On 8 September 2015, a Boeing 777-200ER (G-VIIO) being operated by British Airways on a scheduled passenger flight from Las Vegas to London Gatwick as BA 2276 with an augmented crew rejected take off on runway 07L at Las Vegas in day VMC when a very loud noise was heard from the vicinity of the left engine, a GE90-85B. It soon became apparent that the engine had suffered a catastrophic and uncontained failure as a result of which a fuel fed fire developed rapidly as soon as the aircraft came to a stop. An emergency evacuation of the 170 occupants was successfully accomplished but in the process one crew member was seriously injured and 19 passengers sustained minor injuries. The aircraft sustained substantial but repairable damage but the failed engine was destroyed.
The aircraft fire taking hold. [Reproduced from the NTSB website]
An Investigation into the accident was carried out by the NTSB. The FDR and CVR were removed and their data successfully downloaded as were data from the QAR, which also recorded some parameters not recorded by the FDR including some that allowed the Investigation to determine the status of the left engine before the failure and the performance of the aircraft during the rejected takeoff.
It was noted that the 63 year-old Captain, who was acting as PF had joined the Operator in 1973 and had obtained an ATPL for the first time in 1997. He had accumulated approximately 30,000 total flying hours including approximately 12,000 hours on type. The 30 year-old First Officer had had joined the Operator in 2006 and had accumulated approximately 6,400 total flying hours including approximately 3,100 hours on type. The operating crew also included a 45 year-old First Officer designated as Relief Pilot who was, in accordance with the Operator’s procedures, occupying supernumerary seat in the flight deck for the accident takeoff. He had joined the Operator in 1997 and his licence medical certificate was endorsed to require him to fly as or with a co-pilot. He had accumulated approximately 14,000 hours total flying time of which 10,000 hours were on type.
It was established that the takeoff had commenced from the taxiway A8 intersection with runway 07L which was just over 800 metres from the runway threshold and left 3,500 metres of runway ahead. Nine seconds after the takeoff roll started, the CVR recorded a loud ‘bang’ at 77 knots and the aircraft had veered to the left. Over the next two seconds there was a sound indicative of an engine spooling down followed by the EICAS aural annunciation ‘engine fail’ and the Captain's ‘Stop’ callout. The thrust levers were moved to idle immediately and one second later the aircraft began decelerating from about 77 knots, which was the maximum airspeed achieved. As the A/T had not been disconnected before the thrust levers were retarded and this action was initiated below 80 knots, the thrust levers automatically began to move forward towards takeoff thrust and when the First Officer noticed this as wheel braking began, he realised what was happening and immediately disconnected the A/T.
Just over six seconds after the ‘bang’ was recorded, the aural fire warning occurred and the FDR showed that a left engine fire warning had illuminated. The Captain instructed the First Officer to tell ATC they were stopping (which was done) and called for the engine fire checklist (which was not acknowledged or commenced). The aircraft came to a stop 15½ seconds after the CVR recorded the ‘bang’. The Captain reported that he had not used maximum braking during the rejected takeoff because he had initially suspected that the noise heard was a tyre bursting and the First Officer stated that he had not selected reverse thrust or the ground spoilers (both PM responsibilities) because he had been "distracted by the thrust lever increasing".
Immediately the aircraft stopped, the Captain made a second request for the engine fire checklist to be run and it was commenced immediately by the First Officer and the Relief Pilot asked the Captain if he should make a passenger PA which was agreed with the instruction that passengers “should stay where they were”. The Captain then declared a MAYDAY to TWR and requested the fire service to attend. The Relief Pilot entered the cabin and immediately recognised that although the engine fire warning had gone out, external fire and smoke were visible. He returned to the flight deck and soon afterwards, the Captain made a PA ordering an emergency evacuation and the First Officer advised this to TWR. The evacuate order was given just over half a minute after the ‘bang’ had been heard but before the Evacuation Checklist had been run. The Captain’s attempt to do this from memory led to the inadvertent omission of the second item which was to confirm that both engines were shut down. The First Officer was simultaneously performing items on the same Checklist including unsuccessfully attempting to ensure that the pressurisation outflow valves were set to allow the aircraft to depressurise and did not shut down the right engine until the Relief Pilot queried whether this had been done. This meant that the right (serviceable) engine continued to run for 43 seconds after the Captain had ordered the evacuation which was about 2 minutes after the aircraft had come to a stop. At about the same time as the right engine was shut down, a forward cargo bay fire warning had been annunciated and the Captain armed the cargo fire switch and discharged three of the five cargo fire extinguisher bottles.
