Aeroflot Flight 593 crashed after the autopilot was partially disabled by applying certain force to the control wheel but the only function which was disabled was controlling the ailerons. This partial disengagement went unnoticed for a number of reasons.

I'm reviewing the official accident report (text is in Russian, linked to from the Wikipedia article) and here're my findings from there so far.

The flight manual (cited in the report) is claimed to say that manipulating the control wheel to contradict the autopilot is definitely abnormal and should be avoided. The mechanism for overriding the autopilot is a safety device that engages outside normal operational conditions.

So it looks like there some functionality that allows to override the autopilot and partially disconnect it (so that only ailerons are no longer controlled) and it is claimed to be a safety measure.

What's the purpose of this functionality?

  • 2
    $\begingroup$ If an emergency maneuver is necessary you don't want the pilot to have to find the Off button of the auto pilot before he can do the necessary motions. $\endgroup$ Commented Sep 23, 2014 at 13:14
  • 1
    $\begingroup$ @ratchetfreak: Do the most likely emergency maneuvers need the ailerons only? $\endgroup$
    – sharptooth
    Commented Sep 23, 2014 at 13:32
  • $\begingroup$ @sharptooth: I believe that if you push the column forward with enough force the old system would disengage autopilot pitch control. Basically the system previously allowed the pilot to disengage autopilot control of any individual aspect (pitch, roll, throttle) by using high force on the appropriate controls in the appropriate direction. $\endgroup$ Commented Sep 23, 2014 at 13:36
  • $\begingroup$ @ratchetfreak: That's the reason why autopilot should disconnect when pilot applies force to the control column. But does not explain why each axis disconnects independently. Note that A320 (that accident was A310) and newer don't have that mode any more. $\endgroup$
    – Jan Hudec
    Commented Sep 25, 2014 at 18:28
  • $\begingroup$ Same as hitting the breaks on a car disengages the cruse control $\endgroup$
    – Steve Kuo
    Commented Oct 27, 2017 at 21:03

3 Answers 3


The autopilot will disconnect on large control inputs because this is the most intuitive way to control the plane. On Boeing planes, the controls will physically move along with the autopilot. If the plane starts doing something you don't want it to do, the instinctive action is to grab the controls to correct it. While the autopilot button would normally be the way to disconnect the autopilot, it may not be the instinctive way in an emergency, and it can be helpful to already have both hands on the controls when the autopilot disconnects, rather than having one hand disconnecting the autopilot.

The reason for only disconnecting one part of the autopilot could be to help the pilot. Maybe there is a problem with the elevator controls, or an upset that affects the aircraft primarily in pitch. Maybe the TCAS gives a "climb" command. The pilot wants to take over pitch control to address this situation, but other controls are fine. While the pilot now has control of just pitch, the autopilot is still taking care of the other controls. The autopilot only gives the pilot control of the modes that are being overridden, instead of just turning off and requiring the pilot to deal with all control modes.


Although it doesn't address partial disengagement, this SB gives the reasons why forces on control columns could be used to disengage the autopilot.

Airbus Service Bulletin (sb) No. A300-22-6021 says

2) Objective/Action To provide autopilot disengagement by applying a 15 daN force on the control column in go-around mode above 400 feet (radio altitude) this Service Bulletin recommends to modify the software at both flight control Computers.

The modified Flight Control Computers will also include­ improvements which have been identified from the last standard.

(3). Advantages

Operational benefit and/or passenger comfort by :

  • autopilot disengagement by 15 daN force on control column during go-around above 400 feet (radio altitude),
  • avoidance of unwanted atopilot disengagement when the pilot takes firmly the control column,
  • nose dowm improvement to avoid the pitch attitude increase after main landing gear touch down,
  • improvement of "LEVEL CHANGE" mode; to avoid the "VMO" overshoot,
  • "ALT HOLD" mode improvement in heavy turbulence.
  • improvement of autopilot capacity to counteract strong vertical gust in cruise.

n.b. 15 daN = 15 deca-Newtons = 150 Newtons = ~34 lbs.


The reason for this kind of functionality is that if you are handling the controls with that much force, the autopilot assumes you want to fly and gets out of the way. The forces needed to do this typically are not the kind produced by incidental contact but rather from purposeful contact. In the planes I'm familiar with the aural warning system will start announcing "AUTOPILOT" until you hit a button to silence it. In addition, the autopilot annunciation area will indicate the change in autopilot mode or disconnect. These indications should be monitored at all times the autopilot is flying and the chain of errors here is:

  1. inadvertent autopilot disconnection (partial), and
  2. failure to properly monitor the airplane.
  • 1
    $\begingroup$ 0. Bringing your children onto the flight deck and encouraging them to play with the controls in-flight. $\endgroup$ Commented Sep 24, 2014 at 7:32

You must log in to answer this question.

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