Fatal Go Around at Tatarstan Part Two

4 Aug 17 9 Comments

Last week, we left off as the flight crew of flight TAK363, a Boeing 737-500 inbound to Kazan Airport in Tatarstan, broke off their approach for a go around.

The first officer had finally made visual contact with the runway but much too late.

Captain: “Where do you see it? I don’t. Where is it?”

First Officer: “Here it is, the runway. No. Go around. Go around.”

Captain: “Go around, report we are going around. Position unsuitable for landing.”

At that moment, the aircraft was configured for landing (gear down, flaps set to 30°), travelling at 130-135 knots (IAS) level at 270 metres (900 feet) above the runway maintained by the autopilot. The autopilot was in the process of an S-shape manoeuvre to attempt to capture the landing track.

Standard operating procedure would be to capture the glideslope beacon and select the go-around altitude of 500 metres (1,700 feet). Because the aircraft was much too high, it wasn’t possible to capture the glideslope beacon. The go around altitude was not selected.

The first officer reported the go around to Kazan Tower. The controller responded that they should climb to 500 metres and contact Radar Control.

The standard operating procedure during a go around does not include a radio exchange. It’s a high pressure moment which requires the flight crew’s attention. Although the controller’s response seems straight forward, it added another task to distract the first officer, rather than simply acknowledging that the aircraft was going around.

The first officer responded with confusion. “Er… climbing 600 metres, right? And contact radar at 119 decimal 4, right?”

The controller repeated his instruction, which the first officer read back correctly. The entire exchange took over 20 seconds, during which neither crew member selected the go-around altitude of 500 metres.

The first officer also did not contact Radar Control.

One of the flight crew members engaged the TO/GA mode (Take Off/Go Around) at 15:22:45. This automatically cleared the approach path set in the autopilot. The flaps were retracted from flaps from 30° to 15°. The warning sound of the autopilot disconnecting was recorded on the Cockpit Voice Recorder.

The autopilot disconnection was accompanied by a red light illuminating on both pilots’ control panels as well as the audible warning. The crew did not acknowledge the disconnection nor did they switch the alert sound off (simply a case of pressing the autopilot disconnect button on the control wheel).

The Flight Director pitch bar would have shown a target pitch of 15° to climb away for the go-around.

Once the vertical speed is up to 300 feet per minute, the Flight Director shows the pitch angle required to maintain the appropriate airspeed for the flaps position (in this case 15°). If the go-around altitude selected is greater than the current flight altitude, the Flight Director will transition to ALT ACQ mode (altitude acquired). But the crew hadn’t set the go-around altitude of 500 metres; the selected altitude was still 270 metres (900 feet).

If the selected altitude is lower than the current altitude, the Flight Director pitch bar will continue to show the pitch angle to maintain the appropriate airspeed until the crew select a different flight mode. If the aircraft starts descending in this profile, the pitch bar will try to regain the required climb profile. If the aircraft descends past the selected altitude (290 metres/900 feet above the ground) then the TO/GA mode will automatically disengage while the Flight Director will transition to the ALT ACQ (altitude acquired) mode.

The other task of the Flight Director is to recover (or avoid) pilot-induced oscillations. So although there were significant deviations from the required flight profile and climb speeds, the Flight Director pitch bar didn’t exceed 5-6° while in TO/GA mode.

Note: The analysis of the wreckage and the FDR shows that there were no issues with the elevators or any other control systems in the aircraft. Everything was working (except for the map shift which was hardly relevant at this time), just that no one was doing anything.

FDR at the accident site

The engine power increased. The flaps had been retracted and the autopilot had disengaged. Neither of the crew was holding the controls. As a result, the aircraft went nose up and began to climb.

The indicated airspeed went up to 150 knots. The Speed Trim System moved the stabilizer by about 0.5° nose up. The aircraft deviated 300 feet from the selected altitude of 900 feet (270m), which triggered the altitude alert.

There were no significant column inputs and the pitch angle increased to over 25° nose up. At the same time, the bank angle changed from 7° right to 2° left.

In other words, no one was flying the plane.

