When the French BEA released a partial cockpit voice recorder (CVR) transcript of the Air France Flight 447 accident in late May, pundits wasted no time unleashing pointed analysis implicating the A330’s crew. The Airbus crashed into the South Atlantic, killing all 228 people aboard. Indeed, the edited details of the BEA seemed to offer few other possibilities. To some experts, however, the report actually raised more questions than it answered, leaving many to wonder about the BEA’s motivation in choosing the items it publicized.
Certainly the most telling element of the report was data indicating that not only did the pilot flying lose control of the airplane shortly after the autopilot and autothrottles disconnected on their own at 35,000 feet, but that while the aircraft was falling at some 11,000 fpm, he kept the sidestick pulled back, holding the 452,000-pound aircraft in a full stall. Air France received a series of automated messages before the crash that suggest the aircraft began shedding automation due to the loss of valid airspeed indications related to on-going icing issues with the Airbus’s Thales-made pitot tubes.
The brief BEA transcript reported that everything appeared normal in the flight until the moment the crew prepared to deviate around some forecast thunderstorms about four hours into the flight. Capt. Marc Dubois, the senior pilot aboard, had just left the cockpit for a required rest break and most likely transferred command of the aircraft to David Robert, who had 4,500 hours of experience in the A330. Pierre-Cédric Bonin, the most junior of the trio, had logged fewer than 3,000 hours total flying experience.
Just seconds after Air France Flight 447 began to deviate around the storms, the autopilot and autothrottles shut down as the aircraft rolled to the right. The pilot flying–believed to be Bonin–reacted by pulling back on the sidestick and raising the nose of the aircraft. The stall warning sounded twice as indicated airspeed on the captain’s primary flight display dropped quickly from 275 knots to 60 knots. The standby display showed the same numbers. About 10 seconds after equipment began failing, the microphone of the pilot not flying recorded, “We’ve lost airspeeds,” then “alternate law […]” most likely a reference to the system of Airbus control laws that determines which portions of the aircraft’s operating envelope the computers protect at any given moment.
For example, under the normal law of ordinary flight, computers prevent the pilot from exceeding the critical angle of attack (AOA); they also guarantee high speed, pitch attitude, yaw, load factor and bank angle limitations protection. If the aircraft slips into the realm of alternate law, however, many of the A330’s protections disappear, leaving only low- and high-speed stability, load-factor limitation and yaw damping. This is why–if the crew did maintain significant backpressure on the sidestick as the FDR indicates–the aircraft could indeed have stalled once the critical angle of attack was exceeded.
An A330 check airman who requested anonymity told AIN, “In the event of dual [or triple] ADR [air data reference] failures [which seems to apply in this case], there is no low-speed stability either. The nose attitude is referenced to indicated airspeed [IAS] instead of angle of attack [AOA].” He added, “There would also have been an ECAM message stating ‘alternate law, protections lost,’ as well as an alerting chime. Some green indications on the PFD would have switched to amber as an additional warning. There would also have been additional messages about the air data loss,” he said. Imagine the task the pilots faced of just trying to decide which set of failures and backups to focus on first as the airplane was being tossed around in the night sky.
According to the BEA report, about 50 seconds after the first stall warning sounded, the Airbus began a climb at a rate of 7,000 fpm and then rolled some to the left and right as the recorded speed increased to 215 knots. As the stall warning sounded again, thrust was set to takeoff as the pitch attitude increased to nearly 16 degrees, where it remained until the aircraft struck the water.
Some 90 seconds after the automation began to click off, and as the captain re-entered the cockpit, all recorded speeds became invalid and the stall warnings stopped. At this point, angle of attack exceeded 40 degrees and the vertical speed was about 10,000 fpm down as the pilot flying made inputs to hold the nose of the aircraft up.
Shortly thereafter, the pilot flying pulled the thrust levers back to idle and reduced the angle of attack slightly, although the recording shows it never dropped beneath 35 degrees. At one point, both pilots were trying to add control inputs. Four minutes and 23 seconds after the autopilot disengaged, the A330 hit the water at a groundspeed of just 107 knots and a pitch attitude of more than 16 degrees nose up.
Make the Airplane Fly
The most troubling unanswered questions center around why the crew was unable to recognize that the airplane was not flying, but rather falling like a rock from 35,000 feet. If they did understand what was happening, why were they unable to take the required action to make the Airbus fly again? In the 2009 Colgan crash at Buffalo, neither member of that crew had ever experienced a stick shaker/pusher combination as the aircraft stalled, nor had they ever demonstrated a stall recovery from that altitude.
