By John Croft/Aviation Week & Space Technology

Preliminary accident reports point to a mechanical problem that may have caused one side of the aircraft's split elevator to move trailing edge up, despite input from the control wheel. The action forced the aircraft's nose skyward at least 26.8 deg. just after liftoff, as an astounded crew, thinking the plane's cargo had shifted rearward, invoked a series of roll, pitch and power "escape maneuvers" to bring the porpoising craft back to Mather for landing.

THEIR HEROIC ATTEMPT to fly the aircraft, reconstructed by the National Transporation Safety Board (NTSB) from flight data and cockpit voice recorders, indicated that the crew had knowledge or training experience in the latest industry tools designed to battle the decades-old quandary of how to control a swept-wing jet after an upset. Unfortunately, the outcome underscored the limitations of upset recovery training, a major safety thrust in the airline industry since 1997. One min. 58 sec. after takeoff, Kevin G. Stables, George Y. Land and Russell E. Hicks died when the freighter struck the ground during a chaotic but purposeful track from the departure end of 22L back toward the approach end. Rescuers arrived to a massive inferno in an automobile salvage yard just 2 mi. from the crew's apparent destination.

The obvious question is, would more or alternative training have made a difference, or was the aircraft simply not savable; and how far should training scenarios go in emulating the practically infinite combination of mechanical or control system failures that could occur?

These issues and more are at the heart of new efforts by the airline industry, researchers and regulators to figure out how to deal with upsets--events that are infrequent and improbable but often turn deadly in seconds if control of the airplane is lost. Making the process more difficult is a lack of feedback on such programs. Success often means there's no news, meaning decisive and correct crew actions prevented a mechanically able aircraft from reaching upset conditions, generally agreed to be more than 25 deg. nose-up, 10 deg. nose-down or more than 45 deg. bank angle.

The need for improvement was clear long before Emery 17. According to accident records, loss-of-control was the root cause of 37% of the fatalities in airline accidents between 1987-96, seconded by controlled-flight-into-terrain accidents. A working definition of a loss-of-control condition, according to a safety expert with the Air Line Pilots Assn. (ALPA), is "if you can no longer get the wings level or don't get the wings level." Airlines post-1996 adopted a standard approach to upset training, spurred predominately by the deaths of 132 passengers and crew of USAir Flight 427, a Boeing 737-300 that crashed near Aliquippa, Pa., in 1994.

Investigators found the crash was instigated by a rudder malfunction, and while the pilots acted promptly, they "could not be expected to have assessed the flight control problem and then devised and executed the appropriate recovery procedure," according to the NTSB. Following USAir 427, most airlines included recovery from bank angles over 45 deg. and atypical nose-high and nose-low attitudes into ground school and simulator training.

Upset recovery training is not required by regulation and therefore has no pass/fail criteria, though the FAA over the last four years has been preparing a rulemaking that will mandate certain select maneuver training. FAA officials are unsure when the first draft of the rules will be made public and would not comment on the content, other than to say the rules would affect training requirements in Part 121 subparts N and O in the regulations.

Industry began focusing more intently on the problem after the Gore Commission and the National Civil Aviation Review Commission in 1997 independently issued calls for a reduction in the rate of fatal commercial aviation accidents by 80% by 2007. The FAA and the airline industry responded by forming the Commercial Aviation Safety Team (CAST) to look at the causes of all airline crashes and to scientifically prioritize efforts to combat them. CAST recently ranked combating loss-of-control accidents as one of its highest priorities. During the same timeframe, aircraft manufacturers, airlines, pilot associations, regulators and others came up with a document and video, called the Aircraft Upset Recovery Training Aid, which explained the basic aerodynamics of swept-wing jets and offered generic methods for recovering from unusual attitudes no matter what the origin of the problem.

THOUGH THE AID HAS BEEN widely adopted in the industry, loss-of-control accidents continue--the latest possibly being an American Airlines Airbus A300-600 that crashed in New York in November 2001 after its tail broke off at some point following a wake turbulence encounter.

A recently completed NASA-funded study designed to probe the effectiveness and evolution of upset training was largely inconclusive, though several findings will help guide future analyses. The $500,000 effort evaluated 40 first-year airline pilots with varying experience levels in handling an aircraft near the edges of its operating envelope. The pilots were split into five groups ranging from those with no aerobatic or upset training to those who had received an upset recovery training course in a Veridian Flight and Aerospace Research Group customized Leartjet 25B modified to handle like a large transport aircraft (see p. 58). The object of the study was to test how each pilot would respond in an inflight environment to each of eight well-known crash scenarios that were deemed recoverable, including the icing-induced "aileron-snatch" that led to the crash of an ATR 72 in Roselawn, Ind., the 737-300 uncommanded rudder hard-over at Aliquippa and the 1994 wind shear and microburst encounter that upset a DC-9 near Charlotte, N.C. Scenarios also overlapped with many of the loss-of-control factors anonymously reported by pilots from 1987-95 to NASA's Aviation Safety Reporting System (see bottom graphic).

The results were at the same time promising and discouraging: The new hires were well prepared to handle problems like wind shear or nose-low spirals, possibly because they had received one or more classroom and simulator sessions on how to deal with such situations or because the recovery techniques were straightforward. However, most of the pilots, regardless of background, did not "recover" from six of the eight more difficult scenarios where the recovery actions developed by experts after the accidents may have conflicted with well-learned recovery techniques for similar, but fundamentally different encounters.

