Glossaire Page précédente
Imprimer cette page

Crash de l'Airbus A 320 d'Armavia à Sotchi

Extrait du rapport d'enquête

La technologie Airbus, une usine à gaz pour les pilotes.

Ci-dessous, un extrait (pages 41 à 49) du rapport sur le crash de l'Airbus A 320 d'Armavia à Sotchi, le 3 mai 2006 (rapport complet disponible en pdf sur le site du Bureau d'enquêtes et d'analyses ici).

Prétendre systématiquement que "c'est la faute des pilotes" et se limiter à cette affirmation, même si l'équipage peut être impliqué, partiellement ou totalement, parce qu'il n'aurait pas respecté au millimètre près une procédure, ne résoudra rien. Surtout quand, parallèlement, les équipages sont sous-qualifiés pour la raison loufoque que "les Airbus, ma concierge peut les piloter", comme l'affirme le concepteur de cette technologie, Bernard Ziegler, certainement pas le mieux choisi au vu de ses débuts aéronautiques.

Voici donc ce que les "concierges" doivent être capables de gérer avec la technologie Airbus. On notera le problème posé, une fois de plus, par l'absence de couplage des manches de pilotage.

Ceci concerne un peu plus d'une minute de vol (de nuit, par mauvais temps, à deux heures du matin).

o - O - o

Further on, this section contains a warning about the possibility of reducing speed below VLS
when interrupting fulfilment of a landing approach on autopilot and automatic throttle, if modes
V/S or OP CLB are subsequently switched on.
The go-around procedure for RW 06 in Sochi airport provides for a coupled right-hand climbing
turn at heading 240º in the take-off configuration, with the roll angle at least 20 degrees and at a
speed not exceeding 200 kt, with climb to 600 m (2,060 ft) at the maximum possible vertical
After the controller’s instruction, at 22:11:48:5, the PTLO (PUSH TO LEVEL OFF) button was
pushed to bring the aircraft to level flight immediately. As a result, the autopilot pitch mode
changed to the VS MODE (the preset vertical speed was 0 m/s). The pitch angle increased to 6º.
Autopilot 2 disengaged automatically, since two autopilots can work simultaneously only during
approach. An audible signal advised the crew of the aircraft’s capability of performing ICAO
Category 2 and 3 approaches as downgraded (ILS CAPABILITY DOWNGRADE). The autopilot
lateral-directional mode automatically changed to HDG, with the preset heading equal to the
current aircraft heading of 62º.
At the same time as the VS MODE was activated, the pedals were moved to the -1.4º position
(the minimum pedal force required to move the pedals from the neutral position is ~10 kg). It
should be noted that the A320 does not require the use of the pedals in a normal flight. During a
turn the aircraft co-ordinates its movements automatically.
At 22:11:52 a heading of 172º was set with the use of the selector on FCU. As a result, the
aircraft entered a turn to the right with a roll angle up to 25º, maintaining an altitude of 1,114 ft
(340 m). During the turn the rudder moved to the 2.3º position, i.e. against the pedal travel
direction, and the lateral G was 0.05.
Note: In the course of experiments on the simulator undertaken in Toulouse, it was determined
that when the aircraft is in landing configuration, the turn coordination function works
with rather significant errors. The Airbus representatives admitted this and explained
that this aircraft type was not intended for manoeuvring with high bank angles in the
landing configuration.

