DESCRIPTION
DESCRIPTION
The effects of an unstable oscillation in an aircraft structure is seen as a "flutter mechanism"; a complex interaction of forces and moments, structural dampings, and structural stiffnesses. The change in the intensity and the distribution of airloads with associated structural deformation and restoring motion can result in a cyclic phenomena. Small perturbations about the steady-state flight attitude and small elastic structural deformations create the unsteady flow. Load factor, angle of attack, and the nonlinear lift curve slope have also played in flutter mechanisms. As flutter begins, it is an oscillation of increasing amplitude with each successive cycle. If the oscillation is allowed to persist, the flutter will continue until some structural restraint is reached or until catastrophic structural failure occurs. Figure 1 .(Collar's Aeroelastic Triangle of Forces) shows how all the forces affecting a flight vehicle interact to form the aeroelastic and aeroservoelastic problem. Each effect will be dealt with to some degree in this handbook.
In the case of a multiple degree of freedom system, the flutter mechanism is a coalescence of two or more natural vibration modes of the structure. The phase angles and frequencies of the individual modes shift in a manner that allows the coupling; a change from no net work of the system to positive net work. Even with coupled modes, the structural displacements reach their maxima or minima simultaneously. A coalescence of two modes is called binary flutter. If three degrees of freedom (DOF) contribute to a flutter state then it is termed ternary flutter. However, flutter of more than two DOF is rare. Flutter resulting from the coupling of a structural mode and an aircraft rigid body mode (aircraft pitch, roll, yaw, and plunge or vertical translation) is termed body freedom flutter.
Assuming that the inertia and stiffness of the structure does not change in flight (neglecting heating effects and short term fuel burn), the overall system damping will change with a change in the magnitude and frequency of the aerodynamic forces. The natural frequency of all modes will change as the aerodynamic forces change (increased flow velocity, for example) with the frequencies contributing to the flutter mechanism approaching each other and the phase shifting toward an unstable state. The damping of one of the modes move towards zero. When the two modes finally coalesce to form the "flutter mode" the resulting mode shape will be a combination of both individual modes. The energy from each mode will now contribute to an increase in the net work done by the system.
The change in frequency and damping as velocity increase, and any mode coalescence, can be illustrated in V-G and V-f plots like that shown in Figure 2 (Example V-G and V-f Diagrams). The slope of a crossing of the 0.0G line is important. A very steep crossing implies explosive flutter or that which happens extremely quickly and with little warning because the damping drops very rapidly as the flutter speed is approached. A shallow crossing would provide more warning and may be recoverable once it begins. The airspeeds at which such shallow crossings occur should not be taken as precise. Small errors in the analysis that may raise or lower the modal line will result in relatively large changes in the flutter speed associated with the mode. Modes that cross shallowly and then recross to the stable side of the plot are called "hump" or "dome" modes and may produce Limit Cycle Oscillation but is unlikely to produce explosive flutter. Hump modes are most common in store flutter and may be aggravated by elevated acceleration (g) loads.
Normal aluminum aircraft structures have about 2% of inherent structural damping (G). Composite aircraft structures have about 1% inherent damping. Just the mass of the air and the "centering tendency" of a high velocity air stream will tend to damp aircraft structural oscillations. This is known as aerodynamic damping. The frequency content of unsteady aerodynamics will result in an oscillatory forcing function which may act positively (i.e., add to the system damping) or negatively (subtract from the damping) depending upon modal interaction. The net value of structural damping and aerodynamic damping is called the "aeroelastic" or total (system) damping. Total system damping generally tends to increase with an increase in airspeed until damping in a "critical mode", or the mode with the lowest flutter speed, begins to decrease rapidly. At this point the critical mode begins to extract energy from the airflow and total system damping starts to decrease.
At the flutter speed, the damping of the mode is zero and the oscillation is just able to sustain itself without diverging (amplitude unchanging). Just a small addition of energy to the system by an increase in speed or a small disturbance can trigger the oscillation to diverge. When a flutter mechanism is active within a structure, overall system damping is negative. The aerodynamic damping has completely negated the structural damping and begins to add energy to the system. Once it starts, the motion is self-sustaining and requires no additional external forces. The amplitude of the cycles begins to increase over time rather than decrease as in the case of positive damping. This situation is termed "divergent flutter" and will eventually cause the structural limits of the component experiencing the flutter to be exceeded. The oscillations are very likely to damage the structure, possibly within only two or three cycles. Because of this, flutter is extremely dangerous.
