Air France Flight 4590

Accident Snapshot:
Date:  Tuesday, July 25, 2000
Location:  Gonesse, France (48° 59' 8" N, 2° 28' 20" E)
Persons/Fatalities:  109/109 +4 on the ground
Operator:  Air France
Origin Airport:  Paris, Charles de Gaulle (CDG/LFPG)
Destination Airport:  New York, John F. Kennedy (JFK/KJFK)

Accident Aircraft:
Aircraft Type:  Aerospatiale/BAC Concorde
Registration:  F-BTSC
First Flight:  1975
Total Airframe Hours:  11,989
Cycles:  4,873
Engines:  4 Rolls-Royce/Snecma Olympus 593-610-14-28 Turbojet Engines

Profile view of Air France Concorde © Gregory Maxwell, 2014

Resources:
BEA Final Accident Report - English
BEA Final Accident Report - French
Aviation Safety Network Page
Wikipedia Page
Concorde SST Website
Dateline NBC Segment (42 min)
Air Crash Investigation Special S10 (45 min)

Accident Summary:
During takeoff from Runway 26R at Charles de Gaulle Airport, shortly before rotation, the front right tire (tire No 2) of the left landing gear ran over a strip of metal, which had fallen from another aircraft, and was damaged.  Debris was thrown against the wing structure leading to rupture of tank 5.  A major fire, fueled by the leak, broke out almost immediately under the left wing.  Problems appeared shortly afterwards on engine 2 and for a brief period on engine 1.  The aircraft took off.  The crew shut down engine 2, then only operating at near idle power, following an engine fire alarm.  They noticed that the landing gear would not retract.  The aircraft flew for around a minute at speed of 200 kt and at a radio altitude of 200 feet, but was unable to gain height or speed.  Engine 1 then lost thrust, the aircraft's angle of attack and bank increased sharply.  The thrust on engines 3 and 4 fell suddenly.  The aircraft crashed onto a hotel.

Probable Cause:
High-speed passage of a tire over a part lost by an aircraft that had taken off five minutes earlier and the destruction of the tire.

The ripping out of a large piece of tank in a complex process of transmission of the energy produced by the impact of a piece of tire to another point on the tank, this transmission associating deformation of the tank skin and the movement of the fuel, with perhaps the contributory effect of other more minor shocks and/or a hydrodynamic pressure surge.

Igniting of the leaking fuel by an electric arc in the landing gear by or through contact with the hot pars of the engine with forward propagation of the flame causing a very large fire under the aircraft's wing and severe loss of thrust on engine 2 then engine 1.

In addition, the impossibility of retracting the landing gear probably contributed to the retention and stabilization of the flame throughout the flight.

ACCIDENT ANALYSIS:

Pre Accident Landing Gear Maintenance

July 18, 2000 Air France Concorde, registered F-BTSC, enters the hangar at Charles de Gaulle Airport for routine maintenance.  Concorde due to the stresses incurred during its supersonic flight regime and its unusually high takeoff and landing speeds required regular, extensive maintenance ever few hundred flight hours to ensure the aircraft remained airworthy.  Critical parts especially those associated with the Concorde's landing gear assembly had a useful life of only a few hundred hours due to the tremendous strain placed on them by the weight of the aircraft during takeoff where it reached an average speed of 250 mph before lifting off the ground.  By comparison a 747-400's average takeoff speed is around 180 mph, a full 70 mph slower than Concorde.

Concorde's high takeoff speed is due to its Delta wing configuration which is optimized for high speed supersonic flight, but as a result it produces much less lift at lower speeds than a conventional airliner wing requiring Concorde to accelerate to a much higher airspeed in order to generate enough lift to climb into the air.  Concorde's delta shaped wing in fact produces almost no lift up to the point of rotation, meaning that the entire weight of the aircraft hurtling down the runway at speeds up to 250 mph is supported by the eight tires of the main landing gear assembly and two nose gear tires.  The strain on the landing gear assembly and tires is so great that during the aircraft's initial certification test campaign from 1969-1975 the entire main landing gear assembly had to be redesigned and reinforced to withstand stresses which were well beyond what engineers had predicted.

  Illustration from concordesst.com

The main landing gear assembly of the aircraft is under its greatest strain at the moment of rotation when the the aircraft traveling 250 mph rotates its nose up to a very steep angle of attack of 18 degrees.  During this time the full weight of the aircraft is supported by the main under carriage up until the aircraft achieves enough lift to become airborne.  By comparison a 747 rotates to an angle of about 10 degrees nose up attitude on takeoff and its wings begin to generate a significant amount of lift, thereby transferring a much greater percentage of the aircraft's weight from the undercarriage to the wings well before the aircraft actually leaves the ground.

As a result of the stress imparted on Concorde's main undercarriage components, these parts reach their useful flight cycle limit relatively quickly and must be replaced.  On July 18, 2000 that is exactly what had happened to the Concorde's left main gear undercarriage beam.  The beam is the horizontal tube through which the two wheel axles pass at each end.  In the middle is a low-friction pivot which connects the beam to the vertical leg extending down from the underside of the wing.  The pieces of the pivot that bear the greatest extent of the load are two steel shear bushes.  A spacer made of anodized aluminum about five inches in diameter and twelve inches long sits between the shear bushes to keep them in proper alignment.  When Concorde F-BTSC left the hangar after the completion of repairs on July 21, it did so without the spacer on its front left main gear axle, as the Air France maintenance technicians had failed to follow their own company procedures for replacing the part and the final inspection of the job was either not conducted or the inspector missed the mistake.  

The missing wheel spacer was not an immediate issue and would have gone unnoticed by the crew at first as the shear bushes would have remained in place as designed while Concorde was on the ground with the load bearing shear bushes being positioned horizontally on either side of the beam.  But when the gear was retracted into the wheel well the position of the bushes changed to a vertical alignment with the right hand bush lying on top of the left where normally a spacer in between would keep the two parts separated.  Between leaving the maintenance hangar and the fateful flight of July 25 the Concorde flew two round trip flights to New York and back.  

In that time after each gear retraction the right hand shear bush began to slide down into the gap left by the missing spacer.  By the day of the accident the shear bush had moved a full seven inches out of alignment to the point where the two washers were almost touching.  The movement of the shear bush allowed the beam and wheels to wobble up to three degrees in any direction.  This meant that as the AF4590 taxied out to the runway its front wheels on the left main undercarriage were out of alignment which could potentially have caused the aircraft to pull to the left much like the steering in your car being out of alignment.  Worse still like a supermarket shopping cart there was noting to keep the wheels from jamming and sending the aircraft hurtling off the side of the runway.

Photo from BEA Accident Report

That is exactly the conclusion that long time Air France Concorde pilot Jean-Marie Chauve and flight engineer Michel Suaud reached in their independent investigation.  The two pilots believe that the left front wheels were already dangerously out of alignment before AF4590 began its takeoff roll and that the wobbling of the wheels and potential skidding retarded the aircraft's acceleration during its takeoff roll and caused the aircraft to veer to the left side of the runway.  As evidence to this fact they point to photos taken on the runway shortly after the accident, as seen above.  In the photo at least three independent tire tread marks can be seen on the runway up to the impact with the taxiway light on the edge of the runway.  This would seem to suggest that the front and back wheels were not in alignment.  If the wheels were in alignment then only two wheel tracks would be visible.  

