Understanding Upsets

The crash of Air France Flight 447 into the South Atlantic off Brazil early on June 1, 2009, will forever be considered by safety experts worldwide as one of the classic examples of inadvertently fully stalling a perfectly good aircraft at cruise altitude due to the mismanagement of an onboard malfunction. Both pilots occupying the cockpit seats and the captain standing behind them were clueless as to why this was happening.

During its last minute of flight, the Airbus A330 was moving horizontally at less than 90 kt. but was descending at 10,000 fpm. The sequence from the onset of the malfunction at FL 350 to slamming into the ocean was just under 4 min. If the pilots had realized their predicament and lowered the nose, thereby unloading the wings, at any point during the last few minutes of flight, they would have been able to regain control and fly on to France — shaken perhaps, but unscathed.

Current flight decks are replete with automated systems designed to enhance situational awareness, reduce fatigue, calculate performance and help crews manage their aircraft. If not actively monitored, however, the highly reliable automated system can lull crews into complacency and an over dependence — a condition that, on occasion, could make the aircraft vulnerable to potential upset.

A crew’s ability to recognize and prevent the onset of an upset and to recover should one occur are key factors behind FlightSafety International’s new Upset Prevention & Recovery Training (UPRT) course now offered at its Savannah, Georgia, learning center. Integral to the course is the company’s Gulfstream G550 Full Flight Simulator, which recently became the first such device to receive the FAA’s endorsement for upset training.


Upset replacement FlightSafety UPRT

In his classic paper, “Skills, Rules and Knowledge; Signals, Signs and Symbols, and Other Distinctions in Human Performance Models,” Jens Rasmussen, a world-renowned system safety and human factors professor, outlines his theory of Skill Rule Knowledge (SRK):

  • A skill-based behavior requires little or no conscious control to perform or execute an action once an intention is formed; also known as a sensorimotor behavior.
  • A rule-based behavior employs rules and procedures to select a course of action in a familiar work situation.
  • A knowledge-based behavior represents a more advanced level of reasoning and must be employed when the situation is novel and unexpected.

Dann Runik, FlightSafety’s executive director of Advanced Training Programs, shared his UPRT team’s take on the Rasmussen’s SRK model, noting “We have termed the ability to correlate and fix a problem that you have never seen before as ‘Exceptional Knowledge,’ the highest level of learning.”

While the FAA has approved the FlightSafety simulator for upset training, it has not actually approved the course. It cannot since, as Randy Gaston, vice president of Flight Operations, Gulfstream Aerospace, explained, “The new FAR Part 60 regarding UPRT is not yet published.”

However, Advisory Circular (AC) 120-111, which addresses the issue, was posted this past April and the rulemaking process is currently in its “quiet period.” Once it awakens, the FAA will require FAR Part 121 operators to provide simulator-based UPRT instruction to their flight crews. The AC indicates that this will be accomplished on or before March 12, 2019. FAR Part 91 and Part 135 operators will be unaffected by the eventual rule, but FlightSafety is embracing it well in advance as a higher standard that should be broadly applied across pilot ranks.

AC 120-111 defines an “Airplane Upset” as “An airplane in flight unintentionally exceeding the parameters normally experienced in line operations or training:

  • Pitch attitude greater than 25 deg. nose up;
  • Pitch attitude greater than 10 deg. nose down;
  • Bank angle greater than 45 deg.; or
  • Within the above parameters but flying at airspeeds inappropriate for the conditions.”

Even though Gulfstreams are unlikely to ever be used in Part 121 service, the FAA is clearly impressed with FlightSafety’s ability to re-create viable UPRT scenarios in its simulators — the G650 is now programmed as well — under the guidance of a select group of instructors. And so, once again, business aviation is leading the way in developing additional safety measures and training that far surpass any requirement.

According to Gaston, the effort predates the infamous Air France crash — by a decade. “FlightSafety and Gulfstream have been working on the development of a positive UPRT experience in the sims since 1999,” he said. To which Runik added, “The flight test data developed by Gulfstream during the certification process for the G550 was critical in the programming of the simulator; especially high-speed aerodynamic characteristics and full stall dynamics of the airframe in the hands of the test pilots.”

Simulator vs. Aircraft

Until now, the principal training device for replicating in-flight upset was an actual aircraft, and typically aerobatic models. Such training sessions can be both instructive and fun. However, those in-flight experiences have certain limitations. First, the training aircraft will not be similar to the business jet flown by the student. Second, and more importantly, recognition and recovery techniques can only be demonstrated at an altitude with sufficient safety margins allowing for proper recovery. Low-altitude full stall recovery will not be demonstrated or practiced.

Simulators historically provided stall resolutions and unusual attitude recovery scenarios in line with the regulatory requirements for type ratings and check rides. Accordingly, stall recovery in both clean wing and landing configuration, had to be initiated by the pilot under evaluation at the onset of a stall warning. Pilots were judged on their skill levels using power and attitude in recovering. The object was to minimize altitude loss during the recovery maneuver, while preventing a secondary stall warning. Flying the simulator beyond stall warning was to fly into the unknown since the FAA entertained no data input outside the aircraft’s approved flight envelope, that is, beyond stall or Mmo. Consequently, no approved simulator could replicate such extreme conditions.