It is almost unbelievable but it seems that the captain believed that simply pressing the TO/GA pushbutton was all that he would have to do. He wasn’t prepared for the autopilot to disconnect and it seems that he expected the aircraft to make the go-around and automatically reach the required altitude, although how he expected this to happen is beyond me, as no one had ever even selected the go-around altitude.

Over the course of 25 seconds, there were almost no column or wheel inputs.

As the pitch angle increased, the airspeed decreased to 117 knots by 15:23:15. The Speed Trim System, reacting to the falling airspeed by moving the stabiliser down approximately 1.7° over 13 seconds.

All of these things happened very quickly. The first officer finished the radio exchange with Air Traffic Control and finally began monitoring the aircraft again. He reminded the captain to retract the landing gear, which he then did following the captain’s command. He then (finally) selected the go-around altitude of 500 metres (1,700 feet).

What he didn’t do was monitor the aircraft. The airspeed was frighteningly low and falling, the pitch angle was excessively high and they climbed to over 640 metres (2,100 feet) and were still climbing as he set the go-around altitude to 500 metres (1,700 feet).

The altitude alert sounded again.

This time, it was triggered because the aircraft had deviated 300 feet from the selected altitude. The aircraft was at 2,000 feet by now. This was a clear warning that the selected altitude had been set after they had already climbed past that altitude. Neither of the flight crew acknowledged the alert.

The first officer was not cross-checking the captain’s actions in his role of pilot monitoring. He didn’t call out that they had reached positive vertical speed. He didn’t retract the flaps to 15° as requested by the captain (it looks like the captain did this himself). He didn’t monitor the airspeed. He even called out the wrong go-around altitude to ATC at his home aerodrome. He was clearly not prepared for the go-around.

The stress level in the cockpit increased. The two pilots began mixing English and Russian on their standard callouts. They confused the English words “up” and “down” when discussing the landing gear.

While the landing gear was retracting, the captain started making active control inputs. The speed had fallen and the pitch angle of the nose was more than 25° up. He pushed the control column nose down. The aircraft continued to climb at about 4,000 feet per minute (20 metres/second).

Now, there’s a part of the translation of the report that I can’t make sense of, so I’m going to ask you to help me out.

Personally, I think there’s an error in the text, specifically the same mistake that the pilots made: saying up instead of down. However, in case I’m wrong, here’s the exact text from the accident report. Note that STS is the Speed Trim System. The PIC is the captain who was Pilot Flying.

The PIC started active control inputs during landing gear retraction after the speed had fallen to 125 kt with pitch angle more than 25 degrees nose up. The control column was pushed nose down for approx. 1/3 of the travelling distance from the trimmed position (Figure 12). The actual altitude above runway level was approx. 2000 ft/600 m at that time and kept increasing with a vertical speed of over 4000 fpm (20 m/sec).

The column was kept in that position for slightly over 4 seconds, and then was deflected back to the trimmed position for just 1 second (most probably the PF just released the pushing forces on the column which shifted towards the balanced position due to the load spring). As the elevator was deflected by the column inputs and the stabilizer was trimmed by the STS a significant nose down moment appeared which resulted in a high negative pitch rate. Within 5 seconds the pitch angle changed from 25 degrees nose up to the values required for level flight and continued decreasing. The vertical load decreased to 0.5 g. The altitude almost stopped increasing (it peaked at approx. 2300 ft/700 m). The indicated airspeed dipped to 117 kt, however, the angles of attack were lower than the operational limitations and stick shaker was not activated.

At 15:23:16 the PIC pushed the column nose up again for over half the travelling distance from the trimmed position. The column was kept in that position for about 4 seconds, and then was deflected back to neutral for just 1 second.

As a result of the PIC’s control inputs, by the time the column was shifted back to the trimmed position the aircraft had the following performance: vertical load about 0 g, pitch angle 20 degrees nose down that kept going further nose down, airspeed over 140 kt increasing with a rate of 10 kt/sec, altitude approx. 2000 ft (670 m), vertical speed of descent over 5000 fpm (25 m/sec).

I am pretty sure that the 15:23:16 reference should say that the captain pushed the column nose down because that’s what a push motion does and the reference to “again” doesn’t make sense if he had changed to a nose up direction. But if you can make better sense of this for me, please leave a comment and explain!

