Emirates Flight EK231

Four pilots preparing a long flight

Emirates flight EK231 was scheduled to fly from Dubai to Washington Dulles on 19 December 2021. Boeing 777-300ER A6-EQI carried 354 passengers, 14 cabin crew and four flight crew, giving 372 people aboard. The operating commander flew the aircraft and the operating copilot monitored it. A second commander and copilot occupied the observer seats for departure before providing relief during the long flight.

The departure was at night. Weather, navigation equipment, airport visual aids and communications with air traffic control were normal. The aircraft was heavy: the calculated actual takeoff weight was 350.1 tonnes, about 1.4 tonnes below its maximum. Investigators checked the performance calculations and found them correct. The planned speeds were 170 knots for V1, 177 knots for rotation and 183 knots for V2, with flaps 15 selected.

During preparation, the augmenting crew saw zero displayed in the mode control panel altitude window. The GCAA could not determine who had entered it. With the selected altitude within 20 feet of the displayed barometric altitude, switching on a flight director activated ALT pitch mode. That mode provided guidance to hold altitude, leaving the aircraft set up with inappropriate vertical guidance for takeoff.

An abnormal indication left unresolved

The copilot noticed that his flight mode annunciator displayed TO/GA for roll and ALT for pitch. He continued his preparation tasks intending to address the indication later. He then switched his flight director off and on, but ALT remained displayed. He intended to tell the commander about the indication and failed reset. He told investigators that the busy cockpit distracted him and he forgot.

The reset required a different sequence. With both flight directors on, cycling one switch alone would leave the mode engaged. On the ground, the selected altitude had to differ by more than 20 feet from the displayed barometric altitude, both flight director switches had to be turned off, and either or both could then be switched on. The systems description contained this information, but the FCOM procedures gave no explicit reset instructions.

The required preflight checks included verification of the flight director modes by both operating pilots. The commander recognized the green indications but missed the ALT mode title. Calling the mode names aloud was not required. The before taxi checklist also contained no step ensuring the correct takeoff pitch mode. ALT remained active through taxi and the takeoff roll.

The aircraft had takeoff configuration warnings for conditions inconsistent with takeoff requirements. However, there was no warning for a flight director mode incompatible with the phase of flight. The ground proximity system also depended on its own alert conditions: its loss of altitude warning after takeoff was triggered by a height loss. The GCAA recorded no such warning during the incident. The displayed mode names and the flight path instruments provided the crew with the indications of the developing problem.

The climb flattens as speed rises

The crew changed the takeoff intersection from Mike 13A to Mike 14A to obtain a longer available distance. The aircraft began its takeoff roll at 23:09:43 UTC. Four seconds later the annunciator showed thrust reference, TO/GA roll and ALT pitch. At 23:10:30 the commander began rotating at 181 knots, about 2,280 metres along the runway from Mike 14A. Liftoff followed four seconds later.

About two seconds after liftoff, the pitch attitude reached 10.4 degrees nose up. Airspeed was 197 knots and radio altitude about 30 feet. The pitch then began decreasing. The copilot called a positive climb, and the commander ordered the landing gear up. At 23:10:41, when the gear lever moved up, radio altitude was 101 feet and the climb rate was 880 feet per minute.

The commander saw that the flight director guidance was inconsistent with the expected climb. She nevertheless lowered the nose toward the guidance. She told investigators that she had intended to pitch toward 15 degrees but held approximately 5 to 7.7 degrees nose up for about 15 seconds while trying to understand the indication. She also suspected that the aircraft weight calculation might be wrong.

Looking outside after selecting the gear up, the copilot saw increasing speed and an aircraft still close to the ground. He reported being startled for a couple of seconds before calling for a climb. The augmenting commander called for the pitch to be raised. At 23:10:50, the aircraft was at 160 feet radio altitude and 216 knots, with its climb rate reduced to about 255 feet per minute.

TO/GA and two flap speed exceedances

The commander advanced the thrust levers manually. Between 23:10:54 and 23:10:59 she reduced the pitch from 7.7 degrees to about 3 degrees nose up. Airspeed reached the 230 knot flap 15 limit at 23:10:56, triggering flap load relief. The flaps automatically retracted to position 5 while the flap lever stayed at 15. The protection reduced flap loading as the aircraft continued accelerating.