Because control surfaces are specifically designed to rotate, they can be easily excited by aerodynamic forces or other structural motion. This makes them very important components to consider in flutter prediction and during flight testing. The rotation frequency of the control surface is of paramount concern. For surfaces moved by actuators, the rotation frequency is primarily contributed by the stiffness of this equipment. For surfaces moved by cables or push rods, the mass distribution of the surface will determine the rotation frequency. This can be altered as necessary by changing the balance weight. This is a concentrated mass placed ahead of the hinge line. Mass balancing is necessary every time there is a change to the surface, such as repairs or repainting. Balance weights are almost universally used even on surfaces driven by actuators to ensure against flutter should the actuator pressure decay due to a hydraulic failure.
Standard practice requires analysis to demonstrate, and flight test to verify as much as possible, that the flight vehicle is free of flutter or other aeroelastic instabilities to 1.15VL, or a 15% margin on the design limit flight speed (VL). This provides a safety buffer on the normal aircraft red line in the event of over-speed. The damping (G) of any critical flutter mode or other significant dynamic response must be at least 0.03 at VL. Some configurations may require a reduced flight envelope for flutter-free operation.
PRELIMINARY GROUND TESTS
Because mathematical modeling of the aerodynamics and structural properties of an aircraft are so subject to uncertainties, many ground tests are done with models and the full scale article prior to the beginning of flight testing. The results of such tests are used to improve the computer models and refine the estimates used in planning the flight test. This provides testers with more understanding about the aircraft prior to flight and reduces overall risk. The entire development flow is shown in Figure 3 . (Diagram of Flutter Investigation During Development)
WIND TUNNEL TESTS
An important preliminary test conducted prior to flutter flight test of an aircraft is a wind tunnel flutter program. In the wind tunnel an elastically and mass distribution-scaled model of the aircraft is subjected to an airstream that has also been scaled to the same dimension (i.e. reduced frequency number) as the aircraft. Only a partial model (tail surfaces removed, for example) may be used in an effort to isolated effects. The scale model of the entire aircraft may be "flown" either free or on bungees at speeds corresponding to the envelope of the real aircraft, and the flutter mechanisms verified as much as possible. It is common for the scale airspeed (occasionally produced with a gas other than air) to be taken beyond the normal envelope to verify the flutter margin. Definitely checking for the critical mode by observing the flutter (hopefully without destroying the model) is also done. Testing and data reduction methods are much the same as for full scale flight tests. General tunnel turbulence, upwind oscillating vanes, and external inputs (tapping the surface with a rod) are common excitation means. Mass, stiffness and shape changes can be made relatively easily and the effects tested on the model before being incorporated in the actual design. The wind tunnel data is always subject to errors involved in accurately modeling the aircraft and the airflow. Corrections to the wind tunnel data for tunnel effects and compressibility are typically performed, to include blockage effects (velocity errors).
GROUND VIBRATION TESTS
One of the typical essential preliminary tests conducted on an aircraft is the ground vibration test (GVT). This test may be performed at an early stage on the wind tunnel model discussed in the previous section to ensure that it adequately replicates the desired stiffnesses. A full scale GVT is mandatory for new designs and for any substantial changes to an existing aircraft. This test is used to verify and update the flutter model as well as provide a means of identifying modes from frequencies found in flight test data.
FLUTTER FLIGHT TESTS
Flutter flight tests of aircraft are performed to verify the existence of the flutter margin of safety estimated by analysis. Flight near the estimated flutter speed entails a risk of accidentally encountering the flutter sooner than anticipated, with the potential for a catastrophic structural failure. For this reason, the flight envelope is cleared for flutter in an incremental fashion from a considerably lower subcritical airspeed. The frequency and damping of the vehicle's response to an excitation are measured at successively increasing increments of airspeed. The frequencies permit an identification of the modes of the response by comparison with ground test and analytical data. If sufficient instrumentation is installed, actual mode shapes can be determined to supplement the frequencies in performing modal identification. This is done by comparing phase and magnitude of responses from the transducers across the structure. The damping provides a quantitative measure of how near each airspeed point is to flutter.
