B738, Belfast International UK, 2017
B738, Belfast International UK, 2017
On 21 July 2017, a Boeing 737-800 taking off from Belfast was only airborne near the runway end of the runway and then only climbed at a very shallow angle until additional thrust was eventually added. The Investigation found that the thrust set had been based on an incorrectly input surface temperature of -52°C, the expected top of climb temperature, instead of the actual surface temperature. Although inadequate acceleration had been detected before V1, the crew did not intervene. It was noted that neither the installed FMC software nor the EFBs in use were conducive to detection of the data input error.
On 21 July 2017, a Boeing 737-800 (C-FWGH) being operated by Sunwing Airlines on behalf of Thompson Airways (since renamed TUI Airways) on a non-scheduled international passenger flight from Belfast International to Corfu as TOM1526 began its takeoff from intersection ‘D’ on runway 07 in day VMC. After using almost all the available 2654 metre TORA, the aircraft was observed climbing away at a very shallow angle over generally flat terrain. After just over 2 nm and approximately 800 feet aal, additional thrust was added, a normal rate of climb was established and thereafter the flight was completed without further event. The aircraft was subsequently found to have struck an approach light 36cm tall which was located 29 metres beyond the end of the runway, but to have been undamaged by this impact.
An Investigation was commenced by the UK Air Accidents Investigation Branch (AAIB) after it became aware of the event three days later, by which time relevant data on both the FDR and CVR were found to have been overwritten. It was also found that neither of the two QARs installed on the aircraft were functioning and reconstruction of the circumstances was reliant on time consuming analysis of other data from a range of sources which included the A/T BITE system, the EGPWS memory, EFB data, radar and ADS-B data and airport CCTV recordings. Some data from the Aircraft Condition Monitoring System (ACMS) of use to the Investigation were automatically transmitted via ACARS to the aircraft operator, but when its PCMCIA Card was subsequently removed, it was apparent that data storage had not been programmed.
Although the Canadian aircraft operator did eventually report the event to Canada’s TSB, they did not report it directly to the UK AAIB and the aircraft commander failed to report it at all to either the UK or Canadian Investigation agencies. It was noted that both of these parties, as well as the airline for whom the flight was being operated and the aerodrome authority, “had a legal duty to report the event to the AAIB”. It was noted that Belfast ATC, which, it was considered, should have been sufficiently aware of the event to know that it was likely to be considered a Serious Incident, had merely raised an out of office hours UK CAA MOR in accordance with their corresponding procedures.
It was noted that the 38-year-old aircraft Captain, who was acting as PF for the investigated flight, had a total of 8,234 flying hours of which 2,817 hours were on type. The 45-year-old First Officer had a total of 4,423 flying hours of which 1,219 hours were on type. Both pilots held Canadian licences and had been off duty the day before the flight under investigation, which was their first of the day. They had been employed by the aircraft operator for five and two years, respectively.
During the pre-flight preparations, each pilot independently completed weight and balance and performance calculations on their respective EFBs for a departure from runway 07 entering at taxiway ‘D’, which gave a 2654 metre TORA. After crosschecking, these calculations were entered into the FMC. However, at an undetermined point, a figure of -47°C was entered into the FMC as the OAT without either pilot noticing and since this parameter is used in the FMC calculation of the required takeoff thrust setting, a much lower thrust setting than that appropriate was calculated by the FMS. This calculated rated thrust was then further reduced in accordance with normal procedures by entering a (correctly-calculated) assumed temperature of 47°C.
During pushback, the ground crew saw an area of excessive tyre wear on one of the NLG tyres and it was decided to return the aircraft to its gate and change both NLG tyres. Whilst this was happening, the First Officer re-calculated the takeoff performance data which, after being crosschecked by the Captain, was entered into the FMC by the First Officer using 48°C as the (correctly updated) assumed temperature for the intended reduced thrust takeoff. However, an incorrect value of -52°C was then entered to replace the previously-entered (and also incorrect) value of -47°C as the OAT, again without either pilot noticing. The FMC then used this information to calculate and display a required takeoff thrust setting of 81.5% N1 whereas the EFB calculations made independently by both pilots prior to FMS data input had produced a required N1 of 92.7%. This EFB-calculated figure was not displayed to the pilots at the time, but was established by accessing the two EFBs used during the Investigation.
