I was thinking about Air France 447 recently and had a wonder:

What would happen if the pilots had re-engaged the autopilot and stopped all stick input? (If that's even possible in a stall condition). I know this has since changed, there is the MCAS system now, which will force a stall recovery. But in an older aircraft, like the A330 - 203, what would happen?

  • 4
    $\begingroup$ MCAS on the 737MAX is not an anti-stall system. $\endgroup$
    – Sanchises
    Oct 8, 2020 at 11:14
  • 1
    $\begingroup$ There is a very basic flying skill issue here. A brand new pilot is taught not to be dependent on airspeed indication, doubly so for an IFR pilot. You have speed, pitch attitude and altitude. If airspeed goes wonky, you look to see that the pitch attitude is also doing things that would match the speed indications. If the speed indication is dropping, but the power is on and the pitch attitude is constant, and you don't feel any deceleration, those things outvote the airspeed indication and you just hold the pitch attitude.The two guys on 447 suffered a very basic skill breakdown. $\endgroup$
    – John K
    Oct 9, 2020 at 2:52

2 Answers 2


In the AF447 accident, the autopilot disconnected because of inconsistent airspeed:

Temporary inconsistency between the airspeed measurements, likely following the obstruction of the Pitot probes by ice crystals that, in particular, caused the autopilot disconnection and the reconfiguration to alternate law;

(BEA - Final Report)

As long as the airspeed remains inconsistent, it is not possible to re-engage the autopilot:

Two ADR outputs are erroneous but different and the remaining ADR is correct, or if all three ADR outputs are erroneous but different. The AP and A/THR will disconnect. If the disagreement lasts for more than 10 seconds, the PRIM triggers the NAV ADR DISAGREE ECAM caution. Flight controls revert to ALTN 2 law. The SPD LIM flag is displayed on both PFDs, however VLS and VSW are not displayed. This condition is latched until a PRIM reset is performed on ground without any hydraulic pressure. However, if the disagreement was transient, the AP and A/THR can be re-engaged when the NAV ADR DISAGREE message has disappeared.

(Airbus A330 FCTM - Non-normal operations - Unreliable Airspeed Indications, emphasis mine)

It is not known if the crew attempted to re-engage the AP with the pushbuttons on the FCU because this is not recorded by the FDR:

It is also impossible to see whether there have been any attempts to re-engage the autopilot.

(BEA - Final Report)

The BEA did a study of similar incidents and found that the AP was re-engaged in some of these:

The BEA has studied thirteen unreliable indicated airspeed events involving the temporary loss of this reading, or other anomalies, for which it had access to crew reports, recorded parameters and PFR. [...]

  • The aircraft’s autopilot disconnected in all cases, with no intervention from the crew;
  • [...]
  • In seven cases, an autopilot was re-engaged during the event. In two of these, the re-engagement occurred even though two speeds were consistent with each other, but erroneous;

(BEA - Final Report)

So it is possible to re-engage the AP after it automatically disconnects due to unreliable airspeed, but these events were not fully developed stalls. Your question seems to ask about the situation later on where the incorrect manual inputs had already resulted in a stall. In these cases, re-engagement of the AP would likely not be possible due to the engagement conditions:

AP Engagement

AP can be engaged when:

  • Aircraft speed is within VLS and VMAX.
  • Aircraft pitch angle does not exceed 10° nose down or 22° nose up.
  • Bank angle is less than 40°
  • [...]

(Airbus A330 FCOM - Auto Flight - Flight Guidance - Autopilot)

In a stall the speed or pitch limits are likely exceeded preventing autopilot re-engagement. This is a part of the FDR recording from AF447 (you can see that pitch never exceeded the 22°, but the airspeed went down to 38 kt towards then end of the flight while the F/O pitch command was at the nose-up limit):

(BEA - Final Report)

If the autopilot would re-engage successfully, it will do what it is programmed to do: follow the flight director inputs, which depend on the selected mode on the FCU (therefore just turning on the AP is not enough, one also needs to select appropriate modes on the FCU).

