Kegworth 1989: an accident waiting to happen?
On January 8, 1989, routine domestic flight 092 was enroute from London Heathrow airport to Belfast in Northern Ireland. It was the second flight undertaken by the British Midland Boeing 737-400 that day and the aircraft was close to its landing destination when a combination of mechanical and human error led to disaster.
Preparing to land at the East Midlands airport, the aircraft (tail marked G-OBME) plummeted onto an embankment of the M1 motorway near Kegworth, Leicestershire, killing 47 people and seriously injuring a further 74, including seven members of the flight crew.
In summarising the cause of the accident, The Aircraft Accident Report stated “The cause of the accident was that the operating crew shut down the No.2 engine after a fan blade had fractured in the No.1 engine. This engine subsequently suffered a major thrust loss due to secondary fan damage after power had been increasing during the final approach to land” (AAIB 1980, 35). This much is certainly true, however it was a combination of errors, mechanical, procedural and cognitive, which ultimately caused the aircraft to fail during its final landing phase.
In order to extrapolate the events of that day it is necessary to examine a chain of events rather than to study each constituent error or malfunction in turn. As is often the case with aircraft crash investigation, a sequence of human and operational errors tends to produce a domino effect in which it is the inertia of one event beyond another that results in a catastrophic conclusion (Job,1996; 173). The chronology of these events is therefore particularly important in helping to analyse the failure chain that led up to the crash.
G-OBME was engaged on a double shuttle run between London Heathrow airport and Belfast Aldergrove Airport. The first leg of the journey was uneventful. During the second leg of the shuttle the aircraft climbed initially to six thousand feet where it levelled-off for about two minutes before receiving clearance to climb to a flight level of twelve thousand feet. At 7.58 p.m., clearance was given to climb to thirty five thousand feet. At 8.05 p.m. as the aircraft was climbing through flight level 283 the crew experienced severe vibration and a smell of fire. No fire warnings, visual or audible were alerted by instruments on the flight deck. A later replay of the Flight Data Recorder showed that severe vibrations had occurred in the No.1 (left) engine, together with indications of an erratic fan speed, a rise in exhaust temperature and a low, variable fuel flow (AAIB, 1980; 145).
Captain Hunt took control of the aeroplane and disengaged the autopilot. He later claimed that the engine instrumentation did not give him any clear indication of the source of the malfunction. He also later stated that he thought that the smoke was coming forward from the passenger cabin which, from his understanding of the 737’s air conditioning system, led him to believe that the smoke was in fact coming from the No 2 (right) engine. Consequently the command was issued to throttle back the No.2 engine. As a result of this procedure the aircraft rolled slowly to the left through sixteen degrees but the commander made no corrective movements of either rudder or aileron.
The commander later claimed that reducing the throttle of No.2 engine reduced the smell and signs of smoke and but he later remembered that the significant vibration continued after the No.2 throttle was closed.
After throttling back the No.2 engine, London Air Traffic Control were immediately advised of an emergency situation with appeared to be an engine fire. Forty-three seconds after the onset of the vibration the commander ordered First Officer McClelland to “shut it down”. The shut down was delayed at the First Officer responded to radio messages from London Air Traffic Control asking which alternative airport they wished to land at. Shortly after shutting down No.2 engine BMA Operations requested the aircraft divert to the East Midland Airport (AAIB,1980; 40).
As soon as the No.2 engine had been shut down, all evidence of smoke cleared from the flight deck which further convinced the Commander that he had made the correct decision, not least in that No.1 engine showed no signs of malfunctioning and continued to operate albeit at reduced power and with increased fuel flow.
Passengers were aware of smoke and of smells similar to “oil” or “rubber” in the cabin. Some passengers saw evidence of fire from the left engine, and several cabin attendants saw fire from the No.1 engine as well as light coloured smoke in the cabin.
Despite indication that the fire was emanating from the other engine neither passengers nor cabin crew alerted the flight crew to this fact. This may have been due to general confusion at the time, allied with a belief that the pilot ultimately knew what he was doing.
At 8.20 p.m. at a height of three thousand feet power was increased on the No.1 engine. The aircraft was then cleared to descend to two thousand feet and, after joining the centre line at two thousand feet above ground level (agl) the Commander called for the landing gear to be lowered and fifteen degrees to be applied to the flaps. At nine hundred feet there was a sudden decrease in power from the No.1 engine. As the aircraft dipped below the glidepath and the ground proximity warning system (GPWS) sounded the Commander broadcast “prepare for crash landing” on the cabin address system. The aircraft hit the ground at 8.24 p.m. at a speed of 115 knots.
One survivor, Gareth Jones, described the moment when the plane hit the ground as follows: “There was a shudder, crash, like a massive motor car accident, crunch, blackness, and I was by the emergency hatch.” (BBC, 1989).
The AAIB report (AAIB, 1980; 35) concentrated upon the failure of the flight crew to respond accurately to a malfunction in the Number 1 engine, and highlighted the following operational errors:
1. The combination of engine vibration, noise and the smell of fire were outside their training and expertise.
2. They reacted to the initial engine problem prematurely and in a way that was contrary to their training.
3. They did not assimilate the indications on the engine instrument display before they throttled back the No.2 engine.
4. As the number 2 engine was throttled back, the noise and shuddering associated with the surging of the No.1 engine ceased, persuading them that they had correctly identified the defective engine.
