Accident Overview

History of Flight/Flight Data

Photo of Air France 747-400 on Final Approach for landing
Photo of Air France 747-400 on Final Approach for landing
Photo copyright Manas Barooah - used with permission

Air France Flight 072, from Paris, France to Tahiti, with an intermediate stop in Los Angeles, California, departed Los Angeles on September 13, 1993. The flight to Tahiti was uneventful. Upon arrival in Tahiti, with the first officer at the controls, the flight began the VOR DME approach to runway 22 (VOR DME 22 Approach Plate) at around 8:40 p.m. local time. It was night time and the weather was clear. The airplane was configured for landing, and holding a stable approach speed (149 knots) and descent path. The autopilot was not connected, and the auto-throttles were engaged.

At the missed approach point (approximately 500 feet above touchdown), the automatic flight system (AFS) commanded a go-around. The airplane's pitch attitude did not change because the autopilot was not connected, but the throttles were automatically commanded to full-forward thrust.

The investigation concluded that the pilots were unaware of the throttle increase for almost 20 seconds. When they recognized the throttle increase, the first officer, who was flying the airplane, countermanded the throttle/thrust increase by holding the throttles back at idle with his hand. However, the airspeed had already increased to 189 knots and the descent path had leveled off. With the throttles at idle, the airspeed slowly decreased to 168 knots at touchdown, but the airplane landed long, approximately 3,000 feet down the runway and almost 20 knots faster than the recommended approach speed.

Photo of Final Approach to Faa'a Airport
Photo of Final Approach to Faa'a Airport
Photo copyright tessede - used with permission

This depiction shows the Airplane Runway Trajectory during rollout. An animation of the landing and rollout, culminating in the runway departure, is available at the following link (Landing Excursion Animation). The airplane performance aspects of a long landing are described in the following animation: (Landing Performance Animation). This animation is also available in the Luxair F-27 Mk 050 accident module elsewhere in this library.

Even though the pilot was restraining all four throttles, the auto-throttle system was attempting to advance all throttles to takeoff power. Two seconds prior to touchdown, the number one throttle slipped from the pilot's hand and advanced to full forward thrust. Neither of the pilots recognized the position of the number one throttle, and at touchdown reverse thrust was initiated on the other three engines. The airplane continued to slow. However, with the number one engine at full forward thrust, the AFS prevented spoiler extension and activation of auto-brakes, which had been armed. This subsequently resulted in delayed actuation of manual brakes. It was not clear if manual spoiler extension was undertaken; however, the spoiler handle was in the DOWN position on the flight deck when examined post-accident by the investigators. The lack of spoilers would affect initial braking effectiveness. The investigation concluded that this, in combination with delayed braking, would have an adverse effect on airplane deceleration. The high forward thrust on the number one engine also caused the airplane to drift to the right, departing the runway and coming to a stop in the adjacent lagoon.

Photos of a 747-400 flight deck and 747-400 thrust levers
747-400 Flight Deck (left) – Photo copyright Richier – used with permission
747-400 Thrust Levers (right) – Photo copyright Guillaume Noirot – used with permission

A plot of the engine thrust time history during the landing is available here: Engine Plot

The pilot's were able to shut down engines 2, 3 & 4, but the number one engine did not respond to the pilots commands. The fire department was finally able to get the engine shut down by spraying water into the engine inlet.

Post-Accident photos
Post-Accident photos

747 Flight Control Automation

Diagram of AFDS Architecture
Diagram of AFDS Architecture
View Large

Photo of Flight Director Command Bars on Pilot Display
Photo of Flight Director Command Bars on Pilot Display

The automatic flight control system on the 747-400 consists of the autopilot flight director system (AFDS) and the auto-throttle system (A/T). The mode control panel (MCP) and flight management computer (FMC) control the AFDS and the auto-throttle system to perform climb, cruise, descent, and approach.

The AFDS consists of three flight control computers (FCC) and the MCP. The MCP provides control of the autopilot, flight director, altitude alert, and auto-throttle systems. The MCP selects and activates AFDS modes, and establishes altitudes, speeds, and climb/descent profiles.

The three FCCs, left, center, and right, control separate hydraulically powered A/P control servos to operate flight controls. The A/P normally controls roll and pitch. Yaw commands are added only during a multi-A/P approach. Nose-wheel steering is also added during rollout from an automatic landing. During an ILS approach with all three autopilots engaged, separate electrical sources power the three flight control computers.

