6 ms·

Death in the Atlantic: The Last Four Minutes of Air France Flight 447

39 points by mshafrir 17y ago

andrewcooke 17y ago
worth reading the last line... anyone know which airlines do have "buss"?
- borga 17y ago
  I'd rather know which airlines fly Boeing planes, the ones that let the pilots fly when the computers are off.
  - andrewcooke 17y ago
    that was something i didn't really understand, to be honest. it seemed that the airbus computer goes into a "minimal" (ie "dumb") mode when it detects inconsistent inputs. the implication being (i think) that it does no more than support basic fly-by-wire. so it's not clear to me how that is so different from turning off the computer in a boeing. in either case you're effectively flying "unaided". if the airbus computer had crashed, or had some kind of fault, then sure, better to not have a computer. but that didn't seem to be the case here. and yet, at the same time, it seems that the plane simply stalled and fell out of the sky. when i was fascinated by planes as a kid i learnt that a stall was easily recoverable in airliners, unless they were the old "T tail" design (the rear controls end up in the turbulent wake from the stalled wing and you need to side-slip into a kind of flat spin to recover control). so if they had control, why didn't they recover? and why bother rebooting the computer? there's a suggestion that the pilots were simply incompetent and lost without the computer enhanced controls.
    - nfnaaron 17y ago
      "there's a suggestion that the pilots were simply incompetent and lost without the computer enhanced controls." You fight how you train. They were probably trained to do what they did.
  - jrockway 17y ago
    Does the 777 count? It is a fly-by-wire plane, but has never had a fatal accident.
  - pedrocr 17y ago
    From the info it does seem that for this specific failure mode the Boeing design is better. That may be true for other types of failures too. But from what I've read before the Airbus system of software checks/info on pilot inputs can actually prevent a great deal of failure situations as well. The description of the Hudson landing not long ago included references on how the A320 helped the pilot maintain the optimum glide path and generally fly the plane better even in a much degraded situation. I also remember an article that detailed some accidents or near-misses and the Airbus system got mixed reviews, in that sometimes it was better than a conventional one, sometimes worse. It's easy to have a gut reaction to these things. It sounds like the story for all kinds of advanced technology. Something solves a bunch of common failure modes while adding risk to extreme cases that are hard to test. On balance it may be a positive change and save lives but we are understandably uncomfortable when we see the freak accidents caused. Would you still prefer the Boeing planes if Airbus had a better safety record? They seem to be doing slightly better according to this: http://www.boeing.com/news/techissues/pdf/statsum.pdf http://www.boeing.com/news/techissues/pdf/statsum.pdf Looking at "Hull Loss Accident Rate per Million Departures" (page 21) for recent planes: - 737-300/-400/-500: 0.54 - 747-400: 0.53 - A320/321/319/318: 0.36 - A340: 0.5 (this is a single accident) - A330: 0 (this only goes to 2008, would have to be updated) I haven't done the math to see if this is statistically significant. The upside to all this is that the graph does seem to show planes are getting safer. All the technology does work after all.
barrkel 17y ago
It's disturbing to think that the pilots spent precious time trying to reboot the flight computers, rather than trying to stabilize the plane using the basic thrust, pitch, roll and yaw controls, maintaining level flight and an open throttle. All automated controls like this should be tested with each input in turn (a) disabled and (b) actively malicious, followed by two inputs disabled / malicious, and perhaps three inputs disabled / malicious - with special attention paid to likely cascading failures. Anything the automated system can't deal with should (a) be detectable through sanity checks and (b) have a reliable human response that pilots are trained in. My experience in flight is limited to some skydiving and flight simulators. What I know about stalling from lack of thrust in level flight is that the plane seems to pitch downwards and becomes relatively unresponsive to the small adjustments in the elevators that ought to correct this pitch. A stupidly simple feedback system that doesn't know about the impending stall would simply try to counteract the stall by pitching up further and further, until it ends up in a flat stall / spin. The problem being that stall warnings are based on wind speed, which in turn requires an instrument.
- epochwolf 17y ago
