Experts Challenge Certification Criteria at Flight 587 Crash Hearings

High loading incidents cited as need to raise rudder strength standards
One of the most complex accident investigations in recent memory has shifted from public testimony to assessment of a mountain of data, only about a fourth of which reportedly has been processed more than a year after the accident.

The investigation involves the fatal Nov. 12, 2001, crash of American Airlines [AMR] Flight 587. Recently, the National Transportation Safety Board (NTSB) conducted four days of open hearings into the various issues involved. The first two days of hearings focused on the design of the rudder control system on the accident airplane, an A300-600, and on the training pilots received at American Airlines on upset recovery (see ASW, Nov. 4).

The elemental issue in the first two days might be described as the susceptibility of a powerful yet overly sensitive rudder control system to inputs from pilots whose upset recovery training may have subtly preconditioned them to make excessive use of the rudder. To both issues the main protagonists, manufacturer Airbus and the airline, would say that this was not the case. To Airbus, there are no unexamined issues in the rudder control system. To American, the training stressed coordinated, modest use of the rudder.

On the third and fourth days of the hearings, the design of the composite tailfin was reviewed in detail. The discussion led all the way to the headwaters of that great river known as the certification process.

At this point, the Flight 587 accident investigation is unusually significant for the broad range of issues involved. To highlight just a few, the issues concern basic factors in aircraft structural and control system design, wake separation standards, air traffic control procedures, pilot training, flight system reliability, the sharing of safety- related information, as well as small yet significant details, such as whether it should take one or two clicks to disconnect an autothrottle, and what kind of aural signals or visual queues should be provided to alert pilots to a change in functional status.

In the meantime, highlights of the discussion over the composite tailfin point to what may well evolve into an eventual change in certification standards.

Days 3 & 4 - Composites under scrutiny
The good news, if it can be so characterized in the grim context of a fatal accident investigation, is that the tailfin held until nearly twice its design limit load. The bad news is the number of times tailfins on the A300 and its cousin, the A310, have experienced forces greater than the design limit load.

The disclosures are likely to lead to changes in certification standards for aircraft structural components, whether they are manufactured of metal or composites.

Previous incidents of high tailfin loadings were brought out in the concluding sessions of the NTSB hearings. The composite tailfin on the A300-600 snapped off under the excessive aerodynamic loading in the final seconds of the doomed airplane's brief 103-second flight.

The composite lugs holding the tailfin to the fuselage gave way at nearly twice the worst-case condition envisioned by the airplane's designers, or 1.96 times limit load. Limit load is defined as the greatest combination of shear, bending and torsion loads the airplane would ever experience in service. Consider this limit load has having a value of 1.0. The tail, in this case, must be able to carry the limit load and, if it deforms under the stress, return to its original shape when the load is released. Designers add to the limit load a 50 percent factor to account for unforeseen circumstances. This value, 1.5, defines ultimate load, or the point at which the structure may be able to carry the load, but not without permanent deformation of its shape.

During the original certification trials, the tailfin had gone to 1.83 of limit load before breaking in the test rig. Hence, Airbus engineers regard the fact that the tailfin on the accident airplane went to a load about 7 percent greater as further evidence of the soundness of their design philosophy.

History of high loading events
However, in seven known cases where A300 tailfins experienced unusually high loading in service, four of them were above limit load. In the operational history of the A310, a variant of the A300, in four cases of high loading ultimate load was exceeded once, and limit load twice. In a total of 11 high-loading events for the A300 and A310 combined, two- thirds of the cases exceeded limit load. Three cases exceeded ultimate load: the Flight 587 accident aircraft (event "A" in the table), another American Airlines A300, Flight 903, in 1997 (event "B"), and an Interflug A310 incident in 1991 (event "H").

The number of such incidents gave rise during the hearing to a spirited discussion about whether the calculation of limit load is really taking the worst-case condition into consideration. After all, limit load is supposed to represent the very worst-case of gusts, crosswind takeoffs and landings, loss of engine thrust, and flight control system failures. From some 30 different cases, limit load is based on the most demanding.

One of the most important calculations concerns the loads generated on the tailfin when the rudder moves to full deflection in a stabilized sideslip. The certification requirement involves full deflection, followed by the rudder's return to the neutral position, followed by movement in the opposite direct and return to the neutral position.