French investigators conclude that the impact with the metal strip was what caused the aircraft to pull to the left and that the resulting tire blow out and damage not the missing spacer caused the wheels to go out of alignment.  We will probably never know what the exact sequence of events was but it seems highly probable that the missing spacer played a larger role in the accident than French investigators are giving it credit for.

Concorde's tires are not designed and manufactured like ordinary airline tires.  The tires are specially designed to support the weight of the aircraft as it accelerates at a very high rate of speed during a typical takeoff run.  The tires are under a high degree of strain and as a result the tires don't last very long, requiring replacement after 35 cycles on average, as opposed to a typical jet airliner which can usually go anywhere between 90-300 flight cycles before requiring a tire replacement.

Photo from concordesst.com

Tire Blowout Issue

It is easy to understand given the stresses how tire blowouts would be more common on Concorde than other airliners.  But given the unusual operating regime of the airplane and the stress placed on the tires one would expect that the designers would have considered the foreign object debris (FOD) created by a tire blowout in the design of the aircraft wing.  

Given the close relationship of the landing gear to the engines and the complex arrangement of fuel tanks Concorde would seem vulnerable to damage from a disintegrating tire.  In fact Concorde did have a problem with tire blowouts throughout its service life.  The aircraft between its introduction to service in 1976 and its retirement in 2003 experienced at least 57 documented tire failures.  Six of those tire failures caused serious damage to wing fuel tanks, engines and or the airframe, which resulted in an uncontrolled release of fuel and hydraulic fluid in several cases as well as cutting electrical wires in the process.

Severe Tire Blowout Events

June 14, 1979

Concorde 101 Air France, F-BVFC
While taking off from Washington two tires on the left hand main gear blew. The gear could not be retracted, so the crew elected to return to Washington. Some circuitry was damaged after having been hit by debris from the tires. Debris also caused a fuel and hydraulic leak.

July 21, 1979
Concorde 101 Air France, F-BVFD
On takeoff from Washington-Dulles, a tire blew. At FL270 a compressor stall was experienced probably due to foreign object damage.

September 16, 1980
Concorde 102 British Airways, G-BOAF
A tire blew on takeoff from Washington-Dulles Airport. Upon landing, pieces of tire damaged the engine and airframe. 

February 19, 1981
Concorde 101 Air France, F-BTSD
During takeoff from Washington-Dulles Airport a tire on the left hand main gear blew. The flight diverted to New York-JFK with one engine shut down due to vibration.

July 15, 1993
Concorde 102 British Airways, G-BOAF
During landing roll at London-Heathrow, the right hand main gear tire burst due to brake seizure. Debris caused damage to the wing and hydraulic problems. The no.3 engine was damaged as well, becoming stuck in the reverse position.

October 25, 1993

Concorde 102 British Airways, G-....

While taxing for takeoff (London-Heathrow - Washington) the aircraft suffered a brake lock. This caused a main gear tire to burst. Fragments of the water deflector caused some holes in the fuel tank.

Reading through the incident reports it is plainly obvious that not only had Concorde had dozens of previous tire blowouts before the Paris accident in July of 2000, but at least six of these accidents caused damage to the wing fuel tanks or engines.  The only difference being the tire failure and resulting puncture of the fuel tank did not trigger a fire.  

Dulles Incident

But of all the incidents the most serious by far was the June 14, 1979 incident at Washington Dulles Airport (KIAD).  The aircraft involved, registered F-BVFC, was operating Air France flight AF54 the daily Washington-Paris Charles de Gaulle (LFPG) flight.  As the aircraft accelerated towards takeoff speed the number five and six tires of the left main landing gear burst sending tire debris up into the underside of the aircraft's wing.  The tire fragments and metal shrapnel from the wheel components punctured three separate fuel tanks, severed numerous hydraulic lines and electrical wires.  Worse still several of the fragments passed all the way through the wing and exited the upper skin, ripping a large hole above the landing gear wheel well.  Miraculously the arcing wires and leaking fuel and hydraulic fluid did not cause a fire.  

The crew was not immediately aware of a problem until passenger, Bill Lightfoot, seated in row 23 of the airplane alerted the crew to the large hole in the wing.  Lightfoot had flown previously on Concorde and was familiar with the unique sights and sounds of Concorde.  But he said he knew something was wrong shortly after the aircraft began its takeoff roll on runway 19L.  He stated that shortly into the takeoff roll he felt the aircraft shudder and then he saw something fly vertically past his window.  He leaned over so that he could get a better view out of the very small Concorde windows and could not believe what he saw.  There was a large hole about the size of a coffee table ripped through the top of the wing and fluid was streaming out of the hole.  Up to that point the aircraft had continued accelerating on climb out, and as the aircraft flew faster, more of the top wing skin peeled off.  Once the flight crew saw the extent of the damage they immediately turned the aircraft around and landed back at Dulles.

Photo from NTSB Accident Report of  AF54 Showing the large hole in the top of the wing caused by the tire blowout

The incident was investigated by the U.S. National Transportation Safety Board (NTSB) and the agency was in the process of developing its final report and issuing safety recommendations when a similar incident occurred to another Air France Concorde again at Washington Dulles.  This time parts of the exploding tire were ingested by one of the engines causing a compressor stall.  The similarities between the two accidents resulted in immediate voluntary corrective action by the appropriate authorities. An Airworthiness Directive, issued by the Director Generale de 1’Aviation Civile, and a Technical Information Update, issued by Air France, revised procedures.  

To deal with the tire blowout problem a series of corrective measures were implemented in the early 80's and the rate of tire failures declined.  But the exact number of tire blowout incidents after 1981 are not known by the NTSB as it has been unable to locate incident reports on Concorde between 1988-1996.  British Airways says it recorded 12 tire events during that time period.

So although it appears the tire issue was corrected to some degree through design changes to the tires themselves and operational procedures to deal with a failure, the core issue of how to mitigate damage to the wing and engines from tire debris after a blowout was never addressed.  It is hard to fathom that the near accident at Dulles did not raise alarm bells at Aerospitale-BAC or with the FAA or its French counterparts.  In many ways the damage sustained by Concorde F-BVFC in the June, 1979 incident at Dulles was probably as severe if not worse than that sustained by sister ship F-BTSC on that fateful flight in July of 2000.

Illustration from concordesst.com

The Dulles incident clearly exposed Concorde's flawed landing gear/tire design and the vulnerability of the wing and fuel tanks to withstanding impact from tire debris during a blowout.  But yet nothing was done to fix the issue.  One has to wonder why was it that Aerospitale-BAC didn't see this issue as a serious threat to the aircraft and redesign it to armor critical components like the fuel tanks against FOD damage.  In this context the metal wear strip from the ContinentalDC-10-30 seems to have served only as the catalyst for the tire failure, but it along with Continental can't be blamed for a design flaw that had existed in the aircraft since 1976.

Pre-Flight

July 25, 2000 began like any other day for the staff of Air France responsible for preparing the Concorde for its journey across the Atlantic to New York.  Preparations for the flight began at 9:12 in the morning with the dispatcher logging onto to the computer system to discover that the aircraft originally scheduled to operate AF4590, Concorde F-BVFC, had been reallocated to operate the company's daily flagship AF002 Paris-New York flight, after F-BVFA had been removed from service overnight due to a maintenance issue.  As a result the reserve aircraft, F-BTSC was pressed into service to operate the afternoon charter flight AF4590.

