Twin safety record

[dtuuri]

If I understood this video correctly, this instructor should not be teaching. It appears that he has one prop windmilling (not sim. feather), and orders a go-around from a landing configuration. They are hugging the runway with the instructor calling out airspeed warnings and then says he will simulate feather, removing the drag of the windmilling prop. He was lucky that he didn't end up in a Vmc rollover.

Indeed. The "instructor" (I'm betting he was really a salesman) also said to "keep it above blue line". Well, blue line is a takeoff reference speed, clean, at gross takeoff weight, not a landing weight reference speed, dirty. A lot of CFIs (nearly all?) continue to propagate this misconception. There's no way that plane could climb at that speed in that configuration at blue line.

As for the windmilling prop--an engine loss on final is no reason to mess with anything. The risk of botching either the landing or the handling of the engine failure is just too great in the limited time available. The exercise should have been terminated when the landing was assured. As you point out, no twin engine airplanes are certified for balked landings with an inoperative engine.

dtuuri

 

[coloradobluesky]

If you have an engine that, on average, fails once every lets say 100,000 hours and you fly it for 1,000,000 hours you can expect 10 failures. If you have two such engines and fly both for 1,000,000 hours then you can expect 20 failures, twice as many.

One engine's performance is not dependent on the other's.

So they are independent and there is no statistical 66% sigma stuff going on here that I can see.

 

[mscard88]

Big difference between that and the 250 hour super pilot who just got his ratings at ATP from instructors who were also 250 hour super pilots, and none of them really know what they're doing.

I dunno, most cases I guess. We had two high time experienced IPs flying a Brasilia for an IP checkout. Usually you would get one problem at a time to work. Well these two were having such a swell time the IP decided to load him up with multiple problems (think it was 3 different things) and landed gear up.

 

[flyingron]

If you have an engine that, on average, fails once every lets say 100,000 hours and you fly it for 1,000,000 hours you can expect 10 failures. If you have two such engines and fly both for 1,000,000 hours then you can expect 20 failures, twice as many.

Umm, no. It doesn't even work that way. But it's close enough for the point you're trying to make. But the issue is that the engines aren't running independently. 99% of the time they are operated in lock step. There is a common component: the pilot (there may be others like fuel systems, etc... depending on the plane).

But the operative number is what is the possibility that all engines will fail. That's going to be higher on the single engine plane.

 

[hindsight2020]

Indeed. The "instructor" (I'm betting he was really a salesman) also said to "keep it above blue line". Well, blue line is a takeoff reference speed, clean, at gross takeoff weight, not a landing weight reference speed, dirty. A lot of CFIs (nearly all?) continue to propagate this misconception. There's no way that plane could climb at that speed in that configuration at blue line.

As for the windmilling prop--an engine loss on final is no reason to mess with anything. The risk of botching either the landing or the handling of the engine failure is just too great in the limited time available. The exercise should have been terminated when the landing was assured. As you point out, no twin engine airplanes are certified for balked landings with an inoperative engine.

dtuuri

I not quite sure I follow you. The guy was falling below blue line, that's why the IP was barking. Are you suggesting the guy not climb out at VYSE on a SE go-around? what airspeed should he use then? The climb speed is affected by configuration, noted, but these things won't climb at all anyways dirty, so aiming for the dirty configuration SE-best rate is moot. It's also not published for a reason.

The reason most folks approach at VYSE as a de facto reference speed is not because they don't recognize it's not the correct reference speed (Vsox1.3..usually lower than blue line), but because in anticipation of an engine loss they want to be at blue line immediately. I don't see much wrong with that. Yes, you carry excess energy into the short final and that may translate into longer landing distances, but that's the opportunity cost of flying an airplane you can't go-around on a single engine from.

Your point regarding committing to the landing come hell or high water is noted. If you know you're not gonna go-around anyways, then flying around at blue line isn't as much of a benefit. Me, I figure real world I can keep the gear tucked in and fly fast until I know the school bus ain't coming into the runway then lower gear and land.

 

[brian]]

So, before this goes into they typical twin .v. single debate, I have a question: if you were to do a lot more business utility flying would the picture change a bit? More of a question of wanting to be in a booted twin in potential icing conditions then a booted single. (Switch to TKS if you like.) (Not so sure about flying into known ice.)

