A balanced perspective of mitigating risks of loss of control–in flight (LOC-I) requires an understanding of how defenses have evolved and a willingness to update pilot education and training practices based on collective expertise, says Larry Rockliff, a captain and chief test pilot, Production Flight Test, Airbus China. For 20 years, he has been one of the leaders in the field of airplane upset awareness, prevention, recognition and recovery.
He recently discussed with AeroSafety World a few aspects of government and industry progress in implementing upset prevention and recovery training (UPRT), including one project’s advances during the February meeting of a team representing the world’s major original equipment manufacturers (OEMs) of transport category airplanes. The team has been completing the first draft of Revision 3 of the influential — yet underutilized — Airplane Upset Recovery Training Aid (AURTA). Airbus, The Boeing Company and Flight Safety Foundation had led and/or facilitated the original team behind this reference document.
Rockliff updated the following highlights of his thoughts from September 2013 when ASW staff attended the Captain Ray Jones Lecture sponsored by the Royal Aeronautical Society’s Flight Simulation Group. In that lecture, and the recorded presentation in December 2013 to the International Civil Aviation Organization’s (ICAO’s) Multi-crew Pilot Licence Symposium, he spoke of lessons learned from the industry’s history of counteracting LOC-I, and the way forward.
“We have a very serious problem, but we have to address the problem with the most likely ways of transferring information and giving the pilots skills so that they’ll be able to react if things start to diverge,” he said.
Revision 3 will mirror the UPRT focus on monitoring and immediately correcting pitch, roll and airspeed divergences within the entire airplane operating envelope, and not over-emphasizing recovery methods at extreme edges of the envelope, he said (see “Progress Report: Revision 3 of the Airplane Upset Recovery Training Aid”). Regarding global airlines’ current UPRT implementations, Rockliff told ASW, “I have not seen any erroneous technical conclusions of late. However, I have seen a great deal of misplaced training to emphasize recovery from extreme attitudes instead of intervening during the onset of a developing upset. It doesn’t make sense to expect the pilot to be out of the loop until an airplane is fully upset, yet this is where a lot of training time is being focused.”
Figure 1 — Non-UPRT Training Confined to a Small Area of the Aerodynamic Flight Envelope After Airplane Development
OEMs = original equipment manufacturers of transport category airplanes; MD = maximum flight-demonstrated Mach number; MMO = maximum operating Mach number; UPRT = upset prevention and recovery training; VD = maximum flight-demonstrated airspeed; VFE = maximum flap extended speed; VMO = maximum operating airspeed
Notes: In his December 2013 presentation, Larry Rockliff referred to this chart of generalized flight test data to illustrate initial flight-test stages and the need to focus UPRT within the entire operating envelope rather than at its boundary or beyond. “In the very first few test flights, we go to the stall, and we go right up to MMO–MD and VMO–VD to determine that the airplane meets its design criteria,” he said. “Then we fill in the flight data in the entire operating envelope to ensure that there are no unusual characteristics, and to identify how the airplane is going to handle. After that, we start developing data where all the [non-UPRT] training is done at the air carriers and OEMs — that small portion of the envelope is all that pilots typically are exposed to.”
Although the contributors to official, global UPRT guidance have acknowledged the AURTA as an authoritative reference,1 some stakeholders have told Rockliff they ignored it because it seemed too unwieldy or existing programs seemed satisfactory.
The idea for the AURTA emerged from an early-1995 meeting at the Air Transport Association of America (ATA; now Airlines for America) in Washington. Two U.S. air transport accidents in the early 1990s prompted speculation among ATA-member airlines that related rule making could be imminent. The meeting had convened most training organizations on the Training Committee of the ATA’s Air Transport Group.
A subtext of Rockliff’s 2013 lecture was that aviation safety specialists, many now implementing UPRT, should not forget or ignore lessons from the AURTA’s 1998–2008 evolution. High on the list is realizing that, because of their investment and pride in proprietary training programs, airlines and aviation training organizations (ATOs) may insulate themselves from outside expertise. A related lesson is that committing to the highest level of consensus on contentious issues has a downside. This can be perceived as an inferior approach for the airlines and ATOs convinced that their individual programs are superior.
