Flight Safety Report: A fatal crash using automatic pilot.


the online flight safety digest

Text Box: automatic pilot = automatic crash?

the gremline digest — automatic pilot = automatic crash?

Automatic Pilot = Automatic Crash?

A reliable autopilot in a GA aircraft can be a valuable aid to flying with its ability to maintain present parameters accurately and thus assist single-pilot operation by reducing the workload. However, pilots using an autopilot need a complete understanding of the operation and limitations of the particular autopilot fitted to their aircraft and of the aircraft/autopilot interface. They must not rely on the autopilot to assume command of the flight.



The Mooney M20J, with four people on board, was en route from Sherburn-in-Elmet Yorkshire to Texel, Holland. The aircraft established RT contact with Church Fenton Approach soon after departure and received a clearance to climb to FL55. The flight proceeded under a Radar Advisory Service and had climbed to 2000 feet before being handed over to Humberside ATC. The cloudbase was between 500 and 800 feet with cloud tops between 1800 and 2000 feet. The handover to Humberside ATC was without incident and the aircraft did not transmit any indication of a problem on board.
      Ground witnesses heard the sound of ‘an engine in trouble’. The aircraft was seen to descend out of cloud in a steep nose-down attitude, turning slowly to the right, before impacting the ground. Others described the sound of an engine running but cutting out four or five times, followed by the sound of the impact. A column of smoke rose immediately and the intense fire prevented any rescue attempt. The emergency services were on the scene within minutes of the accident but the four occupants of the aircraft had died on impact.


Radar Recordings

Recorded radar returns from two radar units show a significant reduction in groundspeed followed by a loss of climb rate indicative of a power loss or power reduction. The derived groundspeed from the recorded radar plots show speed variations from 138 knots at 1000 feet to a minimum of 60 knots as the aircraft reached a maximum altitude of 2400 feet, before it began a steep, slow speed descent to impact. It is noticeable that the groundspeed fluctuated erratically, by as much as 40 knots during the climb, with the speed dropping below 70 knots as the aircraft cleared the lower cloud layer.
      The aircraft apparently stalled, descended into cloud and may have been spinning when it appeared below cloud and impacted the ground.


The Airframe

The aircraft was upright when it struck the ground at an angle of 42° below the horizontal and on a southerly heading, having been tracking easterly while airborne. The rear fuselage suffered compressive buckling and the tail unit whipped around to the left, the left stabiliser tip contacting the ground. This suggests rotation to the right at the moment of impact. An intense post-impact fire consumed the cabin area and the inboard wing sections including the fuel tanks. The engine was buried in a 1 metre deep crater that filled with water.

      The post-impact fire destroyed the cockpit instrumentation. The propeller RPM, mixture and throttle control knobs were all fully forward i.e. maximum RPM, mixture fully rich and throttle fully open. The cowl flap lever was in the fully closed position. It would be normal for the cowl flaps to be selected at least partially open during the climb to control cylinder head temperatures.

      The flying controls did not show any evidence of pre-impact disconnection. The horizontal stabiliser trim screwjack was found with eleven threads exposed. Fourteen threads are exposed at full nose up trim setting; three threads are exposed at full nose down trim. Eight exposed threads give approximately neutral trim.


The Engine
The engine was a Lycoming IO-360-13B6D piston engine with a fuel injection system. There was no evidence of pre-impact failure. However, both inlet valve cam lobes on the camshaft were severely worn, with approximately 0.140 inches removed from the cam peaks. This would have reduced the inlet valve lift by the same amount, producing an unquantifiable reduction in power output. The overhaul agent stated that worn inlet valve cams were occasionally seen, but usually only one cam would be affected. Engines so affected were invariably in the workshop for reasons other than symptoms of reduced power.
      The problem of cam wear has been addressed by Lycoming and is usually associated with engines that are run relatively infrequently. During a period of inactivity, the oil drains away from the camshaft that is located at the top of the engine. The cam lobes can then suffer corrosion pitting, with the engine suffering wear on start-up resulting from metal-to-metal contact between cams and cam followers, before oil reaches this are of the engine. The inlet valve operating cams are more likely to be affected as each cam operates the inlet valves of two opposing cylinders, and thus work harder than the exhaust valves, that operate only one valve each.


The Autopilot
The aircraft was fitted at manufacture with a King KFC 200 Autopilot system. The autopilot was most often used by the pilots of this aircraft in the heading and attitude hold modes. It also had an altitude hold facility and could interface with the GPS and other navigation systems in the NAV mode. In the attitude hold mode the autopilot maintains an aircraft attitude selected by the pilot by means of a switch on the autopilot control panel. While in use the autopilot will trim out any control forces after a few seconds so that the aircraft will be in trim following autopilot disengagement.
      In addition to the mode control panel, the autopilot components include pitch and roll servo motors, a pitch trim servo motor and the autopilot computer. The pitch servo is located in the rear fuselage and consists of an electric motor driving a cable drum attached to the elevator. Engaging the autopilot operates a solenoid that engages the servo motor to the cable drum. Out-of-trim forces, generated by moving the elevator, take the form of tensions in the cables that register as a torque on the cable drum axis. This triggers one of two torque switches, depending on whether the tension is in the ‘elevator up’ or ‘elevator down’ side of the cable drum. This electrical contact is detected by the autopilot computer which signals the pitch trim servo motor to operate the trim screwjack until the cable tension is relieved and the torque switch ‘unmakes’.
      The autopilot can be used to make the aircraft climb by operating the vertical trim switch in the UP direction which then drives the flight director V-bar. The pitch servo then moves the elevator until the aircraft adopts the selected attitude.
      The pitch and roll servos, together with the autopilot computer and the pitch trim servo from the crashed aircraft were bench tested. A defect was found in the pitch servo in that the ‘nose down’ torque switch made only intermittent contact. This would have resulted in only intermittent nose down trim screwjack operation. Thus out-of-trim forces would have been held by the servo and would not have been apparent to the pilot until the autopilot was disconnected. The intermittent nature of this defect probably meant that a severely out-of-trim condition would gradually reduce with the autopilot in use for a long time.