In any event, it’s clear that the captain wasn’t able to assess the situation and work out how to recover. His initial control input was correct for the recovery of the upset aircraft but as the aircraft returned to level flight, further control inputs were made. He must have assumed that the initial input was not enough to recover the aircraft and so he tried again, now over-correcting.

The first officer asked “What’s up?” in an anxious voice. From the FDR data, it seems he may have also pulled the control column to show the appropriate action to the captain. The captain did not react to the question at all. By now he was totally disoriented.

The aircraft was descending at high speed.

At this stage, it could still have been recovered. The Extended Ground Proximity Warning System sounded with PULL UP, PULL UP.

At 15:23:21 the accident report refers to “a third column node down input” which furthers my belief that the captain had previously made two nose down inputs and no nose up inputs.

Certainly, they were now descending at speed as a third input was made on the column, pushing the control forward to the stop. The captain was now pitching the aircraft down as hard as he could. A simulation of the accident showed that the pressure on the column by the captain exceeded 50 lbf (22 kg).

The nose pitched down to 55°-60° and the negative vertical load increased to 0.9 g. Now the first officer could not take control of the aircraft even if he wanted to. He called out the captain’s name twice.

The captain responded with “What?”

There was no further communication between them.

The further control actions are described as “chaotic” including a bank to the left, reaching 35°.

As a number of simulator experiments revealed the last moment when it was possible to recover from the nose down pitch with an acceleration of 3.0 to 3.5 g (exceeding the operational limitations with no structural damage to the aircraft) was apparently at 15:23:23 (the aircraft was descending with a nose down pitch angle of 40 degrees, the height was 1900–2000 ft (580-610 m) with an airspeed of 175 to 180 kt. The height loss would have been 1600–1700 ft (about 500 m) meaning the aircraft would have been recovered at a height of 200 to 300 ft. However, neither at that moment nor further did the crew regain situational awareness or initiated any recovery actions.

It cannot be excluded however that during the last 20 seconds of the flight the pilots could have experienced a nose up somatogravic illusion. The possibility of such an illusion is confirmed by the pertinent analysis (Section 0).

Somatogravic illusion is a general form of vestibular illusion or wrong perception. Somatogravic illusions can lead to spatial disorientation. Significant longitudinal aircraft acceleration can create a nose up pitch illusion. In this case the pilot can instinctively push the control column in order to prevent the increase of perceived pitch angle. Rapid deceleration of the aircraft, on the contrary, leads to a nose down pitch illusion and the pilot can instinctively pull up resulting in pitch angle increase.

The captain literally flew the plane into the ground.

The Boeing 737-500 impacted the ground travelling at a speed of 245 knots (450 km/hour / 280 miles/hour) at an angle of 75° nose down. I try to avoid over-dramatising crashes but this aircraft simply smashed on to the grass next to the runway.

The accident site

So what the hell happened? Tune in next week… No, I’m kidding, we’ll keep going now.

First of all, investigators were confident that there were no technical issues with the aircraft or failure of the controls. The only fault was the map shift.

The layout of the fragments of the aircraft

It looks like what happened was that when the flight crew aligned the Inertial Reference System at Moscow before the flight, they entered the incorrect coordinates.

Certainly, the map shift was wrong from the point of departure from Moscow’s Domodedovo airport, as you can see from this image when they took off.

Map-shift during takeoff at Domodedovo Airport

Because of the way that the navigation data is corrected, it is not possible to see if the crew used navigational aids to correct the Flight Management Computer data during flight.

It seems unlikely. The recorded flight path and crew communications make it clear that the aircraft position according to their flight instruments was very close to the uncorrected IRS position. Also, the first officer commented early on the flight that there was an IRS NAV ONLY error. This warning means that the aircraft’s navigation accuracy is less than required for that phase of flight.

It’s hard to understand why they didn’t correct their position, though. The Kazan VOR/DME frequency was tuned in; the station’s callsign can be heard on the Cockpit Voice Recorder. The aircraft was within range of updating the aircraft’s position and there’s no apparent reason as to why the position update didn’t happen.

