Air France 447 Thoughts

by philg

Friends have been asking for my thoughts on how Air France 447 crashed. Without the flight recorder and cockpit voice recorder it will be tough to know. Here’s a guess..

  1. it was the middle of the night and bumpy; the airplane is on autopilot, just like any other airliner in cruise flight
  2. some of the airspeed and attitude instruments disagreed slightly, either because one was defective or conditions were so turbulent that readings differed substantially on the left and right sides of the airplane
  3. the avionics did what they always do in this kind of situation… disengage the autopilot and dump the airplane back into the pilots’ laps: “I can’t determine what’s going on, despite my massive electronic brain, so you try to figure out what to do with this airplane.”
  4. the airplane immediately started pitching and rolling from the turbulence, thus presenting the tired and startled pilots with an “unusual attitude recovery” challenge
  5. the pilots failed to meet the challenge and their control inputs were not helpful in stabilizing the airplane
  6. the airplane came apart from being oversped, overstressed, etc.

How could this happen? Those same pilots would have had unusual attitude training in a Cessna 172 and they did fine. There are a few important differences between a Cessna 172 and an Airbus. The unusual attitude training was 20 minutes into a flight during the daytime. The pilots were prepared for it. It takes a long time to push the Cessna 172 over its speed limit or beyond its stress limits. Pushing the nose down on a jet, by contrast, builds up airspeed at a frightening pace. The Cessna is very tough to spin and can be easily recovered from a spin. A multi-engine jet need not demonstrate spin-resistance or spin recovery. The assumption is that the plane will spend its entire life within a normal envelope of flight attitudes and airspeeds. The Cessna 172 is built to withstand nearly 4Gs and can handle more at the cost of some bending. An airliner is designed to withstand 2.5Gs and the Airbus planes have sometimes had trouble even meeting that standard (if you built an airliner as strong as a four-seat airplane you wouldn’t be able to carry as many passengers).

This explanation of the problem does not require the plane or pilots to have done anything unusual. The Airbus had some sort of problem with its very complex set of sensors, gyros, and computers. That is a very common occurrence on a plane that has three of everything. The autopilot tripped off in response to a failure or disagreement. This is normal behavior, though much more common in light airplanes than in jets. A couple of pilots who were tired and deprived of a natural horizon by the darkness, open ocean, and clouds, turned out not to be heroes, at least not this time.

There is probably more to it, but this is my best guess.

27 thoughts on “Air France 447 Thoughts”

  1. http://trueslant.com/milesobrien/2009/06/08/the-coffin-corner-and-a-mesoscale-maw/

    The Air France 447 mystery may never be solved beyond a shadow of doubt, but there are some telling, tragic clues to consider based on what we know about the airplane systems and the extreme weather and aerodynamic conditions it encountered before it went down a week ago.

    First, a bit of aerodynamics: The doomed Airbus A-330-200 was flying ever so close to its maximum altitude – in a zone pilots call the “Coffin Corner”. It refers to the edge of so-called “flight envelope” of an aircraft. At this altitude, the air is much thinner and that significantly narrows the swath of speed at which the airplane can safely operate.

  2. I’ve been asked similar things by friends and family (though I’ve got nowhere near the hours and experience that you do). I hadn’t considered the possibility of an unusual attitude developing so early in the progression of events though, and it seems just as if not more plausible than anything else.

    Brings up a question of recurring training for airline pilots, which you may be (if allowed) able to answer. From what I’ve been told by other airline pilots, it seems that airline-supplied recurrent training focuses heavily on SOP (i.e. the same things the pilots do every day, and have at least one other pilot in the cockpit watching and correcting), or “standard” failures that, even if the pilots don’t know is specifically about to occur, can know to expect at some point in the training.

    I’ve never heard of unusual attitude training in the simulator, though I don’t know for sure. If not, is there any reason not to mandate it, even if just to deliver the message about how many situations that were recoverable in the four seat trainers are unrecoverable in an airliner?

  5. I’ve read elsewhere that Airbus does _not_ drop autopilot when things get complicated, on the assumption that its massive electronic brain will do a better job than a human one in even the worst conditions.