While developing the first AURTA, the Upset Recovery Industry Team realized that some airlines “had reached conclusions on what and how to train based upon experience from other airplanes or from simulators, and they were assuming some things to be correct when, in fact, they were not correct,” Rockliff said. “Even as OEMs, we were unable to convince them otherwise. Because of the consensus aspects — that we had to have everybody on board with everything — we used a lot of language that perhaps was not as firm or as provoking as what we could have or should have done.”
Objections were raised by a number of team members during the three-year development of the first AURTA, released Dec. 22, 1998, with no version number. “One of the things that they really were against was identifying upset-recovery techniques that were specific to aircraft types. They said, ‘We want to proceduralize this for all types because, in the training world, if we can make a procedure, then it simplifies the teaching and standardization task.’”
An example of a challenge due to the consensus requirement came from some training managers who were against the use of trim during an upset recovery. “This was based upon their experience in simulators, where a nose-up runaway trim would often result in a nose-low upset because the pilot in training would not release the trim switch,” Rockliff said. “But what they failed to realize was that the moment the flight condition would result in less than 1 g [one times standard acceleration of gravity] on the pilot’s body, she or he would instinctively release the trim. The problem in the simulator world is that you are always operating at 1 g, and this is a serious deficiency of dynamic maneuvering in a simulator.
“Examples of some of the deficiencies were that, in the event of thrust-loss asymmetry, air carriers did not want their pilots touching the rudder at all initially. They were more concerned about pilots selecting the wrong rudder pedal than the effects of just using aileron or roll control initially. Another major carrier was overstating the value of applying rudder at high alpha, high angle-of-attack. Another was promoting the use of rudder to induce roll — in ways that it wasn’t really intended to be used, and it was a big concern for us. Yet another was adamant about not utilizing trim in a recovery — even if there was unusual loading on the control column.”
In 1995–1998, the majority of the AURTA team’s work had focused on how to recover from airplane upsets. In 2015, the consensus of OEMs and others is that upset recovery must be part of UPRT training but is secondary to clear emphasis on upset awareness, prevention, recognition and corrective action within the operating envelope.
Authors of the AURTA and UPRT guidance (ASW, 12/13–1/14) have grappled with a fundamental false assumption about the uniformity of airline pilot knowledge and skills. “We assume that they’re all experienced and proficient throughout the entire operating envelope of the airplane — when in fact, perhaps, they’re not,” Rockliff said. “For pilots unable to perform, say, a raw-data, non-instrument ILS [instrument landing system approach]; to enter a hold without an FMS [flight management system]; or to perform something simple like a 30-degree banked turn at high altitude, it’s going to be quite a leap of faith to anticipate that they’re able to get into extreme conditions and recover.”
He noted that hundreds of thousands of airline pilots — typically spending 90 percent of their time above Flight Level 250 [approximately 25,000 ft] — have never selected the autopilot OFF at these flight levels, felt the aerodynamic damping or noted the elapsed time required to accelerate between two airspeeds or Mach numbers, compared with the faster elapsed times at low and medium altitudes. He said most simulator training today takes place at low altitude.
“If pilots have no concept of how their airplane operates within the entire operating envelope at high altitude and high Mach, without the aid of autopilot and automated systems, it is not realistic to take them outside the operating-envelope domain to teach recovery techniques. There is an enormous amount of effective training that can be completed throughout the operating envelope that, heretofore, pilots have only monitored,” Rockliff told ASW. He had said previously, “We have to question whether it’s reasonable to take the airplane to the extremes of the developed upset versus a developing upset. This was an important discovery for the OEMs.”
The OEMs concluded that any full-motion flight simulator can be used for strongly reminding airline pilots of the differences involved in high-altitude operation versus low-altitude operation, with and without automation. “There are not a lot of pilots, practically speaking, in any of the training programs that exist today, who are exposed to that because they climb at a VC [design cruise speed] with a Mach transition, and then level off at a constant Mach. They may go from Mach 0.78 to 0.79, but that is about it,” he said.
