[Geezer (OP)]

Reading the latest on Air France 447 reminded me of some prophetic statements I heard some time ago. I can't remember who it was that said it now, but their concern was that a lot of automation might result in a situation where the crew is so overwhelmed with data that they forget to fly the aeroplane.

[CliffordK]

I must admit that cockpit photos show a bewildering array of dials and gauges.

Die Hard 2 was about terrorists that captured an airport and apparently changed the elevation readings, and thus none of the planes could land.  That may not be a realistic scenario, but does bring up an risk of having too much automation.

I think for the most part, the automation, sensors, and etc are all good.  I do have to wonder if they are over-sensitive to things like cell phones and computer WIFI, but they can work on better protecting their aircraft electronics.

However, if something goes very very wrong, I would hope there are good pilots on board.

There are many fields where 99% of the procedures are routine.  But, there is that remaining 1% of the time that really needs someone smart at the moment when it is needed.

[Geezer (OP)]

Sadly, based on the little that I have read, I suspect that the crew were so bewildered that they didn't believe any of their instrumentation.

It's been suggested that abberations from the airspeed indicators confused the automatic systems. However, if that was the case, it should not have affected the basic attitude and altitude readings that are essential to flying any aircraft, particularly at night.

There could be a great danger in relying too much on automation. Automatic systems can do a wonderful job in all the test scenarios that proved them worthy. However, as soon as you introduce an unanticipated test scenario, they may well collapse.

There is still an enormous difference between artificial intelligence and human intelligence. Automation is a great tool, but it's no substitute for a human when things go "off the map".

 

[Don_1]

Coming down to Earth somewhat, but still relevant, I recall a company which made power steering units for the auto industry. So confident were they in the automation of their production process, that they began to replace their skilled workers with unskilled workers who they taught to use the machines 'parrot fashion'. The problems soon started to arise, when things went wrong with the production and the company had no skilled workers to identify the problems, let alone put them right.

Automation is all very well, but we still need human skills to keep a check on the machines.

[np_complete]

" I can't remember who it was that said it now, but their concern was that a lot of automation might result in a situation where the crew is so overwhelmed with data that they forget to fly the aeroplane."
I believe the term that you are referring to is "information overload" ;-)

Whilst your question is interesting, one should note that "automation" isn't related to automation - it is related to information acquisition - whether automated or not. In fact, if used correctly, automation can avoid information overload and improve situational awareness of pilots through decision support and/or filtering of incoming data.

Orasanu and Martin wrote an excellent article on the subject: “Errors in Aviation Decision Making: A Factor in Accidents and Incidents” , HESSD, 1999.

[Geezer (OP)]

" I can't remember who it was that said it now, but their concern was that a lot of automation might result in a situation where the crew is so overwhelmed with data that they forget to fly the aeroplane."
I believe the term that you are referring to is "information overload" ;-)

Whilst your question is interesting, one should note that "automation" isn't related to automation - it is related to information acquisition - whether automated or not. In fact, if used correctly, automation can avoid information overload and improve situational awareness of pilots through decision support and/or filtering of incoming data.

Orasanu and Martin wrote an excellent article on the subject: “Errors in Aviation Decision Making: A Factor in Accidents and Incidents” , HESSD, 1999.

Actually, I am referring to automation. Modern aircraft are so extensively automated that the crew don't have much oportunity to actually fly the plane. In fact, one of the main problems they face is boredom. Consequently, when things go wrong, and the automation shuts down, their lack of experience can cause problems.

[CliffordK]

One of the problem that has happened to Toyota is that they are now disconnecting the controls from the actuators. 

In every car I've ever owned, there was a cable, wire, or linkage going from the gas pedal to the carburetor or injection pump.

That is no longer the case.

The gas pedal goes to an analog to digital converter to a computer which then sends a signal to the engine to increase or decrease the fuel.  And, it should have something where if you press the brake, it should send a signal to cut the gas.  But, somehow that wasn't working right.  And, something was sending a signal to the engine "wide open accelerator", when in fact it should have been sending "shut down".

On the older vehicles, there was an ignition switch on the dash.  But, the new ones all have it integrated into the steering column.  And, while one may be able to shut off the engine and still steer, I certainly wouldn't want to test it in an emergency situation.

