A Take-off that used 4,000 metres of runway, and then some
On 15 August 2026, a Vietnam Airlines Boeing 787-9 departed from Munich for
Hanoi and became airborne from grass beyond the end of runway 26L. Runway
26L is 4,000 metres long, and the afternoon was clear with light, variable
winds.
The tail struck the ground; the main gear and tail took out 20 approach
lights, and the crew returned two hours later with damaged tyres. All 287
people walked away. The aircraft did not walk away as cleanly; the BFU
classifies it as an accident with substantial damage.
The German BFU published its interim report on 28 September. It names no
cause yet, and neither will I. But the recorder data already tells a story
every airline pilot, training captain, and safety manager should read
closely. Here is my take, as someone who has spent a career on both sides of
the check-ride table.
What the interim report says
The setup was hot and heavy, but nothing exotic. Munich sits at 1,487 ft.
The temperature was 33°C, and the take-off mass was 240,857 kg, against a
maximum of 247,207 kg. The crew calculated V1 at 169 kt, Vr at 175 kt and V2
at 180 kt.
On board were 271 passengers, 12 cabin crew and four pilots. The captain on
take-off was 64, with 27,182 hours, including 7,795 as a 787 captain. The
pilot monitoring had 12,375 hours. The operator's nominated pilot in command
sat in the left jump seat.
The key moments from the flight recorders, in local time:
The tyre tracks ran across the 60 m clearway and ended about 240 m past the
runway end. Stall protection extended the slats automatically 1 second after
V1, with pitch at 12.2° and an angle of attack of 14°. Read that again: this
aircraft reached its decision speed in the air.
What's new since the interim report
Ten days on, there is still no cause and no BFU safety recommendation. As
far as I can tell, Boeing, the FAA, and EASA have not issued any
airworthiness action on the 787 brake system. However, several details have
emerged that sharpen the picture.
How far the pedals moved. Vietnamese coverage of the report gives the
maximum recorded brake pedal angle as 7.2 degrees, on a scale that runs to
12.5. That is more than half the available travel. If it were a foot, it
would be more than a light brush of the toes. If it were a sensor, it would
be a convincing false signal. Either way, it is exactly why hardware testing
matters.
Where the investigation is looking. Aviation Week reports that the BFU is
concentrating on the Brake System Control Units and the Electric Brake
Actuator Controllers. With the take-off audio overwritten, the answer will
have to come from flight data and the components themselves.
The crew. The four pilots comprised three captains and one first officer,
with 63,938 hours between them and 19,675 hours on the 787. Whatever
happened here, inexperience was not the issue.
The response in Hanoi. Vietnam's Ministry of Construction directed the CAAV
to coordinate with the BFU, report findings and corrective measures, and
tighten safety oversight of training, operations and maintenance across
Vietnamese airlines. Vietnam Airlines was told to work with the BFU, Boeing
and the engine manufacturer, and to review its training, operations and
maintenance processes.
Eighteen seconds that ate the runway
The single most important number in this report is not a speed. It is a
duration: roughly 18 seconds, at an airspeed of about 168 kt.
Do the arithmetic. 168 kt is about 86 m/s. At that speed, 18 seconds cover
about 1,550 metres of runway, close to 40% of 26L, without gaining a single
knot. A heavy 787 would normally reach Vr within the last 7 knots, in a few
seconds. You do not need a performance error or a weak engine to explain the
overrun. That plateau alone does it.
Thrust does not appear to be the culprit. One second before rotation, the
pilot monitoring increased the thrust from 100% to 101% N1. The engines were
delivering. Something was taking that thrust away, and the recorders show
what was competing with it: the brakes. Feet on the brakes, or a fault in
the box?
Three seconds into the roll, the recorders logged rudder pedal movement and
both left-seat brake pedals moving together. Tyre marks were later found
near taxiway B4. Left braking was higher than right throughout.
Anyone who has spent time in a simulator as an instructor has seen this
picture. The heels creep up the pedals, the pilot makes small rudder
corrections to hold the centreline, and toe pressure follows the rudder. If
the left brake drags more, the nose wants to go left, right rudder comes in,
and the cycle feeds itself. It is a human, physical, and very ordinary
failure mode.
But I would not hang this on the pilot yet. The 787 uses electric brakes,
and the recorded "pedal movement" comes from position transducers. A
transducer or controller fault could produce a brake command with no foot
pressing anything. The BFU has seized the Brake System Control Units and the
Electric Brake Actuator Controllers, and that hardware will be decisive. The
honest framing today is pilot input, sensor or controller fault, or some
combination of the two.
The V1 trap: This is the part that should keep training departments up at night.
According to Vietnam's preliminary account, the crew reported that
acceleration stopped for about two seconds halfway down the runway. They
then saw speed fall after V1, judged there was insufficient runway to stop,
selected maximum thrust and continued.
The recorders tell a different story. The plateau lasted about 18 seconds,
not two. They never reached V1 on the runway. The crew made a continue
decision at a decision speed they had not yet achieved. I am not criticising
them for that. Time compression under surprise and high workload is well
documented, and I have watched excellent pilots misjudge seconds in the
simulator.