MCP switches select automatic flight control modes. Selected modes are annunciated. The autopilot is engaged by pushing one of the MCP autopilot engage switches. Disengagement is accomplished either through a disengage switch on the control wheel of either pilot, or via the MCP disengage bar. When the autopilot is disengaged, either manually or automatically, a warning message, and aural tone provide an alert to the disengagement.

Diagram of Mode Control Panel
Diagram of Mode Control Panel
View Large
Diagram of Auto-throttle Arm Switches
Diagram of Auto-throttle Arm Switches

The flight director steering commands normally display any time the related flight director switch is turned on. Autopilot and auto-throttle modes are displayed in flight mode annunciation boxes on the pilot displays.

The auto-throttle system can provide thrust control from takeoff through landing. Auto-throttle operation is controlled from the MCP and the flight management computer displays (CDU). The MCP allows mode and speed selection. The CDU allows selection of FMC reference thrust limit. When an autopilot or flight director pitch mode is active, the FMC selects auto-throttle target modes and target thrust values. The auto-throttle can be operated without using the autopilot or flight director.

The auto-throttle can be manually overridden, or disconnected by using either auto-throttle disconnect switch located on the throttles, or by positioning the auto-throttle arm switch to OFF. If either action is taken by the crew, the disconnection is annunciated.

VNAV is an available auto-flight function that computes guidance commands for the autopilot or flight director and auto-throttle to follow the vertical profile of a route, or flight path, programmed into the FMC. VNAV is designed to optimize the airplane's vertical performance capability. A brief overview of the VNAV function, entitled "VNAV for Dummies," is available at the following link: (VNAV Overview).

Air France Flight Profile

In this accident, the co-pilot was manually flying the airplane with auto-throttles engaged, in good weather. The flight director was active, and the flight guidance mode was VNAV. The flight was cleared for the VOR/DME approach to runway 22.

Diagram of VNAV Select Switches on MCP
Diagram of VNAV Select Switches on MCP

The early stages of the approach were routine. As the airplane passed through 4,500 feet, the captain read the flight mode annunciator speed and VNAV path instructions. At ten nautical miles from the runway, the landing gear was extended, and the flaps were set at 30 degrees. During this approach, the airplane roll, pitch, and yaw were controlled by the co-pilot, via the control wheel and rudder pedals, and the auto-throttle was controlling airplane speed by controlling engine thrust levels. The speed commanded for the auto-throttle was the recommended approach speed plus five knots (149 knots). The auto-throttle was employed as a workload reduction strategy, and is commonly used in this manner. The crew monitored the descent path by crosschecking altitude as a function of DME distance, and also monitored the visual approach guidance located near the end of the runway.

All approaches can be categorized as precision or non-precision. The VOR/DME approach in this accident is considered a non-precision approach. At the time of this accident, it was standard procedure for pilots to fly non-precision approaches with the autopilot or flight director engaged in vertical speed mode. The pilots in this accident were using a non-standard configuration by flying the non-precision approach in VNAV versus, for example, the vertical speed mode. In VNAV, the automation is programmed to initiate a go-around at the missed approach point, since VNAV is not an approach mode for the auto-flight system. Though, by initiating the go-around, VNAV in this case performed as designed. Investigators noted that the airplane manufacturer had not made airlines, including Air France, aware of this particular aspect of VNAV. Therefore, Air France had not trained crews to disengage VNAV prior to the missed approach point.

Diagram of Auto-throttle Disconnect Switches on Throttles
Diagram of Auto-throttle Disconnect Switches on Throttles
View Large

At 500 feet the defined "missed approach point", since VNAV was still engaged, a go-around was automatically commanded and engine thrust was increased by the auto-throttles. When VNAV is engaged, the flight path is defined to the point corresponding to a decision height - in this case, the defined missed approach point (MAP) of 500 feet. Once reaching the MAP, if VNAV is still engaged, it is designed to follow the missed approach flight path and initiate a go-around. At the MAP, the approach criteria require that the pilot have sufficient visibility to disconnect the automation and complete the landing. With insufficient visual references, the pilot must perform a go-around. The investigation stated that it was part of the design of the automatic flight system that if an end of descent point is reached (the MAP, in this case), the auto-flight system concludes that insufficient visual references were available, and a go-around is appropriate. When the go-around is initiated, the auto-throttle system increases thrust, and the flight director, if engaged, provides pitch guidance. If the autopilot is engaged, it will fly the airplane through the go-around, maintaining the speed selected on the MCP to the selected altitude.