The Emergency Evacuation and the Fire Service Response
All occupants evacuated the aircraft using only 3 of the 8 exits, those at doors 1L, 1R, and 4L, although only five passengers had used door 1L before it had to be blocked due to the fire outside. Video evidence showed that the evacuation was complete around 2½ minutes after the Captain's initial order to commence it. Other exits were not used because either their slides deployed and inflated correctly but did not end up in a useable attitude (doors 3R and 4R) or because of the external fire hazard (doors 2L, 2R and 3L).
The three available ARFF vehicles were dispatched to the scene after the Captain requested the fire services and video evidence showed that the first one arrived from the nearby (650 metres distant) fire station about 2 minutes later midway through the evacuation and began discharging fire retardant in the area of the left wing. After half a minute the second one arrived and, having positioned ahead of the aircraft and taken up a position to the right of the aircraft "to protect passengers exiting from both sides of the aircraft", began discharging fire retardant from its high-reach extendable turret in the general direction of the fire. This initially only reached the area around the 1R door and led directly to the only serious injury sustained during evacuation. This was to the senior member of the 10-person cabin crew who fell upon reaching the end of the 1R slide because it was wet and slippery after being contaminated with the extinguishing agent. She broke her left forearm and suffered a compression fracture of the first lumbar vertebra in her back. The third ARFF vehicle arrived near the 1L door/slide and the aircraft nose 15 seconds after the second and assisted in extinguishing the fire which was achieved 2½ minutes after the first vehicle arrived on scene which was almost a minute after the evacuation had been completed and 5½ minutes after ARFF attendance was requested.
A group debrief of the cabin crew found that although some passengers had taken their cabin baggage with them, this had not impeded the evacuation as most passengers who retrieved baggage did so after the aircraft came to a stop and before the evacuation order was given after which the cabin crew’s assertive commands had limited further retrieval. The cabin crew manning the two exits used by almost all occupants (doors 1R and 4L) recalled seeing very little cabin baggage at their exits, and neither cited cabin baggage as a problem. However, the Investigation noted that accident aircraft was only 55% full and so whilst not a factor in this evacuation, the NTSB “remains concerned about the safety issues resulting from passengers evacuating with carry-on baggage, which could potentially slow the egress of passengers and block an exit during an emergency”.
Engine and Airframe Damage
Significant explosive and/or thermal damage was caused to parts of the failed engine remaining in situ and some of the explosive debris was ejected. However, none of this debris penetrated either the fuselage or the wing and much of the interior of the engine was not materially affected at all.
The inboard side of the left engine. [Reproduced from the Official Report]
The left inner wing was fire-damaged and the left side fuselage skin was superficially damaged from just aft of the 2L door to just aft of the wing leading edge and the skin in the centre of this was buckled and cracked. The longest cracks were about 12 inches and coincided with the area of most severe thermal damage to the wall panels in the cabin interior. This area also had 13 thermally-crazed cabin windows (see below). The wing-to-body and wing/fuselage fairings on the left side of the aircraft were delaminated and blistered to varying degrees.
The thermally damaged fuselage and inner wing skin and thermally crazed windows. [Reproduced from the Official Report]
The wing-to-body fairing panels on the right hand side of the aircraft were also fire-damaged with some delamination but to a lesser extent than on the left hand side. Near the wing/fuselage join, buckling had of the aluminium skin had occurred and the windows above this area were thermally crazed.
Soot and thermal damage to the right wing root and adjacent fuselage. [Reproduced from the Official Report]
The Cause of the Engine Failure
It was found that the uncontained failure had been caused by a fatigue crack in the High Pressure Compressor (HPC) stage 8 disk. This crack started on the aft face of the disk web and progressed through the web in a circumferential direction. The area where the crack had occurred indicated that low cycle fatigue crack growth (the process of progressive and permanent local structural deterioration occurring in a material subject to cyclic variations in stress and strain of sufficient magnitude and number of repetitions which will eventually produce a detectable crack) had occurred. Previous GE assessments of worst-case conditions for low-cycle fatigue cracking in this part of the engine had estimated that the initiation phase (i.e. the period before a detectable crack appears) would last almost 30,000 cycles. The HPC stage 8-10 spool involved had accumulated 11,459 total cycles and the low-cycle fatigue crack was calculated to have propagated over approximately 5,400 of those cycles so that the balance was the number of cycles for crack initiation - approximately 6,000 cycles and about 20% of the GE-predicted time.
