Now the leading cause of airliner hull losses and fatalities, loss of control is driving improvements in training to help pilots recognize and recover from aircraft upsets in flight.

Including Colgan Air Flight 3407 and Air France 447 in 2009, loss of control in flight (LOC-I) killed 1,841 people from 2001-10 in 20 commercial jet accidents worldwide. This compares with 1,007 fatalities in 17 crashes caused by controlled flight into terrain (CFIT), which was the biggest killer until the installation of enhanced ground-proximity warning systems (EGPWS) was mandated.

“CFIT was the number-one cause, but has been tackled by EGPWS and was down in 2001-10 while LOC-I was up,” says Sunjoo Advani, chairman of the International Committee for Aviation Training in Extended Envelopes (Icatee), organized by the Royal Aeronautical Society. “Stall is the number-one cause of upsets leading to loss of control. Pilots are well trained, aircraft have protection systems and yet we are still getting upsets. Why?” he asks. “Because loss of control in flight is rare, unpredictable and catastrophic—and pilots are not adequately trained.”

Icatee is in the final stages of developing new tools and guidelines for upset prevention and recovery training (UPRT) to address concerns raised by LOC-1 accidents including Colgan Air 3407 and AF447. Fifty people were killed in the Colgan Air crash, 228 in the Air France accident, bringing a sense of urgency to mitigating the threat. “The likelihood of a LOC-I accident is very low, but the likelihood of recovery from a significant upset is extremely low,” Advani says.

On Colgan Air 3407, the stickshaker activated on the approach to warn the crew of low airspeed and impending stall, but instead of pushing forward on the control column to lower the nose, the pilot pulled back, increasing angle of attack (AOA). This caused the wing to stall and the stickpusher to activate, to force the nose down, but the pilot pulled back on the column, fighting the stall-protection system's attempt to decrease AOA. The NTSB cited the probable cause as the pilot's “inappropriate response to the activation of the stickshaker.”

The final report on the investigation into AF477 will not be released before year-end, but in their July interim report French accident investigators noted that the copilot flying made continued nose-up control inputs after the stall warning sounded. “Neither of the copilots had undertaken any training in manual handling of the airplane on approach to stall or on stall recovery at high altitude,” the investigators said.

One contributing factor is the increasing automation in aircraft. As pilots become system managers, their manual flying abilities can degrade over time. Also, up through the 1990s, airlines could draw on ex-military pilots with all-attitude/all-envelope flying experience. But that pool is drying up and most commercial pilots now come through the general aviation pipeline, where training does not include exposure to full stalls.

“We moved away from aerobatic training in the 1980s,” says Advani. And the training pilots do receive can teach the wrong response. Until recently, the accepted rule in stall training was to recover with less than 100 ft. of altitude loss. This often required back pressure on the stick to keep the nose up, and not forward pressure on the controls to reduce angle of attack.

“It was drilled into them to minimize altitude loss to avoid hitting the ground—to pull, and not push,” he says. “The biggest fear of pilots is hitting the ground, but it really should be stall and loss of control. We have to get rid of that rule.” In 2010, the FAA revised its test standard to allow altitude loss during a recovery.

Icatee has identified limitations imposed by four unwritten assumptions built into today's pilot training. “We assume the aircraft is within its normal operational envelope and not in agitated flight, that the pilots have good situational awareness, that licensing training has provided them with good aircraft-handling skills and that their psychological and physiological reactions are predictable and reliable,” says Advani. If one or more of those assumptions is violated, specialized training is required, the group concluded.

Aircraft are normally operated to their maximum lift-to-drag ratio, well before the stall, and back again. “Pilots are trained to go to the initiation of stall in a light aircraft, but never into a fully developed stall,” Advani says. “There are two things wrong with that: They are not in a representative aircraft and they never train in a full stall. How will they ever know how to react? It is blatantly obvious that you push the nose down, but why does it not happen, again and again?”

Icatee believes LOC-I happens because those assumptions on which training is built no longer apply. In an upset “the aircraft is outside its normal operating envelope and agitated,” he says. “There are many cues going on—visual, aural, tactile, vestibular, g-loading. In training the pilot experiences them one at a time. Now many things are happening at once, and we have not trained the pilot to deal with that and correlate all the stimuli accurately,” he says. “The Colgan Air pilot did not recognize the stall and fought against the automation. That shows the limitations of the training,” he says. “We want to train pilots to be prepared for the unexpected, and not to panic. Pilots are even overpowering the aircraft protection systems—we need to teach them how to use them.”