The need to go-around was caused by the aircraft being unstabilised when approaching the runway, as a result of the map shift of about 4 km, failure of the crew to apply complex airmanship skills and navigate with required accuracy, as well as lack of active assistance from the ATC who were watching the long-term deviations of the aircraft from the established pattern.

The report calls out Air Traffic Control specifically for their apparent complacency as they watched the flight on radar. To be fair, they warned the flight crew three times that their positioning was incorrect (twice by Kazan Control and once by the Radar Controller before the base leg) but the flight crew were unable or unwilling to make the corrections.

However, the crew did not analyze the situation as appropriate and took no corrective actions, like turning right for at least 20 degrees to align with the back landing course or requesting vectoring. In fact, if the pilots had compared the on-board DME readings with the ATC information they could have seen their compatibility (9 km or 4.8 nm), while according to the EHSI the aircraft was at a far greater distance (approx. 13 km or 7.0 nm). Instead, the PIC spent 33 seconds swearing [at] ATC (from 15:17:11 to 15:17:44), while he was proceeding to the base turn point. Continuing the flight with the same heading confirms that the pilots must have identified the aircraft position using the EHSI information only and were not aware of their true position. Apparently, the PIC was considering the possibility of capturing the localizer after the base turn or during the maneuver.

The flight path was deviating significantly and realistically, the flight crew needed vectors to get back on track. However, the controller stated that his instructions were to perform vectoring for air navigation assistance at the request of the crew; as the crew never requested air navigation assistance, vectoring was not provided.

However, the fact remains that the map shift and navigational confusion is not what made the aircraft crash into the ground. It may have been the instigating incident but it did not cause the aircraft to impact the runway.

The question, then, is why did an experienced captain of a Boeing 737 so completely lose the plot.

I usually quickly scan through the crew qualifications and don’t bother to mention the details in my summary. Usually, it’s not particularly interesting. In this case, though, it bears a closer look.

The captain worked for the airline as a navigator from 1991 to 2010. At the end of 2009, while working for Tatarstan Airlines as a navigator-instructor, he submitted his commercial pilot licence to the airline in order to be retrained for the Boeing 737 first officer position.

From 2010 to 2013, the captain underwent nine simulator sessions. After four of them, he was recommended to revise his go-around procedures. After two of them, he was recommended to revise NDB and VOR approach procedures. Over the course of three years, the same shortcomings were highlighted over and over in his training but not corrected. The first officer also had repeated recommendations to revise his go-around procedures as well as NDB and VOR approach procedures.

The crew of the aircraft were both weak pilots who were known to have issues with executing go-arounds and the use of navigational aids.

Part of the conversion training for the Boeing 737 that both pilots whent through included English Language Proficiency, which is ranked in ICAO levels. The captain was listed as ICAO Level 3, although an expert evaluation of his examinations showed that that his actual abilitity corresponded to ICAO Elementary Level 2. The fact is, even Level 3 proficiency is not enough to understand the English-language documentation in the aircraft, a failing of the training program.

The first officer appears not to have passed any English Language Proficiency testing. An instructor had noted that additional training was required during type rating simulator sessions.

Both pilots were granted ICAO Level 4 after they passed a qualification test at Tatarstan Airlines Aviation Staff Training Centre, although neither pilot had more than Level 2 ability.

Interesting, isn’t it?

The pilots both passed their theoretical knowledge exams during the type rating. However, the captain had originally failed the test: he got 57% correct over an hour and twenty minutes. He took the test again on the spot and the second time, he spent only 55 minutes and got 89% correct. It seems extremely unlikely that he was suddenly able to pass the test without help.

The first officer took the test three times on three different dates (3rd, 6th and 8th of December). After failing the first two exams, he answered every single question correctly the third time, with a result of 100%.

It’s not possible to know exactly what happened in these two instances but it seems likely that the pilots were given help. There is now a CCTV in the testing room.

The airline was never supplied with any records from the training which meant they had no records of the instructors’ comments.

According to ANEO S7 Training, after the PIC and co-plot’s training was completed only the following was provided to the customer airline: Training Completion Certificate, resolution on the training completion and a copy of the simulator attendance log.

Tatarstan Airlines did not request provision of all the training records.