However, during certification flight trials, test pilots fly aircraft far beyond redline Vmo and Mmo speeds, and conduct full stall series. And their manufacturers capture all the resulting data in detail. It was these data, volunteered by Gulfstream, that FlightSafety programmers analyzed and incorporated into the UPRT simulators. The process took months but bore impressive results.

During that time, the programmers at the company’s simulation center in Broken Arrow, Oklahoma, analyzed the data along with the environmental conditions at the time they were captured to determine the aircraft’s response under a variety of circumstances since those, for example, could cause a stalling aircraft to break right or left or result in a “falling leaf” stall. Despite this “nonlinearity of data,” Runik said the analysts found the common factors that cause an aircraft to behave seemingly inconsistently and unpredictably in extreme flight conditions. Consequently, the simulator performs in the same manner, surprising and confounding pilots who expect predictability.

“They ask, ‘How do you do that?’” Runik laughed. “We’re not saying. It’s FlightSafety’s secret sauce recipe.” And thus proprietary.

In addition, “We invited the test pilots who were involved in the certification of the G550 to fly the UPRT simulator,” Runik recalled. “They indicated that the simulator replicated the feeling and proper controlled recovery from a full stalled condition that was experienced during the aircraft certification flight program. They enjoyed the sim very much.”

The advance of simulator technology has truly been amazing to observe during the past 40 years. The challenge created by AC 120-111 and the inevitable FAA rulemaking that will follow is focused on enhancing the fidelity of Level C and D simulators so as to stimulate the recognition and recovery from upset scenarios.

Ground School: Accident and Incident Discussion and Analysis

FlightSafety’s UPRT program is a one-day, 7.5-hr. immersion — 3.5 hr. in the classroom, complemented by two 2-hr. sim sessions, allowing the pilots to trade seats. It is designed for already paired crews, but single attendees are accommodated in the sim by pairing with another single or a FlightSafety instructor specifically trained in UPRT techniques. Only pilots type rated in the subject aircraft model can take the UPRT program. Although retired from line flying for a few years now, this past July, I was one of those singles.
The training session began with an informative ground school. Runik led a fascinating discussion that began with a detailed review of 16 accidents and incidents that were precipitated by upset scenarios. Many of the more noteworthy, including Air France Flight 447, Colgan Air Flight 3407 and China Air Flight 006, illustrated how over-reliance on automation led to upsets with catastrophic results. The ground school replication of each incident and accident provided me with a more informed understanding of upset onset and recovery techniques.

Accident investigation reports have historically offered lessons about preventing similar mistakes. But the presentations offered during the UPRT program highlighted the importance of early recognition and the urgency of following an upset with implementing “Push, Roll, Pull and Power” to recover from a deep stall. I discovered firsthand that this technique is highly effective during the afternoon simulator session, at altitude and in the landing pattern.


It’s been almost 50 years since I was issued my copy of Aerodynamics for Naval Aviators upon arriving at the U.S. Navy’s primary flight training base in Pensacola, Florida. That is to say, in my subsequent, decades-long — and accident-free — career in military and business flying, I’d not since had to again parse some of the fundamentals that kept me aloft. So, I much appreciated Runik’s well-structured review of aerodynamics — subsonic, supersonic and transonic flight. He helped us revisit the relationship of drag and lift, α — angle of attack, β — angle of sideslip and γ — gamma (dynamic pressure). His explanation of the factors leading to the unbalanced condition termed “Dutch roll” was particularly informative.

The review of flow dynamics associated with an aerodynamic stall reinforced for me the need to first unload a stalled wing, then roll to a wings-level attitude, pull the yoke toward the horizon and, lastly, adjust power. While application of power and proper attitude are taught to be utilized at the onset of a stall, should the wing fully stall, adding power will not be sufficient to recover; we are not flying F-18 Hornets.

Runik emphasized the critical importance of decreasing the stall speed of the wing simply by “pushing” the wing over and thereby effectively lowering the relative angle of attack (AOA) of the airfoil. Once the AOA is decreased, you can initiate a more positive roll rate due to the fact that the ailerons are then more effective. I learned to rely on AOA during my years flying off carriers. Smoothly pulling back on the yoke or stick while properly adjusting power will complete the recovery. This technique allows full recovery from an approach upset at 700 ft. AGL during a base turn to final. If I can do it, you can, too.

Importance of CRM

The opportunity to fly the G550 simulator also was my first experience with a head-up display (HUD). Incorporating the flight path vector (FPV) within my scan as I hand flew the sim was easy and natural. Then, when I learned that the difference between the FPV and the pitch attitude, “side ears,” was the actual AOA, I felt more confident in my ability to fly the simulator properly.

