Effective Date: 15 July 99

AEROSERVOELASTICITY

FLIGHT TESTING

Bill Norton

The sensors for measuring flight parameters such as aircraft motion (normal and lateral accelerations, pitch, roll and yaw rate, etc.) are mounted in the aircraft. Since the aircraft structure is not infinitely stiff, these sensors will also measure acceleration and angular velocities resulting from such structural deformation as fuselage bending and torsion. Control surface rotation and elastic modes will also be picked up via surface position sensors. General structural motion can cause what are perceived as uncommanded aircraft motion or control surface deflections. The signals produced by the sensors will then be fed into the electronic flight control system (EFCS) computer which will, in turn, command control surface deflections (for fly-by-wire aircraft or automatic flight control systems) to counter what are perceived to be aircraft rigid body motions or erroneous control surface positions. If the phase lag from the sensor to the control surface is 180 degrees and the system gain is high enough, an instability and sustained surface motion can result. These sorts of feedback are called structural feedback or structural coupling. It is possible for structural feedback to produce large neutrally damped or even divergent oscillations of the control surfaces, resulting in overall system instability and the possibility of structural failure. Even if control surface motion is not divergent, the response can be at a frequency and amplitude that promotes general aeroelastic flutter. This class of problems are addressed in the field of aeroservoelasticity or ASE.

High loop gain is a principle factor in ASE difficulties. Design features that lead to such high gains are high dynamic pressure, relaxed static stability, leading edge control devices, and forward swept wings. In general, any condition where high control system gain is matched with low structural frequencies can aggravate ASE instabilities.

The effects of structural coupling of the kind just described can be reduced by placing the sensors at ideal locations or "sweet spots" within the structure. These are generally locations with the least motion overall, but may also be where particular structural modes are least likely to create feedback problems. Figure 1 shows how a fuselage second longitudinal bending mode has two points of least angular motion and one practical point of least linear motion. The point of least angular motion (the antinode of the motion) is ideal for a gyro sensor, and the point of least linear motion (at the node) is ideal for an accelerometer. Low order modes generally contain the most energy and sensors are usually placed to minimize influence from them. Figure 2 shows how the best location was found for the F-15 SMTD using this criteria, although practical limitations on the placement must also be considered. All of this was accounted for during the design of the X-29 aircraft but problems were encountered when an "oil canning" effect (panel oscillation) of the bulkhead to which the sensor package was mounted produced adverse feedback.

More typical today, the structural signal is simply filtered at the frequency of concern. Because of the bandwidth of flight control sensors and the low energy of high frequency structural modes, structural coupling will usually involve fundamental or other low order modes. Analysis and testing therefore concentrates on these modes. However, it is difficult to filter out low frequency responses like fuselage first and second bending while maintaining adequate gain and phase margins, and acceptable handling qualities.

Figure 1. is an example of Sensor Placement Considerations for Fuselage Second Vertical Bending.

Another example of elastic structures causing control problems was the oscillations of heavy stores on the wings of the F-111 fighter-bomber. An antisymmetric store pitching occurred near the edge of the envelope which, due to the high inertia of the stores, actually caused the airplane to roll. The motion was at a low enough frequency for the pilot to "get in the loop" or accentuate the motion when trying to stop it. Even when the pilot tried to hold the stick centered, the rolling accelerations caused his body to sway from side to side and lateral stick inputs were still made. Releasing the stick was no good because the pendulum mode of the stick was nearly the same as the rolling motion and the stick simply oscillated stop to stop. The only way to recover was to slow down.

The situation in which a pilot may couple with the aircraft elastic motion, acting through the flight control system, is illustrated in Figure 3. In this case the first vertical bending mode of the YA-12A. The elastic response can produce pilot cues or aircraft rigid body motion which can be enhanced when the pilot attempts to damp the oscillations, and a pilot-induced oscillation (PIO) results. In addition to the hazards of PIO dangerous attitudes and airloads - the addition of the aeroelastic element can also mean dangerous structural dynamics.

Figure 3 is an example of Fuselage First Vertical Bending Mode Shape.