The dispatcher was also made aware of a technical issue with the new aircraft, F-BTSC, when his computer indicated that the thrust reverser on the number two engine was inoperative due to an issue with the secondary nozzle.  The aircraft could be safely operated in this condition but the issue would reduce the aircraft's maximum operating weight by 2.5%.

Concorde's Secondary Thrust Reverser Nozzle in the Fully Closed Position

Based on this information and an anticipated headwind of twelve knots, an atmospheric pressure of 1008 hPa, a higher than normal temperature and the usable runway length of 13,828 feet the dispatcher calculated the aircraft's maximum takeoff weight for this flight to be 177,930 kg (392,268 lbs).  However the anticipated weight with 100 passengers checked in was calculated to be 184,400 kg (406,532 lbs).

At approximately 9:30 the dispatcher informed the duty officer of the weight issue with the aircraft.  The duty officer considered three options to resolve the issue.  First, substitute another aircraft for the flight, secondly to try to resolve the problem with the thrust reverser, third to load the passenger luggage onto another Air France flight to New York.  For his part the dispatcher studied flight plan alternatives which included making an en-route technical stop for fuel or changing the aircraft's load and choosing an alternative routing that would hopefully make the flight feasible.

Following this conversation the dispatcher received a call from the flight crew just before 10:00 am during which he informed them of the weight problem and his attempts to resolve it.  The flight crew said they would submit an order to have the pneumatic motor on reverser number two replaced and asked the dispatcher to file a direct ATC flight plan.  The crew also informed the dispatcher that they were going to take over the flight preparations themselves.

Because the flight was delayed to fix the technical fault with the thrust reverser the aircraft manager did not commence provisioning and loading of the flight until 11:13.  By 11:34 the one hundred passengers and their seventy nine bags had been checked in.  The checked baggage represented a total weight of 1,651 kg (3,639 lbs), but since the loading had not been complete the aircraft manager estimated the final weight of luggage to be 1,700 kg (3,747 lbs).  

In reality the actual weight of the baggage was closer to 2,525 kg (5,566 lbs) a difference of 825 kg (1,819 lbs).  The reason for the discrepancy can be traced to the fact that there was a disagreement of 19 bags between Air France's baggage reconciliation system (GAETAN) and the Baggage Reconciliation System (BRS) which is a security system which ensures that only bags that can be matched to a passenger on the aircraft are loaded.  In total the two systems showed a disagreement of 29 checked bags, but only 19 were eventually loaded onto the aircraft.

The process for clearing the bags at Air France was as follows:

1. The baggage is checked in by the passenger, agents print the bag tags and the tags are identified in the Air France GAETAN system.

2. The GAETAN system sends the passenger baggage information to the BRS system which cross checks the information for security purposes.

3. The information is then stored in the BRS and the GAETAN system is updated in real time.

4. The bags are then scanned by the aircraft manager during loading.  The BRS system then authorizes it to be loaded on the aircraft.

In the case that a bag tag number is not recognized by the BRS system the aircraft manager's display returns a message of "tag unknown".  In the case of AF4590, the BRS system could not reconcile 29 bags.  The bags were not in the BRS system and were unknown to GAETAN because the bags had only been registered in the Air France system for the connecting flight from Dusseldorf to Paris, but not onward to New York.  This oversight required the Air France agents in Paris to manually enter the bag weight and tag number into the system prior to loading onto Concorde.  It appears that this reconciliation of bags by Air France personnel was not done in a systematic manner explaining why the 29 pieces of luggage did not show up in the BRS system.  

Nineteen of the remaining twenty nine pieces of luggage were loaded after the aircraft manager confirmed the passengers were in fact on board the aircraft.  However because the 19 bags had not been entered into the system they were not accounted for by GAETAN nor the computerized load sheet used by the aircraft manager to calculate the final weight of the baggage loaded onto the aircraft.  More importantly the luggage having been put on the aircraft last after all the other bags hand been loaded were placed in the larger rear bulk luggage compartment which affected the aircraft's center of gravity (CG).

Cutaway Illustration showing the rear baggage compartment in the tail of the aircraft - from concordesst.com

Cargo Being Loaded into the Rear Baggage Compartment of Concorde

During this time Concorde took on a total of 94,853 kg (209,115 lbs) of fuel for the projected 3 hour 30 minute flight to New York.  This total included 1 ton of fuel which would presumably be burned off during taxi to the runway.  The individual 13 tanks on the aircraft were considered to be full by the fuel system when they reached 95% of their capacity, or 94% in the case of fuel tank number five, which corresponded to the upper limit of the fuel tank sensors.  However the tanks could be overfilled by a maximum of 1,631 liters between the 13 tanks total.  In the case of AF4590 an overfill of 300 liters of fuel corresponding to roughly 237 kg (522 lbs) was loaded onto the aircraft.  The overfill amount was observed to have been pumped into tanks 1, 2, 3 and 4.  The fuel loader's paper work showed a total fuel weight of 94,800 kg.  The decision to overfill the tanks while a common practice would play a significant role in propagation of the disaster chain after the aircraft impacted the titanium wear strip during its takeoff run.

Concorde Fuel Tank Schematic - from BEA Accident Report

The combination of overfilling the number 1, 2, 3 and 4 fuel tanks and the extra unaccounted weight in baggage meant that as AF4590 prepared to depart the gate at Charles De Gaulle Airport it was at least 2 tons over the maximum takeoff weight.  While aircraft can safely takeoff above their maximum takeoff weight the practice is generally avoided.  Most takeoffs that occur with the aircraft in a configuration beyond its maximum takeoff weight occur because passenger weights and baggage are not exact figures, bur rather are calculated from averages.  

There are industry accepted average weights for passengers and luggage that are used to establish the taxi and takeoff weight for a specific flight.  Since the figures are not exact there are occasions when the aircraft may depart 50 or 100 pounds over weight but taking off more than a ton over an airplanes certified maximum takeoff limit is very unusual and should generally be avoided as this is well beyond the weights considered by the manufacturer during the aircraft's certification program.

Taxi Sequence

As AF4590 pulled away from the gate on July 25, 2000 and taxied slowly towards runway 26R it weighed 186,880 kg (412,000 lbs) as reported by the flight engineer.  The actually weight of the aircraft at taxi, due to the discrepancy in the load sheet caused by the unreconciled bags in GAETAN and the extra fuel loaded onto the airplane was slightly different by about two tons.

The exact difference in weight of Concorde F-BTSC between what the flight engineer reported and the actually taxi weight will never be known as the baggage and passenger weights as stated before are calculated based on averages.  However the BEA in its investigation was able to establish a range between which the actual taxi and takeoff weights most likely fell.  The difference between the BEA's high and low calculated taxi and takeoff weights were about 494 kg (1,089 lbs).  It is important to note that the maximum structural takeoff weight of Concorde is 185,070 kg. (408,009 lbs)  The chart below explains the weight calculations in detail as established by the BEA.

(1)  There were 122 items of baggage loaded, average estimated weight of 20.7 kg each, making a total of 2,525 kg.
       19 items of baggage loaded on board were not taken into account, only 103 items appearing on the load sheet.
(2)  Allowing that the aircraft consumed a ton of fuel during taxiing.
(3)  By applying the fixed average for passengers: one passenger = 84 kg, one child = 35 kg.
(4)  By applying the fixed average for men and women: one man = 88 kg, one woman = 70 kg, one child = 35 kg.
(5)  The EIC corresponds to 60 kg of newspapers.