For those pilots who don’t wish to fly the HUD, the G550 avionics suite has a HUD selection for the PFD called HUD Mode. Keeping the relative AOA within +2/-1 deg. in high-speed flight or +4/-1 deg. during slow-speed flight indicates to the crew that the wing is operating within the normal parameters of flight in those regimes.

Trending toward the outer boundaries of these limitations is a critical indication of an impending stall and/or potential upset. During recovery from an upset, it is much easier to smoothly return to normal flight conditions keeping in mind where the normal AOA limitations are in relation to the situation. This is especially important in pulling out of a descent while resisting the tendency to rapidly pull into a high-g recovery or in a low-altitude, low-airspeed situation, to over-correct into a secondary stall. If the Air France Flight 447 pilots had flown an AOA during the attempted recovery when the airspeed and other flight reference indications were rendered unreliable, they would have survived to fly another day.

Crew resource management is key in the prevention of an in-flight upset. It was interesting to note that many of the accidents and incidents that we discussed and flew would have been preventable if the crew had been exercising proper CRM. Proper monitoring within a busy flight deck is just as important as proper handling of the flight controls; therefore, it takes two professionals to make a great flight crew.

Flying the Sim

As noted earlier, single trainees are paired with another single or a FlightSafety instructor for the simulator sessions. I was fortunate to be paired with Lou Poth, FlightSafety’s chief pilot – Advanced Training Programs. Having him in the right seat was reassuring as I demonstrated some basic — albeit slightly rusty — piloting skills. Once level at 15,000 ft., I conducted some 60-deg. banked turns and went all the way to the stops in aileron control. Next we conducted some clean stall recoveries at stick shaker onset. Following those, we repeated the stall recovery drill in the landing configuration.

Satisfied that I could handle the aircraft under those conditions, we ran quickly, as only a simulator can, to FL 450. Once there the autopilot was engaged at Mach 0.84 and the real fun began.

The UPRT simulator training includes several situations intended to elicit full stall recognition and recovery within the confines of the airport traffic area. All those scenarios are replications of actual accidents that occurred at low altitude that might have been prevented had the crews been alert to the onset of an upset and knew how to respond.

“Many of the situations experienced in the sim are not handled successfully by our students the first time,” Poth said. “The reinforced positive training aspects of teaching specific recognition and recovery techniques is the key to our UPRT program.”

Handling the full seven accident scenarios really tested my recovery skills. Without the hours of ground school and the discipline to follow the upset recovery technique that was stressed by Runik, I would not have been successful in returning the aircraft to normal flight conditions every time, especially at low level. He recommended that upon initial recognition of each upset condition, I should give voice to my actions while conducting the recovery: “Push, Roll, Pull, Power.” Rudder input was highly discouraged, due to the unique characteristics of the Gulfstream 550’s yaw damping system.

During my training session, I was asked not to divulge the accident upsets that I encountered so that each will surprise the unsuspecting student. And believe me, they do surprise.

Some pilots maintain that only in-aircraft training can deliver a true upset experience (see “Keeping Cool Through Upset,” May 2015, page 63) because, in part, simulators cannot deliver any high-g sensation. While the observation about g’s is correct, that may not be much of a shortcoming at all. The G550’s visuals are compelling — there’s ground rush like you never want to see for real — and in my most difficult recovering, the sim’s computerized g-meter never exceeded 2.2 — hardly a heart-pounding experience.

Moreover, the simulator allows you to do low-level upsets that you’d never try in a real aircraft. And with a simulator, you get to experience them again and again until the recovery actions become ingrained and almost automatic.

During my career as a business aviation pilot, my first reaction to an upset more than likely would have been to try and power out of it while adjusting attitude. I could not have imagined reacting by pushing the nose over and unloading the wing. It’s simply not natural, and certainly not at low altitude with the rushing ground filling the windscreen. But after undergoing UPRT, I’m convinced that’s exactly what needs to be done. A lot of pilots and their passengers would be alive today had the crew reacted correctly.

Over the decades I’ve undergone extensive pilot training and I can say, without qualification, that the ground and simulator UPRT program is among the best training I’ve ever experienced as a business aviation pilot. Hats off to FlightSafety and Gulfstream for putting together such a well-structured and realistic experience.

Debrief and Analysis

The UPRT program, currently available in Savannah, is projected to be rolled out to all FlightSafety training centers with Gulfstream simulators. Long term, the company hopes to obtain relevant flight test data from other airframers in order to eventually offer the training in a wide array of aircraft makes and models.

A pacesetter in the program’s expansion, according to Runik, “is to identify and develop an instructor corps within FlightSafety to teach these skills in a unified and positive manner.” An inherent risk of the simulator training, one readily acknowledged by the FAA, is that done wrongly it could result in a negative training experience, one that could instill incorrect techniques in a student’s mind and exacerbate the problem. To prevent any such occurrence, Runik is conducting an internal search for “the best of the best” instructors to conduct FlightSafety’s UPRT training.

I can attest from personal experience that the first cadre is already at work in Savannah. B&CA

Watch in the video from Gulfstream and Jim Cannon on Upset Recovery or go to AviationWeek.com/upset.