Pilot Error in the Air France 447 Accident

Pilots make many errors. It is recognition of this fact, vigilance in looking for errors and willingness to quickly correct errors which makes for safe operations. When pilot errors cause or contribute to accidents it is in understanding why that leads to correction of problems and prevention of future accidents.

When Airbus developed it’s fly-by-wire control system for the A-320 and subsequent aircraft, it was often stated the Airbus system was “just like other aircraft, but better.” Part of the “better” was that Airbus designed protection limits in the system such as high angle of attack (AOA) limits which prevented a stall. The advantages of this feature were extolled in technical reports like SAE 861801.

Dual page magazine adds stated “Taking the danger out of windshear. . . . The pilot then simply pulls back on the side-stick, and the aircraft automatically corrects itself to achieve the maximum lift needed to fly out of danger.” (Newsweek, 9/14/87). Pilots were trained to use this by selecting TOGA (take off/go around) thrust and holding back stick input to achieve what was considered to be maximum performance—not really though because at the limiting AOA an aircraft is far up the back side of the drag curve which limits flight path performance as well as not allowing any extra lift force to be generated to redirect the flight path if it happens to be going in the wrong direction, i.e., down. One feature was that the system would automatically select TOGA.

Certain failures, such as loss of airspeed information, can cause the Airbus flight control system to revert from Normal Law to Alternate Law in which AOA limit and protection is lost. That the PF (pilot flying) of Air France flight 447 thought he was reacting as he should by selecting TOGA and holding full back stick, despite system reversion to Alternate Law, is clear by his statement less than 1 minute before the crash, ”But I’ve been at maxi nose-up for a while.”

Another feature of the Airbus system is auto-trim which, in this case, was functioning in Alternate Law, but apparently none of the three pilots on Air France flight 447 recognized the elevator trim had gone to full nose up. In this condition, it is unlikely that a positive nose down pitch force could have been generated since the elevator trim surface is at least twice as large as the elevator.

According to the accident report (page 182), after the aircraft passed through 31,500 feet going down, there was no valid AOA sensing of less than 35 degrees. One characteristic of a “dynamic stall” identified by S.S. Hoerner in his book on Fluid Dynamics, is that the AOA at which the airflow re- attaches to the upper surface of the wing is considerably less than the AOA at initial stall. Once the aircraft entered the deep stall, there is no assurance that any pilot or method could have recovered, in spite of what the pilots did or did not do. It may have been possible to roll the aircraft into a steep bank to get the nose to drop, but without correcting the elevator trim condition, the aircraft would probably have accelerated into another stall.

The use of TOGA or nearly maximum thrust most of the time did not help. In this case, after the stall if the thrust force is resolved into components parallel and normal to the flight path, it becomes obvious that engine thrust was propelling the aircraft in its downward path, increasing the rate of descent. At two times during the descent, when thrust was decreased, there was a significant reduction in descent rate.

Could the accident have been prevented? Certainly, but that would have required immediate attention to the impending stall condition, pitching the nose down to return to an acceptable AOA and then selecting full thrust. One comment in the accident report (page 187) is “It appears that this absence of positive static stability could have contributed to the PF not identifying the approach to stall.” One of the most obvious differences between the Airbus flight control system and other aircraft is the lack of trim stability as far as the pilot’s feel is concerned, i.e., from a trimmed condition, other aircraft will not diverge far without re-trimming or a control force against the trim. Although the Airbus has trim stability, it is invisible to the pilot. This is because in Normal and Alternate law, inputs to the side-stick which would normally be for pitch control are directions to the automatic system to change the flight path; thrust and elevator trim are adjusted automatically as required.

Every pilot on his first training flight learns the basics of stall recovery which is to immediately reduce the AOA (pitch down) and add thrust or power to obtain adequate flying speed and then begin a climb. The Airbus system allowed pilots to select maximum thrust (TOGA), pull back on the stick and let the automatic system handle the AOA. The addition of maximum thrust without a pitch down and without AOA limiting would tend to accelerate an aircraft into a full stall condition.

In 1944, Wolfgang Langewiesche wrote in his famous book Stick and Rudder “When an airplane is stalled, nothing–absolutely nothing–will help except this one thing: get the stick forward. You might think, for example, that you could regain speed by slamming your throttle wide open; but it will not help as long as the stick is back. It would merely convert the stall into a slightly more vicious power stall, or the spin into a power spin: you would keep going down, out of control, just the same. It is different once the stick has been allowed to come forward and the airplane’s Angle of Attack is reduced: then wide-open throttle makes it possible to complete one’s recovery with quite small loss of altitude. But the stick must come forward first.”