The deeper point concerns what V1 is. V1 is a speed that stands in for a
distance. It only means what we think it means if the aircraft accelerates
as the performance calculation assumed. Once acceleration degrades, the link
between speed and remaining runway breaks, and V1 quietly stops protecting
you. By the time this crew could see the problem clearly, neither stopping
nor going left left any margin for what anyone had planned. We train V1 as a
hard line, and for engine failures that is exactly right. We rarely train
crews to recognise a take-off that is simply too slow, or to judge
acceleration against the distance markers as they pass. Most fleets have no
take-off acceleration monitoring to help them. I expect this to become one
of the central systemic findings.
The rotation, the tail strike and the trip back
Rotation began at 163 kt, 12 kt below Vr, with the aircraft already level
with the far threshold. Rotating that early, at that pitch, made a tail
strike close to inevitable. Given that the alternative was running into
the approach lights at full thrust, I would call that rotation the
least-bad option available. The tail strike was a consequence, not a
separate event.
The return flight was methodical. The crew declared PAN, suspected a
flat tyre, dumped 62 tonnes of fuel over about 45 minutes, and flew two
gear-down passes so the ground could inspect the gear before landing. If
they suspected a tyre failure, why did they retract the gear? Not once,
but thrice!
But a fair question remains. A Swiss newspaper's analysis of the
published radio calls suggests the crew may not have grasped the full
scale of what had happened, even after liftoff. A suspected tail strike
shifts priorities: possible structural damage argues for remaining
unpressurised and landing promptly, rather than jettisoning weight. I
expect the investigation to examine whether the crew recognised the tail
strike and what the QRH required.
Two captains, a commander in the jump seat, and a silent recorder
The flight deck arrangement warrants attention. A very experienced captain handled take-off, while the operator's nominated pilot in command sat in the left jump seat and took control only at 9,000 ft. With two captains up front and the commander not in a pilot seat, who owned the reject-or-continue decision? Authority gradients in augmented crews are a standard CRM line of inquiry, and this event will test that question.
Then there is the cockpit voice recorder. Each recorder holds two hours
of audio. The aircraft landed and remained powered for more than two hours
after take-off, so the take-off audio was overwritten. We will never know
whether V1 and rotate were called, or what was said during those 18
seconds. Under ICAO Annex 13 guidelines, the aircraft operator must take
all necessary precautionary measures to protect and preserve the recorded
evidence, which, logistically, means powering down the units as soon as
the aircraft is safely parked. Many regulators have issued supplementary
orders to the airlines to ensure compliance with this requirement.
VN-A867 was built in 2016, before ICAO's 25-hour CVR requirement for new
aircraft. Every operator flying older types should take this as a
reminder: your procedures for protecting recorders after an event matter,
and the clock starts at the event, not at the gate.
Preventive measures: what can be done now
So far, the official response has been oversight and review. That is
necessary, but it is not a fix. We do not need the final report to act on
what is already evident, and the measures fall naturally into two
groups.
For operators and training departments
1. Foot position on the take-off roll. Make heel contact with the floor an explicit SOP item for the take-off roll, and have instructors and examiners observe and debrief it.
2. Degraded-acceleration scenarios in the simulator. Not engine failures: dragging brakes, a heavy aircraft that will not accelerate, a take-off that does not feel right. Let crews discover how quickly 4,000 metres disappears.
3. Teach V1 as a distance proxy. Crews should know when V1 is no longer valid. Give them a practical cross-check of speed against runway remaining, using distance markers or an operator-defined acceleration gate.
4. A callout for stagnant speed. Train the pilot monitoring to call "speed not increasing" as a defined event, the same way we train "engine failure". A named callout turns a vague unease into a decision point.
5. Know what dragging brakes look like. Brief the cues: a yaw requiring constant rudder, poor acceleration, and rising brake temperatures on the synoptic. On the 787, those cues are subtle, which is precisely the point.
6. Clarify authority in augmented crews. Brief who decides to reject or continue, and what the jump seat commander does, before every take-off.
7. Tail strike discipline. Refresh recognition of a probable tail strike and the QRH priorities that follow, including pressurisation and a prompt landing.
8. Protect the evidence. Make recorder preservation a fixed item after any event that may need investigation and remember the CVR clock starts at the event.
For manufacturers and regulators
1. Take-off acceleration monitoring. Airbus has introduced take-off monitoring functions on some newer types. This event makes a strong case for comparable protection across the fleet, alerting the crew when actual acceleration falls short of the planned profile.
2. Brake-on-take-off alerting. If the brake command rises above a small threshold while take-off thrust is set and no reject is in progress, the crew should be alerted immediately. The 787 already measures the brake command, but no one on the flight deck is alerted.
3. Recorder duration. ICAO's 25-hour CVR requirement applies only to new aircraft. A large in-service fleet still carries two-hour recorders, and this investigation shows the cost of that gap.
4. Brake system scrutiny. If the BFU's testing finds a fault in the control units or actuator controllers, expect rapid action. Until then, a prudent 787 operator would review its own brake-related defect history now, not later.
The last thought is the uncomfortable one. Nothing failed in the classic sense here: no engine quit, no warning sounded, and the weather was benign. The aircraft simply did not accelerate, and our training and technology are mostly built to detect failures, not a take-off that is quietly too slow.
No comments:
Post a Comment