Several scenarios producing pilot-in-the-loop structural instabilities are possible. Aeroelastic structural deformations can produce accelerations or attitude changes at the pilot station which results in PIO when the pilot intentionally attempts to counter these dynamics. Also, aeroelastic structural deformation can produce an aircraft rigid body response which results in PIO when the pilot intentionally attempts to counter these dynamics. In another case, aeroelastic oscillations or mechanical vibrations can produce accelerations at the pilot station which the pilot unintentionally couples with, sustaining or enhancing the dynamics. The latter case has been called pilot-augmented oscillation (PAO). All of the factors contributing to pilot-in-the-loop instabilities are shown in Figure 4.

The lower order structural modes generally contain the most energy, produce the greatest structural deflection, and are the most likely to be within the active bandwidth of the flight control laws and the pilot responsiveness. Since the practical limit of a manual pilot input bandwidth is about 3 Hz, the structural modes contributing to an aeroelastic PIO would typically be restricted to this level as well. However, there have been examples of PIO in which the pilot response and the PIO resonance did not share the same frequency. Because PAO is an unintended input created by a physical vibration of the pilot-stick or pilot-throttle system, frequencies above 3 Hz can come into play in such instabilities. Accelerations as low as 0.01g can be sensed by pilots and thus contribute to coupling. Control system physical as well as virtual dynamics can also aggravate these instabilities. The excited and sensed elastic mode may alter the gain and phase characteristics of the control system such as to allow a PIO to develop at a frequency different from the elastic mode’s frequency.

For the case of the YF-12A example shown in Figure 3, the aircraft experience a small amplitude oscillation during aerial refueling. When the pilot attempted to damp the oscillation manually, he excited the fuselage first longitudinal bending mode. This mode produced even greater motion at the pilot station and a mild aeroelastic PIO resulted. An example of an unintentional coupling was experienced on the C-17A during flight testing. A abrupt application of lateral stick excited the 2.2 Hz antisymmetrical wing bending mode. The wing bending and twisting, combined with engine nacelle pitching, produced a oscillatory roll rate superimposed on the steady-state roll response. This ratcheting motion was felt as a lateral acceleration at the pilot station. The pilot then unintentionally fed the oscillation as the motion shook the arm-stick combination (a limb-bobweight influence), generating aileron inputs.

ANALYSIS

The dynamics of the aircraft (equations of motion) are modeled by control theory methods and run in simulations long before the aircraft ever flies. The simulations play a large role in the overall aircraft design to meet the customer requirements, particularly the flight computer software design. This aspect of the design process becomes very critical for "fly-by-wire" aircraft. An attempt is made to predict structural dynamics at this conceptual design stage so that these may be incorporated into the simulation as elements of the transfer functions representing the aircraft dynamics. This is the earliest stage at which structural filters or sensor locations are considered. The analysis will follow the flow indicated in Figure 5 . There are separate program available to check for aeroservoelastic instabilities in the models. Ground Vibration Tests (GVT), Ground Resonance Tests (see next section), and early flight results may cause changes as required to reduce detrimental structural feedback. The structures flight test engineer must be familiar with all of this work and participate to the largest extent possible to ensure that structural effects are considered in the control design, all necessary testing is completed, and potentially dangerous feedback is accounted for.

Figure 5 Aeroservoelastic Analysis Flow Diagram

ASE ground tests should be proceeded by stability tests of the control system alone using aircraft hardware. This may be done as a bench test or using an ironbird. The ironbird has control surface actuators and representative surface masses, to ensure that there is no confusion as to the source of instabilities. MIL-F-87242 (Military Specification) stated that

The static structural coupling ground test described in the passage is called a Ground Resonance Test (GRT, also known as a Structural Mode Interaction, SMI, Structural Coupling or Structural Resonance Test). It is not a true ASE test since the aerodynamics are not simulated. This testing should not be confused with Limit Cycle Tests in which the aircraft rigid body motions are simulated using the aircraft’s equations of motion and fed into the EFCS for general system stability checks.

There are two types of structural coupling tests: open-loop frequency response and closed-loop gain margin tests. The objective of these tests is to determine the gain margin plus the sensor and overall system susceptibility to structural coupling.

In an open-loop frequency response test, each feedback loop of the system is opened separately and a signal (swept sinusoid being common) is introduced at the "downstream" side of the break. The output is monitored at the "upstream" side of the break. The sine sweep signals (linear or logarithmic) may be produced by a ground unit or as a special function of the flight control computer. Tests with a constant input at both large and small amplitudes should be conducted in case of nonlinear response behavior. The open-loop frequency response is obtained using a dynamic analyzer. Open-loop frequency response plots are made for several different input amplitudes to observe the system’s input amplitude sensitivity. The amplitudes and the phase angle of the input and output signals are plotted (transfer function) as a Bode plot. The same open-loop frequency response results would be obtained using either a swept sine input or a random noise input of equivalent intensity over an equivalent frequency range for a purely linear system.