5. They were not informed of the flames which had emanated from the No.1 engine and which had been observed by many on board, including 3 cabin attendants in the aft cabin.
Many accident reports cite human failure as a primary cause (Johnson, 1998).
However, before looking at the obvious failure in Captain Hunt’s inability to determine which of the 737’s engines had indeed malfunctioned, attention should be drawn to the faulty engine itself. The actual cause of the malfunction was a broken turbine, itself the result of metal fatigue caused by excessive vibration.
The upgraded CFM56 engine used on the 737-400 model were subject to excessive amounts of vibration when operating at higher power settings over twenty five thousand feet. Because this was an upgrade to an existing engine, the engine had only ever been tested in a laboratory, not under actual flight conditions. When this fact was subsequently discovered around a hundred 737-400’s were grounded and the engines subsequently modified. Since the Kegworth crash all significantly redesigned turbofan engines must be tested under actual flight conditions. Arguably then, the inadequately tested CFM56 engine on flight 092 may have been “an accident waiting to happen” (Owen, D. 2001; 132).
The AAIB report concluded that the combination of engine vibration, noise and the smell of fire were outside the flight deck crew’s area of expertise. (AAIB, 1980). This may or may not be a fair assessment since few pilot’s and First Officer’s fortunately ever experience the actual effects of smoke and fire while in command.
Whilst simulators can help train for emergency procedures it is questionable how valuable such procedures may be, particularly if the crew have not been thoroughly trained on the unique procedural and technical requirements involved in flying a particular aircraft variant. Significantly, the flight crew of 092 had little belief in the accuracy of key instrumentation including vibration meters.
Dr Denis Besnard of Newcastle university analysed the Kegworth air crash, concluding “The pilots of the B737 were caught in what is known as a confirmation bias where, instead of looking for contrary evidence, humans tend to overestimate consistent data. People overlook and sometimes unconsciously disregard data they cannot explain” (Besnard D, 2004; 117).
“Confirmation bias”, i.e. the overloading of consciousness by a quantity of bewildering or conflicting data was also established as a primary cause of the crash when investigated by a research team from the University of York and the University of Newcastle upon Tyne. The argument that people tend to over simplify complex situations particularly during crisis has been is both well documented and significant in the causation of the Kegworth air crash (Besnard. D., Greathead, G. & Baxter, G, 2004; 117-119).
Specifically, Captain Hunt had not received training on the new model 737-400 since no simulators for this variant existed in the UK at that time. This is both startling and critical when considering the following points. The captain believed the right engine was malfunctioning due to the smell of smoke, possibly because in previous Boeing 737 models the air for the air conditioning system was taken from the right engine.
However, starting with the Boeing 737-400 variant, Boeing redesigned the system to use bleed air from both engines. Captain Hunt would have been unaware of this fact, which formed a critical part of his decision to shut down the wrong engine. This would prove disastrous.
Apart from the coincidence of the smoke vanishing when the auto-throttle was disengaged, the pilots may have also been in the habit of disregarding the readings of vibration warning meters, since early ones were perceived to be unreliable. The crew of G-OBME do not seem to have been aware that newer ones were, however, more reliable. Should more attention have been paid, therefore, to vibration issues rather than to smoke and the smell of fire, events may well have transpired very differently on the evening of January 8th (Owen, 2001; 131-2).
Subsequent research has critically concluded that “organisational failures create the necessary preconditions for human error” and “organisational failures also exacerbate the consequences of those errors” (Stanton, 1994; 63). The Kegworth air crash was therefore the result of a sequence of failures originating from a mechanical defect.
Additionally, cognitive error on the part of the flight crew enhanced by inadequate flight training compounded the error chain. Finally the flight crew did not verify their interpretation of events by consulting with cabin staff or passengers even though information to suggest the fault lay with the other engine on the aircraft was available at the time.
BBC (1989) On This Day: Dozens die as plane crashes on motorway. [online] available from http://news.bbc.co.uk/onthisday/hi/dates/stories/january/8 [accessed 2 March 2007]
Besnard, D. (2005) International Aviation and Fire Protection Association. [online] available from http://www.iafpa.org.uk/news-template.php?t=4&id=1312 [accessed 1 March 2007]
Besnard, D., Greathead, G., and Baxter, G., (2004) International Journal of Human-Computer Studies. When mental models go wrong. Co-occurrences in dynamic, critical systems, Vol. 60, pp. 117-128.
Job, M. (1996) Air Disaster Volume 2. pp. 173-185. Aerospace Publications Pty Ltd
Johnson, D. 1988; University of Glasgow Department of Computing Science (1980) Visualizing the Relationship between Human Error and Organizational [online] University of Glasgow, 1980. http://www.dcs.gla.ac.uk/~johnson/papers/fault_trees/organisational_error.html [accessed 2 March 2007]
Owen, D. (2001) Air Accident Investigation, 1st ed., Ch. 9, pp. 132-152. Sparkford, Patrick Stephens Limited
Stanton, N.A., (1994) The Human Factors of Alarm Design, Ch. 5, pp. 63-92. London, Taylor and Francis Ltd
UNITED KINGDOM. Air Accidents Investigation Branch (1990) Boeing 737-400, G-OBME, near Kegworth, Leicestershire 8th January 1989, number 4/90. London, HMSO.