The whole and entire point of Airbus’ design and training philosophy is to make the airplane do #3. And every time that malfunctions, the poor sap who was trained that *max thrust plus full aft stick = automatic recovery & total safety *drives the malfunctioning jet into the ground.
It’s apparent (at least to me) that this systemic approach creates a brittle failure mode. The complete man/machine system doesn’t bend and recover; it simply snaps.
There is more and more of this going into the industry, and not just in the Airbus flight control logic. We’re eliminating a lot of the minor screw-ups that lead *towards *accidents and replacing them with fewer larger screw-ups that *are *accidents. What’s lost along the way is the practical hands-on learning that comes from making, or watching the pilot alongside you make, the little screw-ups and the recoveries therefrom.
Again, the explosion in air traffic in this part of the world has come at a cost to pilot training. Shortcuts have been taken and still are. It’s a little scary.
Agreed. The easier you make flying, the less capable we become at handling a difficult situation.
It’s worth noting though that we only hear about the occasions when it goes horribly wrong, we don’t hear about all the times the crew handle the situation well. In fact a lot of the time when the crew does the right thing even they might not remember it as anything special. A flight computer glitch followed by competently flying the aeroplane is a complete non-event.
Page 66 of the report (68 of the linked pdf) has a pic of one of the circuitboards in question in the middle of a section discussing the solder issue (apparently a known risk for years and various correction advisories issued). Not clear to me due to lack of the technical knowledge if what they are talking about are solder failures of connections TO the board or of connections IN the board, it seems the later.
IIRC in the early days of FBW airliner development the fear was that the computer would decide to keep flying straight and level into the mountain, or refuse to increase thrust as the ground came closer, regardless of what the humans wanted. But ISTM now what makes news, is when in some critical cases the meat units are not recognizing until too late when the automation no longer has their backs.
When I was a student pilot doing stalls I got a little carried away. upon the initial snap of the stall I leveled it with the rudder, added power… and pointed the plain DOWN. Well it only takes a few seconds to go from 45 knots to VNE and my instructor tried to pull back on the yoke. Not happening. We’re both alive because he gave me one chance to fix it and he then hit me hard enough to break my grip on the yoke.
This was a plane full of people. The Captain should have knocked him out cold if necessary and that should be part of the training.
True all but we also don’t hear of the stupid things that happen and go unreported.
It’s scary when professional pilots can’t handle a stall. I’m not too keen on the idea of taking a person from zero to commercial airliner without logging a lot of time in the real world.
Also true, though if it is stupid enough the passengers will notice and report it.
Not really. The stick shaker and pusher are intended to mimic traditional stalling characteristics (stick shaker = buffet approaching the stall i.e., stall warning, stick push = nose drop from loss of lift i.e., stall identification.) They are installed on aircraft that don’t naturally have those aerodynamic properties.
They are designed to help a pilot detect and prevent a stall but they can be over-ridden. The Colgan Dash 8 pilot held the stick back against the stick pusher all the way to the ground.
The Airbus FBW aircraft, on the other hand, when operating in “Normal Law” will prevent a stall by effectively saying “I’m sorry Dave, I’m afraid I can’t do that.”
We’re in an Airbus… we have a stall condition. We have inputs from the crew… we have an airliner in flight. Whether 1 and 2 have happened are irrelevant now… (although as practice, there should still be strong diligence for these, not less because of technology.)
What we need now, in flight, is some jump forward for 3.
Correct me as needed, but if an airliner is capable of alerting to a stall, and I believe they are doing just that in these scenarios where the co-pilot or pilot are pulling back, and we’re experiencing inputs that would not rectify the stall, but the system is privy to another set of inputs that would stop the stall (or increase the odds of stopping the stall, since nothing is guaranteed as an outcome), in a stall situation, under X parameters, the control of the airplane can be given to the crew member whose inputs make the most flight sense ?
System sees two conditions. System is alerting to stall. Control granted to member whose inputs are most likely to cancel the stall/warning.
The problem with the Airbus is that once the computer making all these decisions has begun to malfunction, which it did, then how are you the pilot, or you the avionics engineer, going to decide that the next decision it makes about which pilot to trust is itself trustworthy?
In the case of AF 447 they had times when the aircraft thought it was stalled when it wasn’t and times the aircraft thought it wasn’t stalled when it was. It was at least as confused as Bonin, and probably more confused than the other two guys.
The whole and entire point of this systemic problem is that there is heavy training that the correct stall recovery in an Airbus is … stalls are physically impossible; the computer prevents them from happening, period, amen.
And once that computer has malfunctioned you’re left with the pilots’ backup training which insists the way to avoid any crash for any reason is full power plus full aft stick and trust the computer to save the aircraft. Which it duly fails to do.
The fix is to unstack this stupid stack. Not to try to invent another hardware/software patch to add more magical smarts on top of a logically defective and actually malfunctioning stack.