  The problem with the Airbus series of aircraft is that the flight computer has the ultimate authority with flight controls. The flight computer can actually override the pilot if it senses the pilot was attempting to fly outside of programmed limits. Now I would hope that it would release control when it detected a problem, especially when three airspeed sensors are reporting erroneous data. I wonder if this leaves the pilots too dependent on the computer and therefore didn't know how to fix the problems or if the computer itself caused the crash. I could be completely wrong about this. I don't have any experience with flying, I'm just putting out ideas.
  - rickmode 17y ago
    This sounds a heck of a lot like the drive by wire issues Toyota is having. As technologist we should have a healthy distrust of any software. It's simply too complex to predict. Is it corporation group-think that causes them to ignore this? Or hubris? I can't imagine they are that stupid.
epochwolf 17y ago
For several years now, Airbus has offered its customers a special safety program - called "Buss" -- at a cost of €300,000 per aircraft. If the airspeed indicator fails, this software shows pilots the angle at which they must point the plane. Given the relatively common failure of airspeed indicators (according to the article) I can't imagine this feature wouldn't be standard.
- jrockway 17y ago
  Airbus is in the business to make money, and so are the airlines. Maybe 300,000 euros wasn't worth it for the airline, and making it free wasn't worth it to Airbus. Much better to pay the lawyers to fight over how much to pay the families of the 300 dead people...
DanielBMarkham 17y ago
I'm a little confused by this report. If the airplane lost lift and descended at a free-fall rate into the ocean, and the engines were still running? That would suggest to me a deep stall due to ice. You'd think that under FL150 or so they'd point the nose down and try to get some kind of lift from the wings, even if they had to overspeed. But it looks, from the story, like they tried to fight the descent with power and pitch. Were they unaware of icing conditions in the tops of those storms? I find that hard to believe. Did they understand the ASIs were inop? I think they would, and the article assumes they did. So why keep screwing around with the flight computers? Aviate, navigate, then communicate, right? I'm tempted to bash the Airbus FBW system, but that would be too easy. I am concerned, however, that it seems like there were more accidents with the Airbus in regards to pitot icing than Boeing.
- nfnaaron 17y ago
  <I am not a pilot />The article briefly touches on this: when the flight computers on an Airbus are out, handling the plane is sufficiently strange to the pilots that it's dangerous. They are used to the computers being pretty much in control. I also speculate, based on the passage about it being very difficult to look through a manual for angle information while the plane is bouncing, that their training produces a first instinct to go to the manual rather than "aviate ...".
  - lutorm 17y ago
    a first instinct to go to the manual rather than "aviate ..." That's why there are two of them...
  - DanielBMarkham 17y ago
    I'm a pilot, but just little planes and I haven't flown in years. But I do still read up on aviation accidents regularly -- lots of good lessons there. The question I have, which is very nuanced, is does the Fly-By-Wire system in the Airbus lend itself to making it harder to fly by the seat of your pants in partial panel situations than in a Boeing in the same situation? I'm not sure that question even can be answered. Commercial pilots are indoctrinated over and over again to go to the manual for everything. (I have a commercial license) I used to joke that if a UFO landed on the wing of a 747, the copilot would remain calm as he said something like "Looks like an alien craft on the wing, captain. We'll need the checklist for alien contact while in flight" Most of the time the big boys have no trouble at all transitioning from full computers to basic flight instruments, and there are plenty of standbys in a modern cockpit. I don't think there is yet another standby pitot besides the ones they already have, though. And nighttime flight over the ocean in IMC in heavy turbulence is an unusual situation to be in, but these guys should be pros. That's the type thing you should train in the simulators for -- "Let's turn the plane upside down, totally ice it up, and kill all the pitots" See how you guys handle that. Having said that, I don't fly the big birds. There's that deadly coffin corner to consider. Could be that the plane's sudden complex situation simply overwhelmed the ability of the pilots to handle it as quickly as they needed to.
marze 17y ago
I disagree with the Airbus design philosophy of having the computer override the pilot, so I choose Boeing jets whenever possible, even though overall the safety records are comparable. This is clearly a UI related incident.

