21
$\begingroup$

I am thinking of the driver take-over request in automated driving, when the system limits are reached and the back-up driver has to intervene / drive manually. Is there something similar in aviation, e.g. autopilot is engaged and has reached its system limits? Maybe you can give me some examples.

Edit: The reason why I am asking is, that there is a lot of concern about take-over situations in the automobile industry because cars are getting more and more automated. Since automation has a longer tradition in aviation than in automobile industry I assume there is a thing or two to learn from you guys. So basically I want to know what take-over situations there are and what is done to make take-over easy for the pilot (airplane or helicopter pilot).

$\endgroup$
1
  • 16
    $\begingroup$ The most important thing that makes take-over easy is something that can't be duplicated in cars: unexpected situations in airplanes almost never require an instantaneous response. You can afford to take a few seconds after the alarm sounds to establish situational awareness. $\endgroup$
    – Mark
    Sep 9, 2019 at 21:00

2 Answers 2

31
$\begingroup$

I assume in your question the autopilot is engaged. When the auto-flight system reaches the limits of its capabilities it simply disconnects. On Boeing aircraft you get a

> AUTOPILOT DISC

Warning Message on the EICAS and the siren sounds. Now it is all manual flying for the pilots.

An example would be turbulence exceeding the auto-flight system capabilities to maintain the intended trajectory.

After the question was clarified:

The big difference between flying and driving are

  • system redundancy - in an aircraft, everything important is available two times, the very important stuff even three times. If there is a technical problem system performance normally degrades accompanied with appropriate annunciations to the pilots.
  • the environment you are operating in - in the air there is no road to follow or you fall off a cliff. There are mountains, yes, but in cruise flight (similar to driving on the highway) where your vigilance is lower the mountains are below you. In the end, the airway is an imaginary thing. An imaginary line drawn between two waypoints. You won't crash (immediately) if you deviate from it. Even your surrounding traffic is miles away laterally. When the margin for error gets smaller pilots are way more vigilant. During departure/arrival/landing, even when flying fully automated, the concentration is fully on the aircraft. So deviations from the intended trajectory are noticed early and can be acted upon quickly. This even goes so far that with most commercial operators there is a sterile cockpit rule below 10,000 ft where there is no communication in the cockpit which is not relevant to the current situation.
  • the system sophistication - the feat car manufacturers are trying to accomplish is way bigger than an auto flight system for an airliner. The margins for error are way smaller, yet the precision needs to be way higher. In aircraft the autopilot is basically a glorified feedback loop. And even if the autopilot system fails spontaneously due to technical problems the aircraft initially keeps flying without any intervention and only slowly deviates from its current trajectory as aircraft are designed to be aerodynamically stable.
  • training of the operator (pilot/driver) - commercial pilots train for abnormal situations in full flight simulators every six months. Drivers, well, don't. Their last training was most probably when making their drivers license many years ago. Another thing is vigilance and planning ahead. Pilots are trained to anticipate what is coming next. You try to get a mental picture of the traffic around you and where conflicts may arise. And as your aircraft is moving typically pretty fast and there are no brakes you can slam you are flying in a rather defensive way.

And I even have a real world example where the auto flight system performed flawlessly yet human intervention was required:

Departure EDDF Turn Construction by FMS

What you see here is a blurry photo of the incorrect turn construction of a departure. The departure trajectory is supposed to be a left turn to the next waypoint (white star). The flight management system constructed a right 270° turn (magenta line) to match all the constraints of the waypoints (flyover, etc) where a 70° left turn would have been the correct behavior. The aircraft was flown manually to adhere to the departure procedure.

In this case no system degradation or warnings were present. You can see that the Flight Director (FD) on the PFD on the right commands a right turn (vertical magenta line). The autopilot simply follows FD commands and would therefore fly this (wrong) turn.