At 22:11:53 the controller repeated his instruction to go around: “RNV 967, turn right, climb 600,
contact Radar 119.7”. The crew confirmed: “Turn right, climb 600, 119.7, 967”. At that moment
the aircraft was at a distance of 5.8 km from the RW 06 threshold (Attachment 2, Fig. 4 and 10).
At 22:11:57 on the control panel the altitude was selected to be 3,200 ft (975 m), and after that
the OPEN CLIMB mode was activated, with the thrust control levers in the CL position, flaps
and slats fully extended, the landing gear extended, the autothrust in the V/M mode, and the
preset speed at 137 kt. It should be noted that previously the preset altitude was 2,048 ft (600 m),
which corresponded to the go-around altitude. Apparently, the crew unintentionally increased the
preset altitude during the process of mode activation. However, this fundamentally influenced the
flight, due to specific features of activation of the OPEN CLIMB mode due to a variety of
differences between the target and current flight altitudes. The integration of the aircraft pitch
control and engine thrust control, as well as the logical sequence of the OPEN CLB mode are
described below.
The logic of integration of the autopilot/flight director (AP/FD) pitch control and the
autothrust control
If AP/FD pitch mode controls a vertical trajectory (e.g. V/S, ALT), then AT
controls speed.
If AP/FD pitch mode controls a speed (e.g. OP CLB), then AT controls thrust.
If no AP/FD pitch mode is engaged, then АТ controls speed.
Logic sequence of the OPEN CLB mode
for level change more than 1200 ft:
at OPEN CLB mode engagement by the pilot, V/S control with V/S target = +
8000 ft/min (40 m/s) is applied for AP/FD, and SPEED/MACH mode is engaged
for AT
when engine N1 reaches 95% N1CLB mode, AP/FD switches to SPEED/MACH
control law, whereas AT switches to the THRUST mode
Throughout this time the FMA displays THR CLB for AT and OP CLB for
The given scheme of engagement for the OPEN CLB mode ensures the uniformity
of the aircraft response in all configurations and within the whole range of the flight altitudes and
For level change less than 1200 ft:
at OPEN CLB mode engagement by the pilot, V/S control with V/S target = +
1000 ft/min (5 m/s) is applied for AP/FD, and SPEED/MACH mode is engaged
for AT
 Throughout this time FMA displays THR CLB for AT and OP CLB for AP/FD
In this case the climb is in fact performed in the vertical speed control mode.
It should also be noted that if the OPEN CLB mode is engaged less than 30 seconds after the
aircraft level off function is activated, the autopilot is authorized to

use the vertical acceleration at the maximum value of 0.3g, whereas usually it is only 0.15g.
In the automatic flight, when the OPEN CLIMB mode was engaged, the aircraft started climbing
rapidly, at a vertical speed up to 12 m/s, with pitch angle increased to 21º by 22:12:06, maximum vertical acceleration 1.27g, maximum angle of attack 10.7º, and the indicated airspeed reduced to 129 kt (240 km/h), which is 8 kt lower than the target speed. Engine r.p.m. were increased to the maximum possible value for the given position of the thrust control levers (Attachment 2, Fig.4).
At 22:12:04 the aural warning «SPEED, SPEED, SPEED» (LOW ENERGY WARNING) was
recorded. This warning advises the crew that “the aircraft energy is decreasing to the limit, below
which the engine thrust must be increased to regain a positive angle on the flight path”. At the
moment when the aural warning sounded, the aircraft altitude was 1,150 ft (350 m). The correct
crew response to this warning would be to increase engine thrust. These actions are described in QRH Section “ABNORMAL PROCEDURE”.
It should be noted that at the same time as the warning sounded, the engines switched to the
CLIMB mode, and, in accordance with the logic described above, the autopilot switched to the
speed-hold mode in the pitch channel and deflected the elevator in a nose-down direction, in
order to decrease the climb rate and to increase speed.
After the warning sounded, the thrust control levers were set in the TOGA position, therefore the
crew response was fully in accordance with the QRH recommendations. The control input on the
thrust control levers resulted in activation of the go-around modes (PITCH GO and ROLL GO),
the autothrust was deactivated (AT engaged, but not active), and the ECAM page changed from
WHEEL to ENGINE (Attachment 2, Fig. 4, 5 and 6). At 22:12:06:5 disengagement of autopilot 1
was recorded, and the rest of the flight was performed in director mode, with the Captain’s and
co-pilot’s flight directors engaged. Judging from the single annunciation of the autopilot
disengagement, the autopilot was disengaged normally, with the push-button on the side stick. At the moment when the autopilot was disengaged, the flight altitude was 1232 ft (370 m), indicated airspeed 130 kt (240 km/h), roll angle +25º, pitch angle +21º, angle of attack 10.2º, and the rate of climb about 11 m/s.
Analysis of the given flight segment shows that the crew carried out the controller’s instruction
literally, meaning that they stopped descent and started performing a right-hand turn and then
climb. Not a single action of those required in the go-around procedure (setting takeoff power,
flaps retraction from the landing configuration, landing gear retraction) was performed by the
crew. This fact shows that the crew‘s work was far from optimal, and that they were unable to
evaluate the current situation adequately. It can also be suggested that the aircraft behaviour
while manoeuvring and activation of the LOW ENERGY WARNING were unexpected for the
crew. Most probably, the autopilot was disengaged intentionally, because of doubts about
whether it was functioning correctly.
Simulation of the given segment of the automatic flight on the engineering simulator, which fully
reproduces the control laws and indications of the accident aircraft, showed perfect convergence
of the results. During the manoeuvre the autopilot worked in accordance with the established
logic. Based on the FDR readout and the results of mathematical simulation, it was determined
that there was no external influence on the aircraft. It should be noted that the A320 FCOM does
not describe this mode (climbing turn in the landing configuration).
In the course of experiments on the simulator the following was also established:

• Provided that the FCOM procedures are followed, the aircraft performs the goaround
manoeuvre with no difficulties, with a maximum pitch angle of 15-16º, the aircraft speed
not less than VAPP, and the LOW ENERGY WARNING «SPEED SPEED SPEED» not sounding.
In the case where the autopilot is disengaged during go-around, the procedure can still be
completed safely and does not induce any specific stress on the pilot, provided that he follows the command bars.
• In the case where the pilot moves the thrust control levers to the TOGA position
at the beginning of the manoeuvre (after the OPEN CLIMB mode is engaged, with a pitch angle
of +5°…+10º), in FULL configuration, with the landing gear extended, then the aircraft in fact
performs a standard go-around manoeuvre, with υmax = 16-17º and no decrease in speed. This,
therefore proves the correctness of the FCOM recommendations provided in the GO AROUND
and MISSED APPROACH procedures, which require setting the thrust control lever in the
TOGA position first. This action ensures a safe go-around manoeuvre, even with the landing gear
extended or if the aircraft configuration is not changed by one step.
• In case the autopilot remains engaged, while the aircraft is performing a manoeuvre similar to that in the accident flight, the autopilot normally completes the go-around procedure, with the maximum pitch angle not exceeding 21.5º, the short-time decrease of speed not exceeding 10-12 kt, with activation of the «SPEED SPEED SPEED» warning, and without activation of the α – FLOOR function.
The flight segment from the moment when the autopilot was disengaged to the moment,
when right-and-forward control input on the Captain’s side stick was recorded after the
segment of stabilized turn
For the purpose of analysis of the this and the next segments, it should be noted that the crew was probably in a state of psycho-emotional strain caused by a combination of the following factors:
• The necessity to carry out the go-around procedure, which is quite rarely
encountered in actual flight conditions;
• The night flight and the lack-of-sleep state of the crew (at least, of the co-pilot);
• The unstable weather conditions at the destination airport during the whole flight
and the crew being uncertain that they would land there in Sochi;
• The unexpected aircraft behaviour during automatic flight on the segment
described above.
The change in the crew’s intonation during their further conversations indicates that their stress
levels were increasing.
After the autopilot was disengaged, the Captain made a control input on the side stick forward to
9.7º (the maximum deflection angle is 16º), which resulted in decrease of the pitch angle to +4º
(while the angle of attack decreased to 2.2º…2.7º), and sideward to 17º (the maximum deflection angle is 20º), which decreased the aircraft roll angle to 7º…8º. At the same time the indicated airspeed increased to 140 kt (260 km/h), and the rate of climb decreased to 1…2 m/s. At the same time as the autopilot was disengaged, a short input on the pedals was recorded. As a result, they moved to a -2.3º position and then came back close to the neutral position. At 22:12:19 the side stick was moved to a position of -9º (increasing the right bank) and then returned to the neutral position (Attachment 2, Fig. 4, 7 and 8). As a result, the aircraft was stabilized on a right-hand turn with a roll angle of about 20º (which corresponds to the minimum roll angle in the go-around

