I Too Found This Puzzling

The other day I was reading an article on the crash of Air France flight 447, discussing recovery of the black box (two years after the crash).  It was suspected that the air speed measurement devices may have failed, thus impairing the automatic pilot, but it was not understood why the pilots were unable to fly the plane manually.  Was something else going wrong?  Did automatic systems, operating off bad data, override manual controls somehow?

The article said that the black box showed the plane went into a stall, and the pilots spent much of the fall pulling back on the yoke to regain altitude.  This made zero sense to me.  A stall occurs when the wing is angled to steeply.  The wing generates lift because the air on top of the wing must follow a longer curve than the air on the bottom of the wing, and thus must move at higher velocity.  This higher velocity results in lower pressures.  In effect, the plane is sucked up.  When the wing is too steeply angled, the air on the top of the wing breaks away from the surface, and lift is lost.

It is therefore absolutely fundamental in stall situations to drop the nose.  This does two things -- it decreases the wing angle out of stall territory, and it increases speed, which also increases lift.

Apparently, the experts are just as befuddled as I was reading the data.  Dropping the nose in a stall is on the first page of pilot 101.  This is not some arcane fact buried on page 876 of the textbook.  This is so basic I know it and I don't have a pilots license.  But apparently the pilots of 447 were yanking up on the nose through the whole long fall to Earth.


  1. Doug:

    Hadn't the pilot also retired to the rear of the plane for some sleep time, and was not in the cockpit during all of this?

  2. Evil Red Scandi:

    Too soon to ask if they were surrendering to gravity?

  3. anoNY:

    Perhaps the airspeed indicator failure means that the airspeed indicated was higher than the true airspeed. If the pilots thought they had enough speed, would they assume they were in a stall?

    I personally have no idea since my only flight time was as a teenager on Top Gun for NES.

  4. Artemis Fowl:

    One of the things we studied in computer science at my college was the numerous ways software developers cause plane crashes by creating one of two problems.

    1. Systems that feed pilots the *wrong* data
    2. Systems that feed pilots the right data in such an obscure non-traditional fashion (ex: important details like altitude in the corner of a screen that's not default).

    I don't know about piloting much. For instance, I do not know if, in an airliner, it is possible to determine you're in a stall when the computer is telling you that you're not. What I do know is that there is a HUGE incentive for airlines, manufacturer's, and government to blame all crashes like this on pilot error and not faulty computers (since a pilot error is restricted to the now dead pilot)

  5. Jerry:

    Perhaps the copilots were mislead by the airspeed indicator, and didn't know they were in a stall? My (very layman's) understanding is that the angle that produces a stall is dependent on airspeed, and that the normal symptoms of the stall (buffeting and sluggish response from the controls) could have been masked by the storm they were flying in. According to wikipedia, most of the automatic stall detection and warning systems sound like they're tied into the airspeed indicator too.

  6. None:

    "The wing generates lift because the air on top of the wing must follow a longer curve than the air on the bottom of the wing, and thus must move at higher velocity. This higher velocity results in lower pressures. In effect, the plane is sucked up."

    A very common error (I was also taught it on my uni physics degree, and in my PPL theory course). The simplest counter example is that aircraft can easily fly upside down.

  7. Gloobnib:

    There is a decent NOVA special on this topic available on netflix streaming: http://www.netflix.com/Movie/Crash-of-Flight-447-Nova/70148706?strkid=143970750_0_0&lnkctr=srchrd-sr&strackid=673a4c2c70126d4b_0_srl&trkid=222336

    What I learned from the (I am not a pilot, but do fly RC planes and spend some time in FlightSim) special is that standard procedure for loss of all 3 pitot tubes (and hence airspeed indicators) is to increase throttle to x (85% of max?) and pitch up y (5 degrees, I believe). This puts the plane in a sort of "can't fail" flight characteristic where you will stay within the designated performance envelope for that aircraft; you cannot fly too slow because the engines are at 85%, you can't fly to fast because of the +5 degree pitch. I suspect the pilots were trying desperately to get that +5 degree pitch that the training manual called for.

    Also stated in the special is that the pilots were likely totally overwhelmed with alarms. So yes, AF, you are probably right that the software killed all of those people by not making the important things obvious.

    Finally, there is a section of the documentary that puts three highly-experienced trainers into a simulator and they re-create the sequence of events. It is probably key that the most senior pilot was asleep and the two less experienced pilots were at the yoke.

  8. L Nettles:

    Malcolm Gladwell wrote about pilot disorientation and "choking" in the context the the John F. Kennedy Jr. crash in What the Dog Saw. It was a very interesting piece.

  9. Bret:

    Yeah, the "equal transit-time" theory has been disregarded.


