The Scoop on the A330 Accident

From:         Peter Ladkin <Peter.Ladkin@loria.fr>
Date:         09 Jul 94 16:58:51 
Followups:    1 2 3 4
Next article
View raw article
  or MIME structure

Air et Cosmos, 11-24 Juillet 1994, p15, contains an extensive report
on the A330 accident of 30 June 1994 by Jean-Pierre Casamayou.  The
general story has been reported by Peter Mellor (RISKS-16.19). The new
info is highly relevant, and implies that control of the aircraft was
lost while the aircraft was under automatic control.  This is the
first case, to my knowledge, in which this has been proved to have
happened to Airbus aircraft, without any concomitant pilot error.
Sadly, the test pilots allowed the departure from control to continue
for up to 12 seconds in order to analyse the incident. This delay was
gallant but fatal. That's the English for you (RIP Capt. Nick Warner).

The autopilot was using experimental software. This A330 was
undergoing a flight test required for certification of the autopilot
for Category III operations with Pratt and Whitney 4168 engines (the
other A330's already in operation use CF6-80E1's, and such equipment
has already been through this particular flight test sequence).
Category III operations mean use of the autopilot for landing, up to
and including main gear on the runway, and requires special
certification of both aircraft and crew. It follows that a Category
III operation can potentially be aborted, i.e. the pilots can select
go-around while under autopilot contol, with the main gear on the
runway, and in the worst case an engine can fail at this point. One
can see why it's required to conduct this test from an actual takeoff,
rather than at altitude.

The flight was supposed to test the mode SRS (speed reference system)
of the autopilot, which should control the speed and angle of attack
(AoA) of the aircraft in case of an engine-out. AoA is defined to be
the angle that the wing makes with the undisturbed airflow in front of
the wing. The test was performed at rearmost center-of-gravity.  

Following is a translation of a continuous fragment of the article. I
have included the originals of phrases I am unsure of.  Many thanks
also to Pete Mellor for some clarifications. I don't have a dictionary
of French aeronautical terms (although such exist, and they're quite
large).  It refers to the following `V-speeds', defined in FAR
Subchapter A Part 1 Para 1.2 for those in the US.  V_1 is takeoff
decision speed (the speed at which the decision is made to abort or to
continue takeoff in the case of engine failure); V_R is rotation speed
(the speed at which the pilot commands nose-up); V_2 is takeoff safety
speed (the speed at which the airplane may takeoff safely, even with
one motor out); V_{mca} is the minimum single-engine control speed
(the speed at which control of the aircraft may be maintained with one
engine out). QNH is indicated altitude with altimeter set to sea level
mean pressure, and QFE is indicated altitude above highest point on
the airport with altimeter set to airport mean pressure.

[begin translation] 

The takeoff (V_1 = V_R = 126kts and V_2 = 135kts) took place at
136kts, 25 seconds after full power was arrived at (`la mise en plein
piussance des moteurs'], then the aircraft maintained its speed of
climb of 150 kts. After the takeoff, an altitude of 600m QNH (roughly
460m QFE) was selected on the flight director FCU [the Flight Director
on the A330 is called the FCU. pbl] This means that the aircraft
should restore level flight [`retablir en palier'] at 450m from the
ground.

Conforming to the test order, the pilot attained a speed of 150 kts,
and 28 degrees AoA in order to maintain this speed. Six seconds after
takeoff, the autopilot was engaged, then the left engine retarded and
the corresponding hydraulic pump cut to simulate a complete failure of
the left engine. As predicted, the AoA began to diminish and passed
from 29 degrees to 25 degrees, the limit authorised by the FMGES
(Flight Management Guidance and Envelope System) which protects the
flight envelope. But quickly, because of the low altitude selected on
the FCU, the autopilot departed from mode SRS and entered mode
ALT-STAR, the mode for acquisition and retention of altitude, in which
mode the autopilot tries to attain altitude as quickly as possible,
without taking into account the limiting conditions that the airplane
was in: rearmost CoG, one engine retarded and the other at full power,
high `incidence' [another word for AoA. pbl] [this is not a good
explanation of ALT-STAR mode. pbl].  Result: the AoA started to
increase again, and the speed decreased extremely quickly
[`brutalement'].

The flight team noted the anomaly immediately, but purposely let the
situation degrade for about 12 seconds, in order to analyse it better,
as is their role. The AoA attained 33 degrees with speed decaying to
100kts, which is 18kts less than V_{mca}, the minimum single engine
control speed . At this moment, the pilot disconnected the autopilot
and took over control.  But the speed continued to decrease. At about
90 kts, 28kts less than V_{mca}, the aircraft departed [`part en
decrochage'] to the left in a stall with angle of bank [`angle de
roulis'] attaining 110 degrees.

The pilot reacted quickly and well in retarding the right engine
then bringing the wings horizontal. Unfortunately, because of the low
altitude and fast rater of descent, he couldn't avoid impact with the
ground, 35 seconds after takeoff.

[end translation]

Peter Ladkin