Thus, the training process was like a “black box” where the airlines put their wishes as to the selected type rating program and the fee and then after some time received a pilot holding a Training Completion Certificate.

The result was that despite all their certificates and type rating, neither pilot appears to have understood the theoretical knowledge required to fly the Boeing 737 and neither pilot had enough English to be able to understand the training material in the class room, never mind the documentation in the cockpit meant to keep them safe.

On the day of the crash, the initial failure of the crew was their failure to identify their position and their ability to deal with complex airmanship.

The investigation thus then looked into the Boeing 737 type rating, which both pilots were confirmed as attending, to understand why the pilots did not appear to have been able to cope with the situation. The training centre responded to the query.

The Boeing 737 type rating training program developed by aviation experts and approved by Rosaviatsyia does not include map shift identification training.

… Every pilot, holding a license, shall … use information from all available sources to identify and correct the aircraft location. Such skills are acquired by flight crews during initial training at civil aviation flight colleges.”

So the assumption is that the flight crew would have learned to deal with such issues as position identification in their initial training and there was no need to revisit this.

Tatarstan airline requires a three-month check interval for Pilots in Command (i.e. captains) who have less than one year experience in their position. The captain’s most recent check ride was the 1st of September 2013, shortly before his previous check validation period had expired. The pilot-instructor signed off the captain, noting “can continue flights as PIC” and “excellent”. There was just one problem: the check ride had never been performed. The record was only in the pilot’s logbook with no record of an actual flight ever taking place.

Thus, as per the airline’s Flight Crew Training Manual, the captain was not entitled to perform flights at all.

But more interesting is the commercial pilot licence which the captain received from ANEO Sibir Airlines Aviation Training Centre. He was registered as undergoing training from 1 October 2009 to 28 March 2010, yet he claimed to the airline to have finished at the end of 2009. And on top of that, for a chunk of his time there, the Training Centre did not have a valid licence from the Ministry of Education of Moscow and then had their Aviation Training Centre Certificate revoked.

It seems that Rosaviatsyia, the Russian aviation authority, had continued to issue pilot licences to the training centre even after ANEO’s aviation training certificate was revoked. This is a huge systemic weakness as airlines and training organisations reasonably expect that a pilot’s licence demonstrates that the holder is qualified to fly and has been confirmed as safe. If the training centre is not trusted, then how can the training be trusted?

Generally speaking, if a training organisation is closed for shortcomings or noncompliance, all pilots who received certification or licences from that organisation should be reviewed, to ensure that they are properly qualified as aviation personnel. Certainly, training supposedly completed after the training centre’s training activities had been suspended should not result in a pilot’s licence.

The more the investigators looked into the captain’s documentation, the worse it got.

The captain’s commercial pilot licence was issued by the suspended training centre eight months later than the date given by the captain. Worse, investigators could not find any reference to his initial pilot training at any certified training organisation. His documents which confirmed his conversion training were forged. And finally, there’s no record of any flying experience during the period in which he was supposedly training.

The thing is, the airline were in a perfect position to notice this. He had worked there as a navigator since 1992. His paperwork showed that he did all his training and exams while he was working for the airline full-time and couldn’t possibly have been doing training at the same time. When he was approved for the Boeing 737 type rating, no one reviewed his documents or noticed that the dates given did not make sense. Finally, no one checked his starting knowledge at all to ensure that he had basic skills and knowledge of flight.

Knowing all this, the captain’s confusion in the cockpit suddenly makes a lot more sense. He was missing basic situational awareness skills which would have been covered during initial training–training which he never had. He wasn’t able to work out quickly and decisively which issues needed his attention. He didn’t know how to analyse the flight data. Based on his aggressive movements, he didn’t know how to fly the aircraft manually.

He clearly had some airmanship skills, as he legitimately went through type rating training for the Boeing 737 and flew the 737 for three years. But once things got rough, it wasn’t anything like enough.

The accident report concludes that the captain appeared to have believed that pressing the TO/GA pushbutton was an “end-all solution”. He wasn’t prepared for the autopilot to disconnect and expected the aircraft to automatically go around and somehow magically reach the required altitude. When that didn’t happen, he had no idea what to do.