  10. Another guy named Dan:

    @Doug - Contrary to what you are implying, the first officers on the flight should have been capable of flying the aircraft even in an emergency situation. Aircraft Captains and first officers (most airlines do not use the terms pilot and co-pilot expressly for this reason) are trained to the same standards nad have the same qualification requirements, at least as far as the operation of the aircraft is concerned. The differences in duties primarily come from the fact that in addition to the operation of the aircraft, the captain is responsible for the conduct of the cabin crew as well as the comfort and safety of the passengers.

    There have in the past been reports and rumors that the flight control software used by Airbus aircraft would not accept inputs from the pilots that exceeded certain operational parameters, much as the ECM in your car will not allow you to drive your car at 140 miles an hour, even if the engine is capable of it. This is supposedly one of the factors that led to the famous A-300 crash at the Farnborough air show years ago.

    In the old days, a pilot would be able to determine that his plane was about to stall by the fact that the turbulent airflow caused by the boundary layer separation would cause the control surfaces on the wings and tail to vibrate, which could be felt through the stick. Since today's fly-by-wire systems have removed any direct mechanical connection from the controls to the control surfaces, the control systems actually contain servos designed to replicate this effect when the computer determines the aircraft is near a stall condition. The fly-by-wire systems also make it nearly impossible to "turn off the auto pilot", as
    the flight computer is needed in al regimes to translate control movement into control surface movement
    The primary indicators used to calculate if a stall situation is imminent would be airspeed and angle of attack (the difference in angle between the wing meeting the air and the direction the plane is travelling).

    I'm curious if the plane had an out-of-range CoG condition in addition to the loss of airspeeed indication, or if the auto-trim control stayed active during the event.

    Scenario: pito tubes ice over, and airspeed indication is lost. No electronic stick shaker cue or audible stall warning to alert aircrew to stall condition. Auto-throttle control reduces power to engines to prevent apparent overspeed condition as pilots pull back stick to try to gain altitude. By the time by the time the aircrew realizes that the flight control system is recieving bad data from the sensors, they have neither the airspeed or altitude to prevent the flight path from intersecting terrain.

  11. NormD:

    I followed this for a while on airliners.net, where professional pilots hang out. It seems that for large airliners at high altitude, stall recovery is not always the same as for small planes at low altitude. The solution is not always to put the nose down. Gaining too much speed can cause the wings to fail. I read all kinds of back and forth in the forum. What did seem clear is that the pilots did not understand that they were in a stall. These guys had thousands of hours of flying time and lots of time in simulators practicing all kinds of recoveries and yet they missed this simple fact about their situation. Clearly the crew got tunnel-vision. No one stepped back and thought for 30 seconds. They were flying through a lot of turbulence at night. They suspected that they were getting incorrect data from their flight instruments. I suspect that lots of alarms and warnings were going off.


  12. joshv:

    "“The wing generates lift because the air on top of the wing must follow a longer curve than the air on the bottom of the wing, and thus must move at higher velocity. This higher velocity results in lower pressures. In effect, the plane is sucked up.”

    A very common error (I was also taught it on my uni physics degree, and in my PPL theory course). The simplest counter example is that aircraft can easily fly upside down.

    Yep, lift is generated because the wing deflects air downwards. Every action has an equal and opposite reaction, thus the plane is deflected upwards.

  13. Flyfish:

    The pilots of the airbus are trained to believe that the aircraft cannot stall. Obviously they believed this.

  14. Dave:

    I can't claim to be an expert on this crash, but from what I know of previous crashes, it wouldn't be the first time that a sensor failure might have presented the pilots with conflicting information that might have confused them.

    For example Birgenair Flight 301, which crashed after take off from the Dominican Republic, the pilots received overspeed warnings(which because of a blocked pitot tube was reporting the plane was flying too fast), shortly followed by stall warnings (stick shaker) that the plane was flying dangerously too slow. When receiving conflicting warnings, it's not hard to see that a pilot is either totally confused or chooses to believe one warning over another.

    Having listened to cockpit voice recordings in situations like this I can say I'm glad to never have been in a situation where I've had to make a split second choice as to which warning to believe.

  15. Mark:

    I think this is similar to the steer into the skid principle. We are all taught to steer into the direction of the skid when driving to gain back control, but visually if you do this, it will look like you are about to steer the car into the snowbank on the side of the road. So instead of doing what is right and gaining control so you can get out of the situation, one instinctively turns in the opposite direction.

    Another case of this - in the days before ABS was Pump the brakes, cuz you slow down much slower in a skid - but you see that car/ wall coming up to you, it is very difficult to lift your foot from the brake and do the right thing.

    These guys - even with years of training to do the opposite, could probably tell you that when in a stall you need to point the nose down to gain speed and get out of the stall situation. But their instinct took over, and the primitive brain said, your falling, pull up, pull up. and that is what they did.