(The difference here is not the weekend family trips, but more of a "we will get there - even if we buy an airline ticket or drive" utility.)

Seems to me that this time of year the probability of finding and icing airmet is pretty high. Picking a path through that airmet showing no ice is possible (so I hear), but this is a game of probabilities (even without the icing airmet).

 

[dmspilot]

The "instructor" (I'm betting he was really a salesman) also said to "keep it above blue line". Well, blue line is a takeoff reference speed, clean, at gross takeoff weight, not a landing weight reference speed, dirty. A lot of CFIs (nearly all?) continue to propagate this misconception. There's no way that plane could climb at that speed in that configuration at blue line.

I agree that the instructor is wrong, but am curious to know what the appropriate speed would be then--lower or higher than blue line?

As for the windmilling prop--an engine loss on final is no reason to mess with anything.

I'm not sure what you're saying. I think Kristin's point was that, if the instructor is bent on simulating a SE go-around, he or she should probably do it with simulated feather, not windmilling. If you are saying that one should not bother feathering a failed engine on final, I see your point.

 

[dtuuri]

I not quite sure I follow you. The guy was falling below blue line, that's why the IP was barking. Are you suggesting the guy not climb out at VYSE on a SE go-around? what airspeed should he use then? The climb speed is affected by configuration, noted, but these things won't climb at all anyways dirty, so aiming for the dirty configuration SE-best rate is moot. It's also not published for a reason.

I agree that the instructor is wrong, but am curious to know what the appropriate speed would be then--lower or higher than blue line?

Slower would result in optimal performance, but since it isn't published the precise value is only a guess. Keeping in mind certain published speeds for accelerate go on the takeoff roll should help narrow the range--there your gear is also down and in some planes you may even have some flaps (not landing flaps though). On landing, you probably have a better weight for climbing than takeoff, but best of all you have altitude which you don't have on takeoff, so you can afford to be slower than during a takeoff "go" scenario. Zooming in at blue line seems to me to be excessive when 1.3 Vso works quite well as a minimum when altitude is on your side. When you don't have altitude, stubbornly holding a speed that takes you down toward the earth will only leave a bigger smoking hole.

I'm not sure what you're saying. I think Kristin's point was that, if the instructor is bent on simulating a SE go-around, he or she should probably do it with simulated feather, not windmilling. If you are saying that one should not bother feathering a failed engine on final, I see your point.

Yes, that's what I was saying and Kristen too, no disagreement meant with her. I was only trying to amplify her counsel. ( )

dtuuri

 

[Ted]

So, before this goes into they typical twin .v. single debate, I have a question: if you were to do a lot more business utility flying would the picture change a bit? More of a question of wanting to be in a booted twin in potential icing conditions then a booted single. (Switch to TKS if you like.) (Not so sure about flying into known ice.)

(The difference here is not the weekend family trips, but more of a "we will get there - even if we buy an airline ticket or drive" utility.)

Seems to me that this time of year the probability of finding and icing airmet is pretty high. Picking a path through that airmet showing no ice is possible (so I hear), but this is a game of probabilities (even without the icing airmet).

It really depends on the airplane. The only de-iced single I've flown is the Malibu, and it is terrible in icing in my opinion. Additionally, it has less than stellar climb performance to start out with, which only gets worse when you add ice (climb rate is the first to go). Twins tend to do a bit better in this department because of more excess horsepower, but if you throw a turbine on the front, then that changes the story because you suddenly have lots of excess horsepower and a good climb rate.

There are plenty of businessfolk who fly around in Malibus, Cirri, etc. I don't think the argument changes much from the standard single vs. twin for the same mission. What I find is that businessfolk usually do stick to a mission time more, but even when on business travel they'll more likely wait a day than blast off into something that will kill them. Depends on the pilot, etc.

I'll give you one example. Some coworkers recently took the Baron (which is very nicely equipped) on a trip that was 5 hours each way. On the way out, they made a fuel stop. On the way back, they ran out of duty day time (which includes time at the meetings) and so they stopped for the night. Then they had to make another fuel stop on the way home the next day because they'd be cutting it too close otherwise.

