MTBF comes up occasionally in the forum, but how to interpret it is a bit of a mystery to me.
A few questions and hopefully some engineers pitch in.
MTBF for GA piston engines has been suggested as 50,000 hours, presumably turbine is an order of magnitude better with airlines operating in a 1 in a million hour occurrence model? It would be helpful to know the numbers, and if they are part of certification standards.
MTBF uses a bathtub curve for expected failure – are there rule of thumb statistics which might help you understand what the probability of failure might be for a given MTBF? I am trying to understand if the curve is converted to a bell type curve with standard deviation etc
I used to know this stuff, but time goes fast. MTBF of 50,000 hours for a GA piston engine sounds about right, it could be twice also and some more, or much less. Bathtub etc has little relevance here. Bathtub curve is only for one off production or very complex parts, like a whole factory or a powerplant. An airliner probably also has a bathtub curve due to its complexity. Bathtub curve for engines must be seen related to production batches. The first batch is full av faults, the next one is much better etc. But the MTBF will rise at the end of life, or end of TBO. MTBF is used slightly different on different things.
To get the MTBF on series production, the method can be to test them until they fail. For instance, 10 engines runs continuously or according to some start-stop procedure. They run for 1000 hours and two of them fails, then the MTBF = 1000*10/2 = 5000 hours. Then say you test 100 engines and still only 2 fails within 1000 hours, then MTBF = 1000*100/2 = 50,000 hours. The time when it fails means nothing here, and you can get 50,000 in different ways.
But then you test them for 3000 hours and suddenly 20 fails, the MTBF = 3000*100/20 = 15,000. By closer inspection it is found that 18 of those failures happened between 2000 and 3000 hours, so the TBO or TBR is set to 2000 hours which gives an MTBF of 100,000 hours. For instance.
This means the probability of failure is roughly 1 in every 100,000 hours, but only with the constraint that it’s service life is 2000 hours (TBO). I guess instead of testing hundreds of engines, it is better to start with a TBO of 500-1000 hours, and extend it when data for engines in the field becomes available.
LeSving thank you for a clear and very useful post.
My daily runner is a ’72 1300 Beetle with a magnesium alloy crankcase air cooled engine – it would be interesting to know the MTBF as mine is still on its original engine!
I am told the oily engine bay is by design to ensure the engine inhibits the magnesium crankcase against corrosion.
The ’51 L18C also on its original C90-8F, (note a trend) although it has been overhauled a couple of times, and runs very sweet – and likes to spray a bit of oil through the breather (they typically use a quart every six or so hours).
I am just missing a Cessna 195 with a 275 hp Shakey Jake (around a quart every hour or two) in my collection of somewhat oily but reliable engines. It would also mean I could actually plan to join a EuroGA fly in and mean it.
My daily runner is a ’72 1300 Beetle with a magnesium alloy crankcase air cooled engine – it would be interesting to know the MTBF as mine is still on its original engine!
You can of course count the hours between it actually has failed and take the average. Or take the total hours run and device it by the number of failures. Chances are that it may never have failed, or failure was due to worn out parts that were scheduled to be replaced years ago, so you will not get a number, at least not a number that means anything.
My old Peugeot that I replaced some years ago, ran for 350000 km. Never once did the engine fail (but lots of other things 
) That is an equivalent of 5-6000 hours with no failures and no overhaul.
Apologies for thread drift, again – my brother’s ’59 MB 180D (hormiga negra in Spanish, or black ant) managed 2 million km before giving up the ghost (sold to a restorer a few years back). His current ’92 190D is going strong but is still a spring chicken at around half a million km.
I was quite fond of the 180D as it was fifty fifty which way it would swerve when you applied the drum brakes.
Using your approach for estimating MTBF our mixed fleet of pistons has an MTBF of around 9,000 hours – on average they must be on their fourth or fifth overhaul so a bathtub effect may be occurring?
Posts moved here from the homebuilt avionics thread
Easa is probably going to see the light and think “if we can allow a non TSO non TC drone to fly over inhabited neighborhoods, maybe a non TSO non TC plane can do just as well without crashing”.
Despite popular lore even non TSO stuff is quite reliable nowadays. You could even get ETOPS 120 certification for a diesel DA-42 in spite of it being powered by two car engines.
The “car engines” are highly modified and in reality there’s not many parts left from the original engine …. casing, crankshaft, pistons … none of that is original Mercedes
Any evidence backing that other common piece of popular lore?
Evidence for what?
i’ve been in the factory and I watched the machines make the parts. Is that enough evidence?
Thielert replaced a lot of parts to save weight. Austro is mostly unmodified.