In any linear system the frequency response is not a function of input amplitude. However, in real systems numerous non-linearities are present. These can be such factors as break-out forces, hysteresis, non-linear response, or rate limiting of actuators. An intentionally designed nonlinearity is scheduled gain changes in the control laws. The effect of these non-linearities on system stability can be observed by obtaining the frequency response for several different input amplitudes. Most often, the greatest structural coupling has been found to occur at relatively small amplitudes. Therefore, the amplitudes chosen for use in the frequency response usually represent only a small percentage of that which is possible.

Closed-loop tests involve structural excitations to directly test for structural feedback. The EFCS may be used to excite the structure through control surface rotations using the swept sine signals or similar functions described above. The signal can be sent directly to the control actuators themselves for a purely structural input. Manually induced stick raps and rudder kicks may also been used in this portion of the testing. The closed-loop gain margin test is performed by inserting a variable gain at an appropriate location in the EFCS. The loop gain is increased until a condition of lightly damped or undamped response occurs. The frequency and gain for this condition are recorded and their correspondence with data from the open-loop tests checked. Since normal practice requires a 6 dB gain margin, or twice the nominal gain, without instabilities (see Table 1 for precise gain and phase margin requirements), a simple demonstration of this without actually going to instability is generally sufficient. However, it is best to find the actual gain to produce instability in the event that gains are altered in the course of flight testing - a common occurrence.

Table 1

These tests should be performed in all the different control modes, including degraded modes or failure scenarios. It will probably be necessary to create the conditions of flight artificially with a pneumatic ground test unit attached to pitot-static sensors. Hydraulic power and electrical power to the aircraft will also be required. Every effort should be made to ensure that the hydraulic flow rate is not less than that normally supplied by the aircraft pumps or the control surface responses and actuator stiffnesses will not be faithfully duplicated. The test is often performed with the aircraft in the same arrangement as for a GVT such as flotation and sensors. In fact, the GVT and Ground Resonance Test are often done back-to-back since much of the test and analysis equipment is common to the two tests. The use of electrodynamic shakers to excite the structure is not advised because the shaker armature does not move very far and will also resist control surface motion commanded by the EFCS. This can result in the shaker thruster being driven through the surface.

Tests at different aircraft configurations (stores, fuel, wing sweep, etc.) may be necessary. It is also important that the control laws and structure be as close to the final production form as possible for the test to be valid. If significant changes are made to these in the course of flight testing, it may be necessary to repeat portions or all of the test. It is also critical that a single emergency cut-off switch be available to open all feedbacks and kill the artificial input signal in the event of a divergent control surface oscillation that may damage the aircraft.

Prior to the first flight, the aircraft should undergo taxi tests to greater and greater speeds. This will test system responses to actual operational mechanical inputs (taxiing over the ramp tar strips, engine vibrations, etc.) and observe the effect of automatic gain changes as the airspeed increases.

ASE flight testing is a cross between flutter and basic flight controls (flying qualities) testing. Some of the instrumentation is similar to that used for flutter testing; strain gages or accelerometers for structural response and displacement transducers for control surface motion. These serve to warn the test engineers of any undesirable response due to structural feedback. And, some of the maneuvers are the same as for flying qualities testing.

Testing should begin with the gains at half that found to produce an instability in the GRT or the nominal gain, whichever is lowest. The gain can then be incrementally increased in flight to the nominal condition. This naturally requires a means of changing the gains in flight - a feature usually only available on prototype or research aircraft. A means of rapidly opening the feedbacks loops (or at least reducing the gains if the aircraft is unstable without feedback), as in the GRT, is essential for safe recovery in the event of an instability. In the event of a pilot-in-the-loop instability, a recovery would normally include slowing down and having the pilot either freeze the controls or taking the pilot out of the loop completely by release the controls. Transfer function plots of surface motion and other aircraft responses (rates and attitudes) can be produced point-to-point or flight-to-flight to determine phase and gain margins and for comparison with model and simulator results. These plots will assist in identifying any unusual characteristics that may contribute to or are directly attributable to structural feedback.

