Do pilots in modern jets have failsafe instruments to fall back on? (Air France 447 revisited)

[Broomstick]

@Whack-a-Mole

So, apart from the glass cockpit are there no simple mechanical devices that work no matter what in the cockpit? (Thinking of that little ball floating in liquid which shows the pitch and roll of the plane.)

If I understand which “little ball” you’re talking about in fact it says nothing about pitch, just roll. It’s called the turn and bank indicator.

Wouldn’t that floating ball have told an undeniable truth to the pilots (i.e. the nose of the plane is pointed up)?

Nope.

Even though it’s a “simple mechanical device” there are instances where a T&B can give inaccurate or misleading information.

Just wondering why a simple fallback that simply cannot break or at least works on simple physics that won’t go wrong isn’t used (or if there then why was it not paid attention to).

There are such instruments but, because they aren’t the main ones used in modern flying, usually aren’t terribly conveniently located in the cockpit of a large airplane like that. Also, one group of such devices, called “pitot tubes”, had iced over and failed, precipitating the chain of events leading to the accident.

@md2000

Several other things -
apparently the pitot or other controls were frozen over with icing(?) which can happen.

True. This is arguably the precipitating event for the subsequent chain of events leading to the accident.

the autopilot got confusing readings from the frozen-over controls, and slowed the airplane into a stall, then dropped off and let the pilots handle the mess.

Um… more or less. When certain parameters are exceeded autopilots are designed to disengage and let the humans handle the problem because humans are better at handling unanticipated outliers than machines are (machines excel at routine flying)

@Machine_Elf

You have no airspeed measurement, and you can’t see the real horizon outside. This plane, unfortunately (and astonishingly, IMHO), is not equipped with an angle-of-attack indicator, which would unambiguously tell you what you need to know, i.e. the aircraft is in a stall so deep (AOA = ~40 degrees) that no one has ever flown a situation like this in the real world or in a simulator.

My information is that the airplane did have an angle-of-attack indicator but that it only communicated directly with the on-board computer, it did not provide a display to the pilots. Why? Damn if I know. It’s a bit of a mystery why direct angle-of-attack indicators are not more common on all sorts of airplanes. Typically pilots rely on the airspeed indicators for indirect AoA information but with the pitot tubes frozen-over they weren’t getting good/reliable airspeed information.

@clairobscur

Why would it have to be so complicated? Wouldn’t a mere lamp switching on when the other pilot is using the joystick have been sufficient in this case? The pilot would then have noticed that the co-pilot was doing something and could have just asked him what.

The problem is that when the shit hits the fan people have a tendency to try to focus in on some information and ignore what the brain thinks is less important, whether it actually is or not. Cockpits already have a plethora of warning lights, buzzers, bells, and such, to the point pilots have reported all the alarms as more distracting than useful.

Having the pilot’s controls directly reflect what the other pilot is doing with his controls, as happens with mechanically linked controls such as the more traditional yokes, is a little harder to ignore than one more light among many.

@Leaffan

Richard, I’m familiar with your credentials. It seems somewhat intuitive to a non-pilot like me that a plummeting altitude might possibly be a symptom of a stalled aircraft. In your opinion, why wouldn’t the co-pilots have recognized this after some of the futile corrections they attempted?

The problem – one of them – might have been that by the time the pilots figured out they were stalled and falling it might have been too late or too extreme a condition to do anything about it.

@GusNSpot

Have they reproduced the flight in the sims? Was it recoverable & how long did they have to start the recovery? Have they loaded up a real plane, assuming the problem was recoverable, and gone poking around in the corners of the box?

Gus, based on the black box information I looked at a few months ago, it looks like they had pitch angles exceeding 35 degrees and alternating left and right banks up to 40 degrees. It might be hard getting volunteers for that experimental flight in a non-aerobatic airplane, and I have doubts it was recoverable at all. Admittedly, I’m no expert on that category of airplane (arguably, not an expert on any airplane) but it does strike me as rather extreme attitudes for a passenger airliner.

