Slashdot Mirror
Computers Key To Air France Crash
Michael_Curator writes "It's no secret that commercial airplanes are heavily computerized, but as the mystery of Air France Flight 447 unfolds, we need to come to grips with the fact that in many cases, airline pilots' hands are tied when it comes to responding effectively to an emergency situation. Boeing planes allow pilots to take over from computers during emergency situations, Airbus planes do not. It's not a design flaw — it's a philosophical divide. It's essentially a question of what do you trust most: a human being's ingenuity or a computer's infinitely faster access and reaction to information. It's not surprising that an American company errs on the side of individual freedom while a European company is more inclined to favor an approach that relies on systems. As passengers, we should have the right to ask whether we're putting our lives in the hands of a computer rather than the battle-tested pilot sitting up front, and we should have right to deplane if we don't like the answer."
It's not surprising that an American company errs on the side of individual freedom while a European company is more inclined to favor an approach that relies on systems.
How fond Americans are of reductionist dualities that are unhelpful, misleading and frequently downright dangerous: American pilot with The Right Stuff in an American plane would have saved everyone; dangerous European plane and computer killed hundreds. Oversimplified sniping, or childish fantasy?
If I want real facts on flying, instead of wild-assed pseudo-political trollery, I'll go read Peter Ladkin or Patrick Smith: "The gist of the accident appears pretty clear: Air France Flight 447 was victimized by a terrible storm."
you had me at #!
What a dumb phrase. Do you only want former airforce pilots who've actually seen combat flying commercial planes? How exactly is that going to keep you up in the air in a civilian airliner experiencing an electronic or mechanical malfunction?
And if what you really mean is experienced pilots, what about some pilot who's been flying for years and has never had an emergency situation and then makes a mistake and then (s)he makes a judgement error in a critical situation? Are you then going to call for the iron calm of a computer rather than a fallible human pilot?
No, the answer is statistics. What's safer and more reliable in the long run? How many crashes have we had due to computer error rather than human error given x hours flown by each?
The very wording of this ridiculous post presupposes an answer. And in the future it is very likely the wrong answer. Sure computers will make errors. But in general people will make them more often, and computers are just going to get better.
And casting this as some kind of bizarre collectivist vs. individualist ideology debate is ridiculous as well. What does towing some ideological line have to do with safely getting to your destination in an airplane?
This Slashdot article is full of simplistic drivel designed to provoke ideologically based knee-jerk responses instead of any kind of reasoned debate.
The linked to text is much, much better, even though offering people a choice is problematic given how the whole non-refundable ticket system and airline logistics systems currently work, not to mention that making a choice at the gate when you get on the plane will throw off your schedule.
Need a Python, C++, Unix, Linux develop
"What are you doing Dave?"
Be you Admins? nay, we are but lusers!
The Continental flight that crashed in Buffalo on the 12th of February crashed because the inexperienced pilot pulled up when the plane stalled. A computer controlled system might have nosed down to get airspeed and saved 50 lives. Of course I doubt a computer controlled system would be able to make a flawless landing in the Hudson.
What a dumb phrase. Do you only want former airforce pilots who've actually seen combat flying commercial planes?.
Who wouldnt want to be on a commercial flight where random barrel rolls, climbs and dives occur?
Be you Admins? nay, we are but lusers!
It's not surprising that an American company errs on the side of individual freedom...
Eh? You mean the freedom to work under-paid pilots 14-16 hours a day like Colgan Air? And the FAA let them slide because Colgan had friends in that office? Some of their pilots could make more flipping burgers. Like the pair that were tired, under-paid and not paying attention who turned Continential flight 3407 into a giant lawn dart.
This isn't political. I don't care if it's human, machine or a trained goat. Whatever gets the aircraft down in one piece is what I want managing the control surfaces.
That's our life, the big wheel of shit. - The Fat Man, Blue Tango Salvage
"If the Gimli Glider or Flight 1549 had been on an Airbus, there would have been a lot of dead people"....
fyi- Flight 1549 was an Airbus A320. Perhaps you would like to rethink your conclusion?
Summary states:
Boeing planes allow pilots to take over from computers during emergency situations, Airbus planes do not.
According to this link, the Airbus does, in fact, have a manual override mode.
Which would make the argument as presented a moot point.
China Airlines Flight 140 was a route from Taipei, Taiwan to Nagoya, Japan. On April 26, 1994, the Airbus A300 on the route was due to land at Nagoya Airport. The Airbus A300 was completing a routine flight and approach, however just before landing, the First Officer pressed the Take Off/Go-Around button (also known as a TOGA) which raises the throttle position to the same as take offs and go-arounds.