As in 1995, some operators and governments today are concerned that the flight simulation training devices they now use are only 50–60 percent suitable, or perhaps even unsafe, for UPRT. “From an OEM perspective, we can do very valuable upset prevention and recovery training — 85 percent-plus in existing simulators today,” he said. Rockliff told ASW that, in 2015, this is a point of complete agreement among all major Western OEMs, and this will be illustrated in Revision 3 of the AURTA.
Airplane vs. Simulator
UPRT conference presenters often have urged airlines to analyze instructor qualifications, simulator standards, course accuracy and focus before selecting any UPRT training provider. Rockliff told ASW, “Training providers that use fully developed upsets as their point of focus have missed the practical point altogether. Additionally, there is a high degree of risk in using simulators for dynamic maneuvering. The absence of g-loading is a serious deficiency. This is negative training at its fullest. Past accident investigations bear that out. Equally, I am not at all convinced that flying a fully aerobatic airplane has the value that some promote. Lack of inertia, rapid aircraft response, necessity for use of rudder and cockpit visibility are all areas that do not serve well to replicate larger transport aircraft.”
Rockliff participated as a test pilot in UPRT-relevant Airbus research into how closely simulator behavior and airplane behavior match during approach-to-stall and full-stall conditions. Flight data and subjective observations were generated from dozens of actual stalls in an A320 and an A340-600, and from corresponding full-motion simulators of about the same age. “These flights have influenced my perspectives of the complexities involved and how training exercises can mitigate LOC-I risk,” he said.
“The stall warning indications and the speeds at the stall warning were quite well matched between aircraft and simulator, but the buffet cues felt in the simulator were not the same. The buffeting experienced in the simulators was more like turbulence, which is reasonably modeled. And, even though aircraft flight data are provided to model onset buffet, the buffet doesn’t seem to be modeled in the simulators any differently from turbulence.”
In the testing, he also observed differences in the amount of forward stick required to break the stall in the airplane versus the corresponding simulator, or in some cases, the amount of side stick–release to break the stall, followed by the forward stick necessary to recover.
“Practically speaking, these differences didn’t matter because there could be a wide range of center of gravity, CG, that aircraft could be operated at. What mattered was that when the pilot, me, did the correct action — which was to stop the angle-of-attack from increasing — the simulator did what it was supposed to do: It broke the stall,” Rockliff said. “The objective is not to get into the stall; it’s to avoid getting there by recognizing the warning signs and preventing the event itself.”
He paraphrased the first advice in wording issued in 2012 by an aviation rulemaking committee,2 saying that as soon as a flight crew detects “any stall indication — which could be aural warning or buffet — then initiate a recovery.” “It didn’t say, ‘Go past that, try and climb up the CL–alpha [lift coefficient versus angle-of-attack] curve until you experience all of these indications of stall.’ If you make it to the g-break, you may have a wing drop this time, maybe it won’t the next time; you may be able to detect it, maybe not,” he said. “Even more important, all this is done in a ‘quasi-static’ environment — and every stall is going to be different to some extent in a transport category airplane. We want the pilots who unintentionally get into that predicament to be aware of it, and as soon as they get a cue, to initiate recovery to avoid stalling — period. We don’t want them disregarding cues simply to satisfy training objectives to discover where they shouldn’t be in the first place.”
In summary, he said airplane OEMs conduct in-flight tests up to the stall angle-of-attack to prepare data packages for use by simulator OEMs. “We are trying to get some form of repeatable data that can practically be introduced into the simulator. As such, the scenarios flown are deliberately ‘quasi-static,’ and thus, far different from what unintentionally would occur during airline operations. It’s a deceleration at a deliberate rate because — as test pilots — we are trying to derive clean reference data,” Rockliff said.
“We have the tools today to execute the training. The objective is awareness and avoidance. The airplane should not end up in extreme gyrations before the pilot does anything. If the airplane is diverging from what the pilot wants, then the objective is that the pilot intervene and counter the divergence.”
- For examples, see “Graduated Approach,” ASW, 6/12 and “Enough Talk,” ASW, 7/13.