For the first time in my life, I've managed to get a couple of vehicles with automatic transmissions (given to me).  One thing I've noticed is that it is impossible to coast-start the vehicles.  If the engine stalls for whatever reason, one MUST put it in park to get it going again.  I.E.  if one is rolling, one can't pop the clutch, and one must fully stop the vehicle before even restarting with the key.

Anyway, with the small Cessna - type planes, one might have a physical cables connecting from the cockpit controls to the rudder and elevators.  I believe that is done electronically in the newer planes, with the potential for more and more distance between the pilots and the controls.

Here are some notes about the rudder power control unit malfunctioning and causing crashes (presumably).
http://en.wikipedia.org/wiki/Boeing_737_rudder_issues

Of course, you couldn't expand a Cessna into a 747 without some additional electronic control aids.

Ahh... here is what I was looking for.  The "Jackscrew" problem.
http://en.wikipedia.org/wiki/Alaska_Airlines_Flight_261

I suppose my question is whether there is adequate feedback between the cockpit controls and the ailerons and controls.  With the Cessna, if the controls partially fail, it would likely give immediate tactile response to the pilot, although, not necessarily early enough to remedy the problem. 

Here are some more notes about control systems and elevator failures.
http://www.airlinesafety.com/faq/faq10.htm

The last note from that page is particularly distressing as it failed to have a seamless transition from autopilot control to pilot control, nearly causing a crash.

1985, February 19:   

A B-747 SP, flown by a China Airlines Capt., suffered an engine failure while cruising at 41,000 ft.  The Capt. left it on autopilot too long.  The autopilot tried to maintain that altitude, which was ultimately impossible at that weight, with only 3 engines functioning.  As it approached the stall, because the speed kept decelerating, the Capt. finally disconnected the auto pilot.  He was not prepared, because he had failed to trim in rudder to compensate for the asymmetrical thrust condition; the autopilot was maintaining wings level by the use of aileron and spoilers only.

[Autopilots normally do not control the rudder in climb, cruise, or descent.  They use only the ailerons, spoilers, elevators and horizontal stab trim.]

When he hit that disconnect switch, the plane rolled rapidly and entered a dive.  Although the plane exceeded the speed of sound, tearing parts off and causing major structural damage, the Capt. was able to make a recovery at a few thousand feet over the Pacific Ocean, after he broke out of the clouds and could see his attitude via outside visual reference.  There were, incredibly, only two serious injuries to the 274 passengers and crew.

I suppose, as the planes get larger, and more complex, the more automation is necessarily required.  But, it certainly must be approached with caution.

[graham.d]

You sound like a bunch of Luddites :-) I agree with np_complete that there is some confusion in the discussion between full automation and efforts to reduce complexity of information presentation. The facts are that air accidents have trended down significantly since a peak around 1980 despite the increase in air travel. Also most accidents are down to pilot error. Of course caution is needed when introducing automation; in fact caution is needed when designing any safety critical device. The reason Airbus Industries have introduced greater automation is that they have a view that in emergency situations the flow of data can be too great for a human to respond to. Crudely, the pilot can still take over but the default is with the automated systems. As has been discussed before, it is doubtful that the pilot who landed safely on the Hudson River, following a bird strike, would have been able to do so without the automated flight controls afforded by the A320 aircraft.

I agree that pilots should still get practice and be able to fly when needed and they do, and I believe are required to, manually fly and land aircraft periodically. I reckon I can tell the automated landings because the approach requires fewer corrections and the landings are usually much smoother putting less stress on the aircraft's landing gear!

And frankly, when car driving is fully automated (and it will be eventually) I would expect a significant decrease in accidents despite faster journey times. Bring it on!

[np_complete]

@Geezer: I respectfully disagree :-) As graham.d (and pretty much all scientific literature on the topic of aviation accidents) pointed out, the primary cause for accidents is human error (of which one causation tends / tended to be information overload - which is due to data acquisition not automation).

Actually, I am referring to automation.

"Modern aircraft are so extensively automated that the crew don't have much oportunity to actually fly the plane. In fact, one of the main problems they face is boredom."
In fact even on autopilot, pilots are a busy out: for example RyanAir (and all other European airliners) require complete cross-checks to be made every 15 minutes. Not too much time for boredome.