Doublets not required
The certification requirements do not include rudder doublets, or the rapid movement of the rudder to one side and back to the other, without pausing at the neutral position. In five of the seven cases where limit load was exceeded, rudder doublets occurred.

John Clark, NTSB head of aviation accident investigations, raised the obvious question: "If rudder reversals can take the plane past limit load, why are they not part of the certification requirement?"

Airbus officials kept pointing to the certification requirements. "There is no need for more strength," insisted Erhard Winkler, a senior composites specialist at Airbus. "We cover our requirements," he said, adding, "The requirement is 1.5 and we achieved 1.9." He was referring to the 1.96 multiple of limit load estimated to have been the amount of force that broke off the tail on the Flight 587 airplane.

Capt. Don Pitts, safety committee chairman for the Allied Pilots Association (APA), the union of American Airlines pilots, picked up on the same theme. "Did Airbus consider the pilot moving the rudder in the wrong direction in an engine out [situation] and then correcting?"

Ewe Kerlin, an Airbus structural loads expert, replied, "No ... just the requirements."

Pitts directed his next question to a Federal Aviation Administration (FAA) witness. "With seven examples of this design exceeding limit load, where are we with respect to regulatory requirements?" he asked Dr. Larry Ilcewicz.

Ilcewicz, chief FAA scientific and technical advisor for composites, replied, "The reason we design structure to limit load and then to 1.5 is that we recognize there will be exceptional cases."

"For limit load, you want to be good for extremely rare events," he said. For ultimate load, he said, "I want to be sure of a full factor for safety."

Pitts pressed his point. "If we see cases exceeding limit load, don't we need to assess if that load factor is accurate?"

The response was intriguing. "I don't care to comment," Ilcewicz said. Then, after a brief pause, he said, "I have confidence that most of the composite structure out there is safe." However, given the cases where the A310 and A300 were outside of their design envelope, "being a stress analyst, that makes me uncomfortable."

Dr. Bill Ribbens, who teaches airplane design at the University of Michigan, expressed his discomfort with certification standards that do not include the aerodynamic loads involved with doublets. "You should make airplanes so you can't break them, especially in the yaw axis," he asserted.

Wrinkling a stimulus to worry
NTSB member John Goglia expressed his discomfort from a maintenance perspective.

"How do we assure airworthiness for composites," he asked. The wrinkled skin of an aluminum structure is a sure sign that limit load has been exceeded, Goglia pointed out. But for composites, no such telltale wrinkling will occur and, indeed, damage could be hidden from view.

As an example, hidden damage in the tailfin of the American A300 that experienced an upset in 1997 was missed twice - once during a visual inspection shortly after the event, and again when the tailfin was reinspected visually after the Flight 587 accident. It was not until the fin was subjected to ultrasound inspection that interior damage was found. The tailfin, having been subjected to a force greater than ultimate load, remained in service for another five years before finally being removed from the Flight 903 airplane.

"What today would prevent the Flight 903 airplane from being returned to service?" Goglia asked. Ilcewicz referred to the airworthiness directive (AD) requiring tailfin inspections whenever an A300 experiences a lateral loading in excess of 0.3 G.

NTSB member George Black inquired, "So, if a B777 was involved, we wouldn't look at it?" The B777 also has a composite tailfin.

Even though the AD does not apply to the B777, Ilcewicz said the problem is understood within the industry. "This kind of loading event is something maintenance personnel would bring forward for further action," he assured.

But Goglia stressed the conundrum. If damage is not visible on a composite structure, unlike the wrinkling of overstressed aluminum, maintenance personnel will be reluctant "to turn themselves in," he said. By implication, other airplanes with composite structures could experience loadings outside their design envelope and, absent visible damage, could remain in service for years.

--------------------------------------------------------------------------------

Upset Recovery Training and the Potential for Negative Learning

Three issues are woven through a history of American Airlines' [AMR] upset recovery training program. One is whether the instruction encouraged pilots to be too assertive in use of the rudder. Company officials have repeatedly and fervently said no. The second is whether the simulators used to impart the instruction accurately reflected the real-world behavior of the aircraft. And third is the general understanding of maneuvering speed and whether any restrictions need to be applied to the use of flight controls inside the airplane's approved performance envelope.