39 points by mshafrir 17y ago

andrewcooke 17y ago

worth reading the last line... anyone know which airlines do have "buss"?

borga 17y ago
I'd rather know which airlines fly Boeing planes, the ones that let the pilots fly when the computers are off.
- andrewcooke 17y ago
  that was something i didn't really understand, to be honest. it seemed that the airbus computer goes into a "minimal" (ie "dumb") mode when it detects inconsistent inputs. the implication being (i think) that it does no more than support basic fly-by-wire. so it's not clear to me how that is so different from turning off the computer in a boeing. in either case you're effectively flying "unaided". if the airbus computer had crashed, or had some kind of fault, then sure, better to not have a computer. but that didn't seem to be the case here. and yet, at the same time, it seems that the plane simply stalled and fell out of the sky. when i was fascinated by planes as a kid i learnt that a stall was easily recoverable in airliners, unless they were the old "T tail" design (the rear controls end up in the turbulent wake from the stalled wing and you need to side-slip into a kind of flat spin to recover control). so if they had control, why didn't they recover? and why bother rebooting the computer? there's a suggestion that the pilots were simply incompetent and lost without the computer enhanced controls.
  - nfnaaron 17y ago
    "there's a suggestion that the pilots were simply incompetent and lost without the computer enhanced controls." You fight how you train. They were probably trained to do what they did.
- jrockway 17y ago
  Does the 777 count? It is a fly-by-wire plane, but has never had a fatal accident.
- pedrocr 17y ago
  From the info it does seem that for this specific failure mode the Boeing design is better. That may be true for other types of failures too. But from what I've read before the Airbus system of software checks/info on pilot inputs can actually prevent a great deal of failure situations as well. The description of the Hudson landing not long ago included references on how the A320 helped the pilot maintain the optimum glide path and generally fly the plane better even in a much degraded situation. I also remember an article that detailed some accidents or near-misses and the Airbus system got mixed reviews, in that sometimes it was better than a conventional one, sometimes worse. It's easy to have a gut reaction to these things. It sounds like the story for all kinds of advanced technology. Something solves a bunch of common failure modes while adding risk to extreme cases that are hard to test. On balance it may be a positive change and save lives but we are understandably uncomfortable when we see the freak accidents caused. Would you still prefer the Boeing planes if Airbus had a better safety record? They seem to be doing slightly better according to this: http://www.boeing.com/news/techissues/pdf/statsum.pdf http://www.boeing.com/news/techissues/pdf/statsum.pdf Looking at "Hull Loss Accident Rate per Million Departures" (page 21) for recent planes: - 737-300/-400/-500: 0.54 - 747-400: 0.53 - A320/321/319/318: 0.36 - A340: 0.5 (this is a single accident) - A330: 0 (this only goes to 2008, would have to be updated) I haven't done the math to see if this is statistically significant. The upside to all this is that the graph does seem to show planes are getting safer. All the technology does work after all.

barrkel 17y ago

It's disturbing to think that the pilots spent precious time trying to reboot the flight computers, rather than trying to stabilize the plane using the basic thrust, pitch, roll and yaw controls, maintaining level flight and an open throttle. All automated controls like this should be tested with each input in turn (a) disabled and (b) actively malicious, followed by two inputs disabled / malicious, and perhaps three inputs disabled / malicious - with special attention paid to likely cascading failures. Anything the automated system can't deal with should (a) be detectable through sanity checks and (b) have a reliable human response that pilots are trained in. My experience in flight is limited to some skydiving and flight simulators. What I know about stalling from lack of thrust in level flight is that the plane seems to pitch downwards and becomes relatively unresponsive to the small adjustments in the elevators that ought to correct this pitch. A stupidly simple feedback system that doesn't know about the impending stall would simply try to counteract the stall by pitching up further and further, until it ends up in a flat stall / spin. The problem being that stall warnings are based on wind speed, which in turn requires an instrument.

