Understanding Air France 447 (26 page)

Manual resetting of the stabilizer trim may also have been required. In this case, full nose down control was only commanded once, and then only for two to three seconds, until the airplane was passing below 10,000 feet when two other short bursts were input.  Once the stabilizer reached the full nose-up position, it stayed there (or very near there) for the remainder of the flight. 

In roll, the flight controls were in Direct Law. But the airplane was fully stalled, essentially falling like a maple leaf. It is amazing it did not flip over or enter a spin. In the flight recorder tracing below, notice the roll angle (on top in blue) verses the sidestick and aileron position (brown and two shades of blue at the bottom).

It is clear to see the direct relationship between sidestick position and aileron position (demonstrating the Alternate 2 Law roll mode: Direct) and long periods where the sidestick and the ailerons were at full deflection. At the same time, the bank angle was oscillating, showing the relative ineffectiveness of the ailerons. The inboard ailerons are the only ones used in this phase of flight. Outboard ailerons are locked out for pilot inputs when the airspeed is above 190 knots and flaps are up.

During this same time, the yaw damper was driving the rudder fighting left and right yaw motions, which may have been the only thing keeping the airplane close to upright.

Don't Forget to Fly the Airplane

In January 2013, the FAA released a Safety Alert for airline operators titled
Manual Flight Operations
,
⁴²
reminding them that "maintaining and improving the knowledge and skills for manual flight operations is necessary for safe flight operations."

The alert continued: "Modern aircraft are commonly operated using auto-flight systems (e.g., autopilot or auto-throttle/autothrust). Unfortunately, continuous use of those systems does not reinforce a pilot’s knowledge and skills in manual flight operations. Auto flight systems are useful tools for pilots and have improved safety and workload management, and thus enabled more precise operations. However, continuous use of auto-flight systems could lead to degradation of the pilot’s ability to quickly recover the aircraft from an undesired state."

In other words: pilots must continue to hand fly the airplane on a regular basis in order to keep those skills sharp, so that when the automatic systems are not working, they are able to!

The alert urges airlines to incorporate emphasis of manual flight operations into both line operations and training (initial/upgrade and recurrent), and develop policies that ensure there are opportunities for pilots to exercise manual flying skills at appropriate times.

Pilots need to be ready to take over not just in case of failure, but when the automation is not maneuvering the airplane as desired. This may be due to a programming error, that the automation is being too gentle when a faster response is desired, or that it is just easier to click the autopilot off for a few seconds to get what you need. I have delivered the following admonition to my students many times: "If it isn't doing what you want it to, click it off and make it do it!"

However, the more a pilot uses the automation, the less he and his fellow crewmembers may be willing to do without it. In fact, the less likely he may be
able
to do without it.

Just like driving a car with an automatic transmission for decades might cause a driver to lose their skill with a stickshift and clutch, a pilots skills at flying solely on basic instruments can also suffer from non-use, even though he may have been quite skilled at one time.

A study of 30 airline pilots from a major US airline found significant deterioration of raw instrument flying skills over time while flying highly automated airplanes.
⁴³

A wide range of pilots were included in the study, both captains and first officers of wide body and narrow body aircraft with in-seat experience between two and 16 years.

The study had the pilots perform five maneuvers solely by reference to basic instruments (without flight directors, autopilot, autothrust, or moving map display): takeoff, takeoff with engine failure at V
₁
, holding, ILS approach, and a missed approach.

The average grade for each maneuver was below the ATP (Airline Transport Pilot) standards and closer to the instrument rating level. The poorest performance was seen on the holding maneuver. There was no significant performance difference between the pilots of narrow body airplanes, which tend to have more frequent take off and landing operations, and those of wide body airplanes, which tend to have had more experience yet fly fewer takeoff and landings.

The study revealed that the pilots’ actual performance was lower than their expected level of performance in a pre-assessment survey. It may also be worth considering that only pilots who were fairly confident in their abilities might be willing to participate in such a study.

It is clear that when manual instrument flying is required, pilots who are less competent at manual instrument flying will require more concentration to maintain basic control of the airplane. Pilots who are more competent at those skills can spend less attention and cognitive function simply flying the airplane and more time on the problem at hand.

The study highlights the fact that even pilots who were extremely competent at hand flown instrument flight at one time, will see their skills deteriorate over time with non use. For pilots who spent a large portion of their career with advanced automation, those raw skills may not have been well developed to begin with.

Despite the time and effort required in training for pilots to become proficient with the complex automatic systems, there is still some hand flying done in training and checking. However, manual flying in the simulator for hand-flown maneuvers such as stalls and unusual attitude recovery is usually preceded with “OK, now we’re going to do stalls or unusual attitude recovery.” That, of course is not how the real-life crisis situations develop.

One maneuver that comes close to the transition from fully automated to fully manual (or nearly so) is the PRM approach breakout maneuver. A PRM approach is a Precision Runway Monitored approach used at some airports that have approaches on closely spaced parallel runways. Because the approaches are closely spaced, a dedicated air traffic controller monitors the approach and will give breakout instructions to the aircraft on one approach should an aircraft on the adjacent approach start to stray toward them. These breakout maneuvers are practiced in the simulator so that crews are able to perform them at the few congested airports that use them. Pilots will pre-brief the event, go into the simulator, and begin the approach knowing exactly what is to take place.