 

[EdFred]

If you have an engine that, on average, fails once every lets say 100,000 hours and you fly it for 1,000,000 hours you can expect 10 failures. If you have two such engines and fly both for 1,000,000 hours then you can expect 20 failures, twice as many.

One engine's performance is not dependent on the other's.

So they are independent and there is no statistical 66% sigma stuff going on here that I can see.

That is incorrect.

It's like flipping a coin. (heads good engine, tails fails 50% failure rate)
Flipping a coin twice does not guarantee you will flip tails. Just like playing the scratch-off lottery that says chances of winning in 1:4. Buying four tickets does not guarantee a win, nor does buying 8 tickets guarantee 2 wins.

2 engines does not equal twice as likely.

 

[PPC1052]

That is incorrect.

It's like flipping a coin. (heads good engine, tails fails 50% failure rate)
Flipping a coin twice does not guarantee you will flip tails. Just like playing the scratch-off lottery that says chances of winning in 1:4. Buying four tickets does not guarantee a win, nor does buying 8 tickets guarantee 2 wins.

2 engines does not equal twice as likely.

A slightly clearer way of explaining this is to suppose the following hypothetical scenario:

Assume that there is a 50% chance that any particular engine will fail. If you have two engines, the chance that you will have at least one engine failure is not 100%.

 

[hindsight2020]

Gotta love the common core math on these threads. The significant number to bear in mind is that the odds of having an engine fail when you run two at a time is higher than only carrying one. BUT, the chances of losing both engines in the same amount of time is much lower than the single engine failure rate.

The problem arises when the remaining engine doesn't help you maintain a flyable envelope, aka single engine go/engine failure decision to continue; aka non part 25 twin. Or additionally, when the pilot is not proficient at keeping the airplane in parameters in IMC or VMC when confronted with a thrust arm of consequence aka, lack of recurrent training.

That relative costs alters the maintenance behavior of the twin owner/operator is yet another variable which is real but hard to quantify and hard for the population to allocute to. Does this mean the single is safer by proxy? As a matter of semantics, it could very well be. My perception is my reality type of thing. I know people wince at the idea we are not all belonging to the same demographic of pilots, but yeah, turns out we're not all cut from the same cloth just because we're all capable of dying.

 

[Everskyward]

Right! Which means that up until that point, the prop was windmilling. Trying to go-around with gear down and a windmilling prop is great training up until the point that you lose it and crash. Too much risk of a crash to justify making the training that realistic, assuming that it was realistic, which I question.

I don't know what they are flying, but if you are making a single-engine approach, why is the prop not set to simulated feather, as in a real situation you would have feathered the engine. A SE go-around is generally for aircraft that really have the capability of doing that, which is not most light twins, at least not at gross on a warm day. This one apparently did not have much extra thrust as it was not able to overcome the windmilling prop, and the instructor had to set zero thrust, which seems to me should have been done a long time before.

It seems as if he gave the engine failure on pretty short final so in a real situation it might have been a good thing to leave the prop alone if the runway was made rather than to risk any quick, wrong, moves. On the other hand, I like others here would not have tried a real life single engine go-around, much less with the prop windmilling, in a small piston twin.

I don't think we can tell too much from the video. The wide angle lens makes it seem as if the approach and climb out are shallow. Even the first approach before the engine was pulled looks shallow. After they land it looks as if they are sitting on the ground. We also cannot see any of the instruments and can only go by the instructor's commentary.

 

[coloradobluesky]

Look you have ONE engine. It will fail, on average, 10 times in 1 million hours.
Now you have another engine exactly the same. It will fail, on average 10 times in 1 million hours.

The engines failures are NOT dependent on the other engine. They fail independently.

Lets say the engines aren't even in the same plane!
Say they are two engines in two different single engine aircraft!!!!!!

How many times will an engine (either one) fail in 1 million hours?
The correct answer, based solely on the information given, is 20 times.
Why would it be anything else?

That is as simple as I can explain it.

 

[hindsight2020]

A slightly clearer way of explaining this is to suppose the following hypothetical scenario:

Assume that there is a 50% chance that any particular engine will fail. If you have two engines, the chance that you will have at least one engine failure is not 100%.

