On 13 September 2017, a Boeing 737-800 (VH-VUE) being operated by Virgin Australia on a scheduled domestic passenger flight from Melbourne to Adelaide was conducting a high speed descent in day VMC when flight crew response to an indicated airspeed increase to prevent an overspeed led to cabin crew injuries, one of which was serious. The rest of the accident flight was without further event but the seriously injured cabin crew had to remain where they had fallen to avoid the risk of further aggravating the injury and their post flight removal from the aircraft by ambulance crew could not be safely accomplished for over an hour.
An Investigation into the Accident was carried out by the Australian Transport Safety Bureau (ATSB). The FDR and CVR were removed from the aircraft and their data successfully downloaded and the aircraft operator also provided a copy of the parallel QAR recording.
It was noted that the Captain had joined Virgin Australia as a 737 First Officer in 2003 and had been promoted to Captain on type in 2012. They had a total of around 18,000 hours flying experience which included approximately 10,000 hours on type of which in excess of 4000 hours were as Captain. The First Officer had joined Virgin Australia as a Boeing 777 Second Officer in 2010 and qualified as a 737 First Officer two years later. They had a total of around 8,500 hours flying experience which included around 3,500 hours on type. Both pilots were Melbourne-based and the accident flight was the third of four rostered for their days flying duty period and their second operating together.
It was established that the First Officer had been designated as PF for the accident flight, which had climbed to FL360 for the cruise. Prior to departure, the Captain had briefed the cabin crew that the flight would take just over an hour and that the en route weather would be the same as that at Melbourne which the Senior Cabin Crew Member (SCCM) “perceived as cloudy”. In fact, the applicable significant weather forecast chart provided to the crew was found to have shown that the descent to Adelaide would pass though airspace where moderate CAT was expected.
Flight conditions in the cruise and the initial stages of the descent were smooth and clear of cloud. Approaching the top of descent, ATC gave the flight a STAR which would position it for an approach to runway 23 and clearance to descend when ready to FL 250. They then instructed the crew to maintain maximum speed on the descent and the First Officer indicated that he would select 320 KIAS whilst observing that “this might be too fast” to which the Captain responded with “she’ll be right, don’t overspeed”. After initially descending uneventfully with 310 KIAS selected, this was increased to 320 KIAS passing FL 250. After maintaining this speed for a few minutes, the question of when to switch on the seatbelt belt signs was briefly discussed and the First Officer, having raised the subject, was content with the Captain’s view that current conditions “were good” and did not yet require the signs on.
About six minutes into the descent, just after the First Officer had reduced the selected speed back to 310 KIAS, the indicated speed began to increase, at first gradually, then more rapidly as VMO (340 knots) was approached as the aircraft was descending through approximately 17,000 feet. The CVR recorded both pilots expressing “statements of concern and alarm” followed by the Captain calling ‘pull-up’ whilst simultaneously making an abrupt nose-up input through their control column, which registered a force of about 22 kg and a peak vertical acceleration of 2.3g and caused the AP to disconnect. They then “abruptly released” the control column, which led to a minimum vertical acceleration of 0.9g, before making a second slightly less forceful nose up control input which led to a second vertical acceleration peak of 1.95g about 4 seconds later. The maximum recorded airspeed during the episode was subsequently found to have been 1 knot above VMO.
The First Officer noted the AP disconnect audible alert and the Captain’s control column inputs and recognised that they had taken over control and after about 8 seconds, the Captain prompted the First Officer to resume duties as PF and a normal transfer of control was made. Both pilots reported having perceived that they had encountered “severe atmospheric turbulence” and the Captain said “better put the belts on” which the First Officer then did, whilst continuing to fly the aircraft manually for around 30 seconds, making several nose-down inputs at up to a 15 kg force before re-engaging the AP and settling the airspeed at around 280 KIAS.
Soon after the AP had been re-engaged, the cabin crew called the flight deck on the interphone to report that one of them had fallen during the sudden upset and had broken their leg. Whilst the Captain was talking to the cabin crew, the First Officer advised ATC that the flight would not now be able to comply with the next SID height requirement “due to encountering turbulence” and ATC responded by cancelling all further height requirements and advising that they could reduce their speed. ATC subsequently contacted the flight to confirm their report of severe turbulence and the Captain did so, adding that the turbulence event had been “associated with some cloud”. This reference to cloud was subsequently clarified to the Investigation by the First Officer as "thin wispy cloud” and by the Captain as “light cirriform-type cloud” rather than the full cloud cover which had been entered later in the descent as forecast.
