A recommendation to modify the Airbus A300-600 rudder travel limiter (RTL) to prevent excessive aerodynamic loads could be the first of more calls to improve the airplane’s rudder control system.
The National Transportation Safety Board (NTSB) issued a May 28 recommendation based on its findings that the RTL did not limit rudder movement as designed when an American Airlines [AMR] A300-600 involved in a 1997 stall event experienced rapid airspeed changes that outran the limiter’s ability to keep up. As a result, the tailfin was stressed beyond ultimate load, which eventually required replacement of the composite tailfin. Ultimate load is the 50 percent safety factor added to limit load, which represents the greatest force expected to be encountered in service. The incident airplane, operating as Flight 903 at the time, experienced rudder reversals as a consequence of four rudder inputs in opposing directions. Reversals, or rapid rudder deflections from one side to the other without stopping at the neutral position, can induce yawing that places great stress on the tailfin, rapidly building aerodynamic forces the fin was not designed to withstand.
Another American A300-600 operating as Flight 587 experienced a series of rudder reversals in November 2001, leading again to ultimate load and, this time, to fin separation. The airplane crashed into Belle Harbor, N.Y., killing all 260 aboard and five persons on the ground. It was one of the deadliest crashes in North American aviation history. NTSB sources close to the Flight 587 investigation tell ASW that additional rudder design recommendations are likely.
The Flight 587 crash has locked American Airlines and Airbus into a heated dispute. Airbus claims the copilot flying the aircraft mishandled the rudder system by imparting a series of rapid and inappropriate reversals. American counters that the reversals were the inevitable result of an overly sensitive and flawed rudder control system design (see ASW, March 29).
The NTSB recommendation focuses on the Flight 903 event, saying that in cases of rapid airspeed changes the RTL cannot keep up with the rate of change. Therefore, the rudder could travel beyond the 14.5� deflection limit established for an airspeed of 220 knots (the deflection limit decreases with an increase in airspeed). The Flight 903 stall upset involved an airspeed change from 190 to 220 knots in the span of some three seconds. That is an airspeed change rate of 10 knots per second. In the 165-220 knot speed range, the limiter can “handle” airspeed changes of 2.4 knots per second and restrict rudder movement. However, the 10 knots per second airspeed change in the Flight 903 event was a rate some four times faster than the rudder travel limiter could respond.
Because of this, the NTSB said, “The rudder was allowed to travel in excess of its RTL design limit for approximately 20 seconds.”
Therefore, the NTSB wants the Federal Aviation Administration (FAA) to require Airbus to modify the RTL so that it can respond “effectively to rapid airspeed changes.”
A footnote in the NTSB’s recommendation letter said the safety issues emanating from the Flight 903 event were “not a factor in the Flight 587 accident.”
In effect, the NTSB appears to be disassociating Flight 903 and Flight 587 as two distinct and different events. Airbus officials have publicly seconded this point of view, claiming in response to the NTSB letter that the RTL modification has “nothing to do” with the Flight 587 accident.
However, the acceleration parameter is not viewed by all the participants as the dividing line between the Flight 903 upset and the Flight 587 crash. In its statement reacting to the NTSB letter, American Airlines said its rudder training – revised after the Flight 587 crash – “addresses specific characteristics of the rudder travel limiter system … that this NTSB recommendation involves.”
“These limitations, along with others related to the sensitivity of the system, are precisely the concerns American addressed in its final submission and recommendations to the NTSB regarding the Flight 587 accident,” the airline’s statement said.
The Flight 587 accident is unique in the sense that some American Airlines A300-600 pilots have organized themselves into a group, and they have issued various papers and recommendations during the investigation outlining their point of view (see ASW, June 17, 2002). In response to this latest development, the pilot’s group declared, “As for the limiter failure being ‘separate’ from the AA 587 incident, this attempt to ‘parse’ this deficiency waters the eyes it is such a transparent obfuscation.” The A300-600 pilots group asserted that the investigation has uncovered evidence that the limiter failed twice on Flight 587, “including on the fifth rudder movement.”
One of the most contentious issues concerns the series of opposing rudder movements imparted in the space of some seven seconds on the Flight 587 accident aircraft. Airbus has contended that American’s upset recovery training predisposed pilots to make excessive use of the rudder.
This issue of pilot rudder use in the Flight 587 crash may not necessarily be separable from pilot actions in the Flight 903 upset. In that 1997 event, NTSB investigator John O’Callaghan wrote in the aircraft performance study that ailerons were coordinated with the rudder to arrest the airplane’s increasing bank angle, which approached 70�. O’Callaghan’s report said four large left-to-right rudder inputs were made before the first stall “which generated large rolling moments.” However, his report went on to say, “the initial continued right roll against left control inputs prior to the first stall … is inconsistent with the expected behavior of the aircraft and justifies use of the rudder to help stop the roll.”
This statement can be interpreted by the parties involved in two ways: (1) that the Flight 903 pilot’s use of rudder to recover was appropriate to the situation, as was that of the Flight 587 pilot four years later, which is the American position, or (2) that the rudder inputs in Flight 903 were conditioned by American’s training program on rudder use, with fatal consequences for Flight 587, as Airbus maintains.