Correct. That answer would be 75%. P(A)= 50% P(B)= 50% P(A and B)= .50*.50=.25 Probability of either happening (independent, non-mutually exclusive events)= P(A)+ P(B) - P(A and B)= 1-.25= 0.75.

That's 25% higher than for the single engine airplane, which is significant if you are likely to not be capable of negotiating thrust asymmetry and part 23 twin thrust deficiency as a matter of recency or training.

 

[coloradobluesky]

Take your coin flips. Flip coin ONE 10 times in one minute. You get on average 5 heads.
Now take ANOTHER COIN and flip it 10 times in the same minute. You get on average 5 heads.

Now, how many heads will you get, on average, in in that minute????

 

[PPC1052]

Does this mean the single is safer by proxy? As a matter of semantics, it could very well be. My perception is my reality type of thing. I know people wince at the idea we are not all belonging to the same demographic of pilots, but yeah, turns out we're not all cut from the same cloth just because we're all capable of dying.

This is why I fly a single fixed gear.

 

[Checkout_my_Six]

A twin engine Aircraft has a higher probability of engine failure than a single.... The bigger question, is the pilot capable of dealing with that.

 

[hindsight2020]

A twin engine Aircraft has a higher probability of engine failure than a single.... The bigger question, is the pilot capable of dealing with that.

accounting for demographics and median outcomes, part 25 yes, part 23 no.

 

[RotorDude]

If the two engines were truly unrelated and unconnected, e.g. installed on different planes and flown by different pilots in different parts of the country, then you'd have double the failures with both of them compared to a single one.
But if they are installed on the same plane, maintained by the same shop, flying the same routes and operated by the same pilot(s), then those common factors would make the combined failure rate higher than twice the single rate, since most failure causes are related to the operator or installation, or would otherwise affect both engines.
Still, even if more likely to happen, the typical failure would be one-sided, allowing a well-trained pilot to land safely.

 

[PPC1052]

Take your coin flips. Flip coin ONE 10 times in one minute. You get on average 5 heads.
Now take ANOTHER COIN and flip it 10 times in the same minute. You get on average 5 heads.

Now, how many heads will you get, on average, in in that minute????

Just flip two coins, once each. The chance of getting tails (a failed engine) on at least one coin is not 100%.

 

[EdFred]

Look you have ONE engine. It will fail, on average, 10 times in 1 million hours.
Now you have another engine exactly the same. It will fail, on average 10 times in 1 million hours.

The engines failures are NOT dependent on the other engine. They fail independently.

Lets say the engines aren't even in the same plane!
Say they are two engines in two different single engine aircraft!!!!!!

How many times will an engine (either one) fail in 1 million hours?
The correct answer, based solely on the information given, is 20 times.
Why would it be anything else?

That is as simple as I can explain it.

Explain it all you want, but you are still explaining it wrong.

 

[coloradobluesky]

I see your point though. To be precise, you have to say the LIMIT is twice as likely. Run the experiment an infinite number of times and it will be twice as likely. But the reality is, you can't do that, so it is slightly less than twice as likely. (Thats why I said, on AVERAGE).

The reason I brought it up is people use the argument that MOST of the time the remaining engine will get you back to an airport safely and they throw in the likelyhood that loosing two engines in one flight is miniscule (which is true). But they neglect to consider that since you have two engines, you are more likely to have an engine failure than if you had one.

 

[PPC1052]

If the two engines were truly unrelated and unconnected, e.g. installed on different planes and flown by different pilots in different parts of the country, then you'd have double the failures with both of them compared to a single one.
But if they are installed on the same plane, maintained by the same shop, flying the same routes and operated by the same pilot(s), then those common factors would make the combined failure rate higher than twice the single rate, since most failure causes are related to the operator or installation, or would otherwise affect both engines.

How do you know that it won't be lower? What if that pilot is particularly gentle on plane engines, or that shop is particular good at engine repair?

 

[EdFred]

Take your coin flips. Flip coin ONE 10 times in one minute. You get on average 5 heads.
Now take ANOTHER COIN and flip it 10 times in the same minute. You get on average 5 heads.