The two injured cabin crew were standing in the rear galley at the time of the unexpected upset, which they had perceived as “sudden and without warning”. One of them was thrown up towards the ceiling before then falling to the floor and on impact reported having felt their leg snap and thereafter been unable to move. The other reported having struck their jaw on the galley surface and sustained other minor injuries to their body and face which they subsequently reported to the Investigation as having caused “muscular skeletal injuries (for which they) had sought out chiropractic care”. Once normal flying conditions had returned, the leg injury cabin crew was given oxygen from a portable on-board cylinder and the flight crew were informed that they would be unable to be moved to a seat for landing and would stay on the galley floor. They were also told that an ambulance would be required airside on arrival at Adelaide and arrangements were then made accordingly.
Airport RFFS personnel were in attendance for the arrival of the flight at its assigned gate and six minutes later, the requested ambulance arrived and an “extensive discussion” began about how to remove the injured cabin crew member from the aircraft, complicated by the fact that the ambulance stretcher was too wide to use in the aircraft aisle. When RFFS personnel suggested using a scissor lift or catering truck appliance and asked the ground service supervisor to assist, this person responded that “no scissor lift was available” and declined the use of the catering truck. “Considerably more deliberation” then followed before the injured cabin crew member themselves suggested that the ambulance officers could use a slide sheet to drag them along the aisle to the front of the aircraft for transfer to a stretcher. This was then done but by the time the ambulance was ready to leave the airport, it was over an hour since it had arrived at the aircraft side.
Why It Happened
Whilst it was clear that the upset had been only the indirect result of a windshear encounter, it was noted that it had occurred in airspace where the flight crew were aware that moderate CAT had been forecast and that such awareness in this case should have resulted in a corresponding pre-departure briefing for the SCCM and an early selection of the seat belts to on during the descent. The Captain’s view that this was unnecessary appears to have been based on the smooth flying conditions experienced, the absence of any reports of turbulence from other aircraft and, inexplicably, given that it was CAT which had been forecast, the fact that the flight had remained clear of cloud and there had been no weather radar returns.
Had these findings derived from the forecast been recognised, it was considered that the flight crew might have decided that a high speed descent was not appropriate and informed ATC accordingly. In respect of the response to the risk of overspeed, it was noted that neither pilot had deployed the speed brakes as an initial response to the speed increase as recommended in guidance provided in the Flight Crew Training Manual (FCTM) for their use in such situations. It was also clear from the sudden and highly inappropriate reaction of the experienced Captain to a risk of overspeed that it could be associated with an extreme aversion to any such exceedance combined with a lack of any discussion on how the PF could avoid one. Earlier in the descent, the CVR data showed that “when the First Officer mentioned the threat of exceedances, the Captain made a strong informal interjection saying ‘don’t exceed anything…’ (and) the crew did not discuss these threats any further, nor did they discuss a management strategy”. The fact that the Captain’s sudden intervention as overspeed looked likely had been both extreme and involved a takeover of control without the required announcement was assessed as indicative of great concern about an overspeed in the absence of any prior contemplation of appropriate threat management and possibly also to ‘startle’ and/or a perception that “the AP was not controlling the aircraft”. The Investigation also noted that “the Captain indicated not being aware that there was a margin between VMO and the requirement for a maintenance inspection at 359 knots, or that there was a margin between VMO and the structural limitations of the aircraft”.
FDR data was used to derive estimates of the instantaneous local wind speeds during the part of the descent during which the overspeed had occurred. This work showed that immediately prior to the AP disconnect caused by the Captain’s abrupt control column actions, the wind component relative to the aircraft flight track had changed quickly from a negligible headwind component to an approximately 25 knot tailwind component. This was then maintained for about 10 seconds before halving in two seconds immediately prior to the overspeed as a result of a combination of decreased wind speed and a change in wind direction. These conclusions from a relatively simplified approach were similar to those from a more refined analysis carried out by Boeing.
A comparison of the calculated headwind component changes and the computed airspeed during the two minute period when the AP disconnect occurred. [Reproduced from the Official Report]
In respect of the immediate flight crew performance, it was concluded as follows:
“Overall, this was a lost opportunity for the flight crew to effectively manage the risk of overspeed. The pilots were in a position where they had time and space to think about a potential overspeed as a crew. At the top of descent, the pilots could consider the tolerances of the aircraft to airspeed exceedances, and how they could respond safely given the high-energy state of the aircraft. However, as the aircraft continued the descent and operated close to VMO, the time available to the flight crew reduced. When the Captain saw the airspeed suddenly increase, the Captain perceived the crew were in a position where there was no time to discuss the situation with the First Officer, to think about the implications of the overspeed, or to consider the consequences of the pullback control action. In this way, the absence of planning at an earlier stage in the descent affected how the Captain managed the risk of overspeed during the occurrence, and increased the risk of unsafe intervention.”