In an industry where it is often said that basic flight safety issues have been resolved, leaving only the tough esoteric challenges, the Flight 587 case is significant for the fundamentals involved.

The chronology below was compiled from documents released as part of the National Transportation Safety Board's (NTSB) hearings Oct. 29 to Nov. 1 into the fatal Nov. 12, 2001, crash of American's Flight 587 (see ASW, Nov. 4). First officer Sten Molin was the pilot flying; he had received American's upset recovery training in 1997. The airline was among the first in the United States to initiate such training, investing some $6 million in the program.

The program was the result of industry concern that pilots were responding inappropriately to in-flight upsets. In 1991, a United Airlines [UAL] B737 crashed at Colorado Springs, Colo. In 1994, an American Eagle ATR-72 crashed at Roselawn, Ind., and also that year a USAir B737 crashed near Aliquippa, Penn. Both B737s experienced uncommanded rudder reversals. Icing precipitated the crash of the ATR-72 twin turboprop. In all three cases, the aircrews pulled back on the control yoke, holding it there until impact. These loss-of-control accidents fueled a concern within the industry that pilots were inadequately trained to cope with in-flight upsets.

American's program precipitated a debate within the industry about rudder use in upset recovery and the relevance of simulators as tools to illustrate the conditions and techniques of upset recovery. What follows is a chronology or the program's origins and its evolution over time. It shows that American Airlines reduced its original emphasis on rudder use. At the same time, if Molin had over-used the rudder in an earlier flight, as cited below, subsequent reports by pilots who flew with Molin do not portray him as lead- footed on the pedals. Rather, the picture that emerges is one of a decisive pilot with fast reflexes.

Molin's piloting habits aside, the discussion of upset recovery training touches on fundamental issues of what regions of the flight envelope should be covered by such training, the appropriateness of initiating such training at extreme angles of bank, and the appropriateness of suppressing control response until an extreme bank angle is achieved. American's A300 simulator had preprogrammed event buttons that would place the aircraft in a nose high or a rolling condition. Roll control was inhibited until the aircraft was at a 90� angle of bank, at which point the recovery training began. Since the Flight 587 accident, the procedure has been changed to manually place the simulator in an unusual attitude.

Included in the chronology below are some pilots' understanding of maneuvering speed (VA) and what they believe is the freedom within that envelope to make full use of flight controls.

Of perhaps greater significance, the debate over American's program touches upon the significant limitations of computerized flight simulators to accurately reflect the behavior of the aircraft, especially in regions at the outer boundaries of the aircraft performance envelope involving the physiological sensation of G-loads.

1994: Two high-profile loss of control accidents

American Eagle ATR-72 crashes in icing conditions at Roselawn, Ind., and US Air B737 crashes neat Pittsburgh from uncommanded rudder deflection. In both cases, when the upsets occurred the pilots incorrectly pulled the yokes into their laps and held them there, virtually guaranteeing that the airplanes would nose-dive into the ground.

Spring, 1995: American Airlines Vice President of Flight Capt. Cecil Ewell asks Capt. Warren VanderBurgh to head the development of the Advanced Aircraft Maneuvering Program (AAMP).

Aug. 17, 1995: FAA approves American Airlines' AAMP.

Jan. 1, 1997: AAMP training course material (extracts)

Aerodynamic Definitions

The effectiveness of the rudder as a roll control will increase with increasing AOA [angle of attack]. At the higher angles of attack, THE RUDDER becomes the most effective roll control.
Pilot Response to Wake Turbulence

Do not apply any back pressure on yoke at more than 90� of bank. ROLL FIRST - THEN PULL.
High AOA maneuvering - RUDDER

[ASW note: The guidance here is to ROLL wings level, then PULL to align the flight path to the horizon.