We are not at the state of the art to remove pilots completely. Trying ever harder to reduce them to passengers except when HAL becomes grossly (or subtly) confused fails as well.
We’re not at the point where we can remove human pilots. And we’re at a point where we know we are never going to remove the ‘computer’ pilot.
So, it’s more likely we’re going to improve computer systems, kick up safety a notch, but recognize rigor is still involved in training as long as the human pilots are in the cockpit.
Well, yeah… if a system is malfunctioning, of course the system can’t be relied upon for additional responsibility. This applies to both the human system and the flight system.
Computer can be told to never do X, it won’t unless it is broke.
Human can be told to never do X. But. He has been doing X for his whole life and is in a total panic, last ditch effort, he may well revert to his oldest and strongest memory, both mental & physical.
(Yes Martha, the body has a memory beyond … )
And that is when the tail breaks etc., because in the interest of making more $$$, the bean counters override the human ability to function perfectly until it breaks like a computer not realizing the human is not really broke, just different from the computer.
Something I don’t understand about AF447 and all these stalling incidents: Wouldn’t the artificial horizon indicator have shown the pilots that the aircraft’s angle of attack was way too high?
The artificial horizon does not display angle of attack. it displays nose up / nose down versus the horizon.
Without also knowing the flight path there’s no way to derive AOA from that info.
In the case of AF447 the nose was only slightly above the horizon at least a bunch of the time. IOW in a typical cruise pitch attitude. The problem was that at the same the flight path was like 30 degrees below the horizon and they were at very high AOA.
There are horizon indicators that incorporate AOA at least indirectly by showing what’s called a “flight path marker” or FPM. The difference between the FPM’s attitude and the indicated aircraft pitch attitude is AOA.
In the case of AF447 if they had the FPM showing it would have been buried at the bottom of the display while the pitch attitude showed more or less level. In the case of AF447 the AOA system was part of the stuff malfunctioning that triggered the whole accident sequence. So there’s no assurance they had valid info in front of them. And also no assurance they’d have been able to use it correctly had they had it.
To be sure, if you’re in an airliner and you’ve got the nose waay up above the horizon it doesn’t take an AOA gauge or FPM to know that pretty darn soon you’re going to be in a stall (if you’re not already in one).
An element of negative training is that we practice ground proximity escapes and windshear escapes at low altitude. In both cases the procedure amounts to full power, pitch up to 25-30 degrees & hold on as we power away from the ground. Wind shear escape is much more dynamic and may have large pitch excursions due to the insane turbulence & wind shifts. Whereas ground prox escape is smooth and low dynamics as long as you are, in fact, climbing more steeply than the terrain you’re pointed towards. (And even if not it’s still low stress until the rocks come through the windshield :eek:)
The negative training comes in because 10 degrees nose up leads very quickly to a stall at high altitude whereas 30 degrees nose up is safe at low altitude. So most of the “maneuver for your life” practice serves to get you comfortable at attitudes that are acutely and promptly dangerous at higher altitudes.
Pilots who have good “air sense” don’t get fooled. The other folks who went to airplane operator’s school then just sit there for years while HAL drives all day OTOH tend to blindly regurgitate what they practiced whether it’s applicable or not.
In addition to LSLGuy’s post, you can’t ignore the possibility that the angle of attack display may be incorrect. I saw a recent episode of Air Disasters about the crash of an A320 that was on an acceptance flight while being transferred from one airline to another. The plane had been repainted as part of the transfer and had been washed from high-pressure hoses. Water had gotten into the angle-of-attack sensors (which look like this) and once the plane got to altitude they froze in place. The flight computer recognized it was getting conflicting information and disabled much of the automation, but there was no accurate AOA data to display to the pilots.
Not necessarily. It means you’re losing altitude. No more and no less. Every normal descent from altitude has this feature.
You’re right that any and all unarrested descents eventually end at terrain / water level. The ones done properly have runways under them at the end. The others don’t. So it’s easy to tell the difference (at least in hindsight :))
I haven’t read all the details on Air Asia. I have read most of AF447. AFIAK it’s pretty clear both crews understood at a visceral level that they were in a bad and deteriorating situation. They knew they were descending and that those can only go on so long before the ground arrives.
What they apparently lacked was the ability to step far enough back from the scenario to see and understand the big picture, parsing out the accurate instruments from the lying instruments, pay attention to the relevant correctly behaving noisemakers and ignore the irrelevant and/or false noisemakers, then recall (or guess) the correct corrective, then apply it timely and skilfully enough to resolve the situation.
In a combination of startle, fear, and confusion they became overwhelmed and locked into a rigid fixed response. One that didn’t work.
There’s only so much time available for the sim and a lot of it is taken up by regulatory requirements, in other words only the usual unusual circumstances are practiced. Also, to get the very most out of the sim a lot of the sequence is briefed and known in advance so there is minimal “startle factor”. Could sim time be used better? Certainly. Can adequately prepare for every conceivable scenario? Of course not.