Now, how many heads will you get, on average, in in that minute????

I just did that and got 6 heads on the first set and 5 heads on the second set.

You just made my point.

 

[EdFred]

I see your point though. To be precise, you have to say the LIMIT is twice as likely. Run the experiment an infinite number of times and it will be twice as likely. But the reality is, you can't do that, so it is slightly less than twice as likely. (Thats why I said, on AVERAGE).

The reason I brought it up is people use the argument that MOST of the time the remaining engine will get you back to an airport safely and they throw in the likelyhood that loosing two engines in one flight is miniscule (which is true). But they neglect to consider that since you have two engines, you are more likely to have an engine failure than if you had one.

That is correct, but not twice as likely.

 

[PPC1052]

I see your point though. To be precise, you have to say the LIMIT is twice as likely. Run the experiment an infinite number of times and it will be twice as likely. But the reality is, you can't do that, so it is slightly less than twice as likely. (Thats why I said, on AVERAGE).

What you really mean (and I don't dispute your main point) is that as the chance of an any particular occurrence happening goes down to a small percentage (as in the chance of an engine failure) that math works out such that it is functionally the same as doubling the chance when you add a second enging. And for purpose of our discussion, your math is close enough to be correct in every sense that matters.

 

[coloradobluesky]

Yes, I know someone that took a 172 out after it had both its magnetos service and they BOTH failed. (excess grease or some such). But we are talking INDEPENDENT events. In that case, they both got improperly serviced so they were DEPENDENT. And the reality is there are LOTS of dependencies. But those are really hard to quantify.

 

[hindsight2020]

Alex, I'll take "single T-props" for 200. I'm reckless like that

 

[RotorDude]

How do you know that it won't be lower? What if that pilot is particularly gentle on plane engines, or that shop is particular good at engine repair?

Oh, for sure it will have an effect in both directions. But from the statistical POV, you can't possibly do better than having two completely independent engines, so therefore, by the virtue of being dependent, you'd be worse. How much worse is hard to know or measure, since twin failure rates are not reported (unless followed by an accident or incident).
I think there is little doubt that the pilot's proficiency and training level determines the bottom line stats.

 

[Checkout_my_Six]

Oh, for sure it will have an effect in both directions. But from the statistical POV, you can't possibly do better than having two completely independent engines, so therefore, by the virtue of being dependent, you'd be worse. How much worse is hard to know or measure, since twin failure rates are not reported (unless followed by an accident or incident).
I think there is little doubt that the pilot's proficiency and training level determines the bottom line stats.

ding....ding....ding......winner....winner .....Chicken dinner.


even with single engined aircraft it's the pilot who screws up +80% of all accidents....not a mechanical failure.

 

[Checkout_my_Six]

That is correct, but not twice as likely.

yes....you are correct. More exposure (more engines) increases the probability of the failure event occurrence.

Now....put four engines on that bad boy....and guess what....your probability of an engine failure increased proportionally.

 

[hindsight2020]

yes....you are correct. More exposure (more engines) increases the probability of the failure event occurrence.

Now....put four engines on that bad boy....and guess what....your probability of an engine failure increased proportionally.

which is why industry went away from the 4 engine mode into higher efficiency twins, to much success and economy of operation. I know a bit about that trade with 1000 hours in the B-52. That POS splitting 17k per engine when commercial options to adequate clearance and cross section requirements increase the total thrust and reliability at a fuel consumption decrease of game-changing proportions. Even still, I had zero engine failures and one precautionary due to an oil pressure sensor. That's pretty good considering that many units running at once.

The problem with part 23 twins still remain. They're turkeys. Power redundancy in the absence of real power surplus is a false economy. These things are too underpowered, and that comes as a function of the powerplants being too clunky for the power output. I suppose one can make it an indirect effect of fuel prices, but these things were designed underpowered even in the days of a nickel for your gallon gas.

 

[dmspilot]

If the probability of an engine failure was 50%, then the probability of having an engine failure on a twin would be
0.5 + 0.5 - (0.5 * 0.5) = 75%

That's the same as getting at least one heads in two flips of a coin. Except engines aren't coins.