This means that regardless of which of the two weight calculation scenarios are applied F-BTSC as it stood at the end of runway 26R was overweight by between 1,181 kg (2,603 lbs) and 687 kg (1,514 lbs), or about 1 ton.  The fact that AF4590, was overweight by 1 ton was not considered to be critical when investigators calculated the aircraft's takeoff acceleration rate and projected profile, and I would agree with this assessment.

But between the time the aircraft pushed back from the gate and taxied to the end of runway 26R something significant happened which drastically changed the weight calculations.  The wind which was reported as calm at the time of push back had now increased to an 8 knot tailwind.  Pilots don't generally choose to takeoff with a tailwind as the presence of one increases the amount of runway required for an aircraft to get airborne. The increase in takeoff runway length is proportional to the speed of the tailwind.

On the surface 8 knots may not seem like much to be worried about, and in most conventional airliners it wouldn't be.  But Concorde due to its high takeoff speed of 250 mph and the tremendous stress placed on its tires and landing gear, meant that even a small tailwind could have a severe impact on the aircraft's takeoff performance especially at maximum takeoff weight.  The major problem with increasing the takeoff length is that it also increases the takeoff speed of the aircraft.  The stronger the tailwind the more runway the aircraft will consume and the faster it must travel down the runway.  At some point the tailwind component will result in the Concorde exceeding the available runway length or its tires will exceed their maximum speed limit and potential fail before the aircraft obtains enough speed to takeoff.  Besides increasing the takeoff length and speed required for rotation a tailwind also reduces the Regulated Takeoff Weight (RTOW) of the aircraft.

This is a very important point, as F-BTSC was already 1,000 kg (2,204 lbs) overweight before factoring in the 8 knot tailwind.  The RTOW is unique to each Concorde in the Air France fleet and is set according to detailed tables for the specific aircraft, given the temperature, and wind conditions at the time.  When you factor in a tailwind of 8 knots the aircraft was actually almost 6,000 kg (13,227 lbs) over its regulated takeoff weight on that day of 180,300 kg (396,832 lbs).  Just as important to note is the fact that the aircraft was already at its rear Center of Gravity (CG) limit of 54% as announced by the flight engineer prior to takeoff.

Every Concorde has a CG limit gauge which is positioned on the flight engineers fuel management panel.  It indicates the aircraft's current center of gravity as well as the forward and aft limits.  The center of gravity of the aircraft changes during the flight as the aircraft expands due to the heat of supersonic flight and as fuel is burned off from the wing tanks.  Concorde's center of gravity and trim is adjusted by pumping fuel between the 13 individual tanks.  Generally after takeoff fuel is pumped towards the rear of the aircraft to move the center of gravity back to prepare it for supersonic flight as well as throughout the supersonic flight regime.  Upon initiation of descent the fuel is pumped into the forward tanks to shift the aircraft's CG forward to aid the descent.  Upon taxi out F-BTSC's CG was actually at 54.2%, which was beyond the rear limit.  To correct this the flight engineer proceeded to transfer fuel from tank 11 into the forward feeder tanks 1, 2, 3 and 4 until the CG shifted inside the rear limit of 54% and the red "out of limits" warning light on the CG gauge extinguished.

CG Gauge on Concorde's Flight Engineer Panel, Red Light Indicates the CG is Currently out of Limits

Takeoff and Accident Sequence

14:34:38 - ATC cleared the aircraft to taxi to runway 26 right via the Romeo taxiway.  When the “post engine startup” checklist was complete, the crew began taxiing and started the “taxi” checklist.

14:37:10 - A short time afterwards, the checklist was interrupted by the PFC alarm.  The FO stated that the rudder control had already switched from Blue electrical mode to Green electrical mode on two occasions, and he proposed leaving it in the latter mode.  Blue electrical mode was nevertheless re-selected.

14:38:53 - The PFC alarm appeared again and the FE indicated that they should expect a switch to Green electrical mode during takeoff.  He proposed that in that case they would continue the takeoff, knowing that it was possible to re-arm the Blue electrical mode.  The “taxi” checklist was continued.

14:38:14 - The FE announced that fuel transfer was under way, which meant that the CG changed from 54.2 to 54%.  This transfer was made from tank 11 directly to feeder tanks 1, 2, 3 and 4.  When the checklist was again interrupted by the PFC alarm, the crew decided to leave with the rudder in Green electrical mode, which is in accordance with the minimum equipment list.

14:40:01 - The Concorde was cleared to line up as the crew were finishing the “taxi” check list.  At the request of the Captain, the FE indicated that eight hundred kilograms of fuel had been consumed, which in fact corresponds to the expected consumption by the engines since startup.  Based on the final load sheet handed over by the aircraft manager and knowing that the aircraft took off two minutes later, which
corresponds to an additional estimated consumption of two hundred kilograms, it can be deduced that, for the crew, the aircraft weight at which the takeoff was commenced was 185,880 kg, for a MTOW of 185,070 kg. 

14:40:37 - The “pre-takeoff” check list started and finished about forty seconds later.

14:41:55 - The FE announced that the CG was 54%. The transfer of fuel was complete.

14:42:17 - The Concorde was cleared to line up and take off.  The controller announced a wind of 090°/8 kt. This announcement did not result in any comment on the part of the crew, even though, with those wind conditions, the takeoff weight should be reduced to 180,300 kg because of the “tire” speed limit.

14:42:30.4 - The characteristic clicking of the thrust levers in maximum thrust position is heard.  The Captain gave the takeoff “top” one second later.  The aircraft’s center of gravity was around ninety meters from the threshold of the runway.

14:42:54.6 - In accordance with procedures, the FO announced 100 kt.  The recorded airspeed (CAS) was in fact at 100 kt and, as the recorded Nz variation shows, the aircraft had just passed over the asphalt/concrete join on the runway located six hundred meters from the runway threshold. Its track was centered.

14:42:57 - The FO announced four greens.  This announcement refers to the “GO LIGHTS” and confirms correct engine function, including reheat.  The CAS is recorded as 108 kt.

14:43:03.7 - The V1 callout was made.  The acceleration and the distance run were then entirely in accordance with the simulation calculated for the MTOW, and the value of longitudinal acceleration shows full thrust on all four engines, which is confirmed by the parameters on engines 1 and 2 recorded at 14 h 43 min 08 s and 14 h 43 min 09 s.

14:43:09.5 - (FDR time 97600), a slight variation in Ny, uncommanded by the rudder, is noticeable.  The aircraft was then about 1,700 meters from the threshold, in the area where the first parts of the water deflector were found.  It was probably at that moment that tire No 2 ran over the metallic strip.  In the following half-second, a clean, short noise is recorded on the CVR.  The CAS was 175 kt, the distance from the threshold about 1,720 meters.  It is likely that this noise resulted from the damage to the tire.  It was
in fact in this area that the metallic strip and the large piece of tire were found.

14:43:11 - A very clear change in the background noise is heard, the CAS being 178 kt and the distance run 1,810 meters.  The first marks from tire No 2 were noticeable on the runway.  The piece of the lower part of tank 5 then the kerosene stain were found at 1,820 meters.  At 1,850 meters, the first marks of very dense soot were noted.  These observations allow the conclusion to be drawn that a large quantity of fuel leaked out
before the fire broke out and stabilized.  With detailed analysis of the sequence, it appears that the change in the background noise resulted from the ignition and the stabilization of the flame.