It was noted two other high speed descent events which had occurred earlier in 2017 had also involved Boeing 737-800 aircraft and had also then been the subject of independent investigations. The first of these involved a Ryanair aircraft inbound to Manchester UK with the same AP modification as the Virgin Australia aircraft in which one of the unsecured cabin crew was seriously injured after the Captain responded to a windshear-caused overspeed risk by disengaging the AP and manually applying an excessive nose-up pitch input reportedly without recognising at the time that it had been excessive. This response was contrary to the FCTM overspeed response guidance applicable at the time and trained by the operator and was attributed by the Captain to ‘startle’. The second involved a Qantas aircraft descending towards Canberra in which two cabin crew were injured when the Captain responded to an overspeed risk as trained by making a control column pitch input which led to an unexpectedly abrupt AP disconnect. This occurred because the AP had recently been modified in a way that made the operator’s trained and routinely practised but undocumented response to overspeed risk no longer appropriate.
The Investigation formally documented four Contributory Factors as follows:
- During a high-speed descent, a sudden decrease in tailwind associated with windshear caused airspeed to approach and exceed the aircraft maximum operating speed (VMO)
- The flight crew did not apply speed brakes to arrest the speed increase. In response to the airspeed rapidly increasing towards VMO, the Captain (Pilot Monitoring) perceived a need to immediately intervene, and made pitch control inputs without following the normal take-over procedure and alerting the First Officer (Pilot Flying).
- The magnitude of the Captain's control input was probably greater than intended. This was influenced by a perception that the autopilot was not controlling the aircraft so an urgent intervention was required. The magnitude of the control input caused sudden pitch changes, resulting in the injuries to the cabin crew.
- Although the flight crew identified the risk of overspeed during the high-speed descent into Adelaide, they did not consider steps for mitigating that risk, or how they would manage an overspeed during the descent. This reduced the likelihood of the crew effectively responding to the unexpected increase in airspeed.
Two Other Findings were also formally identified:
- Although the weather forecast included moderate clear air turbulence and the aircraft was making a high-speed descent, the pilots perceived that flying conditions were smooth and elected not to activate the fasten seat belt sign. The cabin crew briefings did not mention the forecast clear air turbulence. The cabin crew were not secured prior to the sudden pitch changes, which increased the likelihood of injuries.
- The Ground Handling Supervisor assessed there was a significant fall from height risk associated with the unsecured use of the catering truck. For that reason, the ground handling supervisor did not agree to the emergency services request to use that equipment to remove the injured cabin crew.
Safety Action taken by Virgin Australia as a result of the investigated event whilst the Investigation was in progress was recorded as having included, but not been limited to, the following:
- The Flight Crew Information Manual was updated to include a section on ‘Managing VMO/MMO Exceedances’. This new section stated that aircraft have been tested beyond VMO/MMO, and that these speeds included a margin below the speeds that require maintenance action or threaten the structure of the aircraft. The manual also stated that it is acceptable to refuse an ATC instruction to perform a high-speed descent and identified that disengaging the autopilot to respond to an overspeed may result in abrupt pitch change.
- A Flight Crew Information Bulletin (FCIB) was sent to all 737 pilots identifying that there had been many instances where pilots had manually intervened to respond to an overspeed or possible overspeed, and that several of these events had resulted in serious injuries. The FCIB noted that VMO/MMO are not 'never-exceed' speeds and that it was acceptable to refuse any instruction to conduct a high-speed descent. It also emphasised that it was preferable to accept a temporary overspeed rather than make large abrupt control inputs.
- Updated procedures provided to Company ground handling staff and other Company personnel working airside require that in situations where emergency services are in attendance, their reasonable directions should be followed and if ground handling staff are in doubt, they should liaise with the Captain, SCCM or Airport Manager, as applicable.
- All pilots have been provided with additional non-technical skills training to further educate pilots about the management response to flight safety events and how the Company Flight Data Analysis Program (FDAP) is run to counter any mis-perceptions.
The Final Report was released on 30 September 2020. No Safety Recommendations were made.