If the probability of an engine failure was 0.01%, the the probability of having an engine failure on a twin would be:
0.0001 + 0.0001 - (0.0001 * 0.0001) = 0.019999%

It is so close to double that it's a distinction without a difference.

edit:
Here is a slightly more intuitive way to calculate it. Now that I think about it, this is the way I remember learning it in stats class. In two coin tosses:
P(at least 1 heads) = 1 - P(zero heads)
P(at least 1 heads) = 1 - (0.5 * 0.5)
P(at least 1 heads) = 0.75

If the probability of an engine failure is 0.01%, the probability of the engine not failing is 99.99% (0.9999). The probability of neither engine failing in a twin would be 0.9999*0.9999, and so:
P(at least 1 engine fails) = 1 - P(zero engines fail)
P(at least 1 engine fails) = 1 - (0.9999*0.9999)
P(at least 1 engine fails) = 0.00019999 = 0.019999%

 

[dtuuri]

If the probability of an engine failure was 50%, then the probability of having an engine failure on a twin would be
0.5 + 0.5 - (0.5 * 0.5) = 75%

That's the same as getting at least one heads in two flips of a coin. Except engines aren't coins.

If the probability of an engine failure was 0.01%, the the probability of having an engine failure on a twin would be:
0.0001 + 0.0001 - (0.0001 * 0.0001) = 0.019999%

It is so close to double that it's a distinction without a difference.

I am so glad you posted this. My math skills aren't adequate to argue with the previous posters, but I knew something didn't ring true. You said this very well, thanks.

dtuuri

 

[Kristin]

Indeed. The "instructor" (I'm betting he was really a salesman) also said to "keep it above blue line". Well, blue line is a takeoff reference speed, clean, at gross takeoff weight, not a landing weight reference speed, dirty. A lot of CFIs (nearly all?) continue to propagate this misconception. There's no way that plane could climb at that speed in that configuration at blue line.

As for the windmilling prop--an engine loss on final is no reason to mess with anything. The risk of botching either the landing or the handling of the engine failure is just too great in the limited time available. The exercise should have been terminated when the landing was assured. As you point out, no twin engine airplanes are certified for balked landings with an inoperative engine.

dtuuri

That "maintain blue line until the runway is made" is a dangerous but prevalent orthodoxy. It is one of the reasons for LOC accidents on landing as the airplane is invariably too hot when crossing the runway. An engine failure on final should get no more response then running up both throttles to make it to the runway.

It this case, it appears that the engine failure happened at some distance from the airport and they were making a SE approach and landing. In such a scenario, the engine out should have been a zero thrust, not windmilling.

 

[GRG55]

Interesting thread.

Yup. In about another 5 posts I am going to sell the Aztec and buy a plastic airplane with a red handle...

 

[Kritchlow]

That "maintain blue line until the runway is made" is a dangerous but prevalent orthodoxy. It is one of the reasons for LOC accidents on landing as the airplane is invariably too hot when crossing the runway. An engine failure on final should get no more response then running up both throttles to make it to the runway.

It this case, it appears that the engine failure happened at some distance from the airport and they were making a SE approach and landing. In such a scenario, the engine out should have been a zero thrust, not windmilling.

Yes, it should have been in this particular scenario.

That said, I know this is mostly a small airplane forum so I don't want to derail things.
But just for general knowledge, things such as red line and blue line (heck, yellow arcs and white arcs) don't exist on larger airplanes. That is not say we don't have flap speeds and single engine speeds, it's just that the nomenclature is different.

 

[brian]]

Pfft. I just pulled out my discrete math book for some light reading...

 

[GRG55]

Look you have ONE engine. It will fail, on average, 10 times in 1 million hours.
Now you have another engine exactly the same. It will fail, on average 10 times in 1 million hours.

The engines failures are NOT dependent on the other engine. They fail independently.

Lets say the engines aren't even in the same plane!
Say they are two engines in two different single engine aircraft!!!!!!

How many times will an engine (either one) fail in 1 million hours?
The correct answer, based solely on the information given, is 20 times.
Why would it be anything else?

That is as simple as I can explain it.

You folks are arguing completely different things. Probability theory and measured mean time between failures (above assumption) are quite different animals.