Two flight recorders were installed on board F-BTSC, in addition to a Quick Access Recorder (QAR) which was also interrogated to assist in the investigation.
The flight recorders were found at the accident site by a technical investigator four hours after the accident. They were recovered as soon as conditions at the site permitted. They were placed under seal and taken to the BEA by two police officers.
Cockpit Voice Recorder (CVR)
The outer casing of the CVR showed signs of exposure to fire and impact damage. The serial number was illegible because of marks left by fire. Nevertheless, the CVRıs thermal protection had functioned and the tape was found intact inside its protective box.
In the following days, a transcript of the entire length of the recording was made. The validation of the identity of the voices of the crewmembers was made with Air France Concorde pilots. Access to the recording was then limited to relevant members of some of the working groups, as well as members of the Commission of Inquiry.
Flight Data Recorder (FDR)
The outer casing of the FDR was damaged by impact and showed signs of exposure to fire. After the protective box was opened, the following was noted:
The tape, after extraction, was cleaned with distilled ethyl alcohol. It was strengthened at one point where the beginnings of a tear had been observed.
Readout of the whole of the tape, with simultaneous synchronisation of the signal being read out, was performed with Sundstrand IAE (Incident Analysis Equipment) (PN 960-0145-002).
Because of the condition of the tape, readout of the recording was of medium quality, and this caused a certain loss of signal synchronisation. This first readout made a preliminary analysis possible, but it was decided to seek better quality data at the same time, either by reading out the QAR or by a new readout of the FDR tape with digitisation of the signal so as to improve synchronisation by using algorithms appropriate to a poor quality signal.
The QAR is an unprotected recorder. It contains a copy of the FDR data on a magneto-optical disc and is used by Air France for flight analysis. The write procedure for the disc uses three backup memories whose role is to stock data sent by the Flight Data Acquisition Unit (FDAU) until such time as the vibration conditions detected by an internal accelerometer in the QAR are favourable for writing on the disc. The memories are volatile and must remain powered for the information they contain to be conserved.
The data readout was performed on 1st and 2nd August at Elancourt by personnel from Thomson CSF, the manufacturer of the QAR, in the presence of a judicial expert and a BEA investigator.
The QAR's box was crushed and the magneto-optical disc was deformed. The memory card, visible through the half torn-off casing, seemed to be in good condition. It was therefore decided to concentrate the work on this card. Two of the three memories had been torn off at impact. The third was still in place and was powered.
Tests were performed on check sample cards so as to define a method of data extraction, since this operation had never been carried out before. The method used was to connect a parallel power supply to the memory so as to be able to transfer it from its card to a receiver card. An uninterrupted series of zeros had first been written onto the two other memories of the receiver card.
The content of the third memory could thus be read out and a copy of the disc was given to the BEA. After analysis, it appeared that the parameters of the accident flight were present on the only one of the three memories which had remained powered. The quality of the recording, because of the technology used, was excellent and there was no de-synchronisation. It was not therefore necessary to try to read out the magneto-optical disc nor to proceed with further acquisition work on the FDR tape signal.
A synchronisation of the CVR recording was performed with the radio-communications recording (UTC time) and with the FDR recording.
Ground test were performed on Concorde using a recorder of the same type. A spectral analysis comparison of the recordings helped identify the sound of certain selectors and alarms.
In collaboration with a researcher from the Institute of Sound and Vibration Research in Southampton (GB) an attempt was made to detect, on the sound recording, possible noises of impacts on the aircraft structure by demodulation of the onboard power supply frequency (400 Hz). This technique proved fruitless.
A transcription of the last five minutes of the recording is attached in appendix 2. Of the thirty minutes on the CVR, the following points are of note:
N.B.: the numbers (1, etc.) refer to positions on the track in 9.1 and in appendix 5.
14 h 13 min 46 s, FO "fire protection", FE "tested".
14 h 14 min 04 s, FO "ZFWZFCG", FE "so I have ninety-one nine and fifty-two two".
14 h 14 min 17 s, Captain "the reference speeds so V1 one hundred fifty, VR one hundred ninety-eight, V2 two hundred twenty, two hundred forty, two hundred eighty, itıs displayed on the left".
14 h 14 min 28 s, FO "Trim", Captain "it's thirteen degrees".
14 h 14 min 53 s, Captain "Then the lever is at fourteen and you'll have an N2 of ninety-seven and a bit", FE "ninety-seven".
14 h 22 min 22 s, Captain "right we're going to do one hundred eighty-five one hundred thatıs to say we're going to be at the....structural limits", "structural err fifty-four per cent balance (*) see".
14 h 39 min 04 s, Captain "So the takeoff is...at maximum takeoff weight one hundred eighty tons one hundred which means four reheats with a minimum failure N2 of ninety-eight', 'Between zero and one hundred knots I stop for any aural warning the tyre flash', 'tyre flash and failure callout from you right", "Between one hundred knots and V1 I ignore the gong I stop for an engine fire a tyre flash and the failure callout", "after V1 we continue on the SID we just talked about we land back on runway twenty-six right".