@md2000

One critical point was that the stall warning shut off (IIRC, past a certain speed?).

I think it was more a matter of the data being so confusing the computer gave up and shut down.

@Machine_Elf

So the pitot tube froze up. Shit happens.

Except… having all three pitot tubes freeze up is highly unusual shit. The pitot tube affects the airspeed indicator which is a pretty important instrument for basic flight.

While the pilots do bear some blame and at least in theory could have performed better I think non-pilots discount just how confusing flying through a severe storm is, and how reliant the pilots are on the instruments when flying in a bad storm, at night, over the open ocean.

I’d like to see them try this on a simulator as see if any team of pilots could have coped with the situation, and if so, how.

[md2000]

I have flown in bumpy enough weather (not a storm, just really really bumpy, ground still visible) where it was impossible to get your heading from the (mechanical) gyro or the compass to within 20 or 30 degrees. From an earlier thread, it was suggested that when a plane this big stalls, it’s shaking like a carnival ride. Mechanical backup indicators can only do so much - although it’s surprising they don’t have something basic like a backup angle of attack indicator.

I don’t suggest the sticks should be solidly linked, but if there were some “push” or springy resistance feeling on the stick to indicate “you are pushing against the other stick” they might have clued into the fact that their controls were cancelling each other out. For the safety report to blame pilot error because they did not happen to also look at one indicator (or two?) during a wild carnival ride of shaking, is a cop-out.

Yes, the other most serious (and aircraft-related) failure was the stall warning that turned off. I suppose that detail had been mentioned in passing with a thousand others in training; but in a real-life situation who has time to go though all those details? Basically the stall was off, they put the nose down, the stall warning comes on, so they pull back again it goes off, and they never recover. I suppose that’s a design decision- what’s worse, a stall warning when there is no stall or a stall with no warning? Or a stall warning that comes back on as you get close to recovering? Ah, engineers… unlike pilots, their decisions don’t kill them along with their victims.

[Xema]

@Broomstick

… I have doubts it was recoverable at all.

I think the evidence suggests it was. The “aft stick” control input was at one point relaxed (briefly), and the plane responded by reducing its AoA and increasing its airspeed. This caused the stall warning to again start functioning, and the PF again went to full aft stick.

It seems very likely that without the return to full aft pitch input the plane would have continued to decrease AoA and could have readily resumed normal flight.

[RedSwinglineOne]

@Machine_Elf

The aircraft designers figured nobody would ever put the plane in a flight configuration like this, so they programmed the computer to assume that wild inputs like this were invalid, and so the stall warning did not sound -

This is understandable, but it would have been simple enough to program the stall warning such that once the stall warning comes on, returning to “normal” flight is the only way for it to turn off. Going from stall to “invalid” should not have triggered the stall warning to stop sounding.

As far as the pilot controls, who should have been hands on the controls? You would expect that the pilot and copilot must often decide “whose driving” the plane. In an emergency you would think one should have been at the controls while the other concentrated on the instruments. What is the protocol in this situation?

[Xema]

@md2000

Basically the stall was off, they put the nose down, the stall warning comes on, so they pull back again it goes off

Right. But since when is aft stick a valid response to a stall warning?

An untrained pilot might be excused for reducing the problem to “Stall warning = Bad - let me see if I can get it to shut off.” A competent pilot should be able to do better.

[md2000]

However they got there, they were in a situation where the stall warning was off. At that point, climb above the storm problem was a valid target; full power and pull back are valid ways to do that, unless it sets off the stall warning. Climb (if you are not in stall) is probabaly a good idea if the altimeter says you are going down - or possibly stuck in a really strong downdraft, another possibility in a turbulent thunderstorm.

The copilot was guilty of not seriously thinking out the situation, but the basic reaction is not completely bizarre and in a way makes sense. The biggest mistake IMHO was the stall alarm turning off at the wrong situation, adding to the confusion.