Nine minutes after the first pushback, the flight was again pushed back and taxied to the runway and, having received a corresponding clearance, began its takeoff on runway 07 from the taxiway ‘D’ intersection. The pilots stated that at around 120 / 130 KIAS (the V1 was 144 KIAS), they had realised that the aircraft was not accelerating normally but had continued with the engine thrust unchanged and had reached Vr approximately 300 metres before the end of the 2654 metre-long runway and become airborne immediately prior to its end. Multiple witnesses, including the TWR controller, reported that the aircraft “took a significant time to lift off before climbing at a very low rate”. The controller reported having decided to “allow the pilots some time” rather than immediately transfer the flight to the radar frequency, which was usual practice as he had been “sure they had a problem they were dealing with and that they would call him if they needed anything”. However, when there was no such call, he proceeded with the transfer. The crew checked in as instructed and were subsequently cleared en route without indicating to any ATC unit that there had been anything abnormal about their takeoff or initial climb out. A message recommending an inspection of the aircraft on arrival at its destination was passed to the aircraft operator who subsequently reported that no damage had been found as a result of such an inspection.
Continued concern at what had been seen resulted in the TWR controller suggesting, after a discussion with an airport fireman who had also witnessed the takeoff, that the airport operator should carry out a runway inspection. This found that one of the runway 25 supplementary approach lihts located within the stopway 29 metres beyond the end of the TORA and 36cm high was lying on the ground with substantial damage evident. It was initially assumed by airport staff that the light had been blown over by jet blast, something which had previously been seen at the same airport after takeoffs by much larger aircraft - although in this case usually near the beginning of the runway behind the point where an aircraft had commenced its takeoff roll. However, a subsequent closer examination by the Investigation team found that significant impact damage had occurred to the light, which had also acquired a pattern of black rubber deposits corresponding to the tread on the tyres fitted to the aircraft involved.
Key points derived from the available evidence.
- Examination of the EGPWS memory found that no Mode 3 Alerts had been generated during the shallow initial climb, which had occurred over ground that sloped away from the runway elevation. This confirmed that the aircraft had not descended by more than 50 feet before reaching 500 feet agl.
- In the absence of any FDR, QAR or CVR flight data, the Investigation carried out a complex integration of radar and ADS-B data in order to reconstruct the initial climb profile achieved by the aircraft and was able to create the illustrations below.
From the vertical profile illustration, it can be seen that when 500 metres beyond the end of the runway, the aircraft had only climbed to approximately 40 feet aal, when 900 metres beyond the end of the runway, it was only approximately 120 feet aal and when 1500 metres beyond the end of the runway it was only approximately 220 feet aal. BITE history for the A/T showed that not until the aircraft had reached approximately 800 feet aal over 2 nm from the end of the runway had the crew manually advanced the thrust levers to an N1 greater than the 81.5% used for takeoff. It was apparent that it had been only “the benign nature of the runway clearway and terrain elevation beyond and the lack of obstacles in the climb-out path, which allowed the aircraft to climb away without further collision after it struck the runway light”.
- Boeing calculations indicated that had an engine failed once the aircraft reached V1, it would have been unable to climb if the takeoff had been continued unless the thrust set had been increased on the remaining engine. It was also calculated that had the crew immediately rejected the takeoff at V1 for any reason, the aircraft would have overrun the end of the runway at a speed of 5 knots. However, if the regulatory criteria for rejected takeoffs used in take off performance calculations, which allow for a two second delay to allow for the decision to initiate a rejected takeoff and discount any use of reverse thrust had been used, the aircraft would have crossed the end of the runway at a speed of approximately 80 knots. It was therefore concluded that had an engine failed at V1, “the most likely outcome would have been either a runway overrun or an attempted takeoff leading to ground impact” with either possibility being an extremely serious event with “the potential to cause significant loss of life”.
- Had the crew been alerted to the abnormally low acceleration while still at low speed, the takeoff could have been rejected and the aircraft brought to a halt well before the end of the runway.
- A number of human factors failures which had been identified during the event were reviewed during the course of the Investigation:
- The FMS data entry error and its subsequent non-detection.
- The failure to detect a significantly lower acceleration than normal during the takeoff roll until a late stage and the failure to respond once it was apparent.
- The delay in recognising the reason for the almost imperceptible rate of climb which followed lift off very near to the end of the runway.
Previous Similar Occurrences
It was found that after previous events in which incorrect OAT figures had been entered into the FMC prior to takeoff, Boeing had issued an FCOM Bulletin which highlighted the fact that “an incorrect reduced thrust target may result in slower acceleration to V1, which may invalidate the takeoff performance calculations and/or result in decreased obstacle clearance margins after liftoff” and the specific need for flight crews to verify the OAT entry on the N1 LIMIT page. It was also found that the most recently updated FMC software version would, subject to a required update of the Common Display System (CDS), have provided an automated crosscheck between the OAT entered by the crew and the OAT detected by an external sensor. This optional update had not been installed in the aircraft under investigation whereas if it had been, it would have precluded the possibility of data entry errors like the one made.
