NASA's Contributions to Aeronautics, Volume 1 by National Aeronautics & Space Administration. - HTML preview

CASE

6

Case-6 Cover Image: The Northrop HL-10 turning onto final approach for a lakebed landing. NASA.

The history of aviation is replete with challenges and difficulties overcome by creative scientists and engineers whose insight, coupled with often-pragmatic solutions, broke through what had appeared to be barriers to future flight. At the dawn of aviation, the problems were largely evident to all: for example, simply developing a winged vehicle that could take off, sustain itself in the air, fly in a controlled fashion, and then land. As aviation progressed, the problems and challenges became more subtle but no less demanding. The National Advisory Committee on Aeronautics (NACA) had been created in 1915 to pursue the “scientific study of the problems of flight, with a view to their practical solution,” and that spirit carried over into the aeronautics programs of the National Aeronautics and Space Administration (NASA), which succeeded the NACA on October 1, 1958, not quite a year after Sputnik had electrified the world. The role of the NACA, and later NASA, is mentioned often in the following discussion. Both have been instrumental in the discovery and solution to many of these problems.

As aircraft flight speeds moved from the firmly subsonic through the transonic and into the supersonic and even hypersonic regimes, the continuing challenge of addressing unexpected interactions and problems followed right along. Since an airplane is an integrated system, many of these problems crossed multiple discipline areas and affected multiple aspects of an aircraft’s performance, or flight safety. Numerous examples could be selected, but the author has chosen to examine a representative sampling from several areas: experience with flight control systems and their design, structures, and their aeroelastic manifestations; flight simulation; flight dynamics (the motions and experience of the airplane in flight); and aerothermodynamics, the demanding environment of aerodynamic heating that affects a vehicle and its structure at higher velocities.

Flight Control Systems and Their Design

During the Second World War, there were multiple documented incidents and several fatalities that occurred when fighter pilots dove their propeller-driven airplanes at speeds approaching the speed of sound. Pilots reported increasing levels of buffet and loss of control at these speeds. Wind tunnels at that time were incapable of producing reliable meaningful data in the transonic speed range because the local shock waves were reflected off the wind tunnel walls, thus invalidating the data measurements. The NACA and the Department of Defense (DOD) created a new research airplane program to obtain a better understanding of transonic phenomena through flight-testing. The first of the resulting aircraft was the Bell XS-1 (later X-1) rocket-powered research airplane.

On NACA advice, Bell had designed the X-1 with a horizontal tail configuration consisting of an adjustable horizontal stabilizer with a hinged elevator at the rear for pitch control, at a time when a fixed horizontal tail and hinged elevator constituted the standard pitch control configuration for that period.[1] The X-1 incorporated this as an emergency means to increase its longitudinal (pitch) control authority at transonic speeds. It proved a wise precaution because, during the early buildup flights, the X-1 encountered similar buffet and loss of control as reported by earlier fighters. Analysis showed that local shock waves were forming on the tail surface, eventually migrating to the elevator hinge line. When they reached the hinge line, the effectiveness of the elevator was significantly reduced, thus causing the loss of control. The X-1 NACA–U.S. Air Force (USAF) test team determined to go ahead, thanks to the X-1 having an adjustable horizontal tail. They subsequently validated that the airplane could be controlled in the transonic region by moving the horizontal stabilizer and the elevator together as a single unit. This discovery allowed Capt. Charles E. Yeager to exceed the speed of sound in controlled flight with the X-1 on October 14, 1947.[2]

An extensive program of transonic testing was undertaken at the NACA High-Speed Flight Station (HSFS; subsequently the Dryden Flight Research Center) on evaluating aircraft handling qualities using the conventional elevator and then the elevator with adjustable stabilizer.[3] As a result, subsequent transonic airplanes were all designed to use a one-piece, all-flying horizontal stabilizer, which solved the control problem and was incorporated on the prototypes of the first supersonic American jet fighters, the North American YF-100A, and the Vought XF8U-1 Crusader, flown in 1953 and 1954. Today, the all-moving tail is a standard design element of virtually all high-speed aircraft developed around the globe.[4]

