On 23 September 2012, a Boeing 737-300 being operated by Thomsonfly on a passenger flight from Faro to Bournemouth UK made a go around in night Instrument Meteorological Conditions (IMC) from its first approach at destination which was subsequently found to have involved a temporary loss of control before recovery was achieved. The approach and landing which followed were successful and without further abnormal occurrence. None of the 137 occupants were injured as a consequence of the extreme manoeuvres which occurred.
An Investigation was carried out by the UK AAIB but this was not commenced until the Branch was made aware of the occurrence by the Operator on 5 October 2007, some 12 days after it had occurred. It was noted that “although the commander reported the event to the operator the following morning, his initial Air Safety Report (ASR) contained limited information and the seriousness of the event was not appreciated until the Quick Access Recorder (QAR) data was inspected on 4 October 2007”, At this point, the ASR was submitted to the UK CAA as an MOR and the AAIB then advised the following day. As a consequence of this delay, both the Cockpit Voice Recorder (CVR) and Flight Data Recorder (FDR) from the event had been overwritten and only the QAR data was available.
It was observed that the aircraft involved had not been subjected to an engineering examination to ensure its continued airworthiness in the interval between the occurrence and notification of it to the AAIB and had remained in service throughout this period.
The Investigation reconstructed the flight by linking QAR data to the accounts provided by the pilots involved. It was found that the First Officer had been designated as PF and that the aircraft had been cleared to self position for an ILS approach to runway 26. For the Flap 40 landing briefed, the Reference Speed (Vref) was calculated as 129 KIAS and the Vapp as 135 KIAS. The aircraft was established on the ILS LOC at 180 KCAS with flap 5 set and, with the AP and A/T engaged, captured the ILS GS from below at 2500 feet Altimeter Pressure Settings. At a range from touchdown of 7nm, descent on the ILS GS began and landing gear down and flap15 were selected. The MCP-selected speed was reduced and the A/T, as expected, retarded the thrust levers to idle in response. At 20 seconds after the descent had commenced, the A/T disconnect warning began to sound and soon afterwards the A/T disengaged. At this point, the thrust levers were still at Idle, where they remained throughout the rest of the approach. Neither pilot recognised the A/T disconnect. The airspeed continued to decay at about one knot per second as anticipated by the PF and the AP continued to track the ILS. As the speed reduced below 150 KIAS, flap 25 was selected and the MCP speed selection was changed to 135 knots. When flap 40 was selected shortly afterwards, the speed was 130 knots and continuing to decrease at approximately 1.5 knots per second.
The first recognition that airspeed was too low came when the commander, as PM, noticed it was 125 KIAS. He reported that he had called ‘Speed’ and taken control and called a go around. The QAR data shows that as airspeed continued to reduce below 115 KCAS, the manual A/T disconnect was used to cancel the A/T disconnect warning and stick shaker activation occurred with the AoA almost 20°. Thrust was increased towards maximum and the airspeed reached a minimum of 105 KCAS before beginning to rise as thrust was increased. There was a two second cessation of the stick shaker as the minimum speed was reached but activation recommenced as thrust rose to 99.4%, TO/GA mode became active as AP disconnect occurred and maximum nose down input was made via the control column. By the time of AP disconnect, as the minimum altitude of around 1400 feet (1200 feet agl) was reached, the stabiliser auto trim had progressively advanced to a significant nose up position in an attempt to keep the aircraft on the ILS GS at idle thrust.
As maximum thrust was reached with the control column fully forward, pitch appeared to stabilise at 22º nose up, the angle of attack started to decrease and airspeed reached 118 KCAS. In the presence of some roll instability, flaps were then selected to 15 and the stick shaker ceased but as the flaps travelled, the nose of the aircraft began to pitch up at an increasing rate and by the time flap 15 was reached, pitch was increasing through 27º and the stick-shaker reactivated and airspeed began to decay again. At this stage, both pilots reported holding the control column fully forward but stated that they had had “no pitch control authority”. They were aware that the airspeed had fallen rapidly but neither was aware exactly what it was.
As nose-up pitch continued to increase through 36º, TO/GA mode disconnected and airspeed fell below 107 KCAS. Just as the wings were levelled after recovery from a roll left which had reached 22º, the aircraft entered a full stall with pitch reaching 44º. The QAR data showed that “with no change in elevator position the pitch rate (then) reversed from positive to negative, although angle of attack continued to increase as the aircraft started to descend”. Despite the reduction in pitch, airspeed continued to fall for a further five seconds to a recorded minimum of 82 KCAS. It was apparent that, as pitch had continued to reduce, airspeed had begun to rise rapidly and, when thrust was then reduced to 86% from the 96-98% set since the go around had been initiated, increased elevator authority was gained and the fully stalled condition exited. Approximately 17 seconds after the full stall had begun, the stick shaker ceased. With the aircraft stabilised in a 5º nose-up attitude, airspeed continued to increase and “the commander regained control of the aircraft”. As the airspeed reached 147 KCAS, TO/GA mode was re-engaged. The 11 second full stall had occurred after a climb of just over 1000 feet with pitch attitude in excess of 20º and recovery had led to a height loss of around 500 feet.