The structural response is typically measured with accelerometers or strain gages. Pressure transducers and velocity sensors have also been used. Anything that shows the dynamic response of the structure from which frequency and damping can be determined is suitable. Strain gages have the advantage of naturally filtering high frequencies which may not be of interest, and therefore can provide "cleaner" data than accelerometers. However, this also may make them unsuitable for adequately resolving modes greater than about 7 Hz. They will also not show rigid body aircraft dynamics (pitch acceleration, for example) which the accelerometers will tend to show. A combination of accelerometers and strain gages has often provided the best flutter instrumentation in the past.
EXCITATION
Several methods have been used to excite the aircraft structure to permit the individual response modes to be observed and their frequencies and dampings obtained. Some methods are more suitable than others for certain frequency ranges, nature of the modes sought to be excited, and amplitude or amount of energy to be put into the modes. Experience will also help in selecting of the best method to use. Whatever the method used, it is important that the excitation be applied as close to the surface of interest as possible. For example, an impulse at the tip of the wing may not have much chance of sufficiently exciting a high frequency horizontal tail mode for adequate analysis. The position of the excitation source relative to the node lines of the significant modes is also critical. Applying an impulse directly to the wing elastic axis has no chance of producing any torsional displacement. But, an impulse off of the elastic axis will produce both torsional and bending excitations. An impulse close to the node line will not produce as much displacement as the same amplitude of impulse farther from the node line by virtue of the longer moment arm.
PULSE
Pulses or raps are sudden control impulses made to produce sudden control surface movements that can excite up to 10 Hz modes very well, and occasionally to 30 Hz not so well. The objective is to produce an impulse that most closely approximates a "step" input since such an input contains the highest frequency content possible. Therefore, the sharper the input, the better the data. Longitudinal stick raps, lateral stick raps, and rudder kicks are the most common pulses used. The choice of pulse direction depends upon the modes sought to be excited.
For pulse excitation, the pilot will stabilize the aircraft on condition and make the rap. The stick pulse can be made by striking the stick with the palm of the hand or with a mallet. For irreversible control systems, above a certain level of input abruptness the flight control system or control surface actuator may attenuate the input and give no additional frequency content. Only trial inputs at a benign flight condition will demonstrate the best input to expect. Aircraft with automatic flight controls can be programmed to make these pulses without pilot intervention.
Sometimes a rapid stick movement, with the stick held in the usual manner, (a "singlet") is sufficient excitation. More energy may be generated by a "doublet" which are two such inputs in rapid succession in opposite directions. But, the singlet and doublet can only be done at a low frequency compatible with human limitations - about 4 Hz for stick inputs and less for rudder pedal inputs. A control oscillation is a variation of the doublet but with some specified time between the two inputs. An example may be a one inch aft stick deflection, hold for three seconds, and a return to neutral. A time to produce the control input, a "ramp-up" time such as one second for the one inch stick deflection, may be specified to ensure that the mode is adequately excited. The oscillation is tailored to excite a specific mode, such as a 3 Hz fuselage first vertical bending mode for an elevator input.
The pilot may deflect the control surface and then fly "hands-off" to allow the aircraft oscillations to naturally damp out ("stick-free" pulse) or the pilot may arrest the stick as it returns to the neutral position ("stick-fixed" pulse), depending upon which technique yields the best response from the aircraft. A variation is to hold the stick as the rap is made. For a large aircraft a rap may not produce enough control deflection for sufficient excitation. In this case a rapid, full-deflection displacement of the control followed by a release is useful.
It may be desirable to make several raps and to average either the time histories or fast Fourier transformations (FFTs) of the raps together to remove noise and other corrupting influences. The principle deficiencies of the pulse technique is the non-selectivity of the frequencies to be excited and the generally poor energy content above about 15 Hz.