Resolving the Challenge of Aerodynamic Damping

Researchers in the early supersonic era also faced the challenges posed by the lack of aerodynamic damping. Aerodynamic damping is the natural resistance of an airplane to rotational movement about its center of gravity while flying in the atmosphere. In its simplest form, it consists of forces created on aerodynamic surfaces that are some distance from the center of gravity (cg). For example, when an airplane rotates about the cg in the pitch axis, the horizontal tail, being some distance aft of the cg, will translate up or down. This translational motion produces a vertical lift force on the tail surface and a moment (force times distance) that tends to resist the rotational motion. This lift force opposes the rotation regardless of the direction of the motion. The resisting force will be proportional to the rate of rotation, or pitch rate. The faster the rotational rate, the larger will be the resisting force. The magnitude of the resisting tail lift force is dependent on the change in angle of attack created by the rotation. This change in angle of attack is the vector sum of the rotational velocity and the forward velocity of the airplane. For low forward velocities, the angle of attack change is quite large and the natural damping is also large. The high aerodynamic damping associated with the low speeds of the Wright brothers flights contributed a great deal to the brothers’ ability to control the static longitudinal instability of their early vehicles.

At very high forward speed, the same pitch rate will produce a much smaller change in angle of attack and thus lower damping. For practical purposes, all aerodynamic damping can be considered to be inversely proportional to true velocity. The significance of this is that an airplane’s natural resistance to oscillatory motion, in all axes, disappears as the true speed increases. At hypersonic speeds (above Mach 5), any rotational disturbance will create an oscillation that will essentially not damp out by itself.

As airplanes flew ever faster, this lightly damped, oscillatory tendency became more obvious and was a hindrance to accurate weapons delivery for military aircraft, and pilot and passenger comfort for commercial aircraft. Evaluating the seriousness of the damping challenge in an era when aircraft design was changing markedly (from the straight-wing propeller-driven airplane to the swept and delta wing jet and beyond). It occupied a great amount of attention from the NACA and early NASA researchers, who recognized that it would pose a continuing hindrance to the exploitation of the transonic and supersonic region, and the hypersonic beyond.[5]

In general, aerodynamic damping has a positive influence on handling qualities, because it tends to suppress the oscillatory tendencies of a naturally stable airplane. Unfortunately, it gradually disappears as the speed increases, indicating the need for some artificial method of suppressing these oscillations during high-speed flight. In the preelectronic flight control era, the solution was the modification of flight control systems to incorporate electronic damper systems, often referred to as Stability Augmentation Systems (SAS). A damper system for one axis consisted of a rate gyro measuring rotational rate in that axis, a gain-changing circuit that adjusted the size of the needed control command, and a servo mechanism that added additional control surface commands to the commands from the pilot’s stick. Control surface commands were generated that were proportional to the measured rotational rate (feedback) but opposite in sign, thus driving the rotational rate toward zero.

Damper systems were installed in at least one axis of all of the Century-series fighters (F-100 through F-107), and all were successful in stabilizing the aircraft in high-speed flight.[6] Development of stability augmentation systems—and their refinement through contractor, Air Force–Navy, and NACA–NASA testing—was crucial to meeting the challenge of developing Cold War airpower forces, made yet more demanding because the United States and the larger NATO alliance chose a conscious strategy of using advanced technology to generate high-leverage aircraft systems that could offset larger numbers of less-individually capable Soviet-bloc designs.[7]

Early, simple damper systems were so-called single-string systems and were designed to be “fail-safe.” A single gyro, servo, and wiring system were installed for each axis. The feedback gains were quite low, tailored to the damping requirements at high speed, at which very little control surface travel was necessary. The servo travel was limited to a very small value, usually less than 2 degrees of control surface movement. A failure in the system could drive the servo to its maximum travel, but the transient motion was small and easily compensated by the pilot. Loss of a damper at high speed thus reduced the comfort level, or weapons delivery accuracy, but was tolerable, and, at lower speeds associated with takeoff and landing, the natural aerodynamic damping was adequate.

One of the first airplanes to utilize electronic redundancy in the design of its flight control system was the X-15 rocket-powered research airplane, which, at the time of its design, faced numerous unknowns. Because of the extreme flight conditions (Mach 6 and 250,000-foot altitude), the servo travel needed for damping was quite large, and the pilot could not compensate if the servo received a hard-over signal. The solution was the incorporation of an independent, but identical, feedback “monitoring” channel in addition to the “working” channel in each axis. The servo commands from the monitor and working channel were continuously compared, and when a disagreement was detected, the system was automatically disengaged and the servo centered. This provided the equivalent level of protection to the limited-authority fail-safe damper systems incorporated in the Century series fighters. Two of the three X-15s retained this fail-safe damper system throughout the 9-year NASA–Air Force–Navy test program, although a backup roll rate gyro was added to provide fail-operational, fail-safe capability in the roll axis.[8] Refining the X-15’s SAS system necessitated a great amount of analysis and simulator work before the pilots deemed it acceptable, particularly as the X-15’s stability deteriorated markedly at higher angles of attack above Mach 2. Indeed, one of the major aspects of the X-15’s research program was refining understanding of the complexities of hypersonic stability and control, particularly during reentry at high angles of attack.[9]