- Section 208 of U.S. Public Law 111-216, “The Airline Safety and Federal Aviation Administration Extension Act of 2010” — responding to a safety recommendation for enhanced airline pilot training after the February 2009 fatal crash of a Colgan Air Bombardier DHC-8-400, operating as Continental Connection Flight 3407, near Buffalo, New York, U.S. — prompted the appointment and work of the Stick Pusher and Adverse Weather Event Training Aviation Rulemaking Committee, which focused on responses to approach-to-stall, icing conditions, and microbursts and wind shear events.
Progress Report: Revision 3 of the Airplane Upset Recovery Training Aid
At the conclusion of 2012–2013 meetings of the International Civil Aviation Organization’s (ICAO’s) Loss of Control Avoidance and Recovery Training (LOCART) committee, members of its original equipment manufacturers (OEM) subgroup were invited to consider updating the 2008 edition of the Airplane Upset Recovery Training Aid (AURTA). The request from ICAO cited a need to include all transport category aircraft — straight-wing propeller types as well as swept-wing jets — recalls Larry Rockliff, a captain and chief test pilot, Production Flight Test, Airbus China. Members accepted, and he has been chairing this revision process for about a year, he told AeroSafety World in February, in conjunction with a meeting to review the first draft.
“If we would do this, ICAO would use it as a reference document,” Rockliff said. “The biggest and most obvious differences from the past AURTA revisions will be the following: It will include all transport category aircraft. It will be formatted in three levels of detail that the user can select — need to know, nice to know and instructor level. It will propose what should be taught and what should not be taught in simulators, but will not provide suggestions as to how to teach it.”
The new team has advantages of common understanding from testing airplanes during development and a shared commitment to upset prevention and recovery training (UPRT); these became evident in the LOCART OEM subgroup’s work. “It was an unexpected delight to discover that each OEM ‘sang from the same proverbial song sheet.’ There were no special interests, no objectives to create business opportunities, no training assumptions and no conclusions derived from demonstrations completed in simulators,” Rockliff said. “Therefore, in very short order, a common set of recovery recommendations was made. This was expanded to training recommendations as to what to teach, but not how to teach it.”
This atmosphere contrasts with the difficulty of the original effort. “For the initial AURTA and the first two revisions, it took much longer to deliver the product than it should have, due to the energy spent to try and convince various stakeholders of what was correct and what was not,” he said. “With Revision 3, we have consensus from all the OEMs. It has been amazing to see the progress made when people speak the same language and understand the issues from those who actually fly the airplanes during development. This is a far cry from correlating knowledge and past experience in other types of airplanes, and then coming to conclusions in simulators.”
In a lecture and conference presentation in late 2013, Rockliff had envisioned ways that the AURTA could be updated by a team of experts smaller than the original Upset Recovery Industry Team, which included 20 airlines among the 33 aviation companies, professional associations, consultancies and government agencies. “If ICAO takes a proactive stance to help us improve upon it, the improvements or the changes would be very minimal,” he said then. “Once again, we would recalibrate what’s already in the document, and put it into a more usable, bullet-type form.”
Drawing from that concept, the new AURTA will take advantage of many advancements in informing and instructing today’s airline pilots, such as those of current Airbus airplane operating manuals, which condense high volumes of information into what he called descriptive units — easily learned/referenced information attuned to different needs of individual users. The three levels of detail in the first draft of Revision 3 already reflect this approach.
“The primary complaint toward past AURTAs was that each one was too dense and, therefore, most pilots were too intimidated to read it from front to back,” Rockliff said. “In reality, the actual academic and training sequences were quite condensed.”
While awaiting the release of Revision 3, he said, all aviation safety professionals involved in implementing UPRT will find especially interesting three elements from Revision 2 — the 50-page Section 2, “Pilot Guide to Airplane Upset Recovery,” and accompanying videos titled Part I, “Overview and Aerodynamics” (24 minutes 12 seconds) and Part II, “Recovery Techniques” (20 minutes 44 seconds). All elements are available at no cost from Flight Safety Foundation at <flightsafety.org/archives-and-resources/airplane-upset-recovery-training-aid>.