Aspects of Airplane Design

Crew Training

The airplane automation performed as designed. The investigation concluded, however, that the flight crew did not understand the airplane's response, and reacted inappropriately. The accident report cites several factors that contributed to the accident:

  • The absence of information from the manufacturer regarding the VNAV go-around feature of the AFS.
  • The crew used a non-standard automation configuration for a non-precision approach.
  • The Boeing flight manual recommended that the crew disconnect the autopilot and the auto-throttle before passing the MAP. The accident report stated that Boeing had failed to inform Air France that if the autopilot and auto-throttle were not disconnected by the MAP, the VNAV mode would command a go-around. Subsequently, Air France did not train their pilots to anticipate a go-around at the MAP when conducting a non-precision approach.

As a result of this accident, Boeing modified the automatic flight system logic for non-precision approaches to more closely reflect the VNAV mode performance employed in precision approaches, and made the system upgrade available to the operational fleet.

Autobrakes & Spoilers

The number one throttle moving forward just before touchdown resulted in deactivation of the automatic brakes. The throttle movement also prevented the spoilers from extending automatically. These two events affected the deceleration characteristics of the airplane and increased the deceleration distance (landing roll) of the airplane.

Pilot Procedures

Communication

The accident report cites failure to observe operational procedures regarding call-outs during approach and landing as well as the lack of communication between the pilots, as factors that contributed "greatly" to the accident. The investigation concluded that industry standards and procedures for crew resource management were not followed. The accident report has a detailed list of deviations from these procedures:

  • A go-around plan was not included in the approach brief.
  • Errors in the checklist were not corrected.
  • Several acknowledgements to callouts were missing.
  • Following touchdown, the pilot not flying did not make the required call-outs, including the failure of the number one thrust reverser to engage, the deactivation of the autobrakes, or the failure of the spoilers to extend and the pilot flying did not comment, or correct the matter.

Go-Around Standard Procedures

Industry standards dictate that anytime on an approach, whether instrument or visual, when close to the landing, if any parameters of flight become unstable, i.e., altitude and airspeed, a go-around should be initiated. Furthermore, once a go-around is initiated, it should be continued. Once the throttles are advanced, the airplane is committed to going around and is no longer in a safe position to land.

The accident report stated that the pilots did not realize that an automatic go-around had been initiated for almost twenty seconds. Once realizing that they had accelerated above approach speed and the throttles were at go-around thrust, they failed to follow industry standards and aborted the go-around, attempting to land from an approach that had become unstable.

The Psychology of Visual Meteorological Conditions

As a result of this accident, and several other similar accidents and incidents, the French accident investigating body, the BEA, initiated a study of the human factors associated with pilot decision making in visual approach conditions. The study discovered that pilots flying in visual meteorological conditions failed to properly consider going around when an approach was not stable. The report concluded that pilots had the mind set, perhaps unconsciously, that once the runway is visually acquired, there are no further obstacles to landing, and going around is no longer considered as an option. Even when parameters were very unstable, approaches were continued because pilots thought they should be able to make the landing.

In a post accident interview, the pilot of this accident actually described this mentality about weather conditions as an explanation for his decision to continue with the landing.

Photo of accident airplane in revenue service
Photo of accident airplane in revenue service
Photo copyright Philippe Gindrat - used with permission

Federal Aviation Administration Human Factors Team Report on: The Interfaces Between Flight Crews and Modern Fight Deck Systems, dated June 18, 1996

Many accidents, including this one, highlighted difficulties in flight crews interacting with flight deck automation. The FAA, along with U.S. & foreign authorities, manufacturers and operators, launched a study to evaluate the flight crew/flight deck automation interfaces of current generation transport category airplanes. This report is a culmination of that study. The purpose of this study was to identify opportunities for improvements which could be applied in future regulatory and policy development associated with flight crew/flight deck automation interfaces. (FAA report)

Back to top