The electronic revolution dramatically reshaped design approaches to damping and stability. Once it was recognized that electronic assistance was beneficial to a pilot’s ability to control an airplane, the concept evolved rapidly. By adding a third independent channel, and some electronic voting logic, a failed channel could be identified and its signal “voted out,” while retaining the remaining two channels active. If a second failure occurred (that is, the two remaining channels did not agree), the system would be disconnected and the damper would become inoperable. Damper systems of this type were referred to as fail-operational, fail-safe (FOFS) systems. Further enhancement was provided by comparing the pilot’s stick commands with the measured airplane response and using analog computer circuits to tailor servo commands so that the airplane response was nearly the same for all flight conditions. These systems were referred to as Command Augmentation Systems (CAS). The next step in the evolution was the incorporation of a mathematical model of the desired aircraft response into the analog computer circuitry. An error signal was generated by comparing the instantaneous measured aircraft response with the desired mathematical-model response, and the servo commands forced the airplane to fly per the mathematical model, regardless of the airplane’s inherent aerodynamic tendencies. These systems were called “model-following.”

Even higher levels of redundancy were necessary for safe operation of these advanced control concepts after multiple failures, and the failure logic became increasingly more complex. Establishing the proper “trip” levels, where an erroneous comparison would result in the exclusion of one channel, was an especially challenging task. If the trip levels were too tight, a small difference between the outputs of two perfectly good gyros would result in nuisance trips, while trip levels that were too loose could result in a failed gyro not being recognized in a timely manner. Trip levels were usually adjusted during flight test to provide the safest settings.

NASA’s Space Shuttle orbiter utilized five independent control system computers. Four had identical software. This provided fail-operational, fail-operational, fail-safe (FOFOFS) capability. The fifth computer used a different software program with a “get-me-home” capability as a last resort (often referred to as the “freeze-dried” control system computer).

The Challenge of Limit-Cycles

The success of the new electronic control system concepts was based on the use of electrical signals from sensors (primarily rate gyros and accelerometers) that could be fed into the flight control system to control aircraft motion. As these electronic elements began to play a larger role, a different dynamic phenomenon came into play. “Limit-cycles” are a common characteristic of nearly all mechanical-electrical closed-loop systems and are related to the total gain of the feedback loop. For an aircraft flight control system, total loop gain is the product of two variables: (1) the magnitude of the aerodynamic effectiveness of the control surface for creating rotational motion (aerodynamic gain) and (2) the magnitude of the artificially created control surface command to the control surface (electrical gain). When the aerodynamic gain is low, such as at very low airspeeds, the electrical gain will be correspondingly high to command large surface deflections and rapid aircraft response. Conversely, when the aerodynamic gain is high, such as at high airspeed, low electrical gains and small surface deflections are needed for rapid airplane response.

These systems all have small dead bands, lags, and rate limits (nonlinearities) inherent in their final, real-world construction. When the total feedback gain is increased, the closed-loop system will eventually exhibit a small oscillation (limit-cycle) within this nonlinear region. The resultant total loop gain, which causes a continuous, undamped limit-cycle to begin, represents the practical upper limit for the system gain since a further increase in gain will cause the system to become unstable and diverge rapidly, a condition which could result in structural failure of the system. Typically the limit-cycle frequency for an aircraft control system is between two and four cycles per second.

Notice that the limit-cycle characteristics, or boundaries, are dependent upon an accurate knowledge of control surface effectiveness. Ground tests for limit-cycle boundaries were first devised by NASA Dryden Flight Research Center (DFRC) for the X-15 program and were accomplished by using a portable analog computer, positioned next to the airplane, to generate the predicted aerodynamic control effectiveness portion of the feedback path.[10] The control system rate gyro on the airplane was bypassed, and the analog computer was used to generate the predicted aircraft response that would have been generated had the airplane been actually flying. This equivalent rate gyro output was then inserted into the control system. The total loop gain was then gradually increased at the analog computer until a sustained limit-cycle was observed at the control surface. Small stick raps were used to introduce a disturbance in the closed-loop system in order to observe the damping characteristics. Once the limit-cycle total loop gain boundaries were determined, the predicted aerodynamic gains for various flight conditions were used to establish electrical gain limits over the flight envelope. These ground tests became routine at NASA Dryden and at the Air Force Flight Test Center (AFFTC) for all new aircraft.[11] For subsequent production aircraft, the resulting gain schedules were programmed within the flight control system computer. Real-time, direct measurements of airspeed, altitude, Mach number, and angle of attack were used to access and adjust the electrical gain schedules while in flight to provide the highest safe feedback gain while avoiding limit-cycle boundaries.