    Boeing takes the opposite approach, making Boeing->Airbus converters nervous, and Airbus->Boeing converters perhaps unprepared.

    Personally, I fly kites, with less considerably success than the worst of jet pilots.

  7. philg… A disturbing Airbus trend here… From the 2002 NY crash of an Airbus…

    **************

    “The pilots of American Airlines Flight 587 didn’t know their Airbus plane’s vertical tail fin broke off before the aircraft’s fatal Nov. 12 crash in a New York neighborhood, a U.S. board’s records show.”

    “Molin [co-pilot] acknowledged wake turbulence from a preceding aircraft
    about seven seconds before the plane’s rudder made five large
    swings. Seven seconds after the rudder movements, the tail fin
    sheared off”

    ***************

    Countering turbulence with rudder is standard practice. That the rudder (vertical fin) sheared-off is incomprehensible.

    Paul P.

  8. Paul: Using a lot of rudder in the event of an upset works great in a single-engine piston airplane with a straight wing. The swept wing of a jet changes everything. I asked for an asymmetric deployment of the spoilers in the CRJ simulator so that one wing had the full spoiler and the other was flying normally. This rolled the airplane more than 90 degrees in less than 1 second. I tried using both aileron and rudder to recover. The simulator went into some oscillations from which I could not recover and eventually we got the dreaded red screen.

    We tried it again and this time I simply used full aileron deflection. I was able to hold the simulator more or less level.

    Fast swept-wing fast are different and tricky and as Joshua notes above, the standard jet type rating training does not include unusual attitude recovery.

  9. I’m curious how they came to fly in to such bad weather in the first place…wouldn’t on-board radar and ground-control based weather reports provide enough warning to fly around it?

  10. J: On-board radar shows rain. It doesn’t show turbulence. It doesn’t show rain that might be behind the rain. Over the open ocean you don’t have the benefit of the datalinked NEXRAD (ground-based) weather radar summary picture. Airline pilots do their best to fly around thunderstorms, but sometimes get into pretty ugly stuff nonetheless.

  11. J. Peterson: there are limits to on-board RADAR, like not being able to see beyond something nasty to something even nastier. There aren’t ground based weather reports for hundreds of miles. Unless this was a very unusual weather event, it’s apparently done on a regular basis through the ITCZ.

    On a related note, a lot of people wonder about the lack of radio coverage. Well, even if they could issue a mayday for help, they would still be hours from real help. And you can’t land an Airbus in the middle of the ocean, at night in a thunderstorm. Sully’s landing on the Hudson, in daytime, clear weather, on a smooth river was a miracle, aided by decent conditions.

  12. I’m a retired AAL A-300 captain and flew 587 many, many times, occasionally w/ the F/O who was lost in the accident; prior to AAL, I flew Navy F-4’s, A-4’s, F-5E’s in the early to late 70’s. “Paul” (above) was my RIO, and is tied with Willy Driscoll for the best ACM RIO I’ve ever known. Not many years before 587, AAL developed and instituted unusual attitude, or “upset,” simulator training in response to a couple low altitude upset incidents/accidents that had been recent occurrences. The danger was low altitude, slow speed (thus high angle of attack) wake turbulence encounters on takeoff or approach. Rudder use was stressed in this situation because of the high AOA, which makes the ailerons and spoilers relatively less effective and whose use results in adverse yaw, as opposed to roll in the desired direction. There is a generation of pilots now in the cockpit who have flown jets – thus “feet on the floor” – their entire careers and know only aileron/spoiler for roll control, unless they flew fighters or tactical jets of some kind where operation at extremely high AOA was a routine part of fighting. At extremely high AOA, 1/4″ of lateral stick in the F-4 produced nothing but a truly scary amount of yaw in the opposite direction – you could literally floor the rudder pedal, however, and roll the airplane just great. Philg’s experience of controlling roll with just the spoilers will work well at high airspeed/low AOA – the higher the airspeed, the more control authority the spoilers will have. As far a losing the vertical fin, until the AAL accident investigation, no one taught that rapid reversal of rudder pedal input could overstress the vertical fin – it certainly wasn’t in the flight manual. What certainly WAS in the flight manual was a description of the rudder limiter, which decreases available rudder displacement as airspeed increases. The implication there was: “Don’t worry about putting the rudder pedal to the floor – you can’t overstress the A/C.”