“In the second video, there’s a segment from 1998 in which three chief test pilots speak at length about the fact that you can’t even contemplate recovering from an upset unless you’ve recovered from a stall — if a stall condition is present,” Rockliff said. “They go into detail about what a stall is. It’s a very compelling four minutes. If regulators would have followed the guidance provided by these experts, versus allowing this document to sit and collect dust on shelves, then several accidents between 1998 and 2013 wouldn’t have occurred. It finally took the U.S. Congress to impose the change upon the Federal Aviation Administration after the Buffalo accident [see note 2 in ‘Totally Relevant’].”
Behind the Scenes
Accidents influenced the U.S. predecessors of upset prevention and recovery training.
By Wayne Rosenkrans
In the mid-1990s, out of concern that future regulatory changes might mandate how to prevent loss of control–in flight (LOC-I) accidents, U.S. airlines began working toward a collective, voluntary solution, says Larry Rockliff, a captain and chief test pilot, Production Flight Test, Airbus China. The U.S. National Transportation Safety Board (NTSB) by then had endorsed the upset-prevention programs of two U.S. airlines, he said, recalling his participation in some of the early discussions.
“The NTSB suggested that the programs were excellent, and so it seemed that there was a good possibility that one program — or a mixture of one or two of them — could be selected to form standard training for all of the U.S. carriers,” Rockliff said.
In the previous 25 years, the NTSB also had issued several safety recommendations to the U.S. Federal Aviation Administration (FAA), but the FAA regarded the full flight simulators of the time as probably ineffective for airplane stall/upset recovery–type training. No rule making actually occurred in the 1990s, and voluntary efforts by individual air carriers and aviation training organizations remained the norm.
“There were some very critically flawed components to some of the training being conducted,” Rockliff said, noting an imperative within some carriers to make any such training uniform across aircraft types. “But for transport jets with under-wing engines, for example, there’s a definite thrust/pitch couple that can affect the controllability and pitch-power combinations that pilots will have available to them when they’re in a high-alpha/angle-of-attack condition. That would mean that, in some cases, the flight crew has to actually reduce thrust instead of adding thrust, even when losing energy, increasing angle-of-attack.”
In a meeting in early 1995, the airline pilot training experts shared their companies’ approaches to the stall/upset problem. One realization was the changing demographic factors. “All the military pilots of the world had more than basic exposure to unusual attitude recovery, but in very many of the civil programs, there was no exposure at all. Obviously, the increase of civil pilots and the decrease of military pilots in the airlines was dictating that it was probably going to result in additional LOC-I accident statistics,” Rockliff said.
He and Dave Carbaugh, a captain and Boeing test pilot, were invited to the meeting, and they determined on site that as representatives of two original equipment manufacturers (OEMs) of transport category airplanes, they were in a position to offer a wealth of relevant knowledge and data, and to help build a new consensus about evidence-based versus unproven training practices.
“We committed to developing another training aid” that would be consistent with collaborative precedents such as the FAA’s Windshear Training Aid (1987) and Takeoff Safety Training Aid (1993), and Flight Safety Foundation’s CFIT [Controlled Flight Into Terrain] Education and Training Aid (1995), he said. “We knew we had to get a grip on stalls because the training programs that existed at that time assumed that the airplane wasn’t stalled. Yet we knew statistically that a very large percentage of airplane upsets involve a stall in the first place.
“The industry also could use simulators a whole lot better than we’d been doing. The key point was that we had to use the simulators for what they were designed to be used for — and not beyond — because it’s in this arena that the trainees and, quite possibly, the instructors could become somewhat confused.”
The formation of the Upset Recovery Industry Team in 1995 led to the 1998 publication of the original Airplane Upset Recovery Training Aid (AURTA). Revision 1 to the AURTA, dated Aug. 6, 2004, contained additions that responded to NTSB safety recommendations issued after the fatal crash of American Airlines Flight 587, an Airbus A300, on Nov. 12, 2001, in New York (ASW, 4/10). In that crash, a pilot’s “unnecessary and excessive rudder pedal inputs” during a wake turbulence encounter resulted in extremely high aerodynamic loads that caused in-flight separation of the vertical stabilizer.
This accident remains an indelible reminder to Rockliff of the risk that well-intended, but inaccurate, training elements can infiltrate any airline’s efforts to mitigate LOC-I, he said. “There was such a paranoia that permeated the industry afterwards that people started to question ‘Were we properly applying rudder?’ And they were,” he said.