Although the limit-cycle ground tests described above had been performed, the NASA–Northrop HL-10 lifting body encountered limit-cycle oscillations on its maiden flight. After launch from the NB-52, the telemetry data showed a large limit-cycle oscillation of the elevons. The oscillations were large enough that the pilot could feel the aircraft motion in the cockpit. NASA pilot Bruce Peterson manually lowered the pitch gain, which reduced the severity of the limit-cycle. Additional aerodynamic problems were present during the short flight requiring that the final landing approach be performed at a higher-than-normal airspeed. This caused the limit-cycle oscillations to begin again, and the pitch gain was reduced even further by Peterson, who then capped his already impressive performance by landing the craft safely at well over 300 mph. NASA engineer Weneth Painter insisted the flight be thoroughly analyzed before the test team made another flight attempt, and subsequent analysis by Robert Kempel and a team of engineers concluded that the wind tunnel predictions of elevon control effectiveness were considerably lower than the effectiveness experienced in flight.[12] This resulted in a higher aero-dynamic gain than expected in the total loop feedback path and required a reassessment of the maximum electrical gain that could be tolerated.[13]

Flight Control Systems and Pilot-Induced Oscillations

Pilot-induced oscillations (PIO) occur when the pilot commands the control surfaces to move at a frequency and/or magnitude beyond the capability of the surface actuators. When a hydraulic actuator is commanded to move beyond its design rate limit, it will lag behind the commanded deflection. If the command is oscillatory in nature, then the resulting surface movement will be smaller, and at a lower rate, than commanded. The pilot senses a lack of responsiveness and commands even larger surface deflections. This is the same instability that can be generated by a high-gain limit-cycle, except that the feedback path is through the pilot’s stick, rather than through a sensor and an electronic servo. The instability will continue until the pilot reduces his gain (ceases to command large rapid surface movement), thus allowing the actuator to return to its normal operating range.

The prototype General Dynamics YF-16 Lightweight Fighter (LWF) unexpectedly encountered a serious PIO problem on a high-speed taxi test in 1974. The airplane began to oscillate in roll near the end of the test. The pilot, Philip Oestricher, applied large, corrective stick inputs, which saturated the control actuators and produced a pilot-induced oscillation. When the airplane began heading toward the side of the runway, the pilot elected to add power and fly the airplane rather than veer into the dirt along side of the runway. Shortly after the airplane became airborne, his large stick inputs ceased, and the PIO and limit-cycle stopped. Oestricher then flew a normal pattern and landed the airplane safely. Several days later, after suitable modifications to its flight control system, it completed its “official” first flight.

The cause of this problem was primarily related to the “force stick” used in the prototype YF-16. The control stick was rigidly attached to the airplane, and strain gages on the stick measured the force being applied by the pilot. This electrical signal was transmitted to the flight control system as the pilot’s command. There was no motion of the stick, thus no feedback to the pilot of how much control deflection he was commanding. During the taxi test, the pilot was unaware that he was commanding full deflection in roll, thus saturating the actuators. The solution was a reduction in the gain of the pilot’s command signal, as well as a geometry change to the stick that allowed a small amount of stick movement. This gave the pilot some tactile feedback as to the amount of control deflection being commanded, and a hard stop when the stick was commanding full deflection.[14] The incident offered lessons in both control system design and in human factors engineering, particularly on the importance of ensuring that pilots receive indications of the magnitude of their control inputs via movable sticks. Subsequent fly-by-wire (FBW) aircraft have incorporated this feature, as opposed to the “fixed” stick concept tried on the YF-16. As for the YF-16, it won the Lightweight Fighter design competition, was placed in service in more developed form as the F-16 Fighting Falcon, and subsequently became a widely produced Western jet fighter.

