By: Mark H. Goodrich – Copyright © 2013
Few things have more clearly reflected the technological advancements made possible by modern computers and electronics than the development of aircraft simulators. As a pilot, flight instructor, check airman and base training captain over the past five decades, I have been personal witness to this process, including a progressive misunderstanding by many regarding the inherent limitations of simulators to meet presumed levels of replication.
In its infancy, represented by the venerable Link AT-18 with its stubby yellow wings and blue fuselage, the goal of using a simulator was clear. Although some models had very minor motion functions, there was no intent that its use could replace flying experience, or presumption that it represented handling qualities. Indeed, the panel layout in the early Link did not mimic that of any particular airplane model. It was instead intended to create an environment for the practice of basic orientation and navigation procedures. As the instructor sat at the adjoining table and watched the tracing mechanism work its way across a chart, the ability of the student to properly react to the “A” and “N” transmissions in Morse Code, and ultimately track the “range” inbound and outbound, was laid down for the subsequent debriefing.
In the next developmental iteration, large transport instrument panels, including switch and circuit breaker panels for the various airplane systems, were presented in a large box with the appearance of the flight deck. Flight instruments were operational, but the systems replication was only in the location of the switches themselves, allowing for checklists to be run, but without any significant systems feedback. Once again, the clear intent was to allow for the practice of basic instrument interpretation and procedures without burning fuel. No motion was used, and there was no visual presentation of the outside world. Rather, frosted eisenglass was used to create the appearance of flying “in cloud”. In some models, engine noise was an added feature, although it was ridiculously pale when compared with that of actual radial engines, and propeller tips just outside the flight deck windows.
More lights, sound and systems response features appeared at about the same time in the simulators for the early century-series fighters, but once again absent any meaningful motion or outside visual presentations. Flight instrumentation was more faithfully represented, but the higher level of sophistication was not really intended to hone instrument flying skills. Instead, the goal was to practice radar intercept procedures.
The introduction of more sophisticated airplanes into the airline inventory was a point of quantum advancement in simulator technology, in that a simulator was available for most models, systems presentations were significantly improved, and visual references were used. In most cases, this was restricted to a night-time representation of approach and runway lights only. In some, large models and photographs of approach path detail were mounted along a wall adjacent to the simulator, and a camera was moved along and over the presentation during the approach phase, with the picture transmitted to a screen behind the windshield. Presentations were simple, and significantly misleading due to extreme parallax errors resulting from the camera angles and mirrors used. Indeed, experienced pilots soon learned to simply avoid any focus on the visuals and fly the instrument landing system to fifty feet or so before making any attempt to align or otherwise respond to exterior visual cues. To do otherwise suddenly left one inexplicably right or left of the landing runway.
In this phase, visuals were still treated by carriers as fundamentally insignificant to the training process, since the goal was again procedural practice, especially crew coordination in the context of normal, non-normal and emergency procedures. As such, improved systems presentation and response to switch position movements were the improvement features most significant to training goals.
Advancement to the next generation of simulators represented not merely a technological step in visual presentations and more faithful system representation, but for the first time was accompanied by a change in the way that simulators were used. As a result of several high-profile crashes during training, and the increasing financial pressures of economic deregulation, regulators and carriers began to see simulators as the way to move away from the use of airplanes altogether in the training process. Instead of a laboratory for procedures practice, the simulator was being seen as a replacement for all airplane training, and developmental work was prioritized to simulate reality well enough to be accepted by regulators as a pure substitute for the use of real airplanes. That each stake holder had its own reasons for wanting to eliminate airplanes from the training equation, and that their rationalizations would skew the analysis, seemed to escape the attention of all involved.
Regulators were increasingly removing themselves from flying. Operations inspectors no longer gave licensing or proficiency checks, but were increasingly desk-bound, relying upon designees of various descriptions to not only train, but to check and certify, as well. Ever fewer of those inspectors were themselves trained or experienced in airline transport airplanes, or as instructors. Further, several of the agency’s inspectors had been lost in airline training accidents while riding as observers – in some cases while the airline crew attempted to perform a risky maneuver specified by the inspector himself – and agency management was thus easily persuaded that simulators were the safest way to conduct airline training. On the airline side, it was ever more about money, and ever less about training. The new generation of airline executives did not appreciate that a significant portion of the training costs were not expenses, but rather an investment in the long-term performance of flight crew and safety of operations. To them, the question was not whether the quality of training would be negatively affected if simulators were used for the entire training process, but rather only if costs would be less and the regulator satisfied. One might assume that pilot groups would object, but few airline pilots are test pilots or engineers. The truth is that most were in awe of the video-game nature of the technology, did not understand the inherent limitations, but did appreciate that thinner manuals and standardized checking scenarios made successful completion of training a certainty. All pieces were in place for what has become a progressive deterioration in the quality of airline training, and increasing reliance upon simulators.