Although some media critics expressed shock at the skill of all these pilots, most airlines and corporate flight departments flying large transport aircraft don’t train much past that standard either because it’s not required. In fact, few large-aircraft crews have ever experienced a complete stall in the aircraft they regularly fly. During initial and recurrent training, the standard has always been to recognize the approaching stall and recover before the aircraft stops flying.
Stall recovery in training, as G550 pilot Steve Thorpe confirmed, is actually pretty routine. “You go up to 15,000 feet and perform the stall series. The recovery is full power, don’t lower the nose and you should be able to power out of the stall with minimum altitude loss,” he told AIN. How then, does a crew that has never actually stalled an A330, or a G550 or a Global Express, learn to recognize when their aircraft is actually stalled and recover if they’ve never tried it in training?
Robert Barnes, president of the International Association of Flight Training Professionals in Scottsdale, Ariz., suggests that it’s time to stop blaming pilots outright for failing to do the right thing in a highly unusual situation. “We should start asking ourselves if we are adequately preparing them for their jobs. Do they really have the knowledge, skills and competence required to fly an airplane or are they simply being trained to manage systems?
“I’m concerned that we’re no longer teaching people how to control the airplane at the most basic levels so they develop an instinctive understanding of how an airplane flies. Some of this could be due to aircraft or simulator limitations, but it could also be due to the classic business case needs analysis that concludes, for example, since there’s a low probability an airplane will roll into an unusual attitude, it’s not cost effective to train for it.”
Barnes adds that the very expression “unusual attitude” implies an unplanned, atypical event. “I’ve certainly had my share of surprises in airplanes that didn’t fit within standard operating procedures, and I’m sure most pilots who are nearing retirement age can say the same thing. How did we ever survive?” he wondered.
He suggests a knowledge of basic skills was invaluable to pilots who have successfully dealt with unusual emergency situations. He pointed to the 1989 United Airlines DC-10 crash at Sioux City in which the crew used differential power to control the aircraft after the uncontained failure of the number-two engine had disabled all hydraulic systems. “Or take Captain Sully, a glider pilot who used basic piloting and energy management skills to land his stricken Airbus in the Hudson River. There are many similar situations in which a basic understanding of aircraft control made it possible for the crew to respond in an appropriate way. In the old days, we called this airmanship.”
Did the Air France crew simply fail to fly the airplane, as some claim, or were they the victims of a training system that taught them to rely too heavily on computers right up to the moment the impossible overload occurred, like the HAL 9000 in Stanley Kubrick’s film 2001? No one questions whether or not the Air France crew met the certification requirements in place at the time they received their type ratings. But does the type-rating requirement on an Airbus, or any other large aircraft, go far enough into the actual handling characteristics of the aircraft–especially at high altitude–and especially when multiple computer failures occur?
Flight Safety Foundation president and CEO Bill Voss believes “Maybe the solution isn’t that terribly hard because this isn’t just an Airbus issue. I think we’ve failed to make the quantum leap in training required by the complexity of the airplanes we fly today. We still train like we fly DC-3s. We need to train for high-altitude failures. The new basics of flying an airplane demand that the [pilot flying] triage the airplane to keep it in the air when the automated systems start clicking off.”
Thorpe mentioned a quote by airline pilot Paul Kolisch in a Wall Street Journal article that spoke to the benefits of stall training today–or perhaps the lack of them–in swept-wing airplanes. “It’s a lot like synchronized swimming,” Kolisch said. “It requires a great deal of skill and execution, but in no way teaches you to swim across a river.” Thorpe mentioned a new FAA requirement that stalls be performed closer to the ground in the simulator for more realism.
If we’re going to perform stalls close to the ground because of the Colgan Q400 accident and perhaps high-altitude stalls and recoveries because of the Air France A330 accident, or work harder on complacency issues brought to light in the Turkish 737 accident in Amsterdam, we should also be wondering what other portions of the aircraft’s flight envelope or other computer gremlins yet unnamed are still waiting to jump out and bite us.
No doubt automation has made flying easier, especially when it comes to delivering the smooth ride passengers expect in jets. But has passenger comfort that demands the precision of these automated systems taken our eyes off the ball of the basic airmanship component? Were the Air France pilots simply overwhelmed with more flashing lights and chimes than they could grasp as a stormy night flight came unraveled? Few pilots hand fly their aircraft in training or on the line. Can we really afford that insulation from actual stick-and-rudder flying any longer? Many companies–airlines included–don’t want to spend the time and money to train pilots for those one-in-a-million calamities that could appear in obscure corners of the envelope. Maybe it’s time to re-evaluate that philosophy.