One of the apparent mysteries of the Air New Zealand A320 crash is how an aircraft controlled by a computer designed to override any dangerous manoeuvre attempted by the pilots could pitch upwards at high power so sharply that it stalled and fell into the Mediterranean.
The computer should have overridden any (even completely accidental) pilot action that might cause such a manoeuvre, so the fact that this crash happened this way (confirmed by the interim report on the accident published yesterday) caused my initial concern that the catastrophe might be linked to a software problem in the A320’s computer, similar to the software problem that caused a serious midflight incident last October in a different Airbus model, an A330 flown by Qantas.
But, overnight, I have read and re-read the report by the French agency Bureau d’Enquetes et d’Analyses (BEA) — which I wrote about yesterday — into the November 28 A320 crash off Perpignan. Buried in its pages seems to be the explanation.
What leapt off page 14 at me is that most of the computer override systems are put out of action if the plane’s landing gear is down, which was the case when the Air New Zealand jet executed a low-speed check as it approached Perpignan airport for a “touch-and-go” landing during a test flight.
In short, this “override of the override” may have meant that the computer did not prevent the initial stall that the German pilot responded to by advancing the thrust levers and operating the “sidestick” (that controls an Airbus’s pitch and direction) to lower then lift the nose.
In this context, here is the relevant section of the report:
1.6.6 Flight control laws
The Airbus A320 has fly-by-wire flight controls. The aerodynamic surfaces, which enable airplane control, are not mechanically linked to the controls. The airplane is flown using two sidesticks. The movements of these sidesticks are transmitted in the form of electrical signals to computers that transform them into orders to the actuators of the various surfaces. The laws that govern these transformations are called “flight control laws”. On the A320, in nominal operation, the flight control law is called “normal law”. Under certain conditions, it can be replaced by two reconfiguration laws: the “alternate law” or the “direct law”.
The normal law offers protections in attitude (the pitch and bank values are limited), load factor, high speed and angle of attack (specifically at low speed). Pitch trimming is ensured automatically by the auto-trim. Bank angle is coordinated with the rudder. The sidesticks control the load factor according to the normal airplane axis and the roll rate.
In alternate law, the sidesticks control the load factor according to the normal airplane axis as for the normal law, but with fewer protections. In roll, they directly control, as they do in normal law, the ailerons and the slats. When the landing gear is extended, the pitch control law passes to direct law.
In direct law, there is no automatic pitch trimming, only the load factor protection is maintained. The control surfaces are activated directly by the controls.
In “normal law,” if the jet’s speed falls under what is called “lowest selectable speed,” the computer will activate protections that pitch the nose down (to stop a stall) and increase its speed (to do the same). If deceleration continues, maximum thrust is automatically applied.
But in alternate or direct law, the aforementioned protections are no longer available, only the stall warning remains activated.
It seems from the report that the A320 was operating under “direct law” at the time of the stall, because its landing gear had been extended for the “touch-and-go” at Perpignan. This seems to suggest that the computer might not have been able to counteract the plane’s slowing to stall speed during the “low-speed check” being undertaken, and been unable to override too steep a pitch-up motion of the sidestick.
Page 20 has this comment: At 15 h 43 min 41 s, the Captain positioned the thrust control levers on IDLE and autothrust disengaged. The altitude was 4,080 feet and the speed was 166 kts. The Captain asked for the landing gear extension then said “we do the err the…” and the New Zealand pilot answered “Slow speed yeah”.
Unless I am completely mistaken about the report’s flight control laws section, the ultimate effect of the “landing gear extension” asked for by the captain would have been to put the computer into the least-protective “direct law.”
The report goes on: At 15 h 45 min 05 s, the airplane was at 2,910 ft altitude and a speed of 99 kt. Pitch attitude was 18.6°. The stall warning sounded. In the following second, the thrust control levers were moved to TO/GA [takeoff - go around] position. Auto-thrust changed to armed mode. A symmetrical increase in engine RPM is noticeable up to N1 values of about 88 %. At 15 h 45 min 09 s, the bank angle reached 8° to the left and the speed 92.5 kt. The Captain made a lateral input to the right and a longitudinal movement forwards on his sidestick.
The report says the flight control laws for pitch and roll passed from “normal” to “direct.”
In the next few seconds, the captain attempted to stabilise the jet, which plunged to 2250 feet, by which altitude its speed had picked up from 92.5 knots to 144.5 knots.
The captain called for retraction of the landing gear. Around this point the jet was climbing sharply at high power and over 21 seconds rose from 2250 feet to 3800 feet with the nose pointing up at an extraordinarily steep 57 degrees. Four seconds after the plane reached 3800 feet, the landing gear was retracted and locked. The speed by then had fallen to just 46 knots and the stall warning was sounding. The plane fell out of the sky.
The interim report does not attempt to explain why any of this happened, nor does it in any way attribute any blame to anyone or anything. It simply states what happened, second by second, based on the detailed information gained from the jet’s flight data recorder (which records every pilot action and all other flight information such as height and speed) and cockpit voice recorder (which records everything said in the cockpit).
So all I am doing is speculating, from the information it contains, that the extension of the landing gear might have disarmed some or many of the computer’s safety override systems at a fatal moment.
An alternate hypothesis seems possible from the same evidence. The computer protection that applies maximum thrust if an A320 continues to decelerate below “lowest selectable speed” might have caused the power surge that pushed the jet up at the steep angle that led to the final, fatal stall. But that seems to rely on the computer being in “normal law.”
I’d appreciate any A320 pilots reading this to read the relevant section of the report and comment here what they think.
It is certainly dark news for Air New Zealand that the report makes it plain that the low-altitude, low-speed manoeuvre that led to the crash was done during the landing phase of the flight at the behest of the Air New Zealand pilot seated behind the German captain flying the plane, who indicated he’d prefer to do that test later or not at all.
The airline will be hoping a technical issue, perhaps like the one I canvass above, proves to be the ultimate cause.
1 Comment
February 26, 2009 at 10:13 am
page 45 of the BEA report in French
Airbus Customer Acceptance Manual:
FL 140 – FL 100….During the acceleration observe:
-auto trim
-the alpha floor activation
- Disconnect this alpha floor function at once
Which law was in effect isn’t the factor – performing a maneuver too low was the factor as the A/C was on the way out of the upset maneuver, just too low to escape the bottom of trajectory.
The BEA’s report, although interim, is sufficiently illuminating to see what they’re pointing at.
Fyfes statements smack of managing the positioning of Air NZ and the implied complicity of the kiwi members talking to the cockpit crew in the sequencing and insistence on performing the low speed test.All a bit grim.Time will tell