SWEEP
This is the most attractive and preferred excitation means but is also one of the most complex and expensive to implement. Sweep data and burst data (next section) requires the use of an excitation system to "shake" the aircraft or individual surfaces in flight. Frequencies as high as 65 Hz have been successfully excited by this method. A mechanical "exciter" system added to the aircraft or the vehicle's control surfaces are generally used to make the sweeps. The frequency response characteristics of the exciter system or control surface actuators will determine the modal frequencies which can excite. Hydraulic or electric actuators generally show an attenuation of respond amplitude beyond 30 Hz, but the excitation in this part of the spectrum can lead to responses at high frequencies. This method of excitation requires some time and thus may be unsuitable if the test condition can only be obtained briefly, such as in a dive. Several different types of sweep exciters have been used.
B-1 flight testing included use of a "wand" at the wing and tail extremities (see Figure 4 ). The wand consists of an mass placed at the end of a pivoted arm attached to the extremity of the structure. The oscillation of the mass about the pivot point produces a periodic structural response by virtue of the inertial displacement of the mass. Other such inertia exciters use out-of-balance rotating masses. One drawback to this technique is the disturbance in the airflow if the wand extends beyond the surface of the aircraft. Another disadvantage is that larger masses are needed to excite lower frequencies, and the overall system weight may become prohibitive or alter the modal characteristics of the structure to which they are attached.
The French have done quite a bit of work with electrodynamic exciters. Akin to electrodynamic shakers used in GVTs, a mass is suspended within the structure by springs. The mass incorporates coils that are in close proximity to coils attached to the structure (see Figure 5 Electrodynamic Exciter). By sending an alternating current through the coils, the mass can be made to move within the electromagnetic field, with the frequency of the movement is proportional to the electrical signal. Although successfully used on many European projects such as the Transall and Concorde, the advantages and disadvantages are much the same as those for the other systems discussed here, with the added difficulty of access once the units are enclosed within the aircraft structure. Also, greater moving mass is required to adequately excite lower frequency modes.
Oscillating vanes or "flutterons" mounted on wings and tails have been used many times, including the C-17 test program (see Figure 6 ). The vanes create a varying aerodynamic force that acts on the aircraft structure. Several sizes and masses of vanes may need to be available as the speed and dynamic pressure at different points in the envelope alter the effectiveness of the inputs. The great advantage of this system is that it can easily excite low frequency modes and higher frequencies are only limited by the frequency response of the vane actuator. Below 3 Hz the input may not be sufficient to adequately excite the structure under test. A disadvantage is the change to the mass distribution of the surface upon which the system is attached and the disturbance to the normal airflow that its presence and operation engenders. Interfaces with the normal aircraft electrical and hydraulic systems may also be costly. Also, when commanded to stop, the vanes may 'wind-down' and input forces at undesirable frequencies. An abrupt stop may also be detrimental as it would input an impulse load with broad frequency content. The stopping characteristics of the vane system should be examined during test preparation.
A variation on the oscillating vane is a rotating slotted tube. This system produces oscillating airloads on the tube but does not add as much weight or disturb the airflow as much as a vane. It was used quite successfully on NASA's testing of the F-16XL.
For "fly-by-wire" aircraft, using the flight control system as an excitation means has the attraction of requiring no additional hardware to that of the basic aircraft. The oscillatory signal to the control surface actuator(s) is produced by a special function of the flight control computer or injected into the loop by a separate unit. The pilot is generally provided the means of varying the sweep rate, end frequencies, and amplitudes of the inputs. Random inputs by this means is also possible. There is difficulty in exciting high frequency modes because the attenuation or rate limiting characteristics of the actuator will bound the displacement or level of excitation in this part of the spectrum. This method has been used with great success on the various F-16 test programs and the B-2 bomber.
When sweep data is collected, the excitation system is run through one or more frequency sweeps at a predetermined rate. Common rates are linear, such as 1 decade/sec, or a logarithmic variation. The proximity of modes and the potential for noise problems may effect the selection of sweep rate. Sweep data is normally analyzed on a computer to determine frequencies to use for a burst-and-decay method or to stand on its own. This method is analogous to conducting a GVT in flight, measuring a forced response of the structure. Its greatest advantage is to directly apply energy into modes that may otherwise be very difficult to excite with sufficient energy by other means.