Advanced computer technologies were making highly sophisticated visual presentations possible, and simultaneously allowing incorporation of digitized performance programming based on certification data for the airplane types. That transport airplanes were themselves becoming computerized – that is, using autopilot, autothrust, computerized flight control, flight management and fault analysis systems – likewise facilitated the apparent replication of airplanes in simulator form. Enormous hydraulic jacks were used to create a variety of sensations from acceleration to pitch, roll and yaw, creating replications of turbulence, systems operation and failure noises, responses to power changes, flare, touchdown and braking. To even experienced flight crew, it all seemed like magic.
But there was a dark side to this apparent simulation of reality that was to haunt the training and checking process. The illusion of reality was often too good. Simulators were accepted as substitutes for actual airplanes, with simulator sessions soon accepted as the equivalent of actual flying. That simulator technology had limitations was overlooked. The stage was set to confuse simulation with reality.
In the early days of Level C and D simulator use, most of those certified were already experienced pilots and flight engineers. Most perceived the distinctions between simulators and airplanes, and were not misled. But new hires were more and more the product of an industry that accepted ever less experience for those working towards a professional flying career. These less experienced pilots more easily accepted simulators as a perfect reflection of the airplanes represented, faithfully replicating the ways in which those actual airplanes would behave in flight. In addition, the deregulated airline business was also changing the curricula and methodologies of training more broadly, reducing training footprints under the increasingly accepted theory that more airplane and systems automation requires less training for flight crew. Whether in training or on the line, flights were normally conducted with autoflight features engaged at 400 feet after takeoff, and disengaged at 200 feet before landing, or after an autoland. When coupled with the lower levels of experience they brought to their first airline job, the result was professional pilots with logbooks showing 3,000 hours, when in reality they had more like one hour 3,000 times.
These training system changes were mostly driven by desire to reduce costs. Airline managers saw costs for training as pure expense. Automation and the exclusive use of simulators for training looked like free money to bean counters with little appreciation for operating realities. Airlines began making the manuals thinner, and the training footprints ever shorter, relying more upon computer-based training to minimum standards, and less upon a determination of whether flight crews have actually learned enough to safely and efficiently operate the airplanes to which assigned. Regulators were their co-conspirators in this folly.
So, what are the problems with simulator replications? To say that they are many merely begs the question, but a few specific examples are illustrative.
Computers are just big dumb adding machines, and simulators are just computers. The responses of a computer are a function of its programming, and those programs may be more or less valid based on the knowledge and experience of the people who wrote them. As to airplane systems operation and response programming, the starting point is the design drawings. Yet, one recurring theme within the investigation of airplane incidents and accidents is the frequent manufacture of airplanes differently than specified within design data. Most significantly, simulators are not constructed, wired or plumbed as are the actual airplanes represented. If it is to be the same as the airplane drawings, then the simulator must also be 200 feet long with all wires, junction boxes, relays, fittings, regulators, pumps, actuators and other systems detail perfectly replicated. Consider that the next time you read that an investigation agency is going to use a simulator to determine what failed.
The same is true of programming for performance. To start with, airplanes off the assembly line often reflect significant variations in performance and handling due to the differences in weight, and alignment of airplane structural details. Engine set-up and wear likewise results in different airplane performance. It is for these reasons that flight management systems provide a feature to bias performance output data for variations in aerodynamic drag and fuel flow. Complicating the simulation of reality is that simulators also reflect differences between one unit and the next. Consider that most airline training instructors will prefer one simulator in preference to another because it “flies better”.
Flight test data incorporated into programming is a combination of information extrapolated from test point data recorded and corrected, and mathematical extensions or projections between and beyond actual test points. Therefore, much of the performance data fed into simulator programming is itself averaged, estimated or assumed. The true facts are that even the best simulators generally replicate airplane handling and performance ever more poorly as performance margins are approached.