$\endgroup$
11
  • 5
    $\begingroup$ @Antonie The pilot should not have his attention elsewhere. And the normal instruments should be sufficient to get your bearings in such an event. $\endgroup$ Sep 9, 2019 at 15:01
  • 3
    $\begingroup$ @ratchetfreak, it's easy to say "should not", but humans are unable to maintain passive vigilance for hours on end. $\endgroup$
    – Mark
    Sep 10, 2019 at 2:47
  • 2
    $\begingroup$ @Antonie there might be some startle effect in the beginning. Lets assume you have no idea what the proper trajectory would be. First advantage: the situation is seldom really time critical. And then you get some clues. A general good idea is to keep the blue up and the brown down on your artificial horizon. Assuming the auto thrust system isn't working as well - just dial in a proven pitch & power value. For cruise this most probably would be something around pitch +2.5° and power around 90% N1. Now you won't fall out of the sky. Time to rebuild your desired level of automation. $\endgroup$
    – Chris
    Sep 10, 2019 at 7:36
  • 6
    $\begingroup$ @Antonie the AP disconnect horn is very, very (very) audible and recognizable. Also it rings every time you manually disconnect the AP, so your brain quickly connects this sound to "manual-flying-mode". I only have around 30 hours as PIC and the autopilot disconnected on me once because of some sudden wind that made the PA-28 roll a little bit too much. Despite of my little experience, my hands grabbed the yoke and I leveled the plane as soon as the loud horn started ringing without me having to even think about it. $\endgroup$
    – Quentin H
    Sep 10, 2019 at 8:07
  • 1
    $\begingroup$ Adding to this, part of your system redundancy point: even the pilots are redundant. In airplanes with a flight engineer position, it's even double redundant (three qualified people minimum in the cockpit normally, though typically only two who are qualified to actually fly the plane). Either pilot can call another pilot's attention to an issue, and of the two pilots up front, any one can take over immediately if necessary. Try that in a typical car... $\endgroup$
    – user
    Sep 10, 2019 at 9:56
15
$\begingroup$

There are many possible reasons for the autopilot to disconnect. The details will depend on the aircraft, but here are the criteria for a Boeing 737 (source: 737 NG FCOMv2 4.20.2 Automatic Flight - System Description):

Autopilot Disengagement

The A/P automatically disengages when any of the following occurs:

  • pushing either A/P disengage switch
  • pushing either Takeoff/Go-around (TO/GA) switch with a single A/P engaged in CWS or CMD;
    • below 2000 feet RA or,
    • with flaps not up or,
    • G/S engaged
  • [Option - A/P auto-disengages for TO/GA above 2000 feet RA]
    [Option - Honeywell -708 FCC and on]
    pushing either Takeoff/Go-around (TO/GA) switch with a single A/P engaged in CWS or CMD above 2000 feet RA with flaps not up or G/S engaged.
  • [Option - Fail-Operational Autoland]
    pushing either TO/GA switch after touchdown with both A/Ps engaged in CMD (except with LAND 3 or LAND 2 annunciated)
  • pushing either TO/GA switch after touchdown with both A/Ps engaged in CMD
  • pushing an illuminated A/P ENGAGE switch
  • pushing the A/P DISENGAGE bar down
  • [Option - Fail-Operational Autoland] activating either pilot’s control wheel trim switch (except with LAND 3 or LAND 2 annunciated)
  • activating either pilot’s control wheel trim switch
  • moving the STAB TRIM AUTOPILOT cutout switch to CUTOUT
  • either left or right IRS system failure or FAULT light illuminated
  • loss of electrical power or a sensor input which prevents proper operation of the engaged A/P and mode
  • loss of respective hydraulic system pressure.

As Chris already said in his answer, the A/P disengagement will trigger a warning horn and also illuminate the Autopilot (A/P) Disengage Light (number 1 in the following image from the FCOM): 737 Forward Panel

The pilot flying will immediately take over manual control of the yoke and fly the aircraft. This is typically not as time critical as in a car, where a small deviation from your lane on the highway could result in a collision within seconds. The only phase of flight where A/P disengagement is critical is during an Autoland, where the pilot flying is already fully focused on monitoring the A/P and is ready to take over at any moment.

Re-engagement of the autopilot is possible whenever the autopilot engagement criteria are fulfilled:

Each A/P can be engaged by pushing a separate CMD or CWS engage switch. A/P engagement in CMD or CWS is inhibited unless both of the following pilot–controlled conditions are met:

  • no force is being applied to the control wheel
  • the STAB TRIM AUTOPILOT cutout switch is at NORMAL.

There is also a system on some aircraft requiring some input from the flight crew after a set time, similar to car systems complaining when you drive with cruise control and lane assistant without touching the steering wheel. For example, on the Boeing 777 (source: 777 FCOMv2 15.20.11 Warning Systems - System Description):

Crew Alertness Monitor

The FMC continuously monitors switch action on the MCP, EFIS control panel, display select panel, CDUs and radio transmitter microphone switches. When a predefined time elapses after the last switch action was detected, the EICAS alert message PILOT RESPONSE is displayed.

$\endgroup$
0

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .