INVESTIGATION REPORT

www.bea.aero
@BEA_Aero

Serious incident to the BOEING - 737 - 800

registered 7T-VJM
on 6 December 2019
at Paris-Orly (94)
(1)
Except where Time Around 11:15(1)
otherwise indicated,
the times in Operator Air Algérie
this report are Type of flight Commercial air transport of passengers
in Coordinated
Universal Time (UTC). Persons on board Captain, co-pilot, cabin crew and passengers
One hour should be Consequences and damage None
added to obtain the
legal time applicable This is a courtesy translation by the BEA of the Final Report on the Safety Investigation
in Metropolitan published in October 2021. As accurate as the translation may be, the original text in
France on the day French is the work of reference.
of the event.

Loss of altitude during go-around

1 - HISTORY OF THE FLIGHT

Note: the following information is principally based on the data from the Quick Access
Recorder (QAR) and the radio communication recordings. A written report made by
(2)
Flight Safety the captain was also supplemented by a telephone interview with him, in the presence of
Bureau. the head of the FSB(2) for Air Algérie and a representative from the Algerian civil aviation
authority.

The crew took off at 08:53 from Tlemcen airport (Algeria) bound for Paris-Orly airport.
A little over two hours later, on an ILS approach for runway 25, the crew pre-set an
altitude of 2,000 ft on the Mode Control Panel (MCP), corresponding to the altitude
to be joined in case of a missed approach. The co-pilot was then PF. The controller
cleared them to land at 11:12:02. They disengaged the autopilot (A/P) at 11:14:01
and then the auto-throttle (A/T) at an altitude of around 600 ft to carry out a manual
landing.
At 11:14:10, when the plane was at an altitude of 477 ft (i.e. 13 ft below the decision
altitude (DA), the controller ordered the crew to go around after the activation of
a Runway Incursion Monitoring and Collision Avoidance System (RIMCAS) warning.
The controller did not inform the crew of the reason for this instruction to go around.
The published missed approach path for runway 25 specifies climbing straight ahead
and then, at an altitude of 700 ft, turning left towards the magnetic track 199° and
climbing to 2,000 ft initially.

The BEA investigations are conducted with the sole objective of improving aviation safety and are not intended
to apportion blame or liabilities.
1/11 BEA2018-0696.en/October 2021
 The reason for this
(3)
At 11:14:14, at an altitude of 401 ft and a height of 117 ft, the crew engaged the TO/
Master Caution was GA mode by pressing the TO/GA button on the power levers (point in figure 1
not determined.
below) and set a nose-up attitude. As the A/T was not manually armed after being
(4)
It was not disengaged, it did not engage itself and the crew manually pushed forward the power
possible, based on levers to a position corresponding to around 90 % of N1. The aeroplane reached
the QAR data and
the statement, to a minimum radio altimeter height of 73 ft before regaining altitude. The MASTER
determine with CAUTION light came on for twelve seconds (3). The engine thrust and the automatic
certitude the roles increase in the nose-up trim, despite the nose-down inputs on the control column (4),
(PF or PM) of each
pilot during the increased the pitch to 18°, slightly above the Flight Director (F/D) horizontal bar (15°).
go-around. The vertical speed quickly increased up to 4,000 ft/min.
At 11:14:21, at an altitude of 380 ft, the crew retracted the flaps to the 15° position
and then the landing gear. The control column inputs continued to follow the F/D
command bars. At an altitude of 930 ft, the LNAV mode was automatically engaged,
followed by, at 1,210 ft, the ALT ACQ vertical mode which took the target altitude of
2,000 ft pre-set by the crew during the final approach. The rate of climb was then at
its maximum at 4,100 ft/min.
At 11:14:46, at an altitude of 1,340 ft, the A/T engaged, after being armed by the crew,
following the engagement of the ALT ACQ mode. The MCP SPD mode to maintain a
speed of 171 kt corresponding to the speed at the time of the engagement became
active. The A/P was not re-engaged. The vertical bar of the F/D gave a left correction
cue to join track 199° of the missed approach procedure. The plane was around
150  m to the right of the runway centreline probably due to the wind gradient.
The crew started the left turn. At 11:14:59, the altitude was 2,000 ft and the vertical
speed 1,300 ft/min. The N1 values were 72 % and decreasing. The bank angle reached
(5)
Alert generated by a maximum value of 38° and the “BANK ANGLE”(5) alert was activated. During the
the GPWS when the
turn, the crew retracted the flaps to position 2°. The aeroplane’s speed, which had
bank is more than 35°
and characterized by decreased, was 160 kt, i.e. 11 kt below the target speed of the MCP SPD mode.
the “BANK ANGLE, The  maximum altitude reached during the turn was 2,070 ft. The control column
BANK ANGLE”
inputs were consistent with the F/D command bar cues.
voice message.
With a nose-up attitude of 4.0° and around 65 % of N1, the altitude started to
decrease and the speed to increase. At 11:15:15, the horizontal command bar of the
F/D gave a nose-up cue to reach around 10° pitch. The pitch attitude remained below
5° and the plane came out of the turn on a heading of 200°. The altitude was around
1,900 ft and the vertical speed in descent was more than 1,500 ft/min. At the same
time, the controller asked the crew to climb to 3,000 ft and to turn to heading 160°.
The crew modified the altitude selection on the MCP which resulted in the transition
from the ALT ACQ vertical mode to the Vertical Speed (V/S) hold mode, the target
vertical speed being the speed when the mode changed, i.e. 1,100 ft/min in descent
(point). After this mode reversion, the crew followed the F/D command bar cues
which kept the plane in descent. The engines were at 45 % of N1 and still decreasing.

2/11 BEA2018-696.en/October 2021

 At 11:15:34, the crew read back the altitude of 3,000 ft again at the controller’s request.
The crew disengaged the A/T and progressively increased the thrust to 50 % of N1.
(6)
GPWS alert At the same time, the “DON’T SINK” (6) alert was activated. The aeroplane was at an
characterized by the
altitude of 1,556 ft, a height of 1,260 ft with a vertical speed in descent of 1,200 ft/min
“DON’T SINK” voice
message and the (point). At 11:15:37, the crew engaged the A/T in MCP SPD mode again with a target
PULL UP message on speed of 175 kt. A few seconds later, the controller told the crew that he could see the
the artificial horizon.
plane descending on the radar and asked them to climb to 3,000 ft. From 11:15:49,
at an altitude of around 1,300 ft, the crew fully retracted the flaps, progressively
increased the pitch up to 11° and thrust up to 70 % of N1 by disengaging the A/T
(point). At 11:15:56, the stick shaker was momentarily activated and at 11:15:57,
the crew reduced the pitch. The horizontal command bar of the F/D still gave a
cue to take a pitch to hold a vertical speed in descent of 1,100 ft/min in V/S mode.
The “DON’T SINK” alert was activated again.
In the following thirty seconds, the crew held level flight at an altitude of around
1,300 ft during which the speed increased up to 292 kt. During this acceleration,
the crew extended the flaps to the first detent at a speed exceeding the flap placard
speed by 20 kt (the flaps were retracted a minute later). The crew engaged the
A/T again and then climbed to an altitude of 3,000 ft with a vertical speed of more
than 4,000 ft/min for ten seconds. The horizontal command bar of the F/D gave a
nose‑down cue until the engagement of the ALT ACQ vertical mode. At 11:17:36,
the crew engaged the A/P at 3,000 ft (point) and were then vectored by
the controller for a new approach. They landed on runway 25 at Orly at 11:37.

3/11 BEA2018-696.en/October 2021

 Figure 1: Horizontal and vertical paths of 7T-VJM based on QAR data

2 - ADDITIONAL INFORMATION

2.1 Crew information

The captain held a valid Airline Transport Pilot Licence (ATPL). He had logged
approximately 8,000 flight hours of which 1,400 hours on Boeing 737. He had
obtained his Boeing 737 type rating in May 2018 and had flown 230 hours during the
previous three months.
The co-pilot held a valid Commercial Pilot Licence (CPL) and had logged around 3,700
flight hours of which 1,700 hours on the Boeing 737. He had obtained his Boeing 737
type rating in May 2017.

4/11 BEA2018-696.en/October 2021

Due to the time which had elapsed between the incident and the telephone interview,
the captain could no longer precisely recall the occurrence. However, he indicated
that he and the co-pilot were preparing for the landing and that they had been very
surprised by the instruction to go around. They had not understood what had caused
it as they were around 100 ft from the ground with the runway in sight and clear.
According to him, this was a preoccupation during the go-around. He heard the
warnings but could not remember which ones. He could no longer remember the
conditions which led to the loss of altitude.

2.2 Weather conditions

The 11:00 and 11:30 Orly airport METARs indicated wind from 200°, 10 kt,
visibility 4,000 to 4,500 m, mist, light rain and broken clouds (5 to 7 octas) based at
500 ft.
The data recorded by the aeroplane shows the presence of a wind gradient with a
40 kt wind at 1,500 ft.

2.3 Activation of RIMCAS warning

The RIMCAS warning was triggered by the presence of a bird-control vehicle at
holding point W33 (north of the runway at around 700 m from the threshold). The
crew of 7T-VJM had just been cleared to land when the LOC assistant controller, over
the dedicated frequency, asked the driver of this vehicle to exit the runway safety
area, but the driver did not reply. The LOC controller then asked the crew to perform
a go-around.
The driver indicated that he had not heard the radio message as he was in the process
of scaring off birds using gun fire. Furthermore, he had ensured that he was not in the
runway safety area.
The air navigation services indicated that the RIMCAS warning was generated due
to an erroneous system configuration. On completion of the work carried out up to
2 December on runway 25, four days before the event, the position of holding
point W33 had been modified and moved 12 m closer to the runway centreline. The
RIMCAS configuration, which took into account the position of the holding points,
had not been modified accordingly. The warning was therefore activated when the
bird‑control vehicle had not exceeded the holding point.
The RIMCAS configuration was corrected after the incident and the RIMCAS monitoring
surface now corresponds to a 90 m rectangle on both sides of the runway centreline
and separate from the position of the holding marks, in accordance with the runway
25 safety area.

5/11 BEA2018-696.en/October 2021

 2.4 Management of automatic systems
(7)
This point was
brought to light by 2.4.1 Managing a go-around with low weight or with a low published altitude
the Aeroplane State
Awareness during The climb performance of modern twin-engine jets tends to reduce the time given
Go-Around study to the crew to manage a dynamic manoeuvre with a high workload(7). As mentioned
carried out by the
BEA in 2013 (ASAGA). by Boeing in the Flight Crew Techniques Manual (FCTM), one of the main risks during
See paragraph 2.8. a go-around with low weight or with a low published altitude, is the exceedance of
this altitude. In order to minimize this risk, the FCTM recommends adopting a rate of
Autopilot
(8)
climb of 1,000 to 2,000 ft/min. To do this, the FCTM suggests using the A/T but also
Flight Director
System: aeroplane recommends disconnecting all the automatic systems (A/P, A/T and F/D) and to level
management off manually if the crew considers that there is still a risk of exceeding the altitude.
system composed
In particular, it is explained in this manual, that the AFDS(8) control laws limit the A/P
of the autopilot,
flight director and or F/D attitude variations for the comfort of the passengers which may lead to the
auto-throttle. target altitude being exceeded without crew action.

2.4.2 Altitude acquisition

The ALT ACQ mode of the AFDS is automatically engaged when the plane approaches
the selected altitude in climb or descent. The AFDS then changes to ALT HOLD mode
when the aeroplane is less than 60 ft from the target altitude, provided that the
vertical speed is less than the absolute value of 300 ft/min.
During the flight, 7T-VJM exceeded the altitude of 2,000 ft +/-60ft with a vertical speed
of always more than 300 ft/min. The ALT HOLD mode was thus not able to engage
although the PF had adopted a pitch slightly below the F/D cues. The reduction
in attitude indicated by the F/D horizontal bar cues was insufficient to reduce the
vertical speed and acquire the altitude. It is probable that this was due to the AFDS
limiting the attitude variations (see paragraph 2.4.1).

2.4.3 Mode reversion

When the AFDS is in ALT ACQ mode and the crew modifies the selected altitude,
this results in a V/S mode reversion, the target speed being the instantaneous
vertical speed of the plane when the altitude was modified. The mode reversion
is indicated to the crew by the mode being displayed in a box on the Flight Mode
Annunciator (FMA) for ten seconds, without an audible alert. The system does not
check for consistency between the vertical speed and the selected altitude. During
the incident, the selected altitude of 3,000 ft corresponded to a climb but the AFDS
took the instantaneous vertical speed of the plane as the target speed in V/S mode
which was 1,100 ft/min in descent.

6/11 BEA2018-696.en/October 2021

2.4.4 Display of vertical speed target
The target vertical speed can only be consulted on the vertical speed scale on
the Primary Flight Display (PFD) where it is represented as a double magenta line
(see Figure below).
 

Source: Boeing - Copyright © 2021 Boeing. All rights reserved

Figure 2: Vertical speed bar on B737 PFD.

1: actual vertical speed indicator.
2: target vertical speed indicator (here around -1,600 ft/min).
3: value of actual vertical speed.

2.5 Crew’s management of go-around

The data from the Cockpit Voice Recorder (CVR) was not preserved for read-out due
to the 2 hour recording time and the fact that another flight was scheduled after the
incident flight. Despite requests addressed to the Algerian civil aviation authority to
collect the crew’s statements, it was not possible to organise an interview with the
co-pilot.
This absence of information limited the analysis of the incident and certainly meant
that it was not possible to establish all the safety lessons which could have been
drawn from this serious incident. Nevertheless, it transpires from the information
collected that:
ˆ The high vertical speed, the turn, the retraction of the flaps and the sharp
increase in the headwind during the approach to 2,000 ft could have constituted
stabilization and levelling off conditions which were not conducive to the AFDS
acquiring the altitude.
ˆ The crew did not disconnect the A/T and the F/D as it is recommended in the FCTM
when the target altitude risks being exceeded. They had perhaps not sufficiently
anticipated this risk.

7/11 BEA2018-696.en/October 2021

 ˆ The crew were surprised by the instruction to go around. This gave rise to some
questioning during the phase which followed and may have absorbed numerous
attention resources. The turn and the relatively low initial stabilization altitude of
the missed approach procedure along with the activation of the MASTER CAUTION
alerts probably led to a high workload for the crew. The surprise combined with
the increase in workload may explain certain inaccuracies observed in the piloting
of the plane (deviations from flight path, flight at more than 250 kt at low height
for around two minutes, extending flaps to the first detent 20 kt above the flap
placard speed).
ˆ The crew probably followed the guidance cues given by the F/D command
bars without having checked that the associated modes and target values were
compatible with the flight path to be followed.
ˆ The time taken to react to the “DON’T SINK” alert and to the controller’s instruction
and then requests to climb to 3,000 ft along with the level flight of thirty seconds
at around 1,300 ft can be explained by the crew’s difficulty in assessing or
understanding the situation.

(9)
https://bea.aero/ 2.6 Go-around altitude on runway 25 at Orly
en/investigation-
reports/notified- In the scope of the investigation into the incident to the Airbus A350 registered
events/detail/
incident-to-the-
F-HREV operated by French Bee that occurred on the 4 February 2020 at Orly(9),
airbus-a350- the BEA was able to determine, with the help of the Orly air navigation services,
registered-f-hrev- that out of the eight go-arounds at Paris-Orly in 2019 giving rise to a deviation of
operated-by-
french-bee-on-04-
more than 200 ft with respect to the cleared altitude, six had a stabilization altitude
02-2020-at-orly/ at 2,000 ft. However, the go-arounds with a cleared altitude at 2,000 ft only represent
21% of all the go-arounds in the same period(10).
(10)
The other
go‑arounds were At the end of the investigation into the incident to F-HREV, the BEA recommended to
initially cleared by
the DSNA (the French air navigation service provider) that it study the feasibility of
air traffic control
to climb to a increasing the published missed approach altitude at Paris-Orly airport to give crews
higher altitude. more time to carry out all the tasks associated with a go-around procedure and limit
the risk of a path deviation.

2.7 Similar event

A similar incident to that of 7T-VJM occurred on 1 June 2019 to a Boeing 737-800 on
approach to Bristol airport (United Kingdom). The air traffic controller ordered the
crew to perform a go-around because the plane was below the glideslope. The crew
flew a missed approach with the selected altitude on the MCP at the DA of 1,000 ft
whereas the go-around procedure altitude was 3,000 ft. The plane was slightly above
1,000 ft, the ALT ACQ mode activated to capture the altitude and the PF followed
the F/D cues in manual control. The crew then set 3,000 ft for the selected altitude,
the AFDS changed to V/S mode and thus held the instantaneous vertical speed which
was in descent at this time. The descent continued for 32 s before the crew corrected
(11)
The report is
available on the the flight path. The United Kingdom Air Accidents Investigation Branch (AAIB)
AAIB website page. did not determine the reason why the crew had not detected the descent(11).

8/11 BEA2018-696.en/October 2021

 2.8 Study on Aeroplane State Awareness during Go-Around (ASAGA)
https://www.
(12)
In 2013, the BEA published a study on Aeroplane State Awareness during Go-Around
bea.aero/en/
safety-studies/
(ASAGA(12)).
access-to-studies/
aeroplane-state- 2.8.1 Startle effect and degradation of crew cooperation
awareness-during-
go-around/ It transpired from this study that “carrying out a go-around procedure and the associated
workload can disassociate the two crew members for a too long period of time”. This is
all the more so when the go-around is not on the crew’s initiative, the instruction to
carry out a go-around will then startle and may unsettle the crew.

2.8.2 Interception of go-around altitude

The difficulty of intercepting the go-around altitude is also a characteristic of the
scenarios studied in the ASAGA study, both brought to light by the analysis of the
occurrences and expressed by nearly half of the 831 pilots who replied to the BEA
questionnaire. This difficulty particularly stands out when the published go-around
altitude is low (less than 2,000 ft above the DA) as is the case for the missed approach
procedure on runway 25 at Orly. The interception difficulty is partly due to a
“mismatch” between the construction of the missed approach procedures and the
available thrust on modern twin-engine jets. These planes can produce very high
climb speeds especially when they are light, reducing the available time to carry out
the actions and calls required for the go-around procedure before the interception.
The TO/GA thrust
(13)
On the Boeing 737, the thrust limitation by the A/T(13) limits the risk of an erroneous
remains available by
perception of the attitude (somatogravic illusion) and also offers the crew more time
pressing the TO/GA
switch a second time. to carry out the go-around. However, when the crew manually manage the thrust,
they do not benefit from this limitation.

2.8.3 Management of automatic systems

During the simulations carried out in the scope of the study, it was noticed that
the crew’s reading of the FMA was often deficient during the go-around. Up to ten
undetected FMA mode changes were observed during the same go-around, although
some of these had a direct consequence on the tracking of the flight path. The non-
detection of FMA mode changes by the two crew members is principally linked to
cognitive saturation, time pressure, the absence of a defined visual scan pattern and
the workload associated with a missed approach.
Likewise, on intercepting or selecting the go-around altitude, simulations showed
that mode reversions were sometimes not detected.
Based on these observations, the BEA issued the following recommendation:
(14)
European Aviation “EASA(14), in coordination with the major non-European certification authorities, ensure
Safety Agency.
that aircraft manufacturers modify ergonomics so as to simplify the interpretation
of FMA modes, and facilitate detection of any changes to them; [Recommendation
FRAN‑2013‑037].”

9/11 BEA2018-696.en/October 2021

EASA replied in 2014 that this recommendation had already been considered by
Certification Specifications (CS) 25.1302 and that the Automation Policy developed by
the agency also dealt with this subject. The BEA considered this response inadequate
given the various studies which showed the difficulties of reading and interpreting
the FMA modes, all the more since the modification of the ergonomics on existing
planes was not dealt with by EASA.

3 - CONCLUSIONS
The conclusions are solely based on the information which came to the knowledge of the
BEA during the investigation. They are not intended to apportion blame or liability.

Scenario
When the plane was on final for runway 25 at Orly, a Runway Incursion Monitoring
and Collision Avoidance System (RIMCAS) warning was activated due to the presence
of a bird-control vehicle in the vicinity of the runway safety area. However, the
bird‑control vehicle was behind the holding point and outside the runway safety
area. The erroneous activation of the RIMCAS warning was due to the relocation of
the holding point, after work, not being taken into account in the system parameters.
Due to the RIMCAS warning and the absence of a reply from the driver of the vehicle,
the controller ordered the crew to perform a go-around after they had passed through
the decision altitude (DA).
The crew carried out the go-around by initially displaying a thrust and a pitch attitude
which resulted in a high vertical speed. They then engaged the auto-throttle (A/T)
although the autopilot (A/P) was disengaged. The left turn was started late and the
high bank angle triggered the “BANK ANGLE” warning. The missed approach altitude
was exceeded during the turn. Under the combined effect of a nose-down input from
the PF and the A/T reducing the thrust, the plane next re-descended below 2,000 ft.
The controller then cleared the crew to climb to 3,000 ft. The crew’s selection of this
altitude caused a V/S mode reversion with the instantaneous vertical speed taken as
the target speed (-1,100 ft/min).
The crew then followed the flight director (F/D) cues for around 20 s. During the
descent, the controller contacted the crew three times, the “DON’T SINK” alert was
activated and remained active for 45 s.
The crew stopped the descent at around 1,300 ft and levelled off for 30 s while
accelerating up to 292 kt. They then resumed the climb to 3,000 ft. After stabilizing
the plane at 3,000 ft, the crew carried out a second, uneventful approach.

10/11 BEA2018-696.en/October 2021

Contributing factors
In the absence of a CVR recording and precise statements, it was not possible to
precisely analyse the crew’s actions. The following factors may, nevertheless, have
contributed to the observed deviations from the procedure and tracking of the path
during the go-around:
ˆ The startle effect linked to a go-around ordered by the controller when at low
height.
ˆ The missed approach path with a low published altitude and a left turn in initial
climb which creates a high workload in a short time.
ˆ The crew’s application of an initial high thrust given the stabilization altitude of
the missed approach.
ˆ Piloting based on a hybrid use of automatic systems (A/P, A/T and F/D) which
was not conducive to acquiring the published altitude of the missed approach
procedure.
ˆ A breakdown in crew cooperation which may be explained by the startle effect
linked to the go-around instruction and to the workload mentioned above.
ˆ The display of the vertical speed target value on the PFD which may require a
verification on the MCP. This may have contributed to the crew not detecting that
the target value was not consistent with the desired path.
ˆ The absence of a system check for consistency between the action carried out
(selection of a higher altitude) and its result (mode reversion leading to a descent)
along with the absence of a crew alert.
The controller’s messages to warn the crew of the plane’s descent, along with the
GPWS warnings, probably contributed to the crew realising that they were on an
erroneous path and to them levelling off the plane.
Safety lessons
Management of automatic systems
This occurrence shows that the conclusions of the ASAGA study are still relevant.
In particular, the fact that the crew followed the F/D cues when the AFDS was in
V/S mode with a negative vertical speed following a mode reversion, supports the
BEA’s recommendation FRAN-2013-037 to EASA regarding the improvement to the
ergonomics to facilitate the interpretation of the FMA modes and the detection of
any changes to them.
Effective crew synergy to closely monitor the FMA modes remains essential. Crews
must not hesitate to disconnect the automatic systems when they no longer
understand them.

11/11 BEA2018-696.en/October 2021