An upset is defined as a pitch attitude beyond 25 deg. up or 10 deg. down, a bank angle exceeding 45 deg. or inappropriate airspeed. The normal operating regime within these limits represents just 4.9% of the “all-attitude threat envelope,” which extends out to ± 90 deg. pitch and 180 deg. bank (see graphic). Commercial pilots are trained to a maximum 11.1% of the threat envelope (± 30 deg. pitch and 60 deg. bank), but actual upsets such as Colgan Air “exceed training limits significantly,” says Advani.

Icatee has concluded that integrated academic, inflight and simulator training is required to teach pilots the correct recovery technique, which is to reduce angle of attack to unload the wing at the first signs of an approaching stall. “A modern simulator can do almost 100% of normal operations, but upset prevention and recovery training cannot be done in any one medium,” he says. Academic training, introduction to generic recovery skills in aerobatic-capable training aircraft and development of type-specific skills in flight simulators are required to provide pilots with a mental model that prepares them to handle unexpected events safely, the group believes.

A key goal of UPRT, as defined by Icatee, is teaching the pilot to recognize the signs of stall, including possible buffet and reduced lateral control and stability. “Our focus is not only on recovery: Prevention is 90% of the goal,” says Advani. “Pilots need to learn what a full stall feels like.” An aircraft's behavior in a stall can surprise many pilots, delaying their response and recovery. “How a pilot deals with startle is not predictable or reliable,” says Advani. “We need to give the pilot exposure to a challenging event during training so they are better able to handle surprise in real life.”

Icatee recommends that, before receiving a commercial or multi-crew pilot license, pilots should complete the UPRT academic course and learn the aerodynamic issues of stall recognition and recovery through inflight experience in an aerobatic-capable aircraft. Recurrent line-oriented training in a qualified simulator would then reinforce the appropriate procedures throughout the pilot's working life.

In-aircraft experience will add to the cost of training pilots. “There is definitely a cost, but it should be outweighed by the benefit,” says Lou Nemeth, chief safety officer of training and simulation company CAE. “It's about $3,000-4,500 for a 2½-3½-day course; but put into the perspective of ab initio training costing $100,000-150,000, you are increasing that by 2-4% to tackle the single largest cause of hull loses and fatalities.”

Icatee is likely to recommend that, after receiving UPRT during licensing training, airline pilots repeat the in-aircraft experience every five years, says Paul Ransbury, president of Phoenix-based Aviation Performance Solutions (APS), which provides upset-recovery training in aerobatic-capable Extra 300s. “Recurrent training is important. At APS, the typical [corporate] flight department sends pilots back every two years, but some only do it once in their career.”

For Icatee, repeating the in-aircraft training every five years to avoid skill degradation “is the best solution, but may not be economical or feasible,” Ransbury admits. For that reason, in-aircraft training is likely to remain optional for airlines. “The issue then becomes doing it in the simulator alone,” says Nemeth.

“One thing the simulator can't do is to address the psychological and physiological aspects of upset,” says Ransbury. “It's better to do it in the air first, then the simulator. If you try to train in the simulator first, then the aircraft, pilots are unsettled by the reality—the startle factor—of the upset. That's the value of in-aircraft training, to prepare for the simulator.”

Nemeth says today's Level D full-flight simulators are capable of providing effective and appropriate training to reinforce the pilot's recovery skills. “Current simulator technology is taken into account in the recommendations,” he says. But simulators will be improved to help with UPRT. Upset recoveries in the simulator “can pull more gs than the aircraft can handle, so we need to improve pilot cueing,” says Advani. Stall modeling, the effectiveness of buffet simulation and role of motion cueing are all being looked at. Instructor-operator stations will be modified to provide better feedback on the flight envelope, display any inappropriate control inputs and provide a wider variety of upset scenarios so pilots can still be caught by surprise.

Instructor training and standardization are pivotal to the Icatee recommendations. Both on the aircraft and in the simulator “we see and worry about instruction that is not accurate, and which could have a devastating outcome. We must set standards for how to train the trainer,” says Ransbury. “There has to be instructor standardization to transfer the correct knowledge to pilots,” says Advani.

Formed in the wake of Air France 447, Icatee will shortly begin delivering its recommendations to the International Civil Aviation Organization. The academic materials, an update to an upset recovery training aid published in 1988, are to be ready early next year. There will be sections for the pilot, instructor, training provider and regulator. UPRT simulation requirements, focused on high-end devices, are to follow by mid-2012.

As release of the AF447 final report approaches, there is a sense of urgency in industry. Emergence of LOC-I as the major cause of airline fatalities has exposed “several big holes in training,” says Advani. “Many of the 230,000 pilots flying today have insufficient training in how to recognize, avoid or recover from upsets. That's the reason for the urgency.”