You can read the full report in English here: http://mak-iac.org/upload/iblock/459/report_vq-bbn_eng.pdf

The accident was caused by the combination of the following:

  • The PIC had no initial flight training;
  • pilots not completely meeting the qualification pre-requisites for training, including
    the ELP requirements, were accepted for Boeing 737 type rating training;
  • The training process was methodologically imperfect, the oversight of results and
    quality of training was nominal;
  • the level of flight operations organization in the airline was low, which resulted in
    the failure to eliminate deficiencies in navigation equipment usage, airmanship and
    CRM (during various flight phases including go-around) that had been revealed for
    a long time.
  • there were systemic violations of duty time and rest period requirements and a
    sufficient amount of unused leaves, which could have resulted in fatigue
    accumulation and affected the crew working efficiency;
  • the airline’s simulator training program did not include training for conducting go-
    arounds from an intermittent height with two engines operative;
  • The crew was stressed before the go-around as they had been unable to determine
    the aircraft position with required accuracy for a long time;

In accordance with the Manual of Aircraft Accident and Incident Investigation (ICAO Doc 9756 AN/965), the factors are given in the logical sequence, without priority assessment.

  • the Aviate – Navigate – Communicate concept was violated both by the crew and
    the ATC, which resulted in the crew failing to follow the SOP during the go-around
    as the FO was distracted from his duties and monitoring flight parameters for a long
    time;
  • The crew failed to identify the A/P disconnection and delayed to apply manual
    control, which resulted in nose up upset;
  • The applied upset recovery simulator training programs and its quality assessment
    criteria were imperfect, which resulted in the crew failing to recover;
  • The crew was possibly affected by somatogravic illusions

And this was the result.

Impact Crater

Thanks for making it to the end. I suspect you can see how why I wasn’t willing to try to reduce this down to a normal length post. It’s one of the most tragic accidents I’ve seen as it was so avoidable. I tend to bristle at accusations that modern pilots are so used to automation that they no longer know how to fly the plane. But in this case, the captain had never learned in the first place.

Category: Accident Reports,

9 Comments

  • The preface to the report ends “The criminal aspects of this accident are described within a separate criminal investigation.” Is that report findable? Your discussion points to fraud in at least the written test, and a crude remedy (video of the test room); it would be interesting to know whether any convictions were made as a result of this mess, which seems to have started long before the flight in question.

  • Great article. Kept me riveted right to the end. What a story of fraud, incompetence, lack of oversight and multiple errors. It makes me think twice before flying in Russia and environs.

    • To be fair, the investigation pulls no punches and shows exactly how the aviation authority let the side down, which I see as an important aspect for change.

  • … this is gripping and chilling … when a passenger, it is Speedbird or Shamrock all the way – except where unavoidable …

  • This is not an easy story to absorb and digest.
    So many blunders, a tale of gross incompetence. Not just the crew, but also of the airline and authorities that qualified those two.
    This is not at all funny, but the best comment would seem: “A fine mess..”!
    Airlines like those using callsigns like Speedbird and Shamrock, and others like Scandinavian, KLM, etc., have crew selection- and training procedures in place that would more than likely weed out pilots who underperform to the extant described in Sylvia’s article. The list, of course, is not nearly complete. I just mention only a very few of the (fortunately many if not most ) airlines with very high crew standards.
    And newer companies, like Ryanair, are operating to very similarly high standards. Not in a small measure thanks to the standards set by the longer established companies and adopted by the licensing authorities.

  • A frightening report well dissected. I wonder how many other small airlines have problems like this waiting to be found? Hopefully not many, but unfortunately it wouldn’t be surprising to hear of similar problems in (especially) Africa as well as Central Asia and South/ Central America.

    The more I have read of your posts, the more I am surprised that the accident rate in commercial aviation is as low as it is.

  • The last few paragraphs scared me – how can a pilot with so little actual experience and training be in charge of a large passenger plane…chilling.

    Also I don’t understand how he flew it in to the ground even with somatogravic illusion – surely one glance at the attitude indicator would tell him whether he is straight and level, regardless of visual clues being absent?

Post a comment:

Your email address will not be published. Required fields are marked *

*
*
*