American Airlines, Airbus Position for Flight 587 Final Hearing
Aviation Week & Space Technology
10/25/2004, page 45
Michael A. Dornheim
Los Angeles
Frances Fiorino
Washington
Three-year-long probe ends: American Airlines and Airbus position for final Flight 587 hearing
NTSB Wraps Up AA587
American Airlines and Airbus will soon engage in their final Flight 587 showdown.
The NTSB meets this week in Washington for the final hearing that will determine the probable cause of the Nov. 12, 2001, accident, in which the vertical tail assembly separated from an American Airlines Airbus A300-600 shortly after takeoff from New York JFK International Airport. The crash killed 260 people on board and five on the ground. It was Airbus' first accident in the U.S.
There is much speculation on what the board will deem probable cause--pilot error, lapses in airline training and nonsharing of safety data by the aircraft manufacturer.
The consensus is that the copilot applied the rudder from stop to stop in what may have been pilot-induced oscillations. The PIOs (also called aircraft-pilot coupling, or APC) in turn caused a large sideslip plus rudder deflection that overloaded the tail fin and ultimately resulted in the inflight structural failure.
What led to the PIOs is at the heart of the hearing. Airbus, in its submission to the NTSB on probable cause, states the pilot's rudder pedal inputs were conditioned by American's Advanced Aircraft Maneuvering Program (AAMP), which "advocated aggressive use of rudder for roll control" and reinforced by negative training in simulators.
On the other hand, American's submission says the cause was "the onset of a design-induced, adverse APC event." American's A300 fleet captain-flight operations, Delvin Young, said the pilot was "part of the process, not the cause" of the accident. American believes the broader issue for the board is examination of system safety issues.
American is concerned that the NTSB final hearing might not adequately weigh prior instances involving the A300-600 and A310 in which fin limit loads were exceeded through rudder overcontrol or Airbus' apparent failure to inform operators and regulators about the events. Airbus has claimed it informed American about one such event--a May 1997 inflight upset incident involving American Flight 903 near Miami.
Airbus has stressed that the NTSB should scrutinize American's training procedures. Airbus maintains it warned airlines not to use rudder for roll control--unless as a last resort--in the Upset Recovery Training Aid, a manual generated by both Airbus and Boeing and first released in 1998.
American claims Airbus never informed it about use of rudder for roll control. Further, Airbus has argued that AAMP training emphasized use of rudder for roll control purposes in classroom lectures, in written material and in the simulator. This encouraged pilots to make the kind of motions the Flight 587 pilot made in response to wake encounters with a Boeing 747-400 that had departed prior to the A300-600.
Airbus claims that training emphasis on use of rudder for roll control was unfavorable for several reasons. First, the only way roll control through use of rudder is achieved is to develop a sideslip and the aircraft will roll in response. However, as there's a long delay between control input and response, when the aircraft responds, it does so with alacrity. Secondly, the only experience Flight 587's pilot had with upset recovery had been in the simulator--which likely had given him false expectations of how an airplane in flight would respond to control inputs.
NTSB graph shows that the rudder limiter on the A300-600 makes the pedal more sensitive at higher speeds. At the 250 kt. of the Flight 587 accident, it is six times more sensitive than its A300B2/B4 predecessor.
Airbus said another critical training issue is that American made a simulator modification which had not been cleared with either Airbus or the FAA. The modification was such that in order to teach the pilot to use rudder in a simulated upset due to wake vortex encounters, the instructor would push a button in the simulator cab to stop all roll control. A rolling moment was then induced, and when the aircraft reached a large bank angle, it washed back at a gradual rate to be control-effective. Airbus is saying the modification, in effect, forced pilots to make full control inputs on the roll and yaw axes.
IN ADDITION, Airbus might argue that even American's own staff voiced concerns about upset training. Airbus points to a letter from Paul Railsback, American's managing director of flight operations-technical, at the time of the Flight 903 event. Addressing American's vice president of flight ops, Railsback states his "grave concerns about some flawed aerodynamic theory and flying techniques that have been presented in AAMP . . . [that] I believe . . . are validated by the recent AA903 accident. Pilots are told to use rudder as the primary means of roll control in unusual attitude recoveries. This is not only wrong, it is exceptionally dangerous. . . . American Airlines is at grave risk of a catastrophic upset."
American is also pointing to the NTSB's findings in a prior accident involving an ATR 72 turboprop operated by Simmons Airlines/American Eagle that crashed after wing icing behind the deice boots caused unstable aileron forces of an estimated 60 lb. snatching the control away from neutral. The Flight 4184 accident occurred near Roselawn, Ind., in October 1994, and earlier incidents and accidents had made the problem known within ATR (AW&ST July 15, 1996, p. 41).