procedure) and a rate of climb of 2-3 m/s (the pitch angle of 2.5-3.5º). Experiments on the
simulator showed that the pitch command bar on the pilots’ PFD was considerably above the
aircraft symbol and set a desirable pitch up attitude of about 8º. The FDR readout shows that side sticks of the both pilots were fully released for 7-8 seconds, and the right pedal was pushed
forward a bit (the pedal force was at least 10-11 kg). As previously mentioned, the aircraft type in
question does not require pedal inputs to coordinate a turn. The continuous force applied to the
pedals, while the side sticks were released, may be a sign that the crew’s mental state was far
from optimal.
Note: It could not be determined, based on the FDR readout, which of the pilots made the
pedal inputs. Presumably, it was the Captain who acted on the pedals, as he was pilot
flying the aircraft at this stage of flight.
Based on the character of the pedal inputs during the whole abnormal event, it could be
suggested that the Captain’s actions were inadequate. The forces applied to the right
pedal were at least 12-15 kg. In a normal situation they could not remain unnoticed by
the pilot. During development of the abnormal situation there was no necessity to deflect
the rudder, therefore it could be suggested that the pedal inputs were uncontrolled (the
Captain perhaps did not even realize the considerable forces applied to the right pedal)
and might have been caused by transfer of his knowledge of flying another previous
aircraft type, while under stress.
Starting from 22:12:20, by means of the selector on the control panel, the preset altitude of 3,200 ft (975 m) was gradually (in 5 sec) decreased to 2,048 ft (620 m). This crew action was
accompanied by intensive discussions for 10 seconds, which shows that the crew knew the
altitude that should be reached. However, since the autopilot was disengaged, the aircraft could
not climb to the given altitude automatically. The change of the preset altitude resulted in the fact that the pitch command bar on the PDF came down, and now the required pitch angle was about 4º. The command bar coming down could have produced the illusion for the crew of pitching up.
At 22:12:24, during the change of the preset altitude, the thrust control levers were moved from
TOGA position first to the FLX position and then to the CL position. The altitude at which these
inputs were made corresponded to the default value of thrust reduction altitude (1500 ft + 40 ft
aerodrome elevation). Therefore, it may be suggested that the crew was monitoring the activated
flashing light LVR CLB on the FMA. Most probably, in this case the thrust reduction altitude
coincided with the acceleration altitude, which resulted in automatic engagement of the OPEN
CLIMB mode and the target speed change to the GREEN DOT (202 kt). After the engines were
set in the CL mode, the autothrust was activated in the N1 mode (ensuring the maximum thrust
for CL throttle levers position) (Attachment 2, Fig.4 and 5). By the time the engine thrust was
decreased, the side sticks were in the neutral position, the pedals were in 1.30…-1.70 position,
the aircraft altitude was 1,576 ft (480 m), and the indicated airspeed had increased to 163 kt (303 km/h). The aircraft was performing a right-hand turn, with a constant roll angle of about 20º, and

a pitch angle stabilized at +3º. The vertical climb rate was 2…3 m/s (Attachment 2, Fig. 4, 5 and 10).
At 22:12:28 the controller repeated to the crew: “Arm, Armavia 967, contact Radar 119.7”. The
crew confirmed: “967, roger”.
At 22:12:30 a short control input on the side stick forward to 5.2º was recorded, and 3 seconds
later – a control input sideways -9.3º. These Captain’s actions resulted in a decrease of pitch to
1.4º and an increase of roll angle to 30º. At 22:12:34 the Captain started moving the side stick
forward and thereby pushed the aircraft nose down. The actual reason for such actions by the
Captain could not be determined. However, it can be stated that such inadequate piloting was
caused by a lack of monitoring of flight parameters, in particular pitch and roll angles.
In the course of the investigation the commission considered the following probable causes of
such piloting, though neither of them could be fully substantiated:
• Influence of somatogravic illusions, in particular the illusion of pitching up
experienced by the pilot flying, with a lack of monitoring of the flight indicators and longitudinal
acceleration of the aircraft, at night, with no visible references. This interpretation is substantiated by the inadequate actions of the Captain that were recorded at the moment when the aircraft deviation from the runway heading was more than 90 degrees. That means that the shore and the ground lights that could be seen through the broken clouds disappeared at that moment, although aircraft acceleration continued. On the other hand, there is evidence that shows that the Captain monitored the PFD and read it correctly, at least the speed and FMA indication.
• Specific features of speed indication on the PFD, especially speed limitations for
the given aircraft configuration that are shown as the red bars at the top of the speed indication
strip. One may imagine the influence of the reflex acquired in training, for example, in response
to a TCAS warning when the pilot is anxious to avoid the displayed red part of the instrument
scale, which may result in the instinctive forward movement of the side stick, especially when the pilot is in a state of psycho-emotional strain. This version is substantiated by the fact that the pilot was monitoring the flight speed and its limitations (VFE) that depended on the aircraft
configuration and retracted the high-lift devices in a timely manner, and the control inputs on the
side stick coincided with the moments when the current speed was getting close to the limit
At 22:12:36 the last crew communication with the ground was recorded: “Sochi Radar, Armavia
967…”. The phrase was not completed, since the Captain ordered the co-pilot, who was
communicating with the controller, to retract the flaps.
Segment where the aircraft started its final descent
The maximum altitude attained by the aircraft during the go-around manoeuvre was 510 m
(Attachment 2, Fig. 4).
The Captain’s actions described above resulted in a decrease of the pitch angle, the onset of
descent and continued acceleration at maximum continuous power. If the crew, who did not
follow the FD indications anyway, had switched off the FDs in accordance with the FCOM
recommendations, the autothrust would have switched to the speed-hold mode with the preset
speed 202 kt (GREEN DOT SPEED).
The table below provides the maximum allowable speed values for various configurations.