A factor in the PF’s action may be how stall recovery training has been taught in recent years to airline pilots. During the arguments over the proper technique to overcome a windshear encounter which occurred in the late 1970’s and into the late 1980’s, one method which received widespread popularity was to deliberately fly the aircraft at the limiting AOA, defined by the stick shaker which simulated an impending stall. At the time, in trying to prove the performance of aircraft at the limiting AOA, stall recovery training began with airlines teaching a technique which was supposed to minimize altitude loss. A power off near stall was done, using the stick shaker as a limit, with a power on recovery, emphasizing keeping a high pitch attitude; ignoring the major considerations of immediate restoration of flying quality and the fact that in the worst case such as a severe windshear, the stall would be power on and it would be impossible to recover using this technique.

To settle disagreements over the high AOA procedure, optimal trajectory studies, funded by NASA Langley, Boeing, the state of Texas and the Air Line Pilots Association were conducted by Professor Angelo Miele and the Aero Astronautics Group of Rice University, and proved that deliberating flying at a high AOA in a windshear is a very bad idea until the end of the shear or unless ground impact is imminent. The first report was published in August, 1985, which was followed by many others after hundreds of trajectories had been proven.

Two of last reports (1989) are “Optimal Recovery Techniques From High-Angle-of-Attack Windshear Encounters: Take-Off Problem.” and the same title for the “Landing Problem”. These reports analyzed aircraft performance in strong windshears at low altitude with 2, 3 and 4 engine aircraft put at the stick shaker AOA and say; “Maintaining the aircraft at the stick shaker is a poor strategy in terms of altitude loss and survival capability.”; and “If the pilot accidentally or deliberately increases the angle of attack to the stick shaker value, following the windshear onset, the optimal recovery trajectory (ORT) requires that the angle of attack be reduced quickly to a lower value and then increased gradually in such a way that the stick shaker value is reached again near the end of the shear.”; with substantiating proof that the ORT can recover from a windshear 50 percent stronger than when a high AOA is maintained, and if a high AOA is avoided in the beginning, the aircraft can survive a windshear twice as strong.

However, many pilots are still being taught the misguided high pitch technique for stall recovery which is thought by some as necessary for low altitude stall recovery despite the conclusive proof of the studies reported above that at low altitude in extreme performance limiting conditions, a rapid decrease in AOA offers the best chance of stall recovery and survival. The accident report (page 181) in discussing the PF’s actions states “Moreover, the flight director displays could have prompted him to command a positive pitch angle, of about 12.5°. This value appears in the stall warning procedure for the take-off phase. It is possible that, even though he did not call it out, the PF had recalled this memorized value and then had clung to this reference without remembering that it was intended for a different flight phase.”

The most important lesson to be learned from this accident is there is no phase of flight or condition where immediate reduction of AOA is not the most critical element in stall recovery and/or prevention. A microburst encountered at take off can easily exceed the performance capability of one engine on a twin engine aircraft. Maintaining a high AOA when ground impact is not imminent can result in an unrecoverable stall and/or inability to soften ground impact if the microburst exceeds the total performance capability of the aircraft.

One factor mentioned in the accident report regarding pilot errors is that according to the Airbus Unreliable Speed Indications checklist (page 103) the autopilot, flight directors and auto-thrust should have been turned off. This is incorporated into an Air France checklist (page 104) with these items listed as memory items. The report notes on page 106 that other pilots in similar simulations failed to execute these memory items of the checklist. Reversion to Alternate Law was because of loss of airspeed information which was accompanied by automatically turning off the autopilot and auto- thrust. The flight directors were not turned off but appeared to be because the command bars disappeared briefly, but returned later with unreliable commands.

Lateral oscillations which persist throughout the descent are probably due to the fact that in Alternate Law, roll control changed to Direct Law, but the pitch axis remained in Normal Law (without the protections); a unique and possibly difficult mix for the pilots to master in a short time.

Expectation that AOA protection was available and training in it’s use, plus possibly other stall recovery training which did not emphasize immediate AOA reduction, are most likely the major cause of the PF’s first actions which led to the deep stall condition. It is uncertain if anything could have been done to recover from the deep stall, especially with the elevator trim full nose up. TOGA thrust before AOA reduction helped initiate the stall and aggravated the condition once in a deep stall. As Langewiesche said; “. . . . it will not help as long as the stick is back. It would merely convert the stall into a slightly more vicious power stall, . . . .”

One of the most important recommendations in the report is to provide pilots with direct AOA information (page 205). Recommendations which are not in the report should be: 1) Stall prevention and recovery training for all conditions should emphasize AOA reduction as the most critical element, 2) If AOA protection is not available for an Airbus fly-by-wire control system, auto-trim should be automatically deactivated, and 3) If the system knows an unreliable speed indication exists and the autopilot and auto-thrust are turned off, then the system should also turn the flight directors off, unless they are re-programmed to provide accurate flight path guidance in this circumstance.


BEA AF 447 Accident Final Report .pdf (English 28MB)

BEA AF 447 Accident Final Report Page (English)

BEA AF 447 Accident Main Page (English)