Key Dismukes, chief scientist for human factors at the NASA Ames Research Center, said the results, while preliminary, show that "training in upset recovery is not sufficient to enable the majority of pilots to deal with all situations they might encounter." The evaluation is one of more than a dozen NASA-funded human factors research projects underway at Ames alone. The work is being done as part of a five-year, $500-million NASA aviation safety program managed by the Langley Research Center.

While the pilots' performance demonstrated a need for training "to go further," Dismukes said the number of pilots tested was too small to determine how. Dismukes said about 400 pilots would have to be studied to obtain "statistically significant" results and more demographic information on the participants would be needed, such as what training they may have received in any previous flying jobs.

THOUGH HE ADMITTED it was speculative, Dismukes said the results highlight problems with the generic nature of today's upset training. For instance, he said many pilots instinctively handled stall-like situations using traditional FAA guidance: "Power out" of a stall with a nose-high attitude to minimize altitude loss. For two of the icing scenarios sprung on the pilots by surprise, however, Dismukes said such a technique "would not work." "We have to think about training pilots to deal with specific situations, not generic training," he said.

Dismukes' viewpoint will likely fall on deaf ears in the airline industry when the NASA study is published later this year. "How do you teach someone to fly a broken airplane?" said an airline industry source involved in the topic. "The answer is you probably can't." Instead, he said, airlines will likely continue to focus on generic upset training and on making new aircraft more error tolerant. For instance, many Airbus aircraft have envelope protection systems that prevent outside disturbances, or a pilot, from exceeding preset limits for angle-of-attack, minimum speed, thrust asymmetry and bank angles. The Boeing 777 has similar limits that warn of impending stalls, overspeed or high bank angle though pilots can override the "stops."

"We can't depend on the pilot being 'Sky King,'" the source said. "We have to stop the situation before it gets there."

Investing in envelope protection for new aircraft is one of many flight safety recommendations CAST distributed to airlines last month as part of its plan to reduce accidents. "CAST is good at trying to get away from a response to the 'crisis du jour,'" said the source, noting that early advanced maneuver training programs born after high-profile crashes like Aliquippa and Roselawn may have been knee-jerk reactions and of dubious value because the aircraft were "questionable as to their airworthiness" at the time either due to a mechanical fault or icing. "If [the aircraft] doesn't respond in the predictable fashion because it's broken, then training has no benefit," he said.

For the American A300, it's too early in the NTSB's investigation to tell how the pilot and aircraft interacted during and after the wake encounter or if that was a factor in the loss of the tail. Industry sources reported that engineers have calculated that tail loads reached 1.8 times the limit load, or 20% beyond the 1.5 safety factor. John Cox, ALPA's executive air safety chairman, said the crash demonstrates a need to incorporate aircraft certification limitations into simulators and for additional studies on rudder usage for large aircraft with under-wing-mounted engines. "The training must be right or there can be serious consequences," Cox said.

That's a point widely accepted in the industry. Experts are concerned that unusual attitude simulator work could be conditioning pilots to incorrectly respond to upset events. That's because in extreme maneuvers simulators must extrapolate aircraft response when the aircraft is in a realm not covered by data from actual flight tests or wind tunnel tests. How or if that played a role in the Emery 17 accident is unclear. According to Emery officials, pilots were taught "escape" maneuvers--using bank angle and power to regulate pitch attitude--in ground school and if time permitted, as part of their 4- or 8-hr. yearly stints in DC-8 full-motion simulators.

Engineers at the Langley Research Center are working on expanding the accuracy of ground-based simulators during unusual attitudes. John White, project manager for the single aircraft accident prevention project, said engineers at Langley have generated wind tunnel data using a 5.5% scale model of a Boeing 757 for angle-of-attack and bank angles up to 90 deg. and yaw out to ±45 deg. Simulators generally are programmed with flight-verified data from the aircraft manufacturer, acquired either in the wind tunnel or from flight testing, that covers 20-25-deg. in angle-of-attack, ±10-20-deg. in yaw and virtually any angle in roll. For older aircraft like the DC-8, the valid ranges were likely more limited. White's team will next convert the wind tunnel data to a form that can be used by a full-motion, or Level D, simulator and next year will bring airline pilots to Langley to fly the simulator in upset conditions. Accuracy of the new data will be verified by comparing output of the simulator with flight data recorder output from past accidents and with data taken by a subscale radio-control aircraft.

WHILE EXPANDING the simulator flight envelope will begin to address the "negative" training problem, or learning an incorrect response due to an inaccurate simulator, White said it will not create the definitive environment for training because it's ground-based and therefore cannot generate inflight G-forces. Instead, White said, NASA is focused on visual cues to maximize the experience.

Inflight simulators like Veridian's Learjet 25B could be used to supplement ground-based simulators, though airlines are cold on the idea. "We don't ever want to go back to [inflight] training again," said the airline industry source.

Cox is a believer in inflight simulators, and argued that at a minimum, simulator instructor pilots in the airline industry should be exposed to such training in an appropriate aircraft. Cox, who said he has had a Boeing 737 "in some unusual places" as a pilot participant in the FAA and NTSB's wake turbulence flight tests in the aftermath of USAir 427, said the Learjet was "very representative" during similar unusual attitudes. As part of an FAA program that started this month and is supported by ALPA, about 2,000 commuter and regional airline pilots are slated to take Veridian's training over the next five years.

In the near term, however, no "silver bullet" has been proffered. While accident reduction is a prime motivator, so are the valiant efforts of crews like Emery's 17. "They went from a zero chance of survival to a damned good chance," Cox said. "It was superior airmanship."