Pilot Wang Lo-chi and copilot Chuang Meng-jung[1] attempted to correct the situation by manually reducing the throttles and pushing the yoke downwards. The autopilot then acted against these inputs (as it is programmed to do when the TOGA button is activated), causing the plane to have a very nose-high attitude. This nose-high attitude, combined with decreasing airspeed due to insufficient thrust, resulted in an aerodynamic stall of the aircraft
-- http://en.wikipedia.org/wiki/China_Airlines_Flight_140
As an ex airline pilot and current software developer I would say that an override must be available in any system. Of course computers are much better in quick decision making and collecting all the facts than humans are. In fact with a glass cockpit, the computer knows the data before the pilot does anyway. But there is the occasion that software fucks up. Plain and simple.
From my own personal experience:
1 - Autopilot with suicide attempt
Boeing 737-400 cruising at FL310 everything happy, clear skies. I'm Pilot flying and the captain suggest I have lunch. With the tray on my lap I eat while glancing at the instruments every once in a while. The captain was supposed to have control. So after a particular tasty piece of chicken I look up only to see the horizon at an angle and way too high. I glance across and see the captain reading the news paper. Look at the instruments which indicate a gentle diving turn. The VNav path on the displays indicate nothing out of the ordinary but this Autopilot decided to go for a turn and decent anyway. The whole thing would have only lasted a few seconds but there was absolutely no reason for the computer to do this manouvre. AP disconnect and reconnect sorted it all out.
2 - Lazy plane
Yeah, uh again during my mean and again I had handed control over to the captain while eating. This time at night. Cruising FL330 when auto throttle decides to close the throttles to idle. Auto pilot maintains altitude. WTF to I push the throttles back up. They stay up for a few seconds and yet again move to idle. Got rid of my food and disconnected the auto throttle. Set cruising power manually and checked everything. Nothing wrong. Re-engaged the auto throttle and things were fine.
3 - Dutch roll gone bad
Climbing through 10.000 feet on auto pilot, the plane begins a slight rocking left and right. No more than a few degrees. As we continue to climb the rocking gets worse. 5 deg bank either way. Auto pilot is working hard to compensate or so it seems because the control column moves noticeably. Again my luck to be pf. We thought the Autopilot had gone mad so after strapping ourselves in tightly we disconnected the ap. I tried to hand fly and stabilise but things got out of control rapidly as the plane started to buck left and right well past 10 degrees bank. I was obviously losing control. Nah, let's face it, I had no control and told the captain. He took over and at least was able to not allow it to get worse. Glad I was with this guy because he flicked off the yaw damper that is an automatic control system to stop an aerodynamic effect called Dutch roll. The plane steadied immediately although we were left with the Dutch roll effect but that was not too bad.
So there you go. In all three cases it was not a matter of pilots being better than computers. Overrides are required when the computer goes mad. I always valued having the mechanical controls as a backup in the 737. I travelled in Airbus aircraft and I no longer fly but I would still hesitate to be a servant to a fly by wire system.
what people are forgetting is that the airbus plane DID return nearly full control to the pilot (nearly because there are still limits to things like how much roll one can request, but these COULD be built in mechanically in the absence of fly-by-wire).
The real issue here is that the computer system detected invalid input and handed the control back to the pilots (under "alternate law" which means most safety rules are disabled), but the pilots may not have had enough information to know whether the control was handed back to them in a safe state, and if not, how to correct.
On top of that, the airplane was flying fairly close to the coffin corner (where the airplane is capable of going too fast and too slow simultaneously, and at this point, in this situation, computers are really helpful). One possible issue is that a gust of wind could have caused "mach tucks" if they were going a little too fast (thus causing downward pressure on the nose during gusts). These could have placed significant stress on the airframe until things started to fail. I have some other theories and observations about debris and ACARS messages, but this isn't the time for that now. All I will say is that all indications are the airplane was flying too fast, and there is NO indication that the instructions Airbus has sent to pilots will remedy that problem because it is unlikely that the pilots would have had sufficient information to act on them.
There are two issues involved here that need additional discussion though:
1) Are airplanes built to withstand forces as well as they used to be? Would, say, a DC8 be able to withstand more turbulance than an A330?
2) Do FBW systems provide sufficient feedback for a pilot to feel the plane? Could accidents be avoided in cases like this by adding additional feedback?