Photo of Runway 26R at Charles de Gaulle, Taken Shorty after AF4590 Departed - from BEA Accident Report

14:43:13.4 - This is consistent with the controller’s comment which indicated extensive flames at the rear of the aircraft.  A few tenths of a second after the change in the background noise, the heading began to diminish at a rate of 1°/s, without there being any observable significant variation in longitudinal acceleration, which
confirms that the aircraft had not yet suffered any significant loss of thrust.  This heading change was probably the result of a combination of the tire burst and the aerodynamic disturbance due to the fuel leak and the fire.

14:43:11.9 - Something unintelligible is heard whose origin it has been impossible to identify.  The CAS was then 182 kt and the distance from the threshold was 1,885 meters.  It was at that moment that the Captain began to deflect the rudder to the right, a slight deflection (8° at first followed by stabilization at an average value of 5°), in reaction to the aircraft’s slight movement to the left.  The last nominal Nx value, at 0.268 g, was recorded at FDR time 97602.5.

14:43:12 - 14:43:13, engines 1 and 2 suffered their first loss of thrust.  This loss of thrust is confirmed by the Nx recording at its minimal value of 0.133 at 97603.5, while the FO said “watch out”.  The “GO LIGHTS” for engines 1 and 2 went out.  The absence of any significant damage leads to the explanation that the high loss of thrust on engine 2 was due to ingestion of hot gases whilst the loss of thrust on engine 1 can be explained either by ingestion of debris due to the damage to the tire or by ingestion of hot gases.

14:43:12.2 - The Captain began to pull back on the control column in a moderate way while the CAS was 183 kt and the distance from the threshold was 1,915 meters.  It was in this area that many people noticed an intense luminous phenomenon accompanied by a strong surge noise.

14:43:13.4 - The side slip to the left noted at this time at a rate of 2°/s, resulted directly from yaw movement caused by the high loss of thrust from engines 1 and 2.  The recorded thrust was then no more than 50% and was mainly delivered by engines 3 and 4.  There was no fire alarm in the cockpit at that time.  The lift-off of the nose gear, which occurred a few tenths of a second later, when the CAS was 187 kt and the distance from the threshold was 2,045 meters, is entirely consistent with the elevon deflection.  This could be the result of the crew taking into account an abnormal unidentified situation.  It should be noted that the rate (1°/s) was lower than normal, which suggests that the crew were conscious of the lack of speed.

14:43:15.7 - At the moment when the side slip to the left occurred, a further rudder deflection is recorded.  It reached 20° to the right, when the side slip reached its maximum of 5° (heading = 264°), then it decreased towards 10° and stabilized.  The simulations described in paragraph 1.16.13 explained this phenomenon as well as why the aircraft continued to deviate from its track.  Around the same time engine 1, in a phase of re-acceleration, was producing around 80% of its nominal thrust and an exclamation by the FE can be heard. The CAS was 196 kt.  Thus, during the three seconds when all the events which led to the catastrophe occurred, the crew perceived through a variety of senses a whole group of anomalies: (very) unusual noises, inertial sensations resulting from the violent kick in lateral acceleration associated with the loss of thrust and the sudden loss of longitudinal acceleration and perhaps smells and the luminous flash generated by the ignition and the leak.

14:43:16.1 - 14:43:18.1, the engine 1 “GO LIGHT” came back on.  This meant that the fuel flow in the engine P7 were, respectively, above 20.5 t/h and 39.1 psi and that it was approaching its nominal thrust.  On the other hand, the engine 2 parameters recorded after its loss of thrust show that it was producing thrust hardly any higher than idle, around 3% of its nominal thrust.

14:43:20.4 - The FE announced the failure of engine 2, in accordance with the appropriate procedures, the speed was 203 kt, the distance was 2,745 meters and the pitch attitude was + 9°.  In the following second, readout of the parameters shows that engine 2 re-accelerated slightly and delivered thrust of around 15% of nominal thrust.  The “GO LIGHTS” on engine 1, then on engines 3 and 4 went out, as a normal reaction to the relaxation of the shock absorber on the left main landing gear.

14:43:20.9 - 14:43:21.9, engine 1 suffered a second surge, caused by the ingestion of hot gases and/or kerosene, aided by the change in the aircraft’s angle of attack.  It was producing thrust that was scarcely above the idle level.  As for engine 2, which was re-accelerating, its auxiliary air intake began to re-open, which caused further intake of hot gases and a further surge.  The aircraft was again powered mainly by the thrust from engines 3 and 4. Around the same time, an edge light on the left of the runway was broken by the passage of wheel No 6.  The track deviation continued, the aircraft then being about 22.5 metres from the runway center line.  No components from this light were identified in the debris found during disassembly of the engine.

14:43:21.3 - Movement of a selector is heard, identified as being the switching of a TCU, probably that of engine 2, from MAIN to ALTERNATE.  This procedure carried out by the FE was intended to regain normal function by switching the computers.

14:43:21.9 - Aircraft takeoff was effective.  The speed was 205 kt, the distance from the threshold was 2,900 meters and the pitch attitude was + 10°.  In the following second, the fire alarm was heard, followed by a gong, and the Engine Warning parameter was recorded.  On the radio, “(?) it’s really burning, eh” is heard, probably coming from a crew in a waiting aircraft, and a few seconds later “(?) it’s burning and I’m not sure it’s coming from the engine”.  The first sample of the parameters on engine 1 after the second surge shows that it was only producing thrust slightly above that corresponding to idle.

14:43:24.7 - As for the parameters on engine 2 recorded they confirm its engine surge and also show that it was at idle.

14:43:24.8 - The FE said, “shut down engine 2”.  In the same second, the Captain called for the engine fire procedure.  Less than two seconds later, a noise is heard, which spectral analysis and examination of the HP selectors has shown to be the movement of the thrust lever to the stop position.  Pulling of the engine 2 fire handle, found in the pulled position in the wreckage, occurred in the following seconds.

14:43:27 - The FO drew attention to the airspeed.  The speed was then 200 kt for a V2 of 220 kt (Vzrc on three engines with the gear extended is 205 kt).  In the following second, a selector sound is heard, identified as being the fall of the electrical pitch trim compensator switches.  This is explained by the fact that, since the aircraft had a high angle of attack, the pitch trim compensator was beyond its normal operating range to
counter this angle.  A gong, identified as the alarm caused by the fall of the switches, is also heard.  Subsequently, there was no further movement of the pitch trim compensators.  The engine 2 N2 went below 58%, leading to automatic switching to CONTINGENCY mode for engines 1, 3 and 4.  Engine 1, in a recovery phase after the second surge, only achieved CONTINGENCY rating seven seconds later.  The thrust it was then producing was 5% less than nominal thrust with reheat in CONTINGENCY mode.  This thrust deficit can be explained by damage resulting from the initial ingestion of solid fragments, since ingestion of hot gases or of kerosene would not have led to the later stability of the engine parameters at a reduced level.