Also, as someone pointed out in the other thread, if the stick is held completely back it triggers a mode where supposedly the aircraft will climb without stalling to maximum altitude - but only in the “standard law” mode and they were not in it. Perhaps the copilot remembered that tidbit from his training.

[mnemosyne]

@GusNSpot

Have they reproduced the flight in the sims? Was it recoverable & how long did they have to start the recovery? Have they loaded up a real plane, assuming the problem was recoverable, and gone poking around in the corners of the box?

Reproducing the flight in the sims would reveal some information, but not much, since the entire series of events was outside the flight envelope: the sim would provide merely theoretical behaviours and cockpit “symptoms”, but would be doing so based on mathematical models, not actual aircraft data. When sounds and motion are simulated in a Cat-D simulator, they are derived from the plane’s behaviour on test flights where they might go right up to some buffering and a recoverable stall, but not further. Any simulation beyond that point is, as I said, theoretical.

In cases like these, using a simulator with a preprogrammed failure (like the series of events in this crash) could tell you a lot about how pilots react to all the information being fired at them, but won’t necessarily be a reliable measure of how recoverable a situation could be.

[Richard_Pearse]

@Xema

Right. But since when is aft stick a valid response to a stall warning?

When you think the aircrafts computer will sort out the finer points for you? That’s not an excuse for what happened but it may be a factor in how the pilot was thinking.

Incidentally full aft stick is what the Colgan Dash 8 Captain did when faced with a stall warning. I think there is a real problem in aviation with how stall training is treated. In the Colgan accident investigation it was found that US pilots were being trained in stall recovery with too much emphasis being placed on minimum height loss rather than reducing the angle of attack. I also know from my own experience that stall training tends to be done an unrealistic way. My airline stall instruction has generally been a case of looking at how the stall warning systems work and then practicing the recovery procedure by deliberately slowing the aircraft down in level flight. The problem is that stalls in the real world don’t happen by deliberately slowing down in level flight.

It is very hard to surprise someone in the simulator. For starters you know that something is going to happen, you don’t just go to the sim to fly an uneventful pretend flight from A to B. This means crews in the sim are at least ready for something to happen and will probably have studied likely scenarios in the lead up to their recurrent training. Also realistic stall scenarios generally require the crew to fail in some way rather than the aeroplane, how do you realistically simulate an initial piloting failure? The problem might also be that stalling incidents are so rare that deficiencies in training are rarely highlighted.

[Machine_Elf]

@mnemosyne

Reproducing the flight in the sims would reveal some information, but not much, since the entire series of events was outside the flight envelope: the sim would provide merely theoretical behaviours and cockpit “symptoms”, but would be doing so based on mathematical models, not actual aircraft data.

They’ve got a flight data recorder full of data (except for a short time when airspeed was not available).

[Richard_Pearse]

@Machine_Elf

They’ve got a flight data recorder full of data (except for a short time when airspeed was not available).

That’s only a very narrow set of data that shows what happens in response to the control inputs of the crew at the time. How does the sim know what to do with a different set of control inputs?

The basic problem is that training aircraft with straight wings stall in a nice predictable way while swept wing jets don’t. The stall behaviour in swept wing jets is bad enough that no one wants to go up and fly fully developed stalls, so there is no data aside from the extremely limited data gained from an accident.

Edit: Going against the trend, the Boeing B747 was originally manufactured without a stick pusher because Boeing believed its stalling characteristics to be benign enough that it didn’t warrant it. When the British CAA certified it however, they demanded a stick pusher.

[Xema]

@Richard_Pearse

When you think the aircrafts computer will sort out the finer points for you? That’s not an excuse for what happened but it may be a factor in how the pilot was thinking.

Or not thinking.

It may be increasingly common, but surely it’s a poor pilot who says, in effect, “I don’t need to know how to fly this plane - that’s the computer’s job.”

[coremelt]

@Richard_Pearse

The basic problem is that training aircraft with straight wings stall in a nice predictable way while swept wing jets don’t. The stall behaviour in swept wing jets is bad enough that no one wants to go up and fly fully developed stalls, so there is no data aside from the extremely limited data gained from an accident.

correct me if I’m wrong but aren’t modern computers for flight simulators powerful enough to model the airflow over the wing surface in realistic ways?