Another PIO occurred during the first runway landing of the NASA–Rockwell Space Shuttle orbiter during its approach and landing tests in 1978. After the flare, and just before touchdown, astronaut pilot Fred Haise commanded a fairly large pitch control input that saturated the control actuators. At touchdown, the orbiter bounced slightly and the rate-limiting saturation transferred to the roll axis. In an effort to keep the wings level, the pilot made additional roll inputs that created a momentary pilot-induced oscillation that continued until the final touchdown. At one point, it seemed the orbiter might veer toward spectators, one of whom was Britain’s Prince Charles, then on a VIP tour of the United States. (Ironically, days earlier, the Prince of Wales had “flown” the Shuttle simulator at the NASA Johnson Space Center, encountering the same kind of lateral PIO that Haise did on touchdown.) Again, the cause was related to the high sensitivity of the stick in comparison with the Shuttle’s slow-moving elevon actuators. The incident sparked a long and detailed study of the orbiter’s control system in simulators and on the actual vehicle. Several changes were made to the control system, including a reduced sensitivity of the stick and an increase in the maximum actuator rates.[15]

The above discussion of electronic control system evolution has sequentially addressed the increasing complexity of the systems. This was not necessarily the actual chronological sequence. The North American F-107, an experimental nuclear strike fighter derived from the earlier F-100 Super Sabre, utilized one of the first fly-by-wire control systems—Augmented Longitudinal Control System (ALCS)—in 1956. One of the three prototypes was used by NASA, thus providing the Agency with its first exposure to fly-by-wire technology. Difficult maintenance of the one-of-kind subsystems in the F-107 forced NASA to abandon its use as a research airplane after about 1 year of flying.

Self-Adaptive Flight Control Systems

One of the more sophisticated electronic control system concepts was funded by the AF Flight Dynamics Lab and created by Minneapolis Honeywell in the late 1950s for use in the Air Force-NASA-Boeing X-20 Dyna-Soar reentry glider. The extreme environment associated with a reentry from space (across a large range of dynamic pressures and Mach numbers) caused engineers to seek a better way of adjusting the feedback gains than stored programs and direct measurements of the atmospheric variables. The concept was based on increasing the electrical gain until a small limit-cycle was measured at the control surface, then alternately lowering and raising the electrical gain to maintain a small continuous, but controlled, limit-cycle throughout the flight. This allowed the total loop gains to remain at their highest safe value but avoided the need to accurately predict (or measure) the aerodynamic gains (control surface effectiveness).

This system, the MH-96 Adaptive Flight Control System (AFCS), was installed in a McDonnell F-101 Voodoo testbed and flown successfully by Minneapolis Honeywell in 1959–1960. It proved to be fairly robust in flight, and further system development occurred after the cancellation of the X-20 Dyna-Soar program in 1963. After a ground-test explosion during an engine run with the third X-15 in June 1960, NASA and the Air Force decided to install the MH-96 in the hypersonic research aircraft when it was rebuilt. The system was expanded to include several autopilot features, as well as a blending of the aerodynamic and reaction controls for the entry environment. The system was triply redundant, thus providing fail-operational, fail-safe capability. This was an improvement over the other two X-15s, which had only fail-safe features. Because of the added features of the MH-96, and the additional redundancy it provided, NASA and the Air Force used the third X-15 for all planned high-altitude flights (above 250,000 feet) after an initial envelope expansion program to validate the aircraft’s basic performance.[16]

Unfortunately, on November 15, 1967, the third X-15 crashed, killing its pilot, Major Michael J. Adams. The loss of X-15 No. 3 was related to the MH-96 Adaptive Flight Control System design, along with several other factors. The aircraft began a drift off its heading and then entered a spin at high altitude (where dynamic pressure—“q” in engineering shorthand—is very low). The flight control system gain was at its maximum when the spin started. The control surfaces were all deflected to their respective stops attempting to counter the spin, thus no limit-cycle motion—4 hertz (Hz) for this airplane—was being detected by the gain changer. Thus, it remained at maximum gain, even though the dynamic pressure (and hence the structural loading) was increasing rapidly during entry. When the spin finally broke and the airplane returned to a normal angle of attack, the gain was well above normal, and the system commanded maximum pitch rate response from the all-moving elevon surface actuators. With the surface actuators operating at their maximum rate, there was still no 4-Hz limit-cycle being sensed by the gain changer, and the gain remained at the maximum value, driving the airplane into structural failure at approximately 60,000 feet and at a velocity of Mach 3.93.[17]

As the accident to the third X-15 indicated, the self-adaptive control system concept, although used successfully for several years, had some subtle yet profound difficulties that resulted in it being used in only one subsequent production aircraft, the General Dynamics F-111 multipurpose strike aircraft. One characteristic common to most of the model-following systems was a disturbing tendency to mask deteriorating handling qualities. The system was capable of providing good handling qualities to the pilot right up until the system became saturated, resulting in an instantaneous loss of control without the typical warning a pilot would receive from any of the traditional signs of impending loss of control, such as lightening of control forces and the beginning of control reversal.[18] A second serious drawback that affected th