By 23:11:00 the climb rate was about 64 feet per minute. About a second later, the commander pressed TO/GA. At activation the recorded radio altitude was 211 feet and calibrated airspeed was 240 knots. The flight mode annunciator changed to thrust reference, TO/GA roll and TO/GA pitch. The flight director pitch bar moved from a nose down command to climb guidance, and the commander began raising the nose.

Increased thrust combined with the initially low pitch caused further acceleration. Airspeed exceeded the 245 knot flap 5 limit at 23:11:03 and reached 262 knots during the recovery. Pitch increased from 3 to 14 degrees between 23:11:00 and 23:11:16. The autopilot was engaged at 23:11:11 as the aircraft climbed through 459 feet radio altitude at about 2,610 feet per minute.

The recorder also showed several large roll control inputs during the shallow climb, although the bank angle stayed below about 2.5 degrees and heading remained about 300 degrees. There were brief speedbrake lever movements, but the speedbrakes did not deploy. No ground proximity warning occurred. At 23:11:20 the flap lever was moved from 15 to 5; subsequent retraction brought the flaps fully up at 23:13:01.

The onward flight and investigation evidence

After top of climb, the augmenting commander offered the commander a short break. She said she had recovered from the surprise and continued as pilot flying. The augmenting commander later returned to the cockpit and stayed for an hour. He also asked her to contact the operations office through ACARS to check the load calculations and the aircraft condition after the overspeed. The flight landed uneventfully at Washington Dulles at 13:30:28 UTC. There were no injuries or aircraft damage.

The flight data recorder was downloaded successfully. The cockpit voice recording had been overwritten because more than two hours had elapsed since the event. Consequently, investigators could not precisely time the verbal interventions described by the pilots. The GCAA reconstructed the flight path from recorded parameters and used crew statements for the cockpit exchanges.

A simulator session reproduced the incident settings and control inputs. The simulation showed the inappropriate ALT guidance, flap load relief and the restoration of TO/GA pitch guidance after the switch was pressed. No ground proximity warning occurred during the simulation. The report describes a slight simulated descent when the incident control inputs were reproduced; the recorded incident sequence gives a climb rate of about 64 feet per minute near TO/GA activation.

Findings, procedures and safety action

The GCAA identified the cause as pitching below the attitude required for a positive, safe climb while following guidance from the wrong flight director pitch mode. Its contributing factors covered the AFDS design allowing takeoff in that mode, the commander missing the mode and altitude entry, the copilot failing to resolve or communicate the mode problem, and the absence of explicit reset procedures.

The investigation found that both operating pilots were qualified and fit for the flight. Fatigue was not a contributing factor. Neither pilot had received training in recognizing and handling indications from incorrect flight director settings. The GCAA found the copilot's intervention limited and his response to taking control insufficiently assertive. It also considered startle and the barriers to taking control in operational circumstances when assessing the delay.

The training manual used the attitude indicator as the primary pitch reference after liftoff and called for crosschecking pitch, airspeed and vertical speed. Investigators examined the commander's nose down control inputs against the aircraft's trim behavior. The primary flight computers had set a trim reference speed consistent with the normal initial climb, so pushing the nose down required increasing force against the pitch trim response. The report associated the large roll inputs and incomplete speedbrake lever movements with stress and confused control actions during the overspeed.

Emirates had removed an additional takeoff review procedure in July 2019. Investigators examined fleet data covering 31 takeoffs in ALT mode between January 2018 and December 2021. Statistical comparisons found no significant difference between the rates before and after removal, so the GCAA could not conclusively link the apparent increase to that procedural change.

The final report, issued on 22 July 2024, recorded a new Emirates FMA check requiring an announcement as well as verification, additional monitoring of AFDS errors and shallow climbs, and a recurrent training module on intervention. Boeing issued fleet information and began developing software changes to inhibit transitions from takeoff pitch mode to other modes on the ground. Recommendation SR10/2024 asked Emirates to tell pilots to avoid setting the MCP altitude at runway elevation, reducing the risk of that entry carrying into the next departure.