Configuration Slats/Flaps VFE (kt)
1 18/0
Intermediate approach, take-off
2 22/15 200 Take-off and approach
3 22/20 185 Take-off, approach and landing
FULL 27/35 177 Landing
By 22:12:37 the indicated airspeed reached the VFE value established for the FULL
configuration. At that moment the Captain ordered flap retraction, and the wing high-lift devices
started retracting from the FULL configuration to configuration 2 at once. During retraction, at
22:12:41 activation of the MASTER WARNING accompanied with the aural CRC warning was
recorded, which means that the maximum allowable flight speed for the given aircraft
configuration had been exceeded. This warning continued on all the time until the end of the
flight, except for some short breaks. At this time the flight altitude had decreased to 1,626 ft (495 m), the indicated airspeed increased to 186 kt (344 km/h), the aircraft pitch angle was -5º (nose down), with roll angle +33.5º.
At 22:12:45 the high-lift devices were retracted further to configuration 1. Though configuration
1F provides for flaps extension to 10º, the flaps were fully retracted automatically, as the flight
speed 210 kt was exceeded. At this moment the Captain made a control input on the side stick
forward to 11º, which resulted in further increase in the nose-down pitch, the descent rate and the indicated airspeed. A control input on the side stick that resulted in an increase in right bank was also recorded. Simultaneously with the control inputs on the side stick, the pedals were moved to a position of -3.4º…-4º, and the lateral G went up 0.02…0.04.
At 22:12:47 the GPWS Alarm sounded, and at the same time the co-pilot’s phrase: “Level off”
was recorded. Altitude decreased to 1358 ft (414 m), the indicated airspeed increased to 211 kt
(392 km/h), the pitch angle decreased to 11…12º nose down, and the roll angle increased to
37…39º. The aircraft descent rate was -20…-22 m/s. At this moment the co-pilot intervened and
moved the side stick to the left stop position (20º) to counter the increasing right bank, while the
Captain continued making control inputs to increase the right bank. Apparently the co-pilot was
trying to counter the bank only. However, while moving the side stick sideways to the stop
position, he had made forward control inputs on it as well (a “parasitic” input produced by high
sideward forces applied to the side stick in the stop position) (Attachment 2, Fig. 4,7 and 8).
While intervening, the co-pilot had not pressed the take-over pushbutton, therefore both pilots’
control inputs were added and averaged. Such dual piloting is prohibited by the A320 FCOM. At
that time the actions of the two pilots were not coordinated. The Captain twice moved the side
stick half-way backwards , possibly, reacting to the EGPWS, but at the same time the co-pilot
was inadvertently making nose-down inputs, which might have led the Captain to believe that the
aircraft response to the control inputs in the pitch channel was not adequate.