LedgerSMB: Open source Accounting/ERP
the simple fact is we think this particular French Airbus crashed do to design failure, specificity computer related and not human failure and this only strengthens the case for humans to still be able to fully operate complex machines.
The autopilot disengaged at 23:10, 4 minutes before the last automated messages indicating failing cabin pressure were sent by the plane.
provide us with 2 examples where the pilot decided to manually override the flight computer and crashed.
Here is 1: http://en.wikipedia.org/wiki/Bashkirian_Airlines_Flight_2937.
Basically, the on-board computers gave the correct advice, but one of the pilots "disregarded the TCAS instruction to climb and instead began to descend, as instructed by the [air traffic] controller, thus both planes were now descending."
The controller was later assassinated by someone who had lost wife and children in the accident.
From where I'm sitting, it seems boeings fall out the sky with more often and with more devastating results than Airbuses - http://news.bbc.co.uk/2/hi/in_depth/2008892.stm
I particularly liked when the A320 came down in the Hudson how, it was "all thanks to the pilot"...and yes, in part it was, but the minute another airbus falls out the sky and it's fatal this time (as crashes often are) it's clearly because of poor design philosophy?
Meh, this whole thing stinks of US vs EU chest-bashing.
throw new NoSignatureException();
More accurately, there are very few people on Slashdot who have any fracking clue how a fly-by-wire system works. It's evident from the comments; just like everything else, people go spouting off and making claims about what they think things are like based on just a tiny bit of knowledge and their own prejudices, rather than looking at the facts and finding out what things actually are.
(Full disclosure: My day job is developing and testing a new FBW system, and I took an entire course dedicated to this in college.)
FBW systems are not the autopilot. They are not autonomous AIs, they do not make their own decisions, they do not just arbitrarily "decide" to go do something against the pilots' wishes. FBW systems are, in essence, little more than simple feedback control loops, similar to the familiar PDI controllers we all remember from control theory classes. All they do is compare the current state (pitch/yaw/roll angle and rate) with the one commanded through the stick, and try to make the two match by moving the control surfaces. The biggest difference is the presence of limiters which will prevent the aircraft from exceeding certain parameters (usually G load and angle of attack). That's it. That's all there is to a fly-by-wire system. It's just a controller.
In fact, let's compare a "traditional" manual system with a (simplified) FBW one from the pilot's perspective. In a traditional system, the stick/yoke in the cockpit is directly mechanically connected to the control surfaces through pushrods, bellcranks, cables, pulleys, etc. A given deflection of the stick will always result in the same deflection of the surface. For our purposes, we'll assume it's roughly linear, so Dsurface = K * Dstick. Now, let's look at the airplane as a whole. A given deflection of a control surface will not always achieve the same result--at low speeds, you need more deflection for a given response than you do at high speed. The net effect is that, at low speeds, the pilot needs to make large deflections of the stick make a given maneuver. At high speeds, he only needs to move the stick a little bit. It's kind of like your car--the steering gets more sensitive the faster you go; you wouldn't use the same inputs on the freeway as you do in a parking lot. Matching the desired response with the control input is the pilot's job--he's the feedback loop connecting control surfaces with the desired flight path.
A FBW system, on the other hand, doesn't have mechanical connections between stick and surface. Instead, the stick uses force or deflection sensors to read the pilot's input. That input is fed to the FCC, which then sends signals to the actuators on the control surfaces. Instead of commanding a given control surface deflection, the pilot's input will usually command something else, eg. roll rate or G load. Rather than varying according to speed and aircraft position, this will be constant--in other words, the command for 20 deg/sec will be the same at really low speeds as it will at high speeds. Basically, the pilot is telling the aircraft "do this", and the FCC figures out how to achieve that by moving the surfaces.
A FBW system will also often have limiters, which prevent the aircraft from exceeding certain parameters. Most common are angle-of-attack and G limiters. Angle of attack (AOA) is the relative angle of the aircraft to the oncoming air. Imagine sticking your hand out the window of your car, palm downwards. As you slowly rotate your hand so your palm faces forwards, notice that your hand wants to go up--you're making lift, and the angle of your palm to the airflow is your hand's AOA. Notice, though, that once you rotate too far, you stop generating lift--that's a stall. On a wing, the amount of lift generated is roughly linearly proportional to AOA, at least up to a point. Past that critical AOA, the air stops flowing smoothly over the top of the wing and gets all jumbled, causing a loss of lift. That's what a stall is.
The meek may inherit the earth, but the strong shall take the stars.