ADS-B Analysis of FlyDubai 981: A Case of Somatogravic Illusion?

On Saturday 19th March, I had just stepped outside my house to grab a taxi to go to a weekend meeting when I received a phone call from Moscow, regarding the FlyDubai flight had crashed in Rostov-on-Don with over 50 killed. I could not go on air immediately due to the meetings but after reading a bit on the wreckage, ADS-B information published by FlightRadar24, several aspects became apparent and at 0831UTC voiced I my opinion.

I did write about fatigue somewhere but it seems that I can’t remember where I said it, but… At 0900UTC I went on air at Russian Times TV voicing that opinion (including fatigue), as I did not agree with the speculation that there was an engine failure, or that the aircraft had hit the ground prior to the go-around, because the facts available then, did not support such theories.

I decided to get to work on this accident. It took a bit to understand the mechanics involving the accident, and much work still needs to be done, but after one week, I feel that I owe my blog readers and social media followers what I found on this accident.

My view remains that spatial disorientation compounded or caused by fatigue is a major cause of the accident (amongst others).

Flight summary:
FZ981-Plot1
Simplified position plot from ADS-B data for FZ981

 

  • Flight departed at 18:38UTC, made first attempt to land at aborted at 22:42UTC.
  • 1st approach was rather a simple one. They went in, according to the STAR ER22A via Manichsky NDB and then turned to the ILS for runway 22, then they decided to abort the approach.
  • After the missed-approach, the aircraft’s path was rather interesting after reaching FL080, but the aircraft eventually settled in a hold at Manichsky at  FL150.
  • The aircraft carried fuel for 8.5-hours of flight.
  • Began 2nd approach at 00:27UTC
  • The accident occurred at 03:43 local time in Rostov (00:43UTC)
  • The crew had been on duty for about 8 hours (from 16:45UTC to 00:43UTC).
  • Weather was turbulent, rainy and gusty at 240⁰ at 28 knots gusting to 44 knots and temporarily 34 knots to 50 knots during the hold.
FD-981-ROV-ILS22
ILS Approach Chart for runway 22, made by Lido. Emirates is a user of Lido charts, and FlyDubai is likely to be one too.
FD981-ROV-STAR
Standard Arrival ER22A via Manichsky NDB was used by FZ981.
Fatigue issues

Fatigue on it’s own should not have caused the accident. Something else must have caused the accident. Let’s have a look at other factors first.

Today’s High Performance Airliners – Hotrods that can surprise you!

The aircraft would be about 55 – 60 tons in weight, which would give a Thrust-to-Weight is about 0.43 to 0.47. While this doesn’t sound like a top-of-the-line fighter jet, it is:

  • Better than a BAe Hawk trainer at full weight (0.325), and the same as Bae Hawk on a typical short mission.
  • The same as a fully loaded T/A-50 high performance jet at max thrust without afterburner.
  • Better than the thrust to weight ratio of a typical training mission of an Alpha Jet (0.27)

The problem is that these airliners have the engines slung below the wing, so from a low-weight approach speed and trim, during a go-around one could experience rapid nose up movement of aircraft, and or, rapid acceleration or deceleration, depending on the pitch angle resulted. Regardless, it is not rare that one would have to fight the pitch up moment both with the yoke and with the trim switches. Whether the crew had to fight the pitch up moment in this case, needs Flight Data Recorder information. Here’s another eerily similar case in Russia involving a 737-500:

Tatarstan735Kazan
The failed go-around profile of Tatarstan flight 363 (source: Arz1969 via Wikimedia)

This was the case of Tatarstan Airlines flight 363, where upon executing a go-around, the nose-up moment generated by the engine thrust, the aircraft started to climb, reaching the pitch angle of about 25 degrees and the Indicated airspeed began to decrease as the crew retracted the landing gear. The crew did not provide positive control inputs through the yoke initiating the go-around maneuver resulting in decreasing from 150 to 125 knots, the crew then started control actions through the yoke, pitching nose down, which has led to stopping climb then starting descent and increase of the airspeed. Maximum angles of attack have not exceeded operational limits during the flight (aircraft did not stall). After reaching the altitude of 700 meters, the aircraft started a steep nosedive, with the pitch angle reaching −75° by the end of the flight (end of the recording).

FZ981-Plot2
ADS-B recorded Groundspeed and Altitude for both FZ981 approach attempts

But I wasn’t satisfied with this explanation alone, and the ADS-B data revealed a significant acceleration in the aircraft’s groundspeed during the climb. A friend (or more like my teacher), wrote about an acceleration-based sensory illusion several times during discussions of various accidents in the past, and I think we both suspected this could have played a part in this accident.

Somatogravic Illusion
InnerEar
Source: Transport Canada

Somatogravic Illusion typically occur during the missed approach or go-around segment of a flight at night or in IMC when speed is slow, power is rapidly applied and the aircraft then accelerates rapidly. As no visual cues exist, this generates a strong ‘tilt back’ sensation which the pilot interprets (incorrectly) as a rapid pitching up sensation while the aircraft may still actually be in a level attitude or only a slight climb. This is the somatogravic illusion.