Note: During the onset and development of the abnormal situation the co-pilot did not comply
with the duties of non-flying pilot during a go-around phase as stipulated in the FCOM.
In particular, he did not inform the Captain of the pitch angle values below 10 degrees
pitch up and of the descent rate, when it developed.
Note: The DUAL INPUT warning was not activated because of its lower priority compared to
the EGPWS warning.
Note: As was shown by outcomes of experiments on the flight simulators and by the results of
mathematical simulations provided by Airbus specialists, after autopilot disengagement
and change-over to manual control, aircraft movement was entirely determined by
deflection of the control surfaces and the engine modes. Aerodynamic and thrust
performance of the aircraft corresponded to the characteristics of the aircraft type.
There were no external influences on the aircraft (wind shear, etc.)
The backward movement of the Captain’s side stick to -11º was accompanied by its sideways
movement to the right to -16º. At the same time the pedals were moved to -4.8º, and the lateral G went up 0.1…0.11.
Neither of the pilots fully fulfilled the FCOM requirements for crew actions in case of
EGPWS activation stipulated in the QRH “EMERGENCY PROCEDURE” Section. Flying at
night or in difficult weather conditions requires an immediate response to this warning, in
particular to move the side stick backward and hold it in this position, as well as to set the takeoff power mode. The crew’s attention might also have been distracted by a long 20-second
message from the controller regarding a change in the approach procedure, which was recorded
by the CVR along with the EGPWS and CRC warnings sounding in the background. The
controller sent the message in accordance with the controller’s operational manual, after the crew contacted him.
It should be noted that in the course of the investigation of the accident to the A320 A40-EK that
occurred during go-around at Bahrain airport on 23 August 2000, the Investigation Commission
also noted inadequate actions by the Captain, who moved the side stick forward and held it there, despite the EGPWS warning sounding for a long time. In that case the crew response to the CRC warning that was sounding at the same time was also correct (flaps retraction).
At 22:12:49 the preset vertical speed became equal to 25.4 m/s. The change in the preset value
was connected with automatic activation of the VS MODE, with the autothrust switched to the
speed-hold mode. In the case where the autopilot is disengaged, the manual control disagrees
with the command bars and the aircraft speed reaches the VFE value plus 4 kt, the mode is
changed automatically in order to prevent

exceeding the speed limitations and stabilize the speed by means of a thrust reduction, while
maintaining the flight path (FCOM 1.22.30). However, at 22:12:51:5 one of the pilots abruptly
moved the thrust control levers to the idle position and then moved them back, which resulted in
disengagement of the autothrust (Attachment 1, Fig. 4). At that time the aircraft descent rate
became -28…-30 m/s, and the flight altitude decreased to 933 ft (284 m). At 22:13:01 the FDR
recorded retraction of the high-lift devices to zero configuration (slats retracted). By then the
pedals were in -7.2…-5.5 position, and the lateral G increased to 0.18…0.25. The combined
control actions of the Captain and the co-pilot resulted in decrease of the roll angle almost to
zero, insignificant increase of the pitch angle to -6º…-7º and decrease of the descent rate to -22
Despite the actions undertaken by both pilots at the last seconds of flight to deflect the elevator to the nose-up position, at 22:13:03 the aircraft impacted the water at an indicated airspeed of 285 kt (528 km/h), a vertical speed of about -22 m/s, a pitch angle of -4º…-5º nose down and a roll angle of 9º…10º to the right. At the moment of the impact the flaps were retracted, the slats were in a position of about 18º retracted, and the landing gear was extended.
The experiments on the simulator showed that the aircraft could have been recovered at any stage of the flight until 22:12:58, provided the crew acted properly. The simulator experiment allowed data to be obtained regarding the decrease in altitude during recovery of the aircraft in a
configuration 18/0 at V =260-270kt, with pitch angles -4,5…-6,5° and various roll angles from 0
to 39º, and with the side stick in various back positions, including the fully back position. For the
parameters similar to those in the accident flight (V=270kt, pitch angle -5,3°, roll angle about
zero), the minimum altitude, at which the aircraft could be recovered and controlled to climb, was 200-230 ft. Allowing 2 seconds for the crew response to the EGPWS warning, the crew had 9 seconds spare to recover the aircraft.

Retour haut de page Page précédente

Accueil    |   Copyright © 2007, norbert jacquet    |   URL :    |   contact :