epochwolf 17y ago
The problem with the Airbus series of aircraft is that the flight computer has the ultimate authority with flight controls. The flight computer can actually override the pilot if it senses the pilot was attempting to fly outside of programmed limits. Now I would hope that it would release control when it detected a problem, especially when three airspeed sensors are reporting erroneous data. I wonder if this leaves the pilots too dependent on the computer and therefore didn't know how to fix the problems or if the computer itself caused the crash. I could be completely wrong about this. I don't have any experience with flying, I'm just putting out ideas.
- rickmode 17y ago
  This sounds a heck of a lot like the drive by wire issues Toyota is having. As technologist we should have a healthy distrust of any software. It's simply too complex to predict. Is it corporation group-think that causes them to ignore this? Or hubris? I can't imagine they are that stupid.

epochwolf 17y ago

For several years now, Airbus has offered its customers a special safety program - called "Buss" -- at a cost of €300,000 per aircraft. If the airspeed indicator fails, this software shows pilots the angle at which they must point the plane. Given the relatively common failure of airspeed indicators (according to the article) I can't imagine this feature wouldn't be standard.

jrockway 17y ago
Airbus is in the business to make money, and so are the airlines. Maybe 300,000 euros wasn't worth it for the airline, and making it free wasn't worth it to Airbus. Much better to pay the lawyers to fight over how much to pay the families of the 300 dead people...

DanielBMarkham 17y ago

I'm a little confused by this report. If the airplane lost lift and descended at a free-fall rate into the ocean, and the engines were still running? That would suggest to me a deep stall due to ice. You'd think that under FL150 or so they'd point the nose down and try to get some kind of lift from the wings, even if they had to overspeed. But it looks, from the story, like they tried to fight the descent with power and pitch. Were they unaware of icing conditions in the tops of those storms? I find that hard to believe. Did they understand the ASIs were inop? I think they would, and the article assumes they did. So why keep screwing around with the flight computers? Aviate, navigate, then communicate, right? I'm tempted to bash the Airbus FBW system, but that would be too easy. I am concerned, however, that it seems like there were more accidents with the Airbus in regards to pitot icing than Boeing.

nfnaaron 17y ago
<I am not a pilot />The article briefly touches on this: when the flight computers on an Airbus are out, handling the plane is sufficiently strange to the pilots that it's dangerous. They are used to the computers being pretty much in control. I also speculate, based on the passage about it being very difficult to look through a manual for angle information while the plane is bouncing, that their training produces a first instinct to go to the manual rather than "aviate ...".
- lutorm 17y ago
  a first instinct to go to the manual rather than "aviate ..." That's why there are two of them...
- DanielBMarkham 17y ago
  I'm a pilot, but just little planes and I haven't flown in years. But I do still read up on aviation accidents regularly -- lots of good lessons there. The question I have, which is very nuanced, is does the Fly-By-Wire system in the Airbus lend itself to making it harder to fly by the seat of your pants in partial panel situations than in a Boeing in the same situation? I'm not sure that question even can be answered. Commercial pilots are indoctrinated over and over again to go to the manual for everything. (I have a commercial license) I used to joke that if a UFO landed on the wing of a 747, the copilot would remain calm as he said something like "Looks like an alien craft on the wing, captain. We'll need the checklist for alien contact while in flight" Most of the time the big boys have no trouble at all transitioning from full computers to basic flight instruments, and there are plenty of standbys in a modern cockpit. I don't think there is yet another standby pitot besides the ones they already have, though. And nighttime flight over the ocean in IMC in heavy turbulence is an unusual situation to be in, but these guys should be pros. That's the type thing you should train in the simulators for -- "Let's turn the plane upside down, totally ice it up, and kill all the pitots" See how you guys handle that. Having said that, I don't fly the big birds. There's that deadly coffin corner to consider. Could be that the plane's sudden complex situation simply overwhelmed the ability of the pilots to handle it as quickly as they needed to.

marze 17y ago

I disagree with the Airbus design philosophy of having the computer override the pilot, so I choose Boeing jets whenever possible, even though overall the safety records are comparable. This is clearly a UI related incident.