Rudder use is advocated to stop any rapid roll at high AOA]

May 1997: First Officer Sten Molin first flies with Capt. John Lavelle. On July 15, 2002, some eight months after the crash of Flight 587, Lavelle was interviewed. From investigator's notes of that interview, Capt. Lavelle recounted his experience:

Captain Lavelle said he flew two or three trips with Mr. Molin over a 12-month period. He stated that Mr. Molin's flying skills were excellent ... and that he had "hands of silk." However, he had one strange tendency: to be very aggressive on the rudder pedals. Capt. Lavelle stated that during a climb-out in a B727, while the airplane was "dirty with flaps 5 degrees," Mr. Molin stroked the rudder pedals "1-2-3, about that fast." Captain Laval thought they had lost an engine. Capt. Lavelle asked him what he was doing, and Mr. Molin said he was leveling the wings due to wake turbulence. Capt. Lavelle stated that Mr. Molin never leveled the wings and his actions created yawing moments on the airplane ... he did not recall the wings moving, but experienced "sideload, sideload, sideload." [ASW note: On the B727 flying at 250 knots, the rudder pedal has a maximum travel of 1.3 inches under 50 pounds of force to move the rudder 7�. For the A300 at the same speed, the rudder pedal has the same 1.3 inches of travel, but it moves under less force - 32 pounds - and the rudder will deflect up to 10�]

After they cleaned up the airplane, they discussed it further. Mr. Molin told him he was leveling the wings as per the AAMP. Capt. Lavelle told him it was quite aggressive and that it didn't really level the wings. They talked about the AAMP. Mr. Molin insisted that AAMP gave him directions to use rudder pedals in that fashion. Capt. Lavelle disagreed, and said he thought the use of rudder was, according to AAMP, for use at lower airspeeds. He disputed Mr. Molin and told him to be less aggressive and more coordinated using rudder.

Capt. Lavelle said that on two subsequent occasions Mr. Molin modified his wake turbulence maneuver to comply with his wishes. Mr. Molin used rudder during these encounters but did not go to full stop ... He was still very quick.

Capt. Lavelle said ... when he is the non-flying pilot, he follows along on the rudder pedals. He felt Mr. Molin's inputs on the rudder pedals during that first wake turbulence event. He said it is typical for him to fly with his feet on the pedals at critical times when the copilot is flying.

When asked why he remembered the event ... Capt. Lavelle stated that it was a very aggressive maneuver ... he said the wake vortex encounter with Mr. Molin was not much of anything, maybe just some choppy air. He thought that Mr. Molin may have been just responding to the choppy air. The ailerons were kept level and he used just the rudder pedals. (For other pilots' impressions of Molin's flying skills)

May 22, 1997: Letter by facsimile transmission from Capt. David Tribout, American Airlines A300 technical pilot, to Mr. William Wainwright, Airbus chief test pilot (extracts):

American Airlines pilots are attending a course called "Advanced Aircraft Maneuvering Program." It is an attempt to teach and reinforce basic aerodynamics and how to handle large transport aircraft in the event of an upset.

I am very concerned that one aspect of the course is inaccurate and potentially hazardous. As you can see from the handout pages attached with this letter, it states that at higher angles of attack, THE RUDDER becomes the primary roll control. The program infers that aileron application in these situations is undesirable since it will create drag caused by spoiler deflection. The instructor teaches that in the event of a wake turbulence encounter, recovery from stall, ground escape maneuvers, etc., THE RUDDER should be used to control roll.

The chance excessive use of rudder at high AOA might cause a spin or snap roll.

May 23, 1997: Response by Wainwright, by facsimile, to Tribout (extract): "I share your concern over use of the rudder at high angles of attack."

June 13, 1997: Wainwright, internal memorandum to Airbus colleagues (extracts):

Use of rudder: Rudder alone roll due to sideslip, thus its use at very low speed will result in large sideslip angles which will tend to provoke loss of control.

Training flight simulators:

A word of caution on the use of flight simulators to train for upset maneuvers. One must not expect training simulators to be accurate at the edges of the flight envelope.

The data package does not include dynamic maneuvers outside the normal flight envelope (e.g., only stall with 1 kt/sec deceleration rates are normally included) and the maneuvers that are verified are within the normal flight envelope.

Furthermore, simulators are not accurate in non-linear parts of the flight envelope. They are particularly inaccurate for large sideslip angles, and a pilot may draw the wrong conclusion from maneuvers involving use of rudder at low speeds.