Later, the Upset Recovery Industry Team felt so strongly that regulators’ support of their work had not lived up to expectations that its leaders initially balked at the FAA’s request for a Revision 2 that would include a new module on high-altitude simulator exercises. But the team leaders ultimately agreed, and Revision 2, dated November 2008, responded to NTSB safety recommendations and the resulting FAA safety initiative on high-altitude upset training, Rockliff said. These additions emerged from the investigation of LOC-I in the fatal crash of a Pinnacle Airlines positioning flight, a Bombardier CRJ200, near Jefferson City, Missouri, U.S., on Oct. 14, 2004.
About a year ago, when the decision was reached to develop Revision 3 of the AURTA, the reasons behind some subject matter experts’ and airline safety managers’ disregard of the early versions had to be considered.
“Many training experts have talked about the benefits and value of Revision 2 of the AURTA, but when they think a new topic is being raised, they are surprised to discover it is already within the existing AURTA,” Rockliff told ASW. “They either overlooked it, or forgot it. Having said this, there are those who are plotting their own course, and I doubt they used the AURTA very much. It’s really a tragic footnote that this was close to 20 years ago, and that it took the president of the United States signing into law specific training requirements in 2010 to really push the industry from motivation, to requirement, to action.”
Questions about how to use simulators safely and effectively persisted until the International Civil Aviation Organization published new standards and recommended practices in 2014 for upset prevention and recovery training. High among concerns had been the risk of accidents involving negative transfer of training.
Especially when flight maneuvers taught in the simulator are not repeatable in the airplane, this violates the fundamental principles of educational psychology — they are also known as laws of learning in the FAA Aviation Instructor’s Handbook, Rockliff said. The law of primacy1 and the law of intensity2 were violated for many years, for example, by the recently discredited practice of “powering out” — instead of reducing angle-of-attack — as the first step in recovery from approach to stall during simulator training and practical tests.
“If the learner — student or flight crew — is led into a false sense of well-being with the wrong idea or a wrong level of training, then we could develop a very dangerous situation,” he said. He compared a pilot who lacks adequate manual control proficiency for flying in a transport jet’s entire operating envelope — particularly for high-altitude maneuvers — and who likely would have difficulty mastering upset recovery at or beyond envelope boundaries to “a person who is taught to swim laps very well in a swimming pool, and then is tossed over the side of a boat in the middle of a lake and told to do the butterfly stroke to shore.”
Eventually, all major OEMs, as a group, began advising air carriers and aviation training organizations that internally develop stall/upset training: “Develop what you wish, but make sure that it conforms with what the OEM recommends and refer to the OEM if you want to do something different,” he said.
Recently, this group of OEMs behind the development of Revision 3 also endorsed high-value simulator exercises that fully use, but do not exceed, the current technological capabilities. For example, one such exercise involves selecting two airspeeds — for example, 200 kt and 250 kt — that a typical transport jet can attain at sea level, 10,000 ft, 20,000 ft and its service ceiling and having the pilots measure the elapsed time as they accelerate from the lower speed to the higher speed at each altitude/flight level.
“During production test flights in an Airbus A320, to accelerate at low altitude between 200 and 250 kt takes about 15–20 seconds. When you get up to 15,000 ft, it takes 45–50 seconds. If you go to 39,000 ft — where at 200 kt, you’re pretty slow — you’re right back around the L/DMAX [maximum lift/drag] curve. To accelerate to 250 — which, by the way, now is related to Mach versus VC [design cruise speed] — takes seven minutes,” Rockliff said. “Just a simple exercise like that provides a very positive reinforcement for pilots.”
- FAA. FAA Aviation Instructor’s Handbook. FAA-H-8083-9A. 2008. The law of primacy says that the first method taught to a student “often creates a strong, almost unshakable impression and underlies the reason an instructor much teach correctly the first time and the student must learn correctly the first time.”
- FAA. The law of intensity says that “immediate, exciting, or dramatic learning connected to a real situation teaches a learner more than a routine or boring experience. Real-world applications (scenarios) that integrate procedures and tasks the learner is capable of learning makes a vivid impression, and he or she is least likely to forget the experience.”