"Consequently, when things go wrong, and the automation shuts down, their lack of experience can cause problems."
Except that this was never the case - critical systems are designed with redundancy, graceful degradation, n-version development etc etc in mind - there has never been a case in which an ENTIRE critical system failed atomically.

With regards to their "lack of experience": Airline pilots are required by (European) law to undergo simulator tests (which also involve "manual" piloting) every six weeks to eight. Failing one of these tests will cause you to lose your job.

"Reading the latest on Air France 447 reminded me of some prophetic statements "
I don't mean to sound like a *****, but this isn't even remotely releated to your argument with regards to automation :-) Basically what happened was that the pitot tubes iced over during, giving false altitude / airspeed indications. This is a relativly common occurence, and there are well defined procedures for dealing with this. The junior pilot (the captain went to rest - which too, is a common procedure for long distance flight) seems to failed to follow these procedures. So, in essence, the accident is due to an overconfidence bias on behalf of the pilot as opposed to an over-automated cockpit lulling him to a false sense of security / experience. And who knows? Maybe an on-board decision support system could have helped him follow procedures.

Yes, your argument of inexperience stands - but I fail to see how automation correlates with inexperience.


@graham.d: " The facts are that air accidents have trended down significantly since a peak around 1980 despite the increase in air travel. Also most accidents are down to pilot error. Of course caution is needed when introducing automation;"
Very good point - as anectodal evidence we could also point towards the maritime industry - one that is often the least automated transport sector. Even within US and European coastal areas (which are deemed the safest in the world), accidents are skyrocketing. In Europe alonge, 624 merchant vessels (500 - 5000gt category) were involved in 540 accidents (groundings. collisions, fires) - in most of these, a simple heading prediction on part of the VTS would have prevented injuries, loss of lives and millions worth in damages.

[SeanB]

Also remember that in the AF case the system did not fail gracefully, but the erroneous speed and altitude input eventually caused the system to disconnect and leave all input to the pilot suddenly. It still gave false information, and there was no other way for the pilot to determine the correct from the incorrect information. The icing problem was known, but it was low priority, and there was not enough caution given to pilots as to weather that would lead to icing as a thing to be avoided. This was aggravated by the high altitude, where the plane is in a narrow safe flight speed range to stay in, and where the icing caused the aircraft to leave this safe speed, whilst still showing the speed was correct.

A case of automation being good, but that it was driven mostly by wanting to remove the most expensive item in the cockpit, the third pilot/flight engineer.

[Geezer (OP)]

No reason to get so defensive guys  []

The automatic system didn't fail, but it was incapable of handling the particular set of circumstances presented to it, so it handed control back to the crew. It appears they didn't know what to do under the circumstances either. It's always easy to second guess a situation after the fact, but I get the impression that AF447 need not have plunged into the Atlantic.

Call it an over-dependence on automation if you prefer, but it has been a concern in the airline industry for some time. Of course, we can always stick our heads in the sand and pretend everything is just fine the way it is.

[graham.d]

I think the airline industry's concern revolves around the competing marketing of Boeing vs Airbus Industries. Boeing have been playing catch-up and have been trying to knock Airbus Industries' more automated approach.

Nobody would suggest that issues should not be addressed, especially when there has been a major accident, but it would be wrong to blame a whole concept rather than to take an informed approach as to what aspects need to be improved.

[Geezer (OP)]

My question had nothing to do with Boeing v Airbus. I suspect the circumstances that affected AF447 could just as easily have affected a Boeing aircraft.

It's a general question about what happens when things become highly automated, and it probably applies to all sorts of systems. Unless we can take automation to the point where the crew is no longer required under any circumstance, we need to maintain the right balance between the actions of the automatic systems and the actions of the crew.

[graham.d]

You're right about your initial question. These issues do have to be addressed with care and caution. It just seemed that the discussion was tending towards a retro=better approach :-) As I suspect many of us are of a similar vintage, it can appear rather negative to a younger reader. And I do think that, in the case of aircraft systems, much care is taken; I think the improving record regarding aircraft accidents verifies this. In any case, whatever forethought is put in, we all know it can never be 100%.