14:43:30 - The Captain requested gear retraction.  The speed was still 200 kt, the radio altimeter indicated 100 feet and the calculated rate of climb was 750 ft/min.  In the following seconds, the controller confirmed that there were extensive flames behind the aircraft. Engine 1 was then producing 75% of its nominal thrust and the reheat had just cut in.  The FE repeated “the gear” for the FO, who was acknowledging receipt of the
transmission from the controller.  The aural alarm indicating detection of smoke in the toilets was recorded by the CAM.  This alarm can be explained by the fact that the burnt mixture ingested by one of the left engines was used for the air conditioning and circulated to the cabin and the forward toilets, though the possibility of a false alarm cannot be excluded.  The fact that this alarm was recorded by the CAM also shows that the cockpit door was open during the takeoff, which is common practice on Concorde.

14:43:35.5 - The FE repeated “the gear”.  In the following second, a gong is heard which very probably corresponds to the alarm caused by low oil pressure due to the shutdown of engine 2.  The Engine Warning
parameter appeared again on the FDR.

14:43:37.7 - The FE repeated “the gear” and the FO answered “no”.  The red WHEEL light, situated above the landing gear retraction controls probably came on following detection of under-pressure resulting from the damage to tire No 2 and the procedure requires in this case that the gear not be retracted, except where the needs of safety require it.

14:43:39 - The Captain ordered “gear retraction” while the FO acknowledged receipt of a message from the control tower.  Three seconds later, the engine 2 fire alarm was reactivated with its associated gong.  It stopped a few seconds after the FE fired the fire extinguishers (the two extinguishers located in the left wing were found fired in the wreckage).

14:43:45.6 - The FO probably answered “I’m trying” to the order given by the Captain, which can be interpreted as an attempt to retract the landing gear.  At the same time the FE said, “I’m firing”. The System parameter overseeing the integrity of the under-pressure system activated, which indicates that the system was functioning up to that moment.  In the following second, the Captain asked “(are) you shutting down
engine two there” and the FE replied “I’ve shut it down”.

14:43:49.9 - The FO repeated “the airspeed”.  This warning, repeated again about ten seconds later, is explained by the fact that the speed remained at about 200 kt, lower than the normal climb out speed of 220 kt with a failed engine.

14:43:49.5 - 14:43:54.5, the first differences between the aircraft’s attitude and the attitude which should result from inputs on the flight controls can be noted (small roll/pitch and pitch/roll interactions).  These
differences seem to be explained by the consequences of the fire on the left wing, in particular on the inner elevon. The angle of attack was then 13°.

14:43:56.7 - When the CAS was 211 kt, the FO noticed and reported that “the gear isn’t retracting”.  This statement would confirm the interpretation of “I’m trying”.  Breakdown analysis showed that the non-retraction of the gear was due to the non-opening or non-detection of complete opening of the left main landing gear door.  The flame had then been established for thirty-five seconds.  A fluctuation of Nx is observable which might result from a large and brief surge on engine 1, not visible because the parameters were not registered at that moment.

14:43:58.6 s - The engine 2 fire alarm sounded again.  It continued to sound until the end of the flight.  In the following second the GPWS “Whoop Whoop Pull Up” warning was heard on three occasions, with the following parameters:  nose up at 5°, radar altimeter at 165 feet, rate of descent of about 160 ft/min.

14:43:59.5 - 14:44:11.5, a first disturbance on the engine 1 FF and EGT parameters is noted.  A second disturbance was recorded eight seconds later, the CAS being 207 kt.

14:44:01 - The rudder switched to mechanical mode, which led to the loss of yaw auto-stabilization.

14:44:11.5 - The engine 1 parameters show a clear deceleration, due to a severe surge. Only engines 3 and 4 remained in operation.

The angle of attack changed in twelve seconds from 12° to over 25°, the bank to the left went from 2° to 113° (figure recorded four seconds before the end of the recording) and the magnetic heading decreased from 270° to 115°.  Spectral analysis showed that the selector noises which were then heard could be attributed to the movement of the thrust levers to idle stop position.  This reduction in thrust on engines 3 and 4 was probably intended to decrease the strong bank to the left caused by the significant thrust asymmetry and by the destruction of vital control surfaces by fire.  

The decrease in thrust on these two engines was accentuated by a surge due to airflow distortion caused by the angle of attack and the level of yaw reached at that moment.  In these extreme conditions, the combination of lateral and thrust asymmetry and the major thrust/drag imbalance, which could not be compensated for by a descent, led to a loss of control.  This loss of control was probably accelerated by the structural damage caused by the fire.  In any event, even if all four engines had been operating, the serious damage caused by the intensity of the fire to the structure of the wing and to some of the flight controls would have led to the rapid loss of the aircraft.

Accident Chain

The origins of the AF4590 accident actually go back several decades to that serious tire blow out incident at Washington Dulles Airport in 1979, detailed in Part 2 of this piece.  The investigation into the accident conducted by the FAA uncovered a serious design deficiency in the aircraft's tires which were prone to blowouts.  The incident also uncovered a critical vulnerability in the Concorde's wing design, in that it lacked the ability to adequately withstand impact from FOD caused by the disintegration of one or more aircraft tires at high speed.  Investigators were particularly concerned with the lack of reinforcement and armoring around the aircraft's wing fuel tanks, and were somewhat dumbfounded by the fact that a piece of rubber or even the composite water deflector could penetrate and pass cleanly through both the lower and upper surface of the wing and impart such critical damage.  It was only by sheer luck that the leaking fuel caused by the puncture to to the tank did not ignite a fire in the Dulles accident.  

Yet even with these known deficiencies and the recommendations issued by the NTSB, Aérospatiale-BAC took no significant action to better armor the wings to protect the fuel tanks from impact damage by FOD.  The history of tire blowouts was considered and steps were taken which helped reduce the rate of tire blowouts on future Concorde flights.  The Dulles incident was almost identical to the fateful AF4590 flight, with the key exception being the leaking fuel didn't ignite and the pattern of damage was more confined than the Paris accident.  

None the less the Dulles incident should have served as a wake-up call to Air France, Aérospatiale-BAC and regulators but the lessons of that accident were not applied to ensure the safety of future Concorde passengers.  Had they been it is likely the AF4590 accident would not have occurred.  The Dulles accident was without a doubt the first key link in the accident chain.

Photo from NTSB Accident Report of  AF54 Showing the large hole in the top of the wing caused by the tire blowout

The next critical link occurred several days before the accident when Concorde F-BTSC entered the Air France maintenance hanger at Charles de Gaulle Airport to have the left main undercarriage beam replaced.  This critical maintenance item was not performed according to established Air France procedure, as documented by manufactures Aérospatiale-BAC in the Concorde maintenance manual.  Not only was proper procedure not followed but the senior mechanic either did not inspect the maintenance work performed or his examination of the work was insufficient in that the missing wheel spacer went undetected.  

Either way the Concorde was released from the maintenance shop and placed back into line service with a progressive deficiency in its landing gear system.  One which with each takeoff and landing cycle placed the tire and entire landing gear under increasing levels of stress as the tire wobbled at high speed and moved further and further out of alignment.  While the exact contribution of the missing wheel spacer will never be know, the premature wear it likely caused on the tire should be considered as it probably weakened the tire's structural integrity which made it more vulnerable to a blowout.  For its part the BEA says its investigation concluded the missing wheel spacer had no effect on the aircraft, nor did it in anyway contribute to the accident.