The Loss of Control
It is estimated that the aircraft was close to maximum weight and at the aft trim limit at the time of the accident. The manufacturer quotes an indicated stall speed of 61 knots at maximum weight. The stall warning consists of a warning horn only, without any associated warning light. The same horn was used to give an intermittent tone any time the throttle was retarded below about 1500 RPM with the landing gear unlocked. The autopilot disconnect also had a similar Audible warning.
      A similar aircraft was flight tested to evaluate the power/airspeed relationship during an autopilot climb. It was found that a substantial power reduction would stop the climb within 200 feet and cause the airspeed to decay to a speed approaching the stall in 20-30 seconds. The trim also drove to an aft position that would have approached 11 turns of screwjack at the stall. The trim was then manually set to11 turns of screwjack and the aircraft flown without autopilot to determine the control forces required. Unusually strong control forces were required and it was felt that this position would not have been attained using manual trim input during normal flight conditions, without autopilot in operation.

The final minutes of recorded radar returns show a decrease in rate of climb to zero and a reduction in airspeed towards the stall, both consistent with partial or total power loss. The flight test on a similar aircraft and the witnesses’ description of the engine cutting in and out is most consistent with a partial power loss. The worn inlet valve cams would have caused a reduction in maximum engine power. It is possible that there was a marked rise in cylinder head temperature during the climb, because of the closed cowl flaps. This would have caused a further power loss. If the cumulative power loss prevented further climb it is possible that the pilot operated the engine controls in an attempt to identify/rectify the problem. With cloud tops about 2000 feet and the base at the accident site about 500 feet a successful forced landing would have been difficult.
      Recognition and response to the engine failure seems to have been a problem. The commander gave no indication of any problem. The sound of the engine power reduction may have been partially disguised by the constant speed propeller attempting to maintain the set RPM. The noise-suppressing headsets worn by the pilots would have been effective in suppressing external noise cues. With the autopilot maintaining a constant pitch attitude right down to the stall, the normal pitch down following loss of power would not have occurred.
      The evaluation of the trim forces during the post-accident flight test led to the conclusion that the autopilot was engaged right up to the time of the loss of control, and possibly until the time of impact. The usual pre-stall cues of a nose high attitude and sloppy controls would have been disguised by the autopilot. It is probable that their first indications may have been the stall warning horn, but this may not have been immediately recognised because of its similarity to the undercarriage warning horn, which they were used to hearing. The sound of the horn may not have been obvious through the noise-suppressing headsets being worn. If the autopilot had been disconnected at the stall a considerable push force would have been required to lower the nose for recovery, and the aircraft was close to the cloud tops. If the autopilot was not disconnected at the stall the positive pitch input would continue to be applied throughout the stall and subsequent spin, making recovery almost impossible.
      Once the aircraft had entered the incipient spin the pilot would have had limited visual reference with the cloud directly below him. On entering cloud disorientation would have been complete and recovery beyond his ability. There was insufficient height available for spin recovery below the cloud.


The Safety Message
The pre-flight checklists on Private category aircraft are not subjected to UK Civil Aviation Authority (CAA) scrutiny as are the Operations Manuals for public transport aircraft. The checklist for this and similar aircraft call for only a cursory check of the autopilot. A full check would have revealed the pitch servo defect and the pilot could have elected not to use the autopilot on this flight.
      A number of light aircraft in the UK are fitted with autopilots. A handbook normally describes the autopilot functions and controls. It would be unusual to receive specific training in the use of the autopilot, other than the use of the switches. Some pilots using these systems do not fully understand the potential problems when a malfunction of the aircraft or the autopilot occurs. In this accident it appears that the autopilot helped to disguise the loss of power and the impending stall. It is likely that the aft trim position would have impeded a recovery attempt.
      The autopilot is a useful tool for pilots trained in its use, but without proper training it can be misleading. It would appear that there is a need for training in the proper understanding and use of autopilots.
      My personal opinion is that the CAA, as part of its safety promotion programme, should bring to the attention of all private pilots likely to operate aircraft that are fitted with autopilots the need for adequate training in their use.


The facts in this article are based closely on Air Accidents Investigation Branch Field Investigation Reference EW/C99/4/3 which source is gratefully acknowledged. Any opinions expressed are those of the author.




Search Now:
Amazon Logo



Text and Photographs © 2007 Gremline & Hill House Publications, unless otherwise stated.


landing page   about gremline   copyright/conditions/contact   information exchange   glossary

uk emergency diversions   uk links, chirp & gasco   global & misc links   forum

the gremline cockpit — index of articles   the gremline bookshop   top of page