During metallurgical examination of the failed HPC stage 8-10 spool, it was found that part of the surface of the stage 8 disk outer web had lower-than-expected shot peen coverage and examination of other similar spools by GE found similarly-reduced shot peen coverage in the same place. However, although shot peening is carried out to make surfaces more resilient, GE estimates of low-cycle fatigue life do not take this benefit into consideration and so it was considered by them that the lower-than-expected shot peen coverage found could not account for the fatigue crack initiation and eventual fracture of the spool. Rather, the evidence indicated that the crack had been initiated by the sustained-peak low-cycle fatigue failure mode. Cracking under the operational conditions to which the stage 8 disk web was subjected had not been seen prior to this failure and since the disk web was not subject to routine inspections, the crack on the accident had gone undetected. However, when the HPC involved was removed from the engine and routinely disassembled, exposing the stage 8-10 spool, it was estimated that the surface crack would have been about 0.05 inches long and in theory detectable by NDT techniques which are considered capable of identifying cracks as short as 0.03 inches. So if inspection of the disk web had been required at that point, the crack should have been detectable. By 2014, when the HPC had been removed from the engine but not disassembled, it was estimated that the length of the crack would have been 0.19 inches. As a result of these findings, GE began implementing enhanced inspection requirements to detect disk web cracks.
The Propagation of the Fire
The engine fire warning occurred whilst the take off was being rejected and the Captain called for the ‘Engine Fire’ Checklist as the aircraft stopped. The third item on the checklist was to set the affected engine fuel control switch to the cutoff position, which shuts down the respective engine by terminating its fuel flow. FDR data showed that about 28 seconds elapsed between the beginning of the engine failure and the time the fuel flow to it was terminated. Boeing estimated that during this time, about 97 gallons of fuel would have spilled onto the runway. It was evident that if the left engine had been shut down sooner, there would have been less fuel pooling on the runway to feed the fire and hazard the fuselage.
The Probable Cause of the accident was determined to be:
“The failure of the left engine high-pressure compressor (HPC) stage 8-10 spool, which caused the main fuel supply line to become detached from the engine main fuel pump and release fuel, resulting in a fire on the left side of the airplane. The HPC stage 8-10 spool failed due to a sustained-peak low-cycle fatigue crack that initiated in the web of the stage 8 disk; the cause of the crack initiation could not be identified by physical inspection and stress and lifing analysis.”
In addition, a Contributory Factor was formally identified as “the lack of inspection procedures for the stage 8 disk web”.
Safety Action initiated by GE Aviation as a direct result of the accident included the following:
- the addition to the GE90 Engine Maintenance Manual of comprehensive non-destructive inspections for the HPC stage 8-10 spool web at the piece-part level as well as the rotor, module, and engine levels.
- the issue of SB 72-1145 on 24 November 2015 for a one-time on-wing ultrasonic inspection of the stage 8 web of the stage 8-10 spools of all GE90 engines with specified part numbers which was followed by FAA AD 2015-27-01 based on the information contained within it.
- the issue of SB 72-1146 on 12 February 2016 extended the requirement for a one-time on-wing ultrasonic inspection of the GE90 HPC stage 8 web of the stage 8-10 spools to all such spools which had the same part number as the accident spool that were not covered by SB 72-1145.
- the issue of SB-1151 to supersede SB 1151 on 10 June 2016 expanding the inspection population to include GE90 HPC stage 8-10 spools with all other part numbers and additionally recommending a repetitive inspection every 500 cycles until an Eddy Current Inspection (ECI) was performed in place of a one-time on-wing ultrasonic inspection. Based on this SB, the FAA issued AD 2016-13-05 on 24 June 2016 requiring an ultrasonic inspection or an ECI of the stage 8 aft web upper face between 8,000 and 9,000 cycles since new or within 500 cycles in service after the effective date of the AD (July 29, 2016), whichever occurred later.
- the modification of the print drawing for GE90 HPC stage 8-10 spools in production to include a check of the shot peen intensity in the outer web areas which was effective in February 2016.
- the issue of SB 72-1149 on 29 March 2016 to address possible shot peening inconsistencies in all GE90 HPC Stage 8-10 spool web and Spacer arms.
The Final Report of the Investigation was published on 19 June 2018. No new Safety Recommendations were made but attention was drawn to the two Recommendations made in February 2018 as a result of the investigation into a similar uncontained engine failure to a Boeing 767 taking off from Chicago in 2016 which raised identical concerns about the absence of any distinction in crew emergency response drills for an engine fire beginning in the air and one beginning on the ground and the consequent potential for delay in commencing an emergency evacuation in the pre takeoff or post landing cases.