It was found that the 56 year old aircraft commander had only accumulated 420 hours on both the aircraft type and in command after spending the previous 17 years as a First Officer on the same operator’s Boeing 757/767 fleet. The 36 year old First Officer had only slightly more time on type after joining the Company the previous year from a DHC8-300 operator.
In an analysis of the excursion, the Investigation came to the conclusion that since stick shaker activations had begun soon after the decision to make a go around with airspeed around 20 knots below the applicable approach speed and at idle thrust, there had been confusion as to whether a go around or an “approach to stall recovery” was being flown. Although the root cause of the lack of elevator authority was the unappreciated position of the stabiliser trim at AP disconnection (no attempt to alter pitch trim manually was made until after control was regained following the stall), it was noted that there was a difference between the “maximum thrust” required for recovery from an approach to stall and the “sufficient thrust” required for a go around.
The Investigation noted that the generic aircraft type FCTM, which was available to crews but over which the Operations Manual Part B (as usual) took absolute preference, placed more emphasis on the use of stabiliser trim in pitch control in the case of both a go around and the approach to stall recovery drill. The FCTM was also found to similarly place more emphasis on the potential usefulness of stabiliser trim in the case of “upset recovery” - an “upset” being defined as generally accepted as including a pitch attitude in excess of 25° in both documents.
It was noted that the unstable approach which eventually led to the decision to go around had developed because neither pilot had noticed the uncommanded but (it would appear) correctly alerted A/T disconnect or the effects that this then had on thrust/airspeed/pitch attitude. The Investigation could not find any fault in the A/T which would have led it to disconnect and noted that the system had operated normally both before and after the disconnect approach and had been comprehensively tested on the aircraft with no faults found. However, after eliminating other possible reasons for a disconnect, it was concluded that “the only condition that remained was an internal fault within the autothrottle computer”.
It was noted that the A/T is a single channel system which is not required for flight operations and that an uncommanded disconnect is not unusual and it is rare for a fault to be found when instances of this are reported. The QAR data indicated that the A/T warning was active and subsequent function testing was unable to find any fault with the warning system. After looking at other similar A/T disconnect events, the Investigation concluded that “there may be a wider (than the investigated event) issue of the (A/T) warning system not alerting flight crews”.
It was concluded that there was “no evidence of pilot fatigue” being a factor and that the pilots had received training which was in accordance with Regulatory requirements as implemented by the aircraft operator. However, it was considered that “the upset recovery techniques outlined in the QRH, FCTM and the manufacturer’s training aid are effective and (their use) would have resulted in earlier recovery of the aircraft (to controlled flight)”.
In respect of the 11-day delay before the flight data from the Incident flight was viewed (by a pilot representative rather than Company flight operations management) it was noted that in addition to loss of recorded data, this had also resulted in the continued service of both the aircraft and pilots involved. It was found that, since there was no requirement in the agreement between the Operator and the Pilots for the use of OFDM data that it should be de identified, it could actually have been reviewed on the day after the event, upon receipt of the pilot ASR.
The Investigation identified the following Causal Factors:
- the aircraft decelerated during an instrument approach, to an airspeed significantly below the commanded speed, with the engines at idle thrust. Despite the application of full thrust, the aircraft stalled, after which the appropriate recovery actions were not followed.
- the trimmed position of the stabiliser, combined with the selection of maximum thrust, overwhelmed the available elevator authority.
The Investigation also identified the following Contributory Factors:
- the autothrottle warning system on the Boeing 737-300, although working as designed, did not alert the crew to the disengagement of the autothrottle system.
- the flight crew did not recognise the disengagement of the autothrottle system and allowed the airspeed to decrease 20 knots below Vref before recovery was initiated.
It was noted that since the investigated event, Safety Action had been taken by Thomsonfly by making significant changes to their OFDM procedures in particular and to other relevant aspects of their SMS.
Three Safety Recommendations were made as a result of the Investigation as follows:
- that Boeing, in conjunction with the Federal Aviation Administration, conduct a study of the efficacy of the Boeing 737-300/400/500 autothrottle warning and if necessary take steps to improve crew alerting. [2009-43]
- that the European Aviation Safety Agency review the requirements of Certification Standard 25 to ensure that the disengagement of autoflight controls including autothrottle is suitably alerted to flightcrews. [2009-44]
- that Boeing clarify the wording of the approach to stall recovery Quick Reference Handbook Non normal Manoeuvres to ensure that pilots are aware that trimming forward may be required to enhance pitch control authority. [2009-45]
The Final Report of the Investigation was published on 21 May 2009.
Flight Safety Foundation