14 h 40 min 19 s, Captain "How much fuel have we used ?" FE "We've got eight hundred kilos there".
14 h 41 min 09 s, FE "Brake temperatures checked one hundred fifty...".
14 h 42 min 31 s, Captain "top".
1 14 h 42 min 54 s, FO "one hundred knots".
14 h 42 min 57 s, FE "four greens".
2 14 h 43 min 03 s, FO "V1".
Of 14 h 43 min 07 s to 14 h 43 min 13 s, various noises which have not been identified at this stage of the investigation
3 14 h 43 min 13 s, message from the controller informing them of flames at the rear, read back by the FO.
14 h 43 min 20 s, FE "Failure eng... failure engine two".
4 14 h 43 min 22 s, fire alarm
14 h 43 min 24 s, FE "cut engine two".
14 h 43 min 25 s, Captain "engine fire procedure" and one second later, end of fire alarm
14 h 43 min 27 s, FO "Watch the airspeed the airspeed the airspeed".
5 14 h 43 min 30 s, Captain "Gear on retract". Over the following eight seconds, the crew mentioned the landing gear several times.
14 h 43 min 42 s, Fire alarm.
14 h 43 min 46 s, Captain "(are you) cutting engine two there".
14 h 43 min 48 s, FE "I've cut it".
14 h 43 min 49 s, FO "the airspeed".
6 14 h 43 min 56 s, FO "the gear isn't retracting".
Between 14 h 43 min 59 s and 14 h 44 min 03 s, three GPWS alarms are heard and at the same time the FO says 'the airspeed'.
7 14 h 44 min 14 s, FO '"Le Bourget Le Bourget" then a few seconds later 8 "negative we're trying for Le Bourget".
14 h 44 min 31 s, end of the recording.
11.3.1 The Flight
The recordings were decoded with the aid of documents provided by Air France and EADS. Four hundred parameters were recorded. Some of these parameters posed validation problems, in particular for their neutral or reference values. The SAT recording was invalid.
The previous flights, supplied by the Air France flight analysis service, were analysed for the purpose of validation and comparison.
Graphs derived from the recorded parameters for the whole of the flight are shown in appendix 4. Details of some significant parameters are listed below.
The following tables show the recorded values of certain parameters. For a given generated time, the associated parameters are values sampled at a specified moment in the course of the corresponding second. This indication does not appear in the tables. In addition, only the parameters of one engine are recorded each second. Thus, the parameters of each engine appear only every four seconds.
N.B.: The numbers (1, etc.) refer to the track shown in 9.1 and in appendix 5.
1 100 kt callout, generated time 97585
2 one second after the V1 callout, generated time 97595
3 flames reported by the controller, generated time 97604
4 radio altitude positive, generated time 97614
5 request to retract landing gear, generated time 97621
7 FO "Le Bourget Le Bourget", generated time 97665
|no. 2 *||97667||5.60||12.92||182.6°||0||13.88|
|no. 2 *||97675||5.19||12.30||168.9°||0||14.05|
|no. 2 *||97679||5.13||11.98
|| no. 3
This paragraph presents a preliminary explanation of the engine parameters. Note that these parameters are recorded every four seconds.
Note: Data on engine speed (N1 for the LP turbine and N2 for the HP turbine) presents a lag relative to the theoretical value. This lag is identical for all four engines. We may consider that it represents of an error in the recorded values.
The initial acceleration values of the engines and their behaviour during the initial phase of the takeoff (up to 97602) are normal on all four engines. From time 97603 onwards, significant changes appear in the parameters of engines 1 and 2. Speed, temperature and flow parameters indicate deceleration. Within the limitations of sampling error, this behaviour appears simultaneously on both engines. Initially engine 2 is more affected than engine 1. Engine 1ıs parameters return to normal values from generated time 97626.
In the CVR recording, the FE announces engine 2 cut off. A noise similar to an engine shutdown/fire handle being pulled is then heard. The change in the parameters of engine 2 confirms the shut down of the engine. This engine's parameters then appear to be normal for an engine windmilling in these flight conditions.
From time 97620 onwards, the parameters of engine 3 then of engine 4 behave in a manner consistent with a switch from "Take Off" to "Contingency" modes. From time 97626 onwards, after the initial problem, the parameters of engine 1 also show engine speeds consistent with a change to "Contingency" mode. The fuel flow, primary exhaust nozzle area and P7 pressure are consistent with reheat operating on these engines.
From time 97654 onwards, the parameters of engine 1 indicate a deceleration and an increase in EGT. From time 97680 onwards, the parameters of engine 3 and then engine 4 indicate a sharp deceleration.
The aircraft's track (see appendix 5) was calculated by integrating the airspeed and magnetic heading parameters, by positioning the first and last points of their known or estimated position. A reasonable approximation of the ground track was thus obtained. Bearing in mind the method used, the tolerance is of the order of about a dozen metres at the ends of the track. It is at its greatest in the middle of the track (around a hundred metres).