This flight testing is usually conducted in association with the flying qualities engineers. During the envelope expansion build-up, the test engineers should watch for any sign of lightly damped control surface or stick motion, a lightly damped rigid body oscillation of the aircraft, or any significant deviation from predicted response as seen on real-time transfer function plots (usually gain and phase margins) or in post-flight analysis. Using a flutter test exciter system, stick raps, doublets, control sweeps made by the pilot, similar inputs made as a function of the flight control computer, or other stability and control test maneuvers are helpful in exciting the structure and generating any potential feedback. Test instrumentation will be a mix of those used in flutter testing and those used by the stability and control engineers. Testing in certain degraded flight control system modes (failure modes) is important because this may affect certain feedback and coupling characteristics.

The fundamental handling qualities test methods are also suitable for revealing aeroelastic pilot-in-the-loop oscillation susceptibility. These methods normally consist of performing high gain handling qualities during tracking (HQDT) tasks such as air-to-air gun tracking. Other basic stability and control tests such as abrupt pullups and push-overs with an attitude capture on return to steady level flight and abrupt sideslips have a higher potential for exciting structural elastic modes which could contribute to a PIO event. However, such maneuvers are seldom performed during HQDT. The sharp manual control inputs common of flight flutter testing will produce the highest manually-induced elastic response and, although also not normally concurrent with a high gain pilot task, might uncover any inherent aeroelastic/pilot resonance. An opportune gust or the inertia effects of store release also have the potential for producing the elastic response that results in an instability. Therefore, the normal envelope expansion testing consist of concurrent structural dynamic and flying qualities testing in the normal build-up fashion with basic tasks performed as early in the test program as reasonable. The early look at operationally realistic mission tasks may also provide insight in this regard.

PIO or PAO susceptibility may be strongly dependent on individual pilot sensitivities and reactions. Therefore, a single unsatisfactory finding among a sampling of pilots, indicating poor handling qualities, should not be dismissed as anomalous. The damping of the basic rigid body modes of the aircraft should be tracked during the buildup and care taken when approaching low damping. The frequencies of structural modes should be tracked in concert with the rigid body modes during an envelope expansion buildup to provide warning of the potential for the coupling of these modes. Pilot inputs should be monitored for signs of PIO or PAO. This requires the close association of the flying qualities and structural dynamics test engineers. Testing should, of course, include failure cases with stability augmentation system turned off (where practical), reversion to mechanical systems, and other such conditions to ensure PIO-free control in these states. Such systems have the potential for artificially damping an elastic mode when active, so an aeroelastic pilot-in-the-loop oscillation may develop when they are deselected.

Should an undesirable system or aircraft response, or the potential for one, be uncovered, the easiest remedy without disturbing beneficial flying qualities is the application of a notch filter. This is either an analog filter of a digital algorithm which seeks to attenuate a signal (output of the roll rate gyro, for example) at a specific frequency to a level that will no longer produce the detrimental system response. As with any changes to the flight controls, a subsequent flight test is mandatory to verify the corrective change and to ensure that other detrimental effects have not resulted (instabilities or undesirable flying qualities). The control system should be designed with structural filters from the beginning using predicted modal responses for the airframe and control surfaces, and these may need to be changed during the course of testing. The data update rate of the EFCS also serves as a form of self filtering. An update rate of 30 samples per second (sps) would effectively ensure that modes above 30 Hz will not be a problem. However, aliasing may cause responses within the sample range from higher frequency modes unless good anti-aliasing filters are used.

Nagy, Christopher J., A New Method for Test and Analysis of Dynamic Stability and Control, AFFTC-TD-75-4, Edwards AFB, California, May 1976.

D’Azzo, J.J. and Houpis, C.H., Linear Control System Analysis and Design, McGraw-Hill Book Company, New York, New York, 1981.

Flying Qualities, Theory and Flight Test Techniques, USAF Test Pilot School, Edwards AFB, California.

"Flight Control System - Design, Installation and Test of Pilot Aircraft, General Specification for," MIL-F-9490D.

Kirsten, Paul W., "Flight Control System Structural Resonance and Limit Cycle Oscillation," in AGARD Conference Proceedings No. 233, Flight Test Techniques.

REQUIREMENTS

Dynamic Flight Condition Check