SomatogravicIllusion
Forces acting on head (source: Neuroscience 2nd Edition, 2012)

A pilot unaware of the illusion will then push forward on the control column to control this imaginary climb thinking they are lowering the aircraft nose back to level flight, when they are actually lowering the nose into a dive. As the aircraft nose lowers, the aircraft continues to accelerate generating additional pitch up sensations, causing the pilot to lower the nose even further. This illusion ends with the pilot commanding the aircraft into a high speed steep dive, and ground contact could be inevitable.

SomatogravicIllusion2
Perceived vs actual flightpaths due to somatogravic illusion. (Source: MzeroA.com)

If the aircraft nose is simultaneously raised, the pilot has a very strong sensation of climbing. The illusion of false climb tends to lead the pilot to lower the nose and descend. The aircraft then accelerates and the illusion can intensify. Pilots cannot rely on their senses and must confirm the aircraft’s nose-up position using the attitude indicator on the primary flight display (PFD). Fatigue however could result in the pilot following his basic natural instinct rather than training.

FZ981-Plot3
Estimated Longitudinal (along aircraft’s axis of travel) Acceleration

ADS-B Data shows that while fateful second go-around was steeper, the longitudinal acceleration throughout the steep climb was generally higher than the first, opening the risk to somatogravic illusion. As the aircraft turned from climb to dive, we can see the longitudinal acceleration became extreme and may pose difficulty to “snap out” of the illusion.

How does somatogravic illusion and fatigue fit in with the leaked information leaks?

We can ascertain that the crew did a single autopilot approach, which was normal. When they aborted, they pressed the TO/GA switch to put the aircraft in the go-around mode. Since this is done below 2000’ above ground, the autopilot disengages. One can always re-engage it as long as no force is applied to the control yokes and the stab cutout switch is at normal.

The leaks seems to have resulted in confusion regarding the stabilizer trim. I’ve been in situations where translations can cause confusion through the media, so, I looked at a version that to me sounded well translated:

After the plane’s captain decided to attempt a second attempt at landing, the Boeing reached an altitude of 1500 meters.

“Ascent continues automatically for 40 seconds,” the source reportedly said. “For unknown reasons, the captain disengaged autopilot before the plane reached that height. Maybe the plane ran into ‘atmospheric scissors’ (pilot slang), which could severely shake it. And the moment the autopilot was cut off, the Boeing dives to the ground.”

According to the transcripts, one pilot told the other, “Don’t worry, pull!” Reporters cited experts who believe that, after the autopilot disconnected, the captain attempted to level the plane, but for some reason the vertical stabilizer was turned on after the nosedive. It is possible that, during turbulence, the captain accidentally pressed the thumb controller as he was cutting off autopilot. He likely didn’t notice it because he was extremely exhausted. As a result, he couldn’t navigate the plane, and the pilots were unable to determine the problem.

Source: SputnikNews

I think this version can lead to the following chain:

  1. Go-Around, A/P (autopilot) disconnects.
  2. Aircraft climbed for 40 seconds with  A/P possibly re-engaged.
  3. Someone was not comfortable with the situation.
  4. Someone switched A/P off, most likely through the trim thumb switches.
  5. That someone gave nose down trim through the trim thumb switches.

Remember, according to ATC recordings, the crew requested an altitude quite high as their desired altitude after a go-around, so they wouldn’t be in a rush to level the aircraft off. Even at 4000fpm, at 4000′ they still had a minute to reach FL80.

Did it stall?

Recent leaks have indicated that the aircraft may have stalled. Unfortunately these leaks have come through unnamed sources, and while we cannot rule it out as we do not have the flight data recorder information, based on on the ADS-B data, it can be said that a stall was unlikely. Why? Let us revisit the following plot:

FZ981-Plot2

We can see that on the second attempt, for each altitude, the groundspeed after the go-around were generally higher than prior to it. Despite the raising of flaps from Flaps 30 or Flaps 40 for the approach to Flaps 15 for the go-around, a stall or an approach to stall was unlikely.

While various leaked information hinted at stall, none mentioned the stall warning or stick shaker being heard. The latest leak indicating that there was disagreement between the pilots, indicates that one pilot was uncomfortable with the steep ascent and probably the acceleration (hence: somatogravic illusion), and then one wanted to continue the climb while the other wanted to lower the nose.

Surely Somatogravic Illusion is easily solved?

OK, before anyone shouts can’t they train to anticipate this illusion in the the simulator, there are a few points to remember:

  • Simulator motion cannot simulate somatogravic illusion. Simulator motion in itself, is using sensory illusion to create a feeling that an aircraft has banked, climbed or descended.
  • To create an environment rife for sensory illusion, the crew would have to be very tired, and this kind of fatigue that can screw someone’s sensory perception in some may only be generated through simulating a long flight and in conditions similar to this flight; the problem is, simulator time isn’t cheap either.

We must remember that this accident happened during a go-around when the aircraft was very light. The pilots (at least one of them) were probably not used to the rapid acceleration and steep ascent, and one does not usually train for an extreme lightweight go-around.

Conclusion

While Somatogravic Illusion in my opinion forms a strong case on what may have caused the aircraft to dive, there will be other factors being investigated by the accident investigators, MAK. Despite the leaks, MAK have consistently refused to comment on the leaks (as they should), and that they have stated they will likely be making a statement on their initial findings within the next week or two.

Other cases of Somatogravic Illusion:
Recent cases of fatigue and spatial disorientation involving Boeing 737-800s
Comments through Facebook

Leave a Reply