The breakout instructions will include a turn and may also include a climb or descent instructions. The maneuver is to be flown with the autopilot off to ensure that the clearance is complied with promptly. Therefore, the clearance requires the pilot to transition from a fully automated approach configuration, turn off the autopilot and flight director, initiate a turn to a given heading and comply with an altitude clearance. The maneuver requires prompt compliance but only normal bank angles and vertical speeds. It is even suggested that the airplane’s configuration (landing gear and flaps) remain unchanged until established on the new heading and workload is reduced. During this maneuver, the autothrust may remain on, and will automatically maintain the selected speed, further reducing the workload.

However, this simple maneuver proves to be surprisingly challenging. My own informal survey finds that very few found the maneuver easy, and at least half needed to repeat the maneuver one or more times. Overcontrolling pitch or roll, and difficulty attaining and holding a pitch attitude, heading, or altitude were cited by instructors as common issues. Yet, it is only a taste of a real-life failure situation when the automation, including the autothrust, suddenly and unexpectedly shuts off.

Pilots, ask yourselves: How is your instrument scan? How well can you hold and make changes to heading, altitude, and airspeed without the flight director?

Will your hand flying skills be ready on your next flight when it is dark and bumpy, you are half tired, and things stop working?

Automation is great, but do not let yourself become automation dependent. No matter how good you were at one time, those skills have been shown to degrade over time if not practiced.

When hand flying, every pilot should deliberately practice precisely flying the attitude, airspeed, heading and altitude. Chances are, when the automation fails and manual flying is required, it will not be on a clear blue day.

Safe Harbor Concept

The Air Safety committee of the Air Line Pilots Association (ALPA) has introduced the concept of establishing a “safe harbor” technique to cope with “automation exceptions.”

The committee recognized that even highly skilled instrument pilots lose perishable instrument flying skills as a result of continual reliance on automated systems. Also, new entrant pilots trained from inception utilizing glass cockpit instrumentation do not develop deeply embedded skills necessary for basic attitude instrument flying (a method of instrument flying based on setting and adjusting attitude and power setting, without the aid of any automation).

The concept of an automation exception involves not just failures but any situation where the automatic systems no longer provide appropriate guidance. This may include such situations as:

• Pitot-static errors including: icing (AF447), insect debris, mud dauber nests, taped over static ports, water ingestion, hardware failures
  
• Software conflicts
  
• ATC requests, traffic avoidance maneuvers, PRM (Precision Runway Monitored) approach breakouts, go-arounds, etc.
  

In addition to these factors, other factors in loss of control incidents include environmental, systems induced failures, and the greatest contributor: pilot induced issues.

It is recognized that years of automatic operations or practicing hand flying but still following the flight director, have made it almost impossible for the pilot to doubt, disregard, or fly in opposition to displayed guidance. This automation addiction in combination with an automation exception sets up a critical window where the pilot must take positive action to maintain control of the airplane.

The paper,
Defining Commercial Transport Loss-of-Control: A Quantitative Approach
by James Wilborn and John Foster (a joint Boeing and NASA effort) studied six in-flight loss of control accidents and incidents that occurred between 1992 and 1996 due to a variety of causal factors. Whether a fight experienced a loss of control was determined by analyzing if the flight exceeded more than three out of five parameter envelopes rather than just a subjective evaluation. The envelopes related to flight dynamics, aerodynamics, structural integrity, and flight control use. Each of the five envelopes plot two parameters against each other relating the normal ranges of: AOA, sideslip angle, pitch and roll attitudes, load factor, airspeed, and pitch and roll attitudes with the trends of each factored against the pilot’s control inputs.

They defined the interval between the time of the first envelope excursion and the time when control was lost as the critical window. In the six situations analyzed the average time available to the flight crew to provide a corrective response was 7.6 seconds. Therefore, the stabilization of the aircraft was found to be time critical.

Compare this short time window with the flight data from AF447. Within the first two seconds the bank angle had increased to 8° right with no pilot input. At that point manual flying began with up to 3/4 nose-up and up to full left roll input by the pilot flying. The first stall warning triggered six seconds after the disconnection and only three seconds after manual inputs began. Within the first 11 seconds of manual flight, the pitch attitude increased to 11°, the vertical speed to 5,200 feet per minute, and the g load reached 1.6g’s in response to pilot inputs. At that time the airplane was established on a trajectory to rapidly gain altitude and lose precious airspeed. This led to a deeply stalled condition and complete loss of control only a minute later. By my estimation, recovery was possible at least within that one minute. However, as that minute progressed, more and more deliberate corrective action was required.

Within a few seconds of the autopilot disconnecting at 02:10:05, there were repeated excursions of the Dynamic Roll Control envelope, which plots pilot inputs against airplane response.  The envelope was exceeded because the airplanes rolling motions were exactly out of sync with the sidestick lateral commands. In the pitch axis, even though the pilot’s pitch inputs were inappropriate, the airplane was responding to them and therefore was not in violation of the Dynamic Pitch Control envelope. By 02:10:51 the roll had stabilized, but the stall warning activated which violated one parameter on each of two related envelopes (AOA and speed).