Sweeps have a tendency to increase the noise from local nonlinearities. A sweep performed too quickly will also tend to make the modes appear to have a slightly greater frequency and damping value than they really do. If there is no transient response evident following termination of the excitation, the response is termed "deadbeat".
BURST-AND-DECAY
Burst-and-decay data is collected by running the excitation system at a certain frequency and then shutting it off to allow the structure of the airplane to damp itself out. This is typically done at the frequency that will excite the critical modes(s) to specifically check the damping of these responses. Generally, only a few cycles are necessary before stopping to allow the decay. For a mechanical excitation system, the wind-down characteristics - that is its ability to cease movement when commanded without overshoots - becomes important to how clean and faithful the resulting decay data can be.
PYROTECHNIC
The so called "bonker" is a very small pyrotechnic charge that is typically placed externally near the trailing edge of a control surface and detonated electrically. The tiny explosion produces a sharp pulse excitation that comes very close to simulating the ideal step input. A series of such charges may be placed along the surface trailing edge so that many such impulses may be produced on a single flight. The disadvantage of the method is the limited number of excitations that can be produced in a flight and the disturbance of the normal airflow that the presence of the charges and their wiring create. The bonker has seen wide use in Europe on such programs as the Airbus, Jaguar and Tornado.
AIR TURBULENCE
Random air turbulence can be used as a source of excitation. Even if not felt by the pilot, there is always some energy content to an air mass. Generally, low altitudes have a greater magnitude of this excitation. Deserts have many rising columns of air and high winds that, when flowing over mountains, produce a good deal of instability in the air. Weather fronts also produce turbulent air. This technique, while attractive from a cost standpoint, must be used with prudence. There is generally very little energy to be obtained in the spectrum above about 20-30 Hz and some structures will not respond to turbulence as well as others because of differences in wing loading. Also, some turbulence is not "random," but has dominant frequencies.
The normal procedure is to have the aircraft stabilize on condition and collect response data for some pre-determined time period, as much as two minutes. The resulting data (the response of the aircraft to the random inputs) is collected on the computer and analyzed to determine the frequencies and dampings of the observed modes. A "near real-time" computer analysis is required for point-to-point clearance if dampings must be tracked. The approach to an instability may sometimes be seen by a "burst", a brief period of lowly damped, higher amplitude response, or by Lissajous figure indications. This is referred to as incipient flutter.
Another approach using turbulence is to measure the gusts using a "gust probe". This may be very sensitive alpha (angle of attack) and beta (sideslip) vanes on a nose boom, or pressure sensors at the tip oriented to allow measurement of vertical and lateral gust components. An accelerometer or strain gage may be mounted on the boom to allow rigid body aircraft and boom dynamics to be removed. The result is a measure of the input forcing function, with the resultant aircraft structural dynamics measured in the usual way. Knowing the input and output allows a transfer function to be formed. This provides a powerful means of predicting modal response at other points in the flight envelope.
Deployment of speedbrakes, landing gear, flaps or other device on the aircraft that will produce a great deal of buffet vibration, and even flying in the wake of another aircraft, have been also used for turbulence excitation. Such techniques are likely to introduce a dominant response representing the primary vibration frequency of the device (vortex shedding rate off of the speedbrake, for example) which may mask true structural responses.
PROCEDURES
In the normal conduct of a flutter test, a build up procedure is used whereby less critical points are flown prior to more critical ones. This technique allows the engineers to determine damping trends as dynamic pressure and Mach number are increased. These two parameters are normally treated separately. Because high Mach numbers are generally only possible at high altitude, and high q's are obtained at low altitudes, there is usually a considerable altitude difference between the two test conditions. If the critical mode(s) are suspected to be altitude-critical then a build-up in altitude may also be necessary. In any event, a minimum safe altitude for the testing should be established. If q is determined to be more critical than Mach (as is commonly the case), the Mach points may be flown first, although jumping between the two conditions in the course of the testing for the purposes of a concurrent build-up in both parameters is not unknown (see Figure 7 Flutter Expansion Example). Because q is so often the determining factor in flutter susceptibility, equivalent airspeed is generally used for velocity references during flutter flight testing. Testing across the transonic range is important because of the significant aerodynamic changes which occur there. Having the pilot maintain precise airspeed control and using corrected airspeed readings (calibrated airspeed to match a target equivalent airspeed) is essential to avoid over-speeding a test point and creating a potentially hazardous situation. A careful airspeed calibration must precede or be conducted concurrent with the flutter program. However, the error in the damping and frequency values determined from flutter data is generally great enough to make any data corrections in airspeed and altitude for pitot-static errors superfluous.