June 27, 1997: NTSB interview with Capt. Laura Boswell, A300 pilot, regarding the June 12, 1997 Flight 903 upset experienced by another crew (extract):

She really liked AAMP training ... she said there was a lot of emphasis on using rudder and commented that its importance was emphasized in low speed situations, when the ailerons are less effective She said the A300 simulator breaks down each time she has done unusual attitude training.

June 28, 1997: NTSB interview with First Officer Lyman Chisholm, A300 pilot, regarding the June 12, 1997 Flight 903 upset experienced by another crew (extract):

He had attended the AAMP lecture. He said primary roll controls during unusual attitude recovery were a combination of rudder and ailerons. He said the simulator was extremely unstable compared to the airplane. He would guess that the programming is accurate but he said flying it is like a greased ball bearing. He said he would use the rudder in a nose high attitude to bring the nose down.

July 9, 1997: NTSB interview with Capt. Paul Hinton, A300 fleet manager:

He is aware of Boeing, Airbus, and McDonnell-Douglas concerns about use of rudder at high Alpha and said he shares the concern. He said he doesn't think the skills taught at AAMP are detrimental.

Hinton said rudder use was overemphasized in AAMP at first; now they are saying "don't forget the rudder" ... Hinton said he is concerned that the rudder is incredibly powerful and has a long arm to work on. He said they don't emphasize to pilots the concerns of overuse of rudder.

Aug. 20, 1997: Airbus, Boeing, Douglas Products and Federal Aviation Administration (FAA) comment to American Airlines on its AAMP (extracts):

Use of Rudder

The excessive emphasis on the superior effectiveness of the rudder for roll control vis-a-vis aileron and spoilers, in a high angle of attack, is a concern ... The pilots are left with the impression that it must be used for all high angle of attack situations ... additional rudder use information should be provided with emphasis on the consequences of inappropriate use of rudder. Although a simple rule about rudder usage cannot be stated, a more appropriate standard is to first use full aileron control; if the airplane is not responding, use rudder as necessary to obtain the desired airplane response.

Airplane Recovery from Upsets

Roll should be introduced only after exhausting the use of pitch axis controls ... Introducing roll angles at extremely high angles of attack creates sideslip and hence has the same concerns as rudder usage.

Use of Simulators

Artificially manipulating a simulator into an environment that is way beyond the valid engineering data creates a potential for negative learning ... Using a vortex flow in the simulator to induce an upset is a reasonable approach, however, inhibiting aileron inputs as apparently implemented in your training simulators, until the airplane has rolled through 90 degrees of bank will invariably result in large sideslip angles - probably outside the range of valid aero data. Additionally, without any aileron effectiveness during the first 90 degrees of roll, the pilot will probably use rudder in an attempt to roll the airplane erect.

Oct. 6, 1997: The American Airlines response from Capt. Cecil Ewell, chief pilot:

Use of Rudder

Let me say this one more time, we do not advocate the introduction of large sideslip angles when flying at high angle of attack [emphasis in original]... In four different sections of the AAMP, emphasis is focused on the fact that when the airplane is not responding to aileron and spoiler control, you should use smooth application of coordinated rudder to obtain the desired roll response.

Airplane Recovery from Upset

We clearly disagree totally with your position ... At American Airlines, we teach out pilots to fly the airplane first using primary flight controls. If unloading with elevator does not generate an adequate nose down pitch rate, then we will not hesitate to roll the lift vector off the vertical to generate the required nose down pitch rate. This procedure will work on all of our aircraft. Any delay in initiating the roll [if required] could lead to a very tenuous situation.

Use of Simulators

Initially, inhibiting aileron input response on the vortex model simulation was a necessary compromise to achieve both realism and the desired learning objective. However, this does not result in large sideslip angles as you suggest. On your next visit to our Flight Academy, we will be pleased to show you the Beta readouts during this event. [ASW note: this statement is the invitation American officials claimed in the Flight 587 crash hearings was made for Airbus to come and examine their simulator programming.] The AAMP modeling and training in our simulators focuses on maintaining the airplane inside its flight envelope regardless of attitude. It is our belief that the fidelity of our simulators is reasonably good ... We do not accept your statement that we are "manipulating a simulator into an environment that is way beyond valid engineering data."