  16. AnObserver:

    It's hard to fathom trained pilots countering their training and stalling their airframe. I'd guess there's always loads of contributing factors, too. Another example: http://en.wikipedia.org/wiki/Colgan_Air_Flight_3407

  17. T J Sawyer:

    "the wing generates lift because the air on top of the wing must follow a longer curve than the air on the bottom of the wing, and thus must move at higher velocity."

    I was surprised to see that on Coyote's post. Sort of like reading, "global warming, resulting from the burning of fossil fuels..." isn't it?

    But it got me to wondering, when did the general science textbooks in high schools finally drop the "suck the wings up" explanation of lift or have they? I recall my science teacher responding very badly to my bringing a model P-51 to class and asking how it could get off the ground. This was back in the early 60's and would seem to indicate about 20 years of experience with a symmetrical (top and bottom) camber on a wing at that time. If I recall correctly, he told me the model wasn't accurate. Wish I'd though of the "upside down" argument at the time.

    One of many things that has made me suspicious of "Generally known stuff" of any sort.

  18. Brian Dunbar:

    But apparently the pilots of 447 were yanking up on the nose through the whole long fall to Earth.

    They continued to apply the wrong input as the plane flew into the ground.

    Why not? A majority of us know how to apply the right action to correct a government and economy. Yet we collectively insist on applying the wrong input even as we speed into the ground.

  19. Shawn:

    Coyotes explanation of lift is correct. Even upside down the airplane will require a high enough angle of attack to "suck" it up. This is usually a significantly higher nose up attitude than when it is right side up. This depends on the wing design as some airplanes are designed to fly upside down as well as up.

    The captain not being on deck is pretty irrelevant. The crew all receives the same training and generally the only difference is the hire date of the pilots. Most airlines upgrade based on seniority with the company which has nothing to do with skill level.

    It seems impossible to know what really happened. It is hard to believe that any crew would not be able to recognize and recover from a simple stall. It's possible the crew thought they were in mach tuck which has some of the same symptoms (loss of altitude) but has totally different recoveries.

  20. Punkster:

    Not all airplanes can fly upside down. Military craft do it because their wings do not supply much lift in normal operations, and are used more for stability. They are effectively rocket powered darts.

    To go slow (intercepting a Cessna private craft, for instance)they have to extend their ailerons out to an extreme position and fly with their nose point high up in the sky, to get the vector momentum from the engine thrust.

  21. LoneSnark:

    The summary I read said the cockpit voice recorded had the computer verbally issuing stall warnings the crew twice on the way down to the ocean. The airspeed indicators were malfunctioning, although it was not a static failure, as the reported air-speed fluctuated from zero to cruise, which should have screamed to the crew that the air-speed indicator should be ignored. A voice is heard saying "our instruments are wrong."

    It is the case they were in a storm at night and therefore had no visible cues available to fly the plane. After the autopilot shut off, the crew flew the plane to an ever higher altitude. Perhaps this was an attempt to get above the weather, or it is evidence that the crew was trying to fly the plane visually (ignoring all their instruments, not just the bad air-speed indicator) in a situation where no visual flight capability existed.

    A stall is associated with a backward pull to a stop while in free-fall until the falling aircraft reaches terminal velocity, an event which might have been overlooked by a pilot in severe turbulence consciously ignoring his instruments.

  22. Dan:

    Another crash where incorrect instrument readings played a key role was Air Florida Flight 90, which crashed into a bridge over the Potomac River in Washington, D.C., in 1982.

    In this case, cold weather, combined with the pilots' failure to turn on the de-icing system, led to ice covering the air speed indicators, resulting in the pilots getting incorrect speed readings in the cockpit as they sped down the runway for takeoff.

    The instrument readings indicated they had enough speed for takeoff, but the co-pilot told the captain that he sensed something wasn't right (in reality, they weren't going quickly enough). The captain ignored the co-pilot and tried to take off anyway. They immediately stalled and only reached about 300 feet in altitude before plunging into the bridge, killing all but 5 aboard the plane and also killing several motorists.

  23. HB Wise:

    Here is a great explanation of what may have occurred and why from an Airbus driver. http://www.avweb.com/eletter/archives/avflash/1925-full.html#204773

    Like you my thoughts went to why they didn't go to known attitude and power settings. This guy's letter goes a long way to answering that question for me.

  24. DensityDuck:

    "Wings generate lift because the air moves faster over the top surface."
    "No, that's not it, they generate lift by directing the air downwards."
    "Well, what causes the air to be deflected downwards?"
    "It moves faster over the top surface."

    If it's all about deflecting air downwards, then how does stall (flow separation) occur? Air would still be "deflected downwards" by a positive angle-of-attack.

    And "moves faster over the top surface" still applies when the wing is upside-down. It just doesn't work as well. And if the wing is somehow in a condition where it has a positive angle-of-attack relative to the airflow it will still generate "lift" just pointed downwards.