Many aircraft airspeed-altitude envelopes have an inflection point at the high-speed end (the "knee" of the curve). This is normally a good mid-altitude to perform another build-up combining Mach and q effects. The build-up at this altitude would required prior Mach and q clearances at high and low altitudes. An example of envelope clearance flutter test points is shown in Figure 7 . A separate, perhaps more abbreviated, build-up to the point in the airspeed-altitude envelope where damping of a critical flutter mode is predicted may also be advisable.
Generally, the build-up will consist of points of ever-increasing airspeed. The airspeed increments between points will depend upon the proximity to the predicted flutter boundary and the confidence in the flutter prediction. Smaller steps are required when close to a flutter condition or where a rapid decrease in damping is observed. Naturally, some practical airspeed must be selected to begin the buildup since the aircraft must takeoff and climb to the test altitude. The choice of this airspeed must be based upon a conservative review of the predicted flutter modes and flutter margin. A reasonable tolerance on airspeed and altitude must be established and published in the test plan. Naturally, no over-speed at the aircraft maximum airspeed is permissible. The test conductor must remain flexible and be prepared to take smaller airspeed steps than the test cards dictate if low damping is observed. This decision can be facilitated by real-time analysis of the data, either on the stripchart, computer, or both, prior to each successive test point. For the purpose of other testing, the aircraft is restricted to the envelope that has been cleared by flutter testing. This is necessarily much less than the ultimate flight envelope of the aircraft, but will slowly be expanded to this objective as testing proceeds.
The maneuvers used in flutter testing depend upon the type of data to be collected. If sweep, burst-and-decay, or random data is acquired, the aircraft will stabilize on condition (i.e., fly a constant airspeed and altitude) for the time required. Sometimes, when dive test points have to be flown, the aircraft will reach a target airspeed, and the random, burst, or sweep data will be taken between a band of altitudes to prevent significant changes in the air density (i.e., dynamic pressure). In some test programs an airspeed schedule is flown to ensure that the aircraft flies along a given flight profile and does not overshoot its target airspeed, q, or Mach. This is usually done when the test conductor wishes to verify flutter free flight along the redline airspeed curve after clearing to this speed at discrete altitudes. Instead of flying many stabilized points at many different altitudes, the plane is dived along this "line" and brief excitations made to verify that no lowly damped responses are seen. The whole line can be cleared in one test maneuver. This technique is good for demonstration purposes but is not advised if useful damping values are sought. It should only be done after the test team is reasonably confident that flutter will not occur along the line, because the dive does not allow as quick a recovery from a flutter event as a straight and level condition.
If personnel at the ground telemetry station or the pilot(s) see flutter beginning, immediate action must be taken to save the aircraft and possibly the aircrew. The command "Terminate, Terminate, Terminate," or similar words like "Abort" and "Stop" are transmitted to the aircraft. A flutter engineer monitoring the aircraft response should have a direct radio link to the aircraft to reduce the reaction time needed to have someone else communicate a termination command. The immediate cockpit action must be to pull the engine power back to slow the aircraft from the critical speed as quickly as possible. Deploying speedbrakes (airbrake) or similar devices may also be advisable, but caution is advised as some speedbrake locations may aggravate flutter. Assuming a slight nose high attitude will also assist in slowing the aircraft. This last maneuver is not advised when store loadings are involved since g-loading has been known to aggravate store flutter. A normal load factor limit should be specified in the test cards. If a loaded maneuver is in progress, unloading and returning to straight and level flight is required. For aircraft with podded engines, a throttle chop to slow down may impart forces to the wing structure which may be undesirable. These considerations point to a need to analyze the consequences of recovery procedures with regard to the most likely flutter modes(s). Mechanical excitation systems should be immediately arrested if in action. It is important that the pilot(s) have direct means of stopping the exciter, usually with a switch on the control column or throttle. The structural oscillations may not immediately die out when "backing off" the test condition as a chance input may have triggered the response beyond a lower possible onset speed.