American officials do not dispute that the Flight 587 pilot applied four opposing rudder commands initiated as he attempted to arrest the airplane’s increasing left bank during the second wake vortex encounter with a Japan Air Lines B747 that had taken off just shortly before Flight 587 roared down the same runway at New York’s John F. Kennedy International Airport.
But where American differs with Airbus is that the rudder system’s sensitivity made fine modulation of rudder movement impossible. The American argument is that the pilot flying, Sten Molin, attempted to use the rudder to assist in roll control but he got a 9.3� deflection in a split second, and within six seconds of his subsequent input attempts to right the airplane the tailfin snapped off.
Indeed, with 130-140 pounds of pressure on the pedals, pilots can achieve additional rudder movement. Airbus officials have referred to this as “elasticity,” implying not a design deficiency but a heavy-footed pilot deficiency.
Two particulars of the case suggest that the NTSB may have more to say about the A300-600 rudder control system as the Flight 587 investigation moves to its expected conclusion later this year.
O’Callaghan chaired the aircraft performance study in the Flight 587 investigation. His analysis of the A300-600 rudder control system concluded that the pilot can “override” the yaw damper at the rudder travel limits. The yaw damper is designed to attenuate directional oscillations, thereby suppressing the buildup of sideslip angle and resulting aerodynamic forces on the tailfin. O’Callaghan’s report provides an example of yaw damper override. If the RTL limit is 10� right, and the pilot pushes the pedal for 10� right rudder, and the yaw damper commands 2� of left rudder, the right10� right rudder command is reduced to 8�. However, the mechanical linkage is such that the pedal moves back slightly, which in this case gives the pilot another 2� of pedal travel. Therefore if the pilot maintains pressure, the resulting output command to the rudder will be 10� right, regardless of yaw damper input, according to the O’Callaghan report.
Data culled from the accident airplane’s flight data recorder (FDR), O’Callaghan reported, “suggest that this suppression of the yaw damper inputs at the rudder limits probably occurred during the accident flight.”
An NTSB source explained that the design allows for higher aerodynamic loads on the tailfin than would be the case in a rudder control system where the yaw damper cannot be overridden.
O’Callaghan’s team explored an alternative design layout, in which the pedals are limited separately. This architecture was dubbed a “pedal limiter” system. In this arrangement, the yaw damper cannot be overridden and there would be no additional pedal motion to push the rudder back to the limit. A computer simulation of the Flight 587 scenario showed a “noticeable” reduction in the lateral load factor on the airplane. Interpreting the graphic results in the report suggests about a 23-25 percent reduction of sideslip angle and about 20 percent in lateral G loading.
Having entertained an alternative design that reduces potentially dangerous aerodynamic loads on the tailfin, it seems unlikely that the NTSB would fail to recommend implementation.
For another thing, the NTSB commissioned an independent review of the A300-600 rudder system. Conducted by Dr. Ronald Hess, vice chairman of the Department of Mechanical and Aeronautical Engineering at the University of California, Hess concluded that the pedal/rudder sensitivity of the A300-600 was seven times greater than that of the Boeing [BA] B767, a comparable design. In comparison to the A300-600B2/B4 series aircraft, which preceded the A300-600 series, and at the 250-knot airspeed at which the Flight 587 accident occurred, Hess reported “the A300-600 exhibits over six times the yaw acceleration per pound of pedal input of the B2/B4 series.”
Hess concluded that the system’s sensitivity was conducive to pilot-induced oscillation (PIO), which he asserted “was evident in the moments before the crash.” The essential characteristic of a PIO is that the airplane’s control characteristics can cause the pilot’s attempts to restore stability to be out of phase with the aircraft’s response. Instead of dampening yaw excursions, as in the Flight 587 case, rudder pedal movements unintentionally were reinforcing them. The analogy may be the effortless synchrony required to maximize the height-gain of a child’s playground swing.
The French accident investigation body, the Bureau d’Enquetes et d’Analyses pour la S�curit� de l’Aviation Civile (BEA), countered that Hess’s analysis was flawed. “As far as we know, no studies have been undertaken on PIO on the yaw axis since the rudder is not a primary flight control,” the BEA said. Noting that the Flight 587 pilot used rudder inputs in an attempt to arrest the airplane’s left roll, the BEA said “even a light rudder pedal input may create a divergent output,” adding, “This system remains stable simply by using the control wheel with high gain.” This remark captures the essence of the Airbus position that if the pilot had kept his feet off the rudder pedals the accident would not have happened.
Since the NTSB commissioned Hess’s report, and Hess is a recognized expert in aircraft stability, control and PIO issues, it does not seem likely that the board will reject outright his findings and their implications for design changes.
These aspects of the rudder system design from the Flight 587 investigation indicate that the recent recommendation on the RTL emanating from the Flight 903 event represents an effort to put the airspeed change vulnerability to bed. The action could be interpreted as clearing the way to focus on the control sensitivity and yaw damper override issues brought to light in the Flight 587 crash investigation.
The accident pilot and American’s upset recovery training may not yet be absolved by any means, but neither may be the A300-600 rudder control system.