Fast forward to the day of the accident, July 25, 2000.  As the aircraft sat on the tarmac being prepared for its flight to New York/JFK two critical events took place which directly influenced the accident chain.  First there was a discrepancy in the baggage manifest between what was registered in GAETAN and cleared in the BRS systems. The issue had been introduced by agents in Dusseldorf  who didn't properly register 19 pieces of luggage in the Air France system.  The bags were properly tagged through to New York but were registered in GAETAN only for the Dusseldorf-Paris segment of the flight, and not the connecting flight to New York/JFK.  As a result these extra bags were not accounted for in the computerized load sheet used by the flight supervisor and flight crew in establishing the aircraft's taxi and takeoff weights.  Perhaps more critically the 19 pieces of unaccounted luggage were placed in the rear bulk luggage hold which shifted the aircraft's center of gravity event further back.

Aircraft Weight Data from BEA Accident Report

While the baggage issue was being sorted out the aircraft was being fueled for its flight to New York.  During this time the crew made the decision to overfill the number 1, 2, 3 and 4 fuel tanks.  The process of overfilling the tanks was a according to the BEA report a manual process that could only be conducted by the fueling personnel while the aircraft was on the ground.  Page 27 of the BEA Report contains the following statement about the capacity of the Concorde's 13 fuel tanks.

"The capacity of the thirteen tanks is shown in the table below.  These represent maximum capacities, without exceeding the upper level sensors, corresponding to real fill of around 95% (94% for tank 5).  The overfill procedure allowed loading of a maximum of 1,630 liters (431 gallons) extra fuel, compared to the quantities mentioned below.  This operation can only be performed on the ground."

Concorde Fuel Capacity Table and Aircraft Fuel Tank Schematic - from BEA Accident Report

The fuel loader's filling order showed a total loaded fuel weight of 94,800 kg. (208,998 lbs) including the 300 liters (79 gallons) of overfill which were loaded into tanks 1, 2, 3 and 4 according to witnesses on the ground.  The flight supervisor based on fuel load, number of passenger and reported baggage weights from the load sheet calculated the aircraft CG to be 54.2% at Zero Fuel and 54.25% for taxi.  The rear CG limit as previously established for the Concorde was 54.0%.  In order to bring the CG inside the rear limit a fuel transfer of some 800 kg. was required from tank 11.  Later in the report on page 159, as part of the accident scenario reconstruction the following conversation is reported from the CVR.

At 14 h 38 min 14 s the flight engineer announced that the fuel transfer was under way to correct the out of limits indication on the CG meter.  During this time the fuel was reportedly transferred from tank 11 direct to feeder tanks 1, 2, 3 and 4.  Page 144 of the BEA report contains a detailed explanation of the fuel transfer procedure followed by the crew of F-BTSC.

"Before takeoff, the transfer procedure allows the center of gravity to be moved to 54% in case of completely full tanks. To do this the STAND BY INLET VALVES of feeder tanks 1 to 4 are positioned on OPEN and the electric pump selectors for tank 11 are positioned on ON. This allows topping up of the fuel consumed from the feeder tanks during start-up and taxiing with the fuel contained in tank 11. A center of gravity of 54% on takeoff is only authorized if all of the front tanks are full (R1 to 10 and 5A, 7A). This limits the fuel ballast to tank 11 only. The only transfer possible to adjust the center of gravity to 54% is thus a transfer from this tank towards the feeder tanks."

It is important to note that only 800 kg (1,764 lbs) of fuel, equivalent to the amount burned during taxi from the gate to the runway threshold, was able to be pumped forward from tank 11 into feeder tanks 1, 2, 3 and 4.  This means that as the Concorde began its takeoff roll its forward feeder tanks were completely full.  In fact the BEA report makes special mention of the fact that a takeoff with a center of gravity of 54% is only authorized if all the front tanks are full.  This statement from the Concorde Operations Manual clearly underscore's that operations at the rear CG limit should be avoided when ever possible.  The decision to overfill the fuel tanks, at the gate most certainly had an effect on the pattern and severity of the damage caused by the pressure wave from the pieces of tire impacting the underside of the wing during the takeoff roll.

The next key event in the accident chain occurred just minutes after the fuel transfer as Concorde F-BTSC was holding short of the runway awaiting takeoff clearance.

"At 14 h 40 min 02 s, the Loc Sud controller cleared 4590 to line up. At 14 h 42 min 17 s, he gave it takeoff clearance, and announced a wind from 090° at 8 kt. The crew read back the takeoff clearance. The FE stated that the aircraft had used eight hundred kilos of fuel during taxiing."

While the BEA in the final report references the 8 knot tail wind they do not seem to have accounted for this when modeling the aircraft's takeoff performance and theoretical runway length used to reach rotation and takeoff speed.  The final report contains the following notation in Section 1.6.6  Takeoff Performance.

pg 33 - "Since the wind readings at different recording sites show a light and variable wind, the calculations are made with calm wind conditions."

pg 34 - "For a tailwind of 8 kt, the takeoff weight is reduced to 183,300 kg. (404,107 lbs) due to a tire speed limitation." (This appears to be a typo in the report and should read 180,300 kg. or 397,493 lbs)

This statement is not clear, but seems to imply that under the conditions present a tailwind of 8 knots would reduces the regulated takeoff weight to 180,300 kg (404,107 lbs).  However according to the BEA's own calculations from the aircraft weight table in Section 1.6.5 Weight and Balance on page 31 the aircraft's takeoff weight was between 186,251 kg (410,613 lbs) and 185,757 kg (409,524 lbs).  So the BEA based on their own calculations concluded F-BTSC was significantly overweight given an 8 knot tail wind, but they concluded that the data from various wind sensors on the airfield averaged to calm winds so they did not model any takeoff scenarios which included the aircraft taking off almost 6,000 kg (13,228 lbs) over its RTOW with an 8 knot tailwind.  Investigators didn't model this scenario even though the controller given the live wind speed information available to him at the time noted the presence of a tail wind.  For the investigators not to consider, or even model the presence of an 8 knot tailwind given the advisory by the tower controller is something I find hard to understand.

Also of importance is the statement that the takeoff weight must be reduced to 180,300 kg (397,493 lbs) due to a tire speed limitation.  This implies that given the conditions any takeoff commenced by F-BTSC at a takeoff weight in excess of 180,300 kg (397,493 lbs) would result in the aircraft reaching its tire speed limit before rotation.  The decision by the investigators not to model the 8 knot tailwind in their takeoff simulations is not supported by the time line of the flight as printed under Section 1.1 History of the Flight on page 17 of the final accident report.

Additional statements are contained in Section 2.1.2 The Flight Until Engine Power-Up on page 159 of the report.

"At 14 h 42 min 17 s, the Concorde was cleared to line up and take off. The controller announced a wind of 090°/8 kt. This announcement did not result in any comment on the part of the crew, even though, with those wind conditions, the takeoff weight should be reduced to 180,300 kg (397,493 lbs) because of the “tire” speed limit. In reality, the wind was practically zero, as is shown by the Météo France readings and analysis of the track. However, even if the crew had previously noticed this absence of wind, for example by observing the indication given by the windsock near the threshold of runway 26L around a thousand metres away, it is difficult to understand the absence of any comment on their part."

The BEA clearly seems puzzled as to why the controllers statement about the wind conditions, resulted in no reaction from the crew.  The crew were well aware that they were taking off a maximum takeoff weight, with the CG at the rear limit of 54%.  The revelation of the tailwind should at a minimum have caused them to review their performance data again to ensure they were within legal limits to takeoff.  Had they done this and delayed the takeoff it is very likely they would have requested to change runways.  But probably due to time pressures, owing to the fact that they were already a full hour behind schedule the crew commenced the takeoff roll and didn't pause to think about the tailwind.