The mimicked sensations, including forces, noises, handling qualities and performance, are often superficial representations. Simulators are not airplanes. They are not affected by the reality of air molecules impacting aerodynamic surfaces. Thrust, lift and drag are not even present, and gravity in conjunction with hydraulic jacks is used in ways much different than with an airborne vehicle free to be moved about three axis.
When training crews some years ago on a popular transport type, the simulators suffered a recurring flight control anomaly that instructors soon learned could be temporarily corrected by freezing the simulator, selecting all of the flight control computers off, and then selecting them back on. One day the fleet manager inquired, after a line captain responded in flight to a computer fault message by doing exactly what he had seen instructors do repeatedly in simulators, resulting in an inflight upset. The equivalency between the simulator and airplane was so ingrained that he considered his actions logical.
In the early days of Level-C simulators, the regulator was persuaded that substantially all training could be conducted using simulators – the so-called Appendix H Training based on “advanced simulation”. Required elements for one type included a high-altitude approach to stall, and a manual reversion event with loss of hydraulic pressure to the flight controls.
As an engineering test pilot, I had stalled most transport types at a variety of altitudes and with various centers of gravity. In the approved curriculum, taking trainees to stick shaker activation, although still some twenty knots above aerodynamic stall, satisfied the training requirement. But most troubling was that the simulator response at aerodynamic stall was dramatically different than in an actual airplane.
It was much the same in manual reversion. Simulators had been programmed with the presumed airplane response based on certification data, however in practice – where a manual reversion flight test event was then required after any change to the flight control system – maintenance manual standards had been brought forward from a predecessor type. The allowable range for trim cable tension on the shorter airplanes was unchanged from that for the longer models, and any airplane with trim cable tension less than the top 20% of the allowable range was barely controllable in manual reversion. Most significant was that the simulator did not reflect anything close to reality in the training event. There was a tick in the box on the approved curriculum, and the regulator was satisfied, but crews were hardly prepared by that training for an actual manual reversion event.
While conducting initial operating experience with a new captain on a four-engine type, we suffered a catastrophic engine failure. His response was slow, and when the engineer asked if he wanted the engine fire or severe-damage checklist, he declined and asked if the engine could be restarted. I took control, intervened with the engineer, did the immediate action items and then completed the procedure, after which I inquired about his response. Because he had not heard the noise programmed into the simulator for a severe damage failure, he presumed in the absence of a fire-warning that it was a mere flame-out. He had been conditioned by the simulator to react to a noise that did not mimic reality in the airplane.
Years ago, following several loss-of-control accidents, some airlines decided to incorporate “advanced maneuvers” training programs, which were essentially teaching aerobatic maneuvering to recover from upset events. Simulator instructors with no aerobatic or engineering experience found that pushing a rudder pedal to the floor in the simulators significantly increased the rate of roll, and began teaching a “full rudder recovery”, along with large pitch inputs. Management was not interested in hearing about limitations in the design strength of vertical stabilizers, or the g-limitations for transport category airplanes in both pitch and with rolling moments. Their response to such concerns was, “But it seems to work out fine in the simulator”. That a tail or a wing to be broken does not exist on a simulator seemed to escape their comprehension. Ultimately, management was persuaded to install a g-meter presentation at the instructor station to display vertical loadings. On the first upset recovery in the simulator, the recovery technique as it was being taught resulted in well over 5.0g’s, significantly above the structural failure point for transport category airplanes designed with the standard limitation of 2.5g’s in the so-called “clean” configuration, and 2.0g’s with any leading or trailing edge device extended.
Simulators are valuable training assets, and their sophistication has been raised to a level where they can be effectively used for more than merely practicing procedures. But too many in the industry – including regulators, airlines, training centers, flight crew and accident investigators – non-critically accept and believe that simulators faithfully replicate airplanes in ways that they do not and can not.
If simulators are to be used in the most effective ways, their inherent limitations must be understood and taken into account when training curricula are developed, so that they do not detract from the safety equation, but enhance it.
Mark H. Goodrich – Copyright © 2013
Simulating Reality was first published in the February 2013 Issue (Vol 10 No 1) of Position Report magazine.