News for hackers. Stuff that matters.
Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by epochwolf on Thursday February 25, 2010 @05:07PM (#1151708)
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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by jrockway on Thursday February 25, 2010 @09:12PM (#1152178)
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...
Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by barrkel on Thursday February 25, 2010 @04:39PM (#1151654)
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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by epochwolf on Thursday February 25, 2010 @05:05PM (#1151703)
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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by rickmode on Thursday February 25, 2010 @07:40PM (#1152037)
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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by idlewords on Thursday February 25, 2010 @08:57PM (#1152154)
"The problem with the Airbus series of aircraft is that the flight computer has the ultimate authority with flight controls."

Where do you have any evidence for this? As the article points out, once the speed indicators diverged, the flight computer on AF447 put itself into an alternate mode where almost all of the overrides you refer to were turned off.

At that point you're basically flying the plane directly, except that hydraulics are moving the flight surfaces for you.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by ovi256 on Thursday February 25, 2010 @10:23PM (#1152267)
"Where do you have any evidence for this?"

This is a design feature of modern FBW systems. It will not let the plane exit the safety envelope. Simply put, the pilot can put the plane through more acceleration than the frame can handle (the control surfaces are big enough), thus the plane could break apart at high speed without the FBW.

You are right for the rest, the flight computer passed control to the pilots after the loss of the pitot tubes.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by idlewords on Friday February 26, 2010 @12:14PM (#1153558)
No, I meant where do you have evidence for it ever being a problem.

There was a lot of FUD around fly by wire in the early days of Airbus, none of which ever panned out. So I'm asking you to justify your similar statements.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by epochwolf on Friday February 26, 2010 @07:34PM (#1154576)
Here's one article about it. http://www.seattlepi.com/business/boe202.shtml
Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by jac_no_k on Thursday February 25, 2010 @06:51PM (#1151927)
This shows the difference in failure management style between Airbus and Boeing. From the article, it seems that Airbus made assumption that the computers will not fail, and if it does fail, it won't be a survivable event. It does deal with sensor errors but no real way of completely ignoring the bad input.

Hopefully Airbus issues an update that allows the computer to fail gracefully. And make it free so there will be high adoption.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by marze on Thursday February 25, 2010 @10:16PM (#1152262)
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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by DanielBMarkham on Thursday February 25, 2010 @07:23PM (#1151997)
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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by nfnaaron on Thursday February 25, 2010 @08:11PM (#1152093)
<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 ...".

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by lutorm on Thursday February 25, 2010 @08:22PM (#1152105)
a first instinct to go to the manual rather than "aviate ..."

That's why there are two of them...

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by DanielBMarkham on Thursday February 25, 2010 @08:22PM (#1152108)
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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by jrockway on Thursday February 25, 2010 @09:08PM (#1152170)
But keep in mind what the physical FBW controls on the Airbus are like -- it's a gaming joystick that you control with your left hand (if you are the captain). I am shocked that A380s can land in crosswinds at all.

There is also no non-synthetic force-feedback; if the computers don't think there is force, you don't feel any force. (Actually, I'm not sure there is any feedback on the Airbus. But I have only seen video demonstrations of the FBW system.) On mechanical or hydraulic flight controls, you can feel the control surfaces not moving because ... you can't move them. (I believed this helped the pilots of Alaska Airlines 261 realize what was happening. Everyone died anyway, but at least the pilots knew what was wrong...)

I assume if the software totally fails, the plane is simply uncontrollable. I can't believe it would be designed that way, but I can't imagine how you are going to move the control surfaces either. (The software doesn't even have to fail; it could just ignore your input. I believe that some inputs are ignored in normal operation anyway, so this could just be a simple bug.)

Anyway, FBW scares me... but I am not a pilot or airplane designer.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by lutorm on Thursday February 25, 2010 @08:21PM (#1152103)
It does seem pretty weird. If it really hit the water flat, the only thing I can imagine is that they didn't realize the plane stalled until it was too late. If the turbulence is really strong, then I can imagine it would be hard to tell a stall from general turbulence, at least for a while. Since the margin between unsafely high speed and stall speed is very small in strong turbulence, at high altitudes, and maybe the fuel situation made them fly higher than they otherwise would?
Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by andrewcooke on Thursday February 25, 2010 @03:48PM (#1151539)
worth reading the last line...

anyone know which airlines do have "buss"?

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by borga on Thursday February 25, 2010 @05:09PM (#1151713)
I'd rather know which airlines fly Boeing planes, the ones that let the pilots fly when the computers are off.
Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by pedrocr on Thursday February 25, 2010 @09:35PM (#1152215)
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

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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by andrewcooke on Thursday February 25, 2010 @07:02PM (#1151949)
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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by nfnaaron on Thursday February 25, 2010 @08:15PM (#1152098)
"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.

Re:Death in the Atlantic: The Last Four Minutes of Air Franc... by jrockway on Thursday February 25, 2010 @09:16PM (#1152187)
Does the 777 count? It is a fly-by-wire plane, but has never had a fatal accident.