MTBF.
If you are talking about very few items that are very expensive to replace (way more than the equipment cost) then it is another issue.