[williambaskerville]

@md2000

the autopilot got confusing readings from the frozen-over controls, and slowed the airplane into a stall, then dropped off and let the pilots handle the mess.

That’s not true.

Read the final report, from around page 171.

http://www.bea.aero/en/enquetes/flight.af.447/rapport.final.en.php

The airplane was not in a stall when the autopilot disconnected, it was put into one by the control inputs of the pilot after the disconnect.

Had the pilots done nothing whatsoever in the 5 minute interval after the episode started, the plane would likely have landed safely and we woudn’t be having this conversation.

[Richard_Pearse]

@coremelt

correct me if I’m wrong but aren’t modern computers for flight simulators powerful enough to model the airflow over the wing surface in realistic ways?

For starters, flight simulators and the computers in them are generally only as modern as the aircraft they simulate and the A330 is a fairly old design. More to the point, I don’t believe they can practically model actual airflow, but I don’t have any cites for that.

[ZenBeam]

@Machine_Elf

While you are trying to figure this out, it would help to know that your idiot copilot has been pulling back on his joystick the whole time.

I wonder everytime I read something like this. Is there any sort of failure that could have happened that made it seem like the copilot was pulling back on the stick when he wasn’t really doing so.

An example would be if there were opposing springs pulling the stick forward and backwards, and the one pulling forward broke or came off. The black box would record the copilot doing the pulling, but it wasn’t him, it was the remaining unopposed spring.

Otherwise, is there any explanation for why a pilot would ever do that?

[robby]

@ZenBeam

I wonder everytime I read something like this. Is there any sort of failure that could have happened that made it seem like the copilot was pulling back on the stick when he wasn’t really doing so.

An example would be if there were opposing springs pulling the stick forward and backwards, and the one pulling forward broke or came off. The black box would record the copilot doing the pulling, but it wasn’t him, it was the remaining unopposed spring.

Otherwise, is there any explanation for why a pilot would ever do that?

Well, less than one minute before impact, the following exchange took place:

Co-pilot in the left seat: “Climb climb climb climb”
Co-pilot in the right seat: “But I’ve been at maxi nose-up for a while”

(From this source.)

So it looks like the co-pilot in the right seat acknowledged that he had been deliberately pulling back on the stick for some time.

[lawoot]

As an aside, I was chatting with a couple of airline (Delta?) employees the other day, and according to them one of the common sayings they have is “If the plane’s not a Boeing, I’m not going.” They really had NOTHING good to say about Airbus products. (of course, the fact that we were in Seattle may have had something to do with that )

[Baron_Greenback]

@lawoot

As an aside, I was chatting with a couple of airline (Delta?) employees the other day, and according to them one of the common sayings they have is “If the plane’s not a Boeing, I’m not going.” They really had NOTHING good to say about Airbus products. (of course, the fact that we were in Seattle may have had something to do with that )

Boeing products are, of course, fault free.

[Richard_Pearse]

@ZenBeam

I wonder everytime I read something like this. Is there any sort of failure that could have happened that made it seem like the copilot was pulling back on the stick when he wasn’t really doing so.

Not that I know of. In the Colgan crash the flight data recorder recorded a constant strong pull on the captains controls, that can only be caused by something physically pulling back on the yoke.

Otherwise, is there any explanation for why a pilot would ever do that?

They’ve either misdiagnosed the problem or have been poorly trained in stall recovery.

[friedo]

@lawoot

As an aside, I was chatting with a couple of airline (Delta?) employees the other day, and according to them one of the common sayings they have is “If the plane’s not a Boeing, I’m not going.” They really had NOTHING good to say about Airbus products. (of course, the fact that we were in Seattle may have had something to do with that )

Delta’s got over 150 Airbuses in their fleet, though, including more than 30 A330s. I fly on them regularly.