Dec. 19, 1997: American distributes to its pilots an unusual attitude video for home study. It contains language emphasizing smooth application of the rudder with small applications for coordinated use. As a direct result of concerns expressed by the manufacturers about the emphasis on the use of the rudder, an additional segment was added to the video explaining proper use of coordinated rudder and emphasizing small, smooth inputs.

Aug. 12, 1998: Airbus submission to NTSB regarding AA Flight 903 upset (extract):

Although a simple rule about rudder usage cannot be stated, an appropriate standard is to first use full aileron control. Then, if the aircraft is not responding, use rudder as necessary to obtain the desired airplane response.

Sept. 28 - Oct. 2, 1998: Airbus Industrie, notes from 10th Performance and Operations Conference summarizing difference of opinion (extracts):

Differences of opinion on stalling:

The training being given in the airlines ... to recover from excessive nose-up pitch attitudes emphasized rolling rapidly towards 90� of bank. This is fun to do, and it was not surprising to find that most of the instructors doing the training were ex-fighter pilots who had spent a lot of time performing such maneuvers in another life.

Differences of opinion on use of rudder:

Large airliners, with all the inertia they possess, are not like fighter aircraft. Based on our experience as test pilots, we are very wary of using rudder close to the stall. It is the best way to provoke a loss of control if not used very carefully, particularly with flaps out.

We finally got the training managers to agree to play down the use of rudder in their existing courses. But we do not say, never use the rudder at low speed. We say that, if necessary, the aileron inputs can be assisted by coordinated rudder in the direction of the desired roll. We also caution that "excessive rudder can cause excessive sideslip, which could lead to departure from controlled flight.

But why did we have so much difficulty in convincing the training pilots that it is not a good idea to go kicking the rudder around at low speed? Their reply was always the same: but it works in the simulator!

Differences of opinion on use of simulator:

We manufacturers are very concerned over the types of maneuvers being flown in simulators ... Simulators, like any computer system, are only as good as the data that goes into them.

The complete data package includes a part that is drawn from actual flight tests, a part that uses wind tunnel data, and the rest, which is pure extrapolation. It should be obvious that conclusions about aircraft behavior can only be drawn from the parts of the flight envelope that are based on hard data. This in fact means being not far from the center of the flight envelope, the part that is used in normal service. It does not cover the edges of the envelope. I should add that most of the data collected in flight is from quasi-static maneuvers. Thus, dynamic maneuvering is not very well represented.

May 1, 2000: AAMP training course material (revised):

Aerodynamic Definitions

The effectiveness of the rudder as a roll control will increase with increasing AOA. [ASW note: compare to Jan. 1, 1997, entry above and note that the second sentence has been deleted, thereby deemphasizing the original emphasis on rudder use]
Pilot Response to Wake Turbulence

Do not apply any back pressure on yoke at more than 90� of bank. ROLL FIRST - THEN PULL.
High AOA - Coordinated RUDDER [ASW note: again, compare to Jan. 1, 1997, entry above and note the new emphasis on use of rudder coordinated with use of ailerons]
Nov. 12, 2001: Crash of Flight 587 following a wake turbulence encounter, in which and rapid and opposing rudder pedal movements are captured on the digital flight data recorder (DFDR).

Nov. 15, 2001: A300 Captain Paul Sulovski interviewed by investigators. Sulovski flew the accident aircraft the day before, without incident. Sulovski commented on his AAMP training:

He had two nose-high scenarios ... on the extremely nose-high, he used rudder to get down to horizon. Unload the wing and use max power at pilot's discretion. Use full flight control deflection if needed to keep blue side up. There are no flight control limits. [ASW note: Capt. Sulovski is expressing the widespread belief that below max maneuvering speed, VMAX, pilots can use the full authority of their flight controls without endangering the structural integrity of the aircraft.]

Nov. 16, 2001: A300 First Officer Walter Gershoff interviewed by investigators. Gershoff knew Capt. Ed States (captain of Flight 587), and recalled his AAMP training:

At least twice on [the] A300 he had had upset training. It was something he had also seen in the Air Force ... training at AA had taught him to be gentler in an airliner.

Jan. 15, 2002: Investigators interview company management, training pilots and check airmen associated with A300 operations and training. Extracts below of those interviews touch on fidelity of simulator training and the definition of maneuvering speed:

Interview with Capt. Aubrey Landry, manager of flight training.

Q: What do you base your data on when you reach those attitudes over 60� bank or 90� bank or 130� bank?

Landry: It's extrapolated data, I would assume.

Q: How can you say [the pilot] is trained to proficiency when you don't have data?

Landry: I don't think we've ever said that the airplane is going to respond the same way. I think you have to go back to the genesis of this program ... this started with the Roselawn crash, where we had some pilots who got into this upside down situation ... it appears that they did the instinctive thing and pulled themselves right into the ground. And so, based on the fact that we might have pilots who ... given the same situation, would end up the same way, I don't care where the data came from because the goal there is to teach that guy ... and whether the simulator responds exactly the way the airplane did or not is a moot point, because we have no other way to teach him that.

Q: Upon leaving the sessions, is the pilot left with the perception that in the real world, real airplane, if he were to find his airplane in ... 80�, 90� bank, that he would have to use his rudders instead of rolling it with just ailerons only?

Landry: The message that we try to get across ... is that you do everything you can with the ailerons and then you use the rudder as necessary ... and, particularly in a case where the airplane's gone beyond 90� of bank and the nose is going to be following quickly now, [if] ailerons are not getting you where you need to be, then some measure of rudder is called for, certainly.

Interview with John Cook, American Airlines A300 simulator pilot, responding to questions about the fidelity of the simulator and the upset recovery techniques stressed in the training:

Cook: That simulator doesn't - you can increase the bank angle to an extreme amount, and the nose won't fall until you put in more rudder than would probably be required on the aircraft.

What I see is most pilots put in the correct amount of bank angle and rudder ... and then I have to - I'm sitting in the instructor seat ... and I just in a very calm voice tell them, more rudder, more rudder. And then I go through again that I don't believe that the aircraft without some type of structural problem would require that much rudder.

Q: And the lack of enough rudder does what on, say, the nose-high [scenario]?

Cook: It will not bring the nose down. But I never teach, you know, "Put on all the rudder" or anything like that. I always teach coordinated flight. That's the way I learned.

Q: Since you've flown both the B757/767 and also the A300, could you give us a sense of comparison between these two airplane types in terms of rudder effectiveness in unusual attitude recoveries?

Cook: In the simulator?

Q: Uh-huh.

Cook: The Airbus requires more rudder - [in] the simulator. Now, that's only during that pitch-up maneuver the way it's programmed. If they get nose-high in another situation, then it reacts more like I believe the aircraft would.

Interview with Walter Goff, American Airlines simulator pilot instructor (no actual flight time on Airbus aircraft):

Q: Have you ever seen in upset maneuver training excessive use of rudder by a student?

Goff: No, I haven't.

Q: Would it be fair to say they'd probably use less rudder than they would more rudder as a general rule?

Goff: I'd say as a general rule they probably need a little more rudder. They start in with the aileron, and the rudder catches up a little bit later generally.

Q: If I gave you the latitude right now today of making changes or developing a better [AAMP] program or changes to the current one, what could you offer me as suggestions?

Goff: There may be something to be done about the simulator - probably the feel of the sim when it goes into a pitch-up.

Q: What do you feel would be a little bit better?

Goff: Once you push the button to insert a pitch-up, it holds it in there a little bit too long. It takes a little bit of time and a little bit of effort to get it out of that nose-up attitude. And what causes that I don't know. Somehow they do it, enter it in the sim.

Q: You said you teach them to feel the rudder input for the recovery from the unusual attitude ... how do you know if they put too little or too much rudder [into that] in relation to the rudder?

Goff: Usually it's the smoothness of the recovery or the lack thereof. If they don't put in enough, it's very rough or it's very sloppy on this recovery.

Q: Okay, can you feel side loads in the sim when that happens?

Goff: Up to a very small point. It moves a little bit, but not a whole lot.

Interview with Capt. Paul Railsback, American Airlines, managing director of Flight Operations Technical (type-rated on A300, among ratings for nine transport-category aircraft):

Q: And no caution has ever been provided you by Airbus regarding any flight control limitation for any speed?

Railsback: No.

Q: But would it be your opinion that if the rudder load limiter were working properly, that if you were still below the VA speed, that you should be able to fully, in the rudder's case, displace it the limited amount and still stay within structural limits?

Railsback: Yes.

Q: Do you know what the term "doublet" means?

Railsback: Yes, I do.

Q: Did you know it before the accident or after the accident?

Railsback: After the accident.

Q: Welcome to the group ... if you were to have neutral rudder pedals, you were below VA speed, and you put in full deflection of the rudder, followed by an opposite full deflection of the rudder, that that might also apply below the maneuvering speed to protect the airplane?

Railsback: I would expect it to.

March 8, 2002: American Airlines A300 Operating Manual Bulletin, re Upset Recovery/Unusual Attitudes (extract):

If normal control is ineffective, careful rudder input in the direction of the desired roll may be required to induce a rolling maneuver recovery.

CAUTION Only a small amount of rudder is needed. Too much rudder applied too quickly or held too long may result in a loss of lateral and directional control.

April 23, 2002: In response to an NTSB query about how simulators are approved for use in unusual attitude training, the FAA responds: "The maneuvers described as unusual flight attitudes (extreme pitch and roll angles) are not contained in the ACs (advisory circulars) for simulator qualification and the aircraft manufacturer provides no flight test data for these maneuvers. Without flight test data to validate a maneuver, the NSP (National Simulator Program) is not able to assure that the simulator is properly programmed to replicate the respective airplane throughout these maneuvers."

The aerodynamic appreciations of the AAMP program aside, there is a school of thought that simulators cannot substitute for airborne training, because the physiological sensations are part of the exercise and cannot be simulated. For the kind of dynamic training that is needed, adherents of this school of thought maintain that a high-performance light jet is the best teaching tool.

If actual flying is superior to classroom instruction and simulator time, perhaps there is a need for a specialist organization to which airlines could send their pilots, perhaps on a schedule of once every three years, for in-flight unusual attitude practice. Pilots could be exposed to dynamic flight maneuver at lower airspeed and high AOA. With the confidence and quick analytical skills gained from such training, two pilots together at the controls, and with the adrenaline pumping, needn't find their first experience with an unusual attitude to be their last.

'Hands of Silk'
Other pilots' recollections of F.O. Sten Molin's airmanship skills:

Nov. 17, 2001: A300 Capt. Louis Merz interviewed by investigators. Merz had flown with both Capt. Ed States and F.O. Sten Molin, and had last flown with Molin on Nov. 8. [Merz] ranked the F.O. as an "8.5 out of 10."

July 15, 2002: B757/767 Capt. Richard Salomon interviewed by investigators. Salomon had flown with Molin on the B727:

He recalled one landing in Miami while flying with Molin. There was a storm on the far end of the runway when they landed. Molin touched down on the dry end of the runway with some pretty good winds ... he got a great landing out of it. The storm just popped up out of nowhere. Captain Salomon said he had his feet on the rudder pedals that time and felt that rudder use was unremarkable ... Capt. Salomon said he never remembered Molin using rudder or not using coordinated rudder when he was hand flying the airplane during turns.

July 24, 2002: A300 Captain Robert Marinaro interviewed by investigators. Capt. Marinaro had flown with Molin:

Molin's rudder management and use was normal, [he] did not have a "quirk" about the use of rudders.

Aug. 28, 2002: First officer Peter McHale interviewed by investigators:

Over a 3-year period he flew with him a lot. McHale characterized Molin's flying abilities as a "good stick" [meaning] he had a good sense of concentration, always on altitude and never recalled a bounced landing. [Molin] put the airplane where he wanted it. [McHale had been the second officer during the B727 incident involving Capt. Lavelle and Molin] McHale said he did remember that Lavelle definitely had a discussion about a piloting issue ... with Molin. He was not privy to the conversation.

On another occasion, he remembered ... they got into the wake of a B737 while Molin was flying and he made the decision to go around ... There was a heavy jolt and the nose pitched. There was no discussion or hesitation. "I'm outta here." Sten made a fast decision to apply max power to climb out and go around (see cockpit voice recording from the Flight 587 accident, ASW, Nov. 4).

--------------------------------------------------------------------------------