THE BOARD concluded that a contributing factor was ATR's inadequate response to continuing icing roll upsets and failure to provide procedures for these conditions. Factors also include the French DGAC certification authority's failing to ensure ATR 72 airworthiness in icing conditions, and failing to give the FAA timely information developed from previous ATR incidents and accidents in icing. The crew and airline were not mentioned as a probable cause or contributing factor.
American argues that Airbus knew that the A300-600 and A310 were prone to rudder overcontrolling from three or four prior incidents dating back to at least 1991 that were presented in NTSB testimony (AW&ST Nov. 25, 2002, p. 44).
Calculations showed that these rudder motions overloaded the fin beyond limit load in several cases, and beyond ultimate load (1.5 times limit load) in at least two instances--a point at which the fin might be expected to break off. One of those ultimate load violations was on American's Flight 903 in May 1997 and the other was on an Interflug A310 in 1991. Other cases include an Air France Flight 825 incident in December 1999 on an A310 where limit load was exceeded.
A key document, American says, is a June 19, 1997, Airbus internal memorandum stating that in Flight 903's "rear fuselage, fin and empennage the ultimate design loads may have been exceeded." This knowledge did not come to the NTSB or American until the document was revealed after the Flight 587 accident more than four years later.
"With Interflug, you could say it was a one-off," says Bruce Hicks, an American spokesman. "But with 903 you have to connect the dots, and Air France 825 clinched it."
American claims Airbus knew of overloads but did not inform operators or authorities in sufficiently specific language that would catch their attention, nor did the manufacturer convey data about the series of rudder overcontrol incidents on the A300-600 and A310.
An independent study requested as part of the NTSB Flight 587 investigation showed that at higher speeds, where pilots rarely use the pedals, the rudder of these two types is 3.2-10 times more sensitive (in terms of deflection per incremental force) than other transports, with a breakout force several times larger than the incremental force for full deflection (see graph, p. 45). An NTSB poll of Boeing, Douglas and Airbus histories showed that no other type had this history of fin overload.
By analogy with the ATR 72 accident, American argues that it and its copilot should not be a probable cause of the Flight 587 crash because the A300-600 has a unique propensity to rudder PIO and fin overloads at the higher speeds of the accident, and that Airbus did not disclose these characteristics to operators.
A distinction between the Roselawn case and Flight 587 is that the ATR 72 aileron snatch happened by itself, whereas the Flight 587 rudder oscillation started with the copilot applying full right rudder for little apparent reason. Investigators have concluded that the aircraft was crossing the wake of a preceding Boeing 747 at that point, but the flight data recorder (FDR) shows only about 2 deg. of roll increase, from 23 to 25 deg. The FDR samples roll angle only once per second, and some roll acceleration felt by the crew may be missed--but the net result is just a small blip in roll. The copilot had previously been chastised by another captain for excessive rudder usage while crossing a mild wake, apparently in response to his interpretation of upset recovery training given by American.
"Let's say the copilot was way too aggressive and should have waited to see if the aileron worked," Hicks says. "Still, pushing the rudder a small amount shouldn't kill all those people." The initial push is believed to have started the PIO, which took off the tail. "Where in pilot training does it say you can't use rudder to assist roll? Where does it say to 'let go of the controls,' as Airbus suggested in the hearing? You can't say he used 'too much' rudder because it's so sensitive it's an on-off system. Certification requirements say it shouldn't have caused an immediate PIO. You can't have flight controls so unforgiving that one pulse gives a PIO that results in the tail coming off in 6.5 sec."
AS FOR THE ISSUE of rudder sensitivity, Airbus claims that in 16 million flight hours, there have only been two upset events, both involving American Airlines, and both involving large amplitude rudder inputs--Flights 587 and 903. American points to other cases: Air France 825 in 1999 and Interflug Flight 103 in 1991. Airbus has also pointed out that it came to light in previous Flight 587 testimony that First Officer Sten Molin had a history of using excessive rudder and had, in the case of Flight 587, exerted a force of 140 lb. on the initial rudder input.
"The NTSB may say the copilot started the oscillation so it's the probable cause," Hicks says. "Yes, the copilot's use of rudder is a factor, but was it wrong? Not based on his training, widespread misconceptions about maneuvering speed, and how we thought the flight controls worked.
"I think hiding knowledge is a cause--with that knowledge we have no accident. Had what happened in Flight 903, Interflug and Air France Flight 825 been known, Flight 587 wouldn't have happened. If we have 587 again today then it's the pilots' fault because they are now trained. But not back then."