  25. caseyboy:

    Here is info copied from AINonline. Technology in the cockpit is wonderful, but apparently there is a downside when the software snags. Unfortunately there is no time for a reboot.

    One question lies with the trimmable horizontal stabilizer (THS). The critical phase of the flight, from autopilot disengagement to the crash, lasted 4 minutes 23 seconds. During the last 3 minutes 30 seconds, the position of the THS went from 3 degrees to 13 degrees nose-up and then remained unchanged. Yet, from about 2 minutes before the crash, the pilot flying switched to pitch-down inputs.

    So, as French website aerobuzz.fr pointed out, why did the THS stay in such a nose-up setting? This may hint at the flight control law being no longer “normal” but in a mode (“alternate” or “abnormal”) where the autotrim function is deactivated. In that instance, the crew has to trim the stabilizer manually. In an A320 accident that took place near Perpignan, France, in 2008, the crew’s failure to recognize this situation contributed to the catastrophic chain of events. In the case of AF447, the crew did mention “alternate law” in the recorded conversation.

    The information released May 27 is far from complete–for example, still unknown is exactly what the crew could see on the cockpit displays. The number of questions raised by the BEA’s document speaks volumes about the complexity of what happened off Brazil’s coast. The investigators arguably need months to reach a fair understanding.

  26. FormerFlyer:

    Part of the problem with this discussion is that the general public has little understanding of the sheer complexity of modern transport category aircraft, and the nightmare that occurs when lots of problems occur at once, and the immense difficulty involved in sorting out conflicting data sets in the minutes or seconds preceding an accident. We used to joke that the definition of pilot error was, "when the NTSB team takes 4 months in meeting rooms with computer simulations to come to a conclusion about the condition of the plane and a recommended course of action that the pilots were expected to make in 30 seconds, while upside down and on fire in bad weather."

    Doug is also the only one that correctly pointed out that an aircraft stalls based on the angle of the wind relative to the wing, AND NO OTHER FACTOR. At lower speeds the wing must be angled more to generate the same lift, which is why stalling usually occurs at lower speeds, but it has only to do with the angle of the wind to the wing, not the wing to the horizon.

    Doug is also right that for most pilots, dropping the nose is the Pilot 101 response to stalling. It is not, however, the PILOT 401 response to stalling in all conditions. In most modern transport category aircraft pilots are trained to minimize loss of altitude by maintain positive deck angle and to advance the throttles to power out of the descent and eliminate the sink rate, which will fly them out of almost all stalls and avoids the problem of "recovering" from the stall by nosing over into the ground. Different aircraft, different systems, different capabilities all add up to different procedures and different training.

    Whatever happened to 447, these guys were not stupid, were not uneducated and were not unfamiliar with the plane. It is likely that due to instrument or flight computer malfunction or damage to external sensors (or some combination), they had data sets showing different and probably impossibly conflicting conditions that the plane could have been in, selected the data set that made the most sense based on their understanding of the situation, and executed a course of action most likely to safely maneuver them out of harm's way. They picked wrong, and paid the ultimate price.

    This happens. This will always happen. Accidents like this can be made less likely through training, design and automation. Automation alone is not the answer. Ask any experienced pilot how often they have to correct what the computers screw up, and you'll be shocked. Ask the military how many drones they lose when the systems get crossed up and they simply crash, and you'd be similarly surprised.

    Pilot's Prayer:
    "Dear Lord, please don't let me screw up, but if I do then Dear Lord, please don't let me live through it."


  27. FormerFlyer:

    Almost all the lift comes from the wings, and almost all fixed-wing aircraft can fly upside down for a while, aerodynamically anyway. The engines frequently can't run upside down (fuel and oil delivery are the usual culprits), and the wings are biased toward flying efficiently right side up, so there are limitations, but I can't think of a single fixed wing airplane that can't sustain -1g.

    Virtually no modern military aircraft fly like you've described. Maybe a case could be made that the F104 flew like that, sort of, but even then not really. Also, to fly slow they have to extend their flaps and leading edge devices (if any), which changes the wing from a low-drag high-speed configuration to a low-speed and high-drag configuration that provides more lift at any given airspeed and angle of attack. The ailerons are for roll control, and those airplanes that can extend them usually call them flaperons. When the wings fly slower they have to increase their angle of attack to generate the same amount of lift, but the amount of lift generated by engine thrust is usually pretty minor (although not zero). Standing nose-up on your engines and blasting them down is a really inefficient way to fly, and even though the pilots don't spend money out of their pockets for fuel, the more efficient the plane the greater the mission capabilities and the longer the endurance. They really don't fly like that.

    I'll avoid the whole subject of actual vectored thrust as a side issue, as it only represents a small fraction of military fighters or strike aircraft, and is not really used the way you mentioned.