  • 4
    $\begingroup$ Outstanding answer! $\endgroup$
    – Jamiec
    Oct 8, 2020 at 16:01
  • 1
    $\begingroup$ Note that the aircraft's actual airspeed was much higher than 38 kt; the aircraft's extreme angle of attack disrupted the airflow over the pitot tubes, producing falsely-low airspeed indications. (This is not to be confused with the earlier erroneously-low airspeed indications resulting from pitot-tube ice-crystal blockage, which had long since cleared by this point.) $\endgroup$
    – Vikki
    Oct 8, 2020 at 22:47
  • 3
    $\begingroup$ @Sean Agreed, but the autopilot engagement condition is obviously based on that speed, not the actual (but unknown) speed. $\endgroup$
    – Bianfable
    Oct 9, 2020 at 7:02
  • 1
    $\begingroup$ @Bianfable: I know - just noting in case anyone thought you were saying that the aircraft's airspeed actually went down to 38 kt. $\endgroup$
    – Vikki
    Oct 9, 2020 at 20:49

On just about all airplanes the Stall Protection System disengages the autopilot automatically whenever a range of parameters is exceeded. One of them is approaching stall. Generally the Autopilot will be kicked off automatically just before stick shaker onset on a traditional non-FBW airplane, and it can't be reengaged until the parameters that were exceeded (pitch/bank angles, speed, AOA etc) are back in limits.

Low speed flight at high altitude is dangerous because the recovery process requires extreme inputs to recover speed compared to down lower (mass is the same, but the air is thin) - you have to dive aggressively to gain speed so even if you could, just reengaging the AP would be inadvisable.

Since you mentioned it, I want to repeat an important point on MCAS because there appears to be a mountain of misinformation. MCAS has nothing whatsoever to do with stall recovery or stall behaviour.

It was implemented strictly to make the MAX's pitch stick force response in certain regimes match older 737s, to avoid an additional training requirement necessitating Sim time. That's ALL it was intended to do. The airplane was technically certifiable as it was when the problem was discovered (it wasn't dangerous, just different), but would have required sim training, which was considered a marketing disaster from an airline cost perspective.

  • $\begingroup$ Hi John, thanks for correcting me! Can I ask, why don't they program autopilots to recover from a stall? So for example in AF447, they could re-engage AP, which would then automatically perform a pitch down and level off once correct airspeed was achieved? Seems like that would be a lot safer than the human instinct to "climb"! $\endgroup$
    – Cloud
    Oct 8, 2020 at 15:46
  • 1
    $\begingroup$ Autopilots have traditionally not had control authority to do those things. They are designed to maneuver the airplane in normal ways, the way a smooth pilot would. The aren't designed to take over and fly beyond gentle turns and pitch rates. It's expected that a pilot who knows how to fly can take care of that. An autopilot that can do what you suggest is now becoming a "robot pilot" you might say, with all of the powers of the human pilot. If the "robot pilot" gets out of control, you have a problem. This may happen in future if somebody ever tries to certify a single pilot airliner. $\endgroup$
    – John K
    Oct 8, 2020 at 15:52
  • $\begingroup$ Or... a no-pilot airliner :0 $\endgroup$
    – Cloud
    Oct 8, 2020 at 16:50
  • 1
    $\begingroup$ @Cloud I wouldn't call what happened on AF447 "human instinct," at least not for a trained pilot. Every competent pilot is trained from very early on to almost reflexively drop the nose at the first hints of the onset of stall. This will be the natural reaction for any competent pilot. Why (at least) one of the pilots on AF447 was not competent is another question. Maybe it's a human instinct to pull back when you see the ground approaching too quickly (e.g. Asiana 214,) but, at that point in an airliner, it's more a matter of determining how you're going to crash than actually preventing it. $\endgroup$
    – reirab
    Oct 8, 2020 at 23:13
  • $\begingroup$ A correctly functioning autopilot does not need a stall recovery procedure for the same reason that a correctly functioning human pilot does not need one: Neither would put the plane in a situation where a stall was going to happen. Of course human pilots are taught to recover from stalls because (1) they sometimes make mistakes and (2) sometimes, automatic systems do not function correctly. $\endgroup$
    – alephzero
    Oct 9, 2020 at 1:16

You must log in to answer this question.

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