The impact with the titanium wear strip was without a doubt the next major link in the accident chain.  When the Concorde's left main landing gear ran over the wear strip it sliced through the tire causing it to blow out.  As was shown previously in Part 2, Concorde had a history of tire blow outs.  A handful of these blowouts had even caused serious damage to the fuel tanks, engines and other critical components.  But even the most serious of these incidents, the 1979 blowout at Dulles, in which the aircraft sustained serious damage to one of the wing fuel tanks, in much the same way as AF4590 didn't result in the loss of the aircraft.  The main difference between the two accidents is the severity of the damage.  So why did the impact from tire debris cause so much more destructive damage in the Paris accident?

The answer can be traced back to the fact that unlike the Dulles incident in the Paris accident AF4590's forward fuel tanks were completely full and highly pressurized.  This meant that when the tire blew out, liberating large pieces of rubber, a 4.5 kg chunk of which impacted the underside of the wing and tank 5, there was no free space in the tank to absorb the impact forces.  When the fuel was displaced by the impact shock wave the force of the impact was transferred from the liquid fuel to the tank structure, causing an explosive blowout of a portion of the bottom of tank 5.  The piece of wing skin and tank structure separated with explosive force from the aircraft leaving a sizeable hole in the tank from which the fuel was ejected under tremendous force.

Diagram Showing Tire Impact with Fuel Tank 5 and Resulting Shockwave Sequence - from BEA Accident Report

The fracture in the tank caused two serious problems.  First it sparked a fire which weakened the structural integrity of the wing structure as the fire propagated unchecked, secondly it caused a dangerous shift in the aircraft's CG, which was already operating at the absolute rear legal limit of 54%.  As the airplane struggled to stay in the air its CG would have moved further and further aft with each passing second as fuel continued to leak from the tank.  This change in the CG would ultimately contribute to the loss of control of the aircraft in its final seconds of flight before entering an aerodynamic stall and crashing to the ground.  

The BEA investigators established the following about the fuel level in tank 5.  The quantity loaded was 7.2 tons and the gauge indicated two tons after the accident. The flight time between the estimated rupture of the tank and impact was around eighty-one seconds.  The estimated fuel flow rate, apart from the leak due to the small puncture and (the) consumption by engines 1 and 2 (around 350 kg) was therefore around 60 kilograms per second.  In conclusion, the overall flow rate of the leak is several dozen kilograms per second, thus about ten times greater than in the Washington event. The high rate of flow from this leak contributed to the ignition of the fuel since it led to a fuel/oxidizer mixture, which was almost a stoechiometric mixture, thus perfectly flammable.  The smaller rate of fuel leakage in the Washington case is almost certainly what prevented the fuel from igniting.

Piece of Wing Underbody and Attached Fuel Tank Skin Blown Out by Shockwave from Tire Debris Impact

The last link in the accident chain was the crew's reaction to the emergency situation.  The puncture of tank 5, absent the fire would not have resulted in the loss off the aircraft.  Even the fire with all its destructive heat energy couldn't bring the aircraft down without some help.  The crew and their actions in response to the emergency also played a critical role in deciding the ultimate outcome of the flight.  When the leaking fuel ignited and the fire took shape and stabilized it produced a localized area of very hot air which was charged with gasses from the fire and unburned fuel.  Some of these gases were ingested by both the #1 and #2 engines causing them both to surge multiple times at various points during the brief flight.  The surges occurred both in response to changes in the aircraft's pitch, which affected the airflow into the engine and the quantity of hot gases ingested.  

The first surge in the #1 engine could also possibly be associated with FOD ingestion.  The crew's first sign that there was a problem with the #1 and 2 engines came at 14:43:12, when the, "GO LIGHTS", on both port engines went out.   The aircraft had just seconds before accelerated normally through V1 and was committed to the takeoff.  At this point both engines experienced a surge, mostly likely caused by the ingestion of hot gases from the raging fire or in the case of the #1 engine ingestion of FOD material from the disintegrating tire.  Engine #1 experienced a power loss of about 50% while engine #2 was producing power no greater than idle.  

The loss in thrust caused F-BTSC to yaw violently towards the left edge of the runway.  Sensing the aircraft was going to depart the paved surface due to the asymetric thrust condition the captain made the only sensible decision he could, to takeoff.  The captain begins the rotation at 183 knots, or about 15 knots before the scheduled rotation speed.  As the aircraft struggled into the air the left main gear struck a taxiway light on the runway's edge.  Shortly after rotation the flight engineer announced that engine two had failed, in reality the engine was still functioning but it had been producing only idle power for the previous 10 seconds.  Three seconds later, and after being informed of the presence of fire behind the aircraft by the tower controller, the engine fire warning light and associated bell triggered on the panel, causing the flight crew to shut the engine down in accordance with the engine fire/failure checklist.  

The decision to shut the engine down came at a critical time in the flight, when the Concorde was still below the engine out safety speed and below 400 feet in altitude.  Air France Concorde operating procedure specifically required the crew to wait until the aircraft has reached 400 feet before shutting down and engine.  The 400 foot requirement was in place to ensure that the aircraft has reached the engine out safety speed and has established a normal climb out.  Below that altitude any engine issue, including a fire is considered a secondary issue next to flying the airplane.  After all an engine on fire is still producing thrust and any thrust is better than no thrust.  

While the condition experienced was highly unusual and not something for which any crew had been trained to deal with the crew failed to follow one of the most basic responsibilities of a pilot.  No matter what happens, no matter how many alarm bells or warning are going off, your chief responsibility is to fly the airplane.  An engine fire can be dealt with, but only after a stable climbout has been achieved.  By shutting the #2 engine down the crew eliminated any available margin that might have kept the aircraft in the air long enough to reach Le Bourget.

Final Conclusions

So in the final analysis who is to blame for the crash of AF4590?  That is the question French prosecutors in the course of their criminal investigation of the accident attempted to determine.  However the chief concern of any accident investigation should not be to find fault and assign blame but rather understand why the accident occurred and prevent future accidents from occurring.  Attaching liability to an individual or company does nothing to advance the cause of safety.  

The case of AF4590 is simple in that it was a tragic accident, one in which not a single party to the accident, not Continental, Air France, Aéroports de Paris, or Aérospatiale-BAC did anything criminal to bring about the crash of Concorde F-BTSC.  Instead a series of mistakes, lost opportunities and circumstance combined to bring about the accident.

The most important resolution to come out of the investigation of AF4590 was the weaknesses in the design of Concorde that had been identified so clearly in the wake of the Dulles incident and tragically repeated in Paris were corrected.  The fuel tanks received kevlar linings to protect them from impact damage, the Concorde's tires were strengthened so they were better able to withstand impact with FOD.  Air France and British Airways both used lessons from the accident to improve their pilot training and make their crews better prepared to deal with situations like this.  The Paris Airport Authority improved its process for inspecting and maintaining the movement areas at Charles de Gaulle Airport and Continental improved its maintenance oversight and processes to ensure a mistake like this was not repeated.

In the end the BEA did its job and the findings of their report helped make the traveling public safer.  The lessons to be learned from the accident have been shared with the global airline community and the industry and regulators have moved forward applying the wisdom gained from this accident to make commercial aviation safer for everyone.