Air France 447: Smart planes still vulnerable to human error

On the night of June 1, 2009, in a place as remote as any on earth, the human and the digital worlds did battle. Both lost.

With the release of a third interim report by the French accident investigation bureau, what happened to Air France Flight 447, the Airbus that disappeared in the mid-Atlantic two years ago with 228 aboard, is now pretty clear. The big twin jet was inadvertently mishandled by the pilot who was flying it. He got it into an extremely unusual position, and neither he nor the other two pilots with him could figure out what was happening or how to fix it.

For the next three minutes the airplane, its fuselage slightly nose-high as though approaching a landing, dropped toward the sea while the baffled and terrified pilots struggled to make sense of the indications on their instruments. But they never did what they needed to do — get the nose down, so that the airplane would be pointing in the direction it was going — to allow the Airbus, which was perfectly sound and intact, to recover and continue on its flight from Rio de Janeiro to Paris.

The problems started with an unusual accumulation of ice crystals in the airplane’s three pitot tubes, probes that measure speed by capturing the pressure of the oncoming air. Losing reliable airspeed information, the autopilot turned the airplane over to its human minders, as it was designed to do. For some reason — perhaps he thought the speed was too high — the pilot who was flying the airplane pulled it up into a climb, rapidly gaining several thousand feet. This in turn led to an aggravated stall.

A stall in an airplane has nothing to do with the engines. It is a matter of the wings suddenly losing some of their lift and gaining a great deal of drag. A fully developed stall is something that no airliner is ever supposed to experience. There is ample and unmistakable advance warning. In training, pilots merely approach the stall, receive the warning, then “fly out of it.” But for some reason, despite the warning, 447’s pilots never understood the situation. They seemed paralyzed. Was the airplane too slow or too fast? Which instruments were reliable? In fact, except for the probes measuring airspeed, they all were, but unfortunately there was no single integrated display on the instrument panel that could, like Microsoft Flight Simulator, let the pilots view the airplane from outside. Such a display, by making clear what the never-before-seen combination of instrument indications meant, would have saved them.

The Airbus A330, like other new-generation airliners, is controlled by a computer, in theory a sort of super-pilot, never tired or distracted, with lightning-fast reflexes and an encyclopedic knowledge of how best to fly. The human pilot still uses the stick and throttles in the traditional way, but commands go to the computer, which in turn executes them. If the pilot tells the airplane to bank too steeply or fly too slowly or too fast, the computer will not comply. Its “laws” are intended to protect against pilot errors that, far more often than mechanical failures, have led to accidents.

The transition from mechanical to digital flight controls has brought about a shift in the way pilots are trained. Basic flying skills — the ability, for instance, to recover from unusual situations or to intuitively sense what an airplane is doing or is about to do — receive less and less emphasis. Testable knowledge of airplane systems and standardized flight procedures takes precedence.

This is the future; the trend will not be reversed. Modern airplanes fly themselves, and pilots become, increasingly, information managers. Someday airplanes, like elevators, could dispense with human operators altogether, though passengers may always prefer to be greeted by an avuncular four-striper in the cockpit door, however inferior his flying skills may be to those of a well-programmed computer. There will still be accidents, particularly horrifying because they will be mindless and automated, but fewer than in the era of hands-on, seat-of-the-pants pilots.

But we are still in transition, and Flight 447 fell victim to a philosophical inconsistency. The computer was supposed to protect the pilots from themselves, but in a pinch it threw up its hands and abruptly turned over control to a startled and unprepared human crew.

The proper response to a loss of airspeed information in steady cruising flight is to change nothing: Maintain power and altitude, and the airplane will be fine. The computer could easily have done this, but the programmers who designed the A330’s flight management system evidently thought it wiser to let the pilots take charge. They never imagined that a pilot would get it so terribly wrong. Why would he command a sudden climb, hold the airplane’s nose up until it stalled, and keep holding the nose up — exactly the opposite of what was needed — for more than three agonizing minutes?

Why? Because he was a human being. That is what the programmers forgot.

By Peter Garrison, who writes a monthly accident-analysis column for Flying Magazine.

Deja un comentario

Tu dirección de correo electrónico no será publicada. Los campos obligatorios están marcados con *