It was noted that a considerable number of previous investigations by the AAIB and other State Investigation Agencies had involved events attributable to a range of data entry errors which had gone undetected and then led to the use of incorrect FMC-calculated takeoff thrust settings.
The formally-documented Findings of the Investigation included, but were not limited to, the following:
- The low takeoff thrust meant that the takeoff was abnormal in terms of:
- low acceleration
- distance along the runway to achieve V1 and Vr
- low rotation rate
- low climb rate
- marginal ability of the aircraft to stop during an RTO from V1
- inability of the aircraft to continue the takeoff following an engine failure at V1 without increased thrust.
- Once the aircraft began its takeoff run with insufficient thrust, the risk controls in place did not alert the crew to act to recover the situation because, in general:
- Pilots are unlikely to recognise that actual acceleration is below a threshold value for a particular runway.
- The use of autothrust de-couples pilots from the thrust levers.
- Pilots are disposed only to reduce thrust to idle during takeoff (in the case of an RTO).
- Pilots remove their hands from the thrust levers at V1.
- Pilots do not have to increase thrust during a takeoff in the event of an engine failure.
- Safety margins built into take off distance required calculations, which cater for normal variations in operational performance, are rendered unreliable or ineffective when there is a data entry error because of the random (and possibly gross) nature of the effect.
The investigated Serious Incident was attributed to three Causal Factors as follows:
- An incorrect OAT was entered into the FMC, which caused the FMC to calculate an N1 setting for takeoff which was significantly below that required for the aircraft weight and environmental conditions.
- The incorrect OAT was not identified subsequently by the operating crew.
- The abnormal acceleration during the takeoff run was not identified until the aircraft was rapidly approaching the end of the runway, and no action was taken to either reject the takeoff or increase engine thrust.
Three Contributory Factors were also identified:
- The aircraft’s FMC did not have the capability to alert the flight crew to the fact that they had entered the incorrect OAT into the FMC, although this capability existed in a later FMC software standard available at the time.
- The Electronic Flight Bags did not display N1 on their performance application (some applications do), which meant that the crew could not verify the FMC-calculated N1 against an independently-calculated value.
- The crew were unlikely to detect the abnormally low acceleration because of normal limitations in human performance.
Safety Action taken in response to the event whilst the Investigation was in progress was noted as having included, but not been limited to, the following:
- Sunwing Airlines began upgrading the FMCs and CDUs installed in their Boeing 737 aircraft so that the available OAT alerting function would be added. They also updated their EFB software so that it would display the computed N1 and amended their SOPs to add a crosscheck of this figure as displayed on the two pilots’ EFBs.
- The UK CAA amended MATS Part 1 to require that senior aerodrome air traffic controllers must notify the AAIB by telephone as part of their initial reporting actions following an aircraft accident or serious incident and also added a similar requirement to the UK Flight Information Service Officer Manual, CAP797.
A total of four Safety Recommendations were made at the conclusion of the Investigation as follows:
- that the Federal Aviation Administration mandate the use of Flight Management Computer OPS software revision U12.0, or later and the Common Display System Block Point 15 update where this is required, to enable the outside air temperature crosscheck on all applicable Boeing 737 aircraft. [2018-12]
- that Boeing Commercial Airplanes give guidance to operators of Boeing 737 aircraft on how they might verify the FMC-calculated value of N1 against an independently-calculated value. [2018-13]
- that the European Aviation Safety Agency, in conjunction with the Federal Aviation Administration, sponsor the development of technical specifications and, subsequently, develop certification standards for a Takeoff Acceleration Monitoring System which will alert the crew of an aircraft to abnormally low acceleration during takeoff. [2018-14]
- that the International Civil Aviation Organisation note the conclusions of this Report and introduce provisions addressing Takeoff Acceleration Monitoring Systems. [2018-15]
Two earlier Safety Recommendations were included in a Special Bulletin, S2/2017, which was published on 20 September 2017 in order to provide initial information on the circumstances of this Serious Incident and clarify requirements for the reporting of accidents and serious incidents. These both related to FMC software updates, one of which has now been superseded by Recommendation 2018-012 above and the other of which was as follows:
- that the Boeing Company promulgates to all 737 operators the information contained within this Special Bulletin and reminds them of previous similar occurrences reported in the Boeing 737 Flight Crew Operations Manual Bulletin dated December 2014. [2017-017]
The Final Report was issued on 21 November 2018.