[Geezer (OP)]

Oh! Retro would not work. You couldn't hope to fly a modern aircraft without all sorts of automation working properly. The concern is that the crew becomes too detached from the workings of the aircraft and, in exceptional circumstances, they don't have the experience to make appropriate decisions.

More automation is inevitable, and I don't know what the solution is. It may actually be better automation, or a slightly different paradigm. Call me an old fuddy-duddy, but personally, I'd be happier if the instrumentation and controls in the latest aircraft looked more like an aeroplane and less like a video game.

[graham.d]

Here are a few views of aircraft instruments:

This is an A380:
http://www.gillesvidal.com/blogpano/cockpit1.htm

And here is a modern 737-800:
http://www.aviafilms.com/photos/boeing-737-800-cockpit.jpg

This is an older 737:
http://aviafilms.com/photos/old-737-classic-cockpit.jpg

For comparison, here is a Cessna-150:
http://aviafilms.com/photos/cessna-150-instrument-panel.jpg

When I was learning to fly many years ago, I found (even then with the simple craft I was flying) that the hardest part was managing all the information. The actual flying of the aircraft was the easy bit. It got easier with time but it was still tricky if something disturbed the routine. I really can't imagine how pilots manage these beasts with all the data they have to deal with. When I had a go in a Phantom simulator last year, it surprised me how simple the controls were and how very ergonomically they were arranged. I guess that, in the case of a Phantom, this was a distinct policy and also the design is rather old.

Modern aircraft are rather similar to a computer game. But people can get quite good at playing these games too. It's a matter of practice and having a degree of natural ability and, I have to say, being young helps.
 

[Geezer (OP)]

I see the 787 retains the more traditional "steering wheel" arrangement - not that it's mechanically connected to anything!

[Geezer (OP)]

According to this there was a standby airspeed indicator on AF447 that seemed to be working properly.

I wonder if this gent is correct, or does he have an axe to grind?

"The crash illustrates “another aspect of automation confusion,” according to Greg Feith, a Colorado-based industry safety consultant and former crash investigator for the U.S. National Transportation Safety Board.” The A330′s automated systems were “based on the concept that pilots would never get themselves into the position” in which they ended up on Flight 447, Mr. Feith said. So the sequence of events “defies all the logic built into the automation.”

http://uprta.org/wsj-air-france-crash-report-likely-to-alter-pilot-training/

[Geezer (OP)]

Airline pilots are required by (European) law to undergo simulator tests (which also involve "manual" piloting) every six weeks to eight. Failing one of these tests will cause you to lose your job.

Jolly good, but apparently it's only recently become possible to simulate stall conditions. Unless they were trained to recover from a stall with a real aircraft, pilots only have the benefit of the correct theoretical response to a stall.

[graham.d]

I find it hard to believe that they do not train pilots on stall recovery because the pilots "would never get themselves into the position". The systems may have given ambiguous data but there is little doubt (for any pilot) about a stall warning and the procedure is the same on all aircraft - which is to push the nose down and increase the airspeed. The fact that there is a stall warning system means the aircraft has been designed with a potential stall in mind so it is astounding that there is no specific training via a simulator. Airspeed indicators can go wrong in a stall because the airflow over the detectors can become turbulent and periodically disrupted; they were probably falling quite fast but not always so the airflow was parallel to the aircraft and the airspeed indicators will not work properly in this case. There would be no doubt about the altimeter's rate of descent though. I'm also surprised that they could not get alternate data on speed (not exactly airspeed but probably a close enough indication) from a GPS system though I expect that there may have been a degree of panic.

The pilot's responses seem to be poor no matter what the aircraft was. It seems a very elementary mistake to make. A corrollary to arguments about over-automation is that it leads to not requiring skilled operators (pilots in this case). The comment about the need for more training is a key point though it should also be mentioned that it needs people with the right inherent skills too. This statement may considered non-PC today but not everyone is cut-out to do any job and one hopes that the dumbing down of the requirements, in this case to be an airline pilot, does not go too deep.

I am surprised that the aircraft systems could not have recognised a stall and got out of it automatically - there was plenty of altitude to make a recovery. Clearly a major software bug! Or maybe the pilot had switched to manual control on some systems. It will be interesting to see the final report.