  13. Andrew: I’ve talked to a friend who flies the Airbus A320. He says that there are quite a number of situations in which the autopilot will give up, including unusual attitudes and heavy turbulence.

  14. “The coffin corner” is not as big an issue on late model jets as it was on first generation swept wing aircraft.

    A straight wing LearJet is probably one of the few in-service aircraft types that could run into this phenomenon.

    The Airbus FMS equivalent (FMGS) does sometimes recommend maximum altitudes that are too optimistic for either the aircraft’s weight or the current meteorological conditions but any pilot with time on the plane soon develops his/her own rule of thumb for which altitudes are too high at which weights.

    Even if a pilot were to fly the airplane too high it would become readily apparent. You can easily see on the speed tape when Vmo/Mmo and Minimum selectable speed are too close together. If a pilot attempts to fly this way and encounters turbulence the airplane will simply begin to lose altitude.

    The airbus autopilot system is pretty capable and I have yet to see an uncommanded disconnect. Even sans autopilot the airbus is flown in a computer stabilized mode called normal law. It takes either multiple computer failures, multiple hydraulic failures (or a combination of the two) or an upset of more than 110 degrees of roll to cause a degradation into the first of two alternate control law modes. Even in these degraded modes there is some flight envelope protection including g load protection.

    I don’t know where Phil did his jet training but unusual attitude recovery has been a part of every jet training program I have seen. This is especially important as unusual attitude recovery in a Jet is different than other aircraft. Take recovery from a high pitch attitude. In a jet unloading the airframe is not desirable so the proper technique would be to roll the aircraft to between 60 and 90 degrees of bank and let the nose fall through the horizon while maintaining one G of load through the primary pitch controls.

  15. Please read the comments on the Miles O’Brien story. I highly doubt the A330 was flying anywhere near the coffin corner.

  16. No talk about the pitot tube issue? Is bad airspeed data enough to disengage the autopilot?

  17. philg: Indeed, I was conflating autopilot and fly-by-wire. Apparently the Airbus design includes software-enforced limits to manual controls. So the pilot tells the computer what he wants the plane to do, and the computer decides whether the request is reasonable, under current conditions. The concern is that the computer might override a dangerous-but-necessary emergency operating request, but the counter-argument is also obvious.

    The good news, I guess, is that failures in these systems are rare enough that neither Airbus nor Boeing has been demonstratively proven more-correct.

    I’ll stick to kites.

  18. I’m seeing reports now that there was a known issue with the pitot tubes in the A330 being prone to anomalous readings in icing. There was updated hardware being installed, but calls for the unupdated aircraft to be grounded were ignored.

    This possibly resulted in a low indicated airspeed reading in the cockpit, which resulted in the flight crew increasing airspeed, and the flight control software allowing full control surface deflections.

    Coupled with bad weather and rapid or full crew inputs lead to the vertical tail structurally failing.

  19. Re unusual attitudes, I have read elsewhere the following conjecture:

    That the ACARS data transmitted at 0214 including cabin pressure warn is likely to have been transmitted via satellite

    That at least one ADIRU was providing attitude information accurate enough for the airplane to successfully point its satellite antenna (the effective beam width for the sat hardware on AF447 is approx 10°, according to this guy, and if we’re buying part of his story..)

  21. Here, an Iberia pilot describes (in Spanish) an emergency he experienced while flying from Rio to Madrid, near the spot of the Air France accident. He mentions an unexpected rise in air temperature, from -48 to -16 degrees Celsius, which caused a stall (he mentions the ‘coffin corner’) that he recovered from by willfully disengaging the autopilot and diving 4000 feet until he found cold air again, five minutes later. I can’t fully understand what he says, but you guys surely can.

    http://www.elconfidencialdigital.com/Articulo.aspx?IdObjeto=21038

  23. I’m curious why there isn’t a “breadcrumbs” transmission from jets flying over oceans? I’ve read about the amateur radio APRS setup and wonder why something similar isn’t used for large passenger jets. Seems like it would help find wreckage and possible survivors more quickly.

Comments are closed.