A safety chase aircraft is almost always mandatory for flutter testing. Because flutter testing is typically labeled as at least Medium Risk, and usually High Risk, only minimum flight crew is permitted. The airspeed system should have been calibrated so that a confidence of no more than 2 to 3 kts in the airspeed readings exists. The aircrew should be cautioned against over-speeding the test points. Such testing is normally conducted in clear air without turbulence (unless used as an excitation source) since it may corrupt the data and add to analysis difficulties. Since the fuel mass in wing and fuselage tanks can significantly effect modal responses, analyzing and testing the aircraft at various fuel states may be advisable. In such cases, tolerances on fuel quantities in the tanks may need to be dictated.
Since the impulse caused by the sudden release of stores may also excite flutter, this should also be considered in analysis and perhaps flight tested. It is also usually advisable to perform separate flutter testing with any control augmentation system in both the ON and OFF states in the event that these systems are adding damping that is hiding a potential flutter response. Consistent and proper servicing of the aircraft is important to avoid shifts in modal frequencies and dampings. The torquing of bolts holding major components on the airframe, such as external stores, can be an important factor in stiffnesses. Wear of such components as hinges and bearings can change rotational frequencies.
DATA REDUCTION
The results of data analysis are typically airspeed (V, M, or q) versus frequency (V-f) and airspeed versus damping (V-G) plots for the modes being tracked. Clearly, some prior knowledge of modal frequencies are required (from prior modal analysis) to identify modes from the flight test data unless sufficient instrumentation has been installed to allow actual mode shapes to be determined (rarely the case). If some measure of the structural input is available, then more detailed systems analysis is possible using transfer functions.
Data is used to establish a safe flight envelope for the aircraft based upon extrapolation of the damping trends to the case of zero damping (onset of flutter) and application of the 15% safety margin in airspeed below the predicted "flutter airspeed". Usually, the aircraft is not tested at damping values less then 2% damping or 0.02 G (though 0.03 G is often specified). During stores testing of the F-16, damping as low as 0.01 G was permitted because of a frequently encountered Limit Cycle Oscillation (LCO, 0.0 G).
The extent of the analysis of real-time data is at the discretion of the flutter test director. In some cases, an experienced test director may be able to determine the damping by "eye-balling" the stripchart traces without extensive analysis. This would be especially true for aircraft which have already undergone considerable flutter testing, such as a fighter-bomber subjected to many external store clearance tests. Some aircraft or modes are more easily excited than others and stripchart traces alone may be adequate for a point-to-point clearance. In other cases, particularly where more than one mode is present, more detailed analysis, perhaps involving computers, is required. Real-time analysis or even display of every flutter parameter may not be technically or practically feasible. The flutter engineer must apply his or her best judgment as to the most critical parameters. If stripchart traces show damping of 0.05 G or less, or when more than one mode is present, more detailed frequency domain analysis is advisable.
If it is important to track damping of critical modes (both predicted and those found to have decreasing damping) for point-to-point flight clearance, then a "real-time" analysis will be required before clearing the pilot to accelerate to the next test condition. The dominant mode appearing in the data is not necessarily the most important mode to watch. A mode that is not easily excited or seen may still have low damping and is waiting to make life interesting. Typically only one of the two participating modes will show the rapid drop in damping as flutter is approached, so it is important that the frequency shifts and damping of many modes be tracked. Detailed analysis between flights should always be done using all available data to track frequency and damping of significant modes with the best precision available in an effort to predict flutter onset from the trends.
Modal frequencies and dampings can be extracted from the data using any number of methods. By using an exciter system, it is possible to obtain an input function for transfer functions and particularly coherence plots. However, it is important to remember that the exciter signal or output is not the true or even the only forcing function acting on the aircraft. The changes to the airflow on the aircraft or the impulsive inertia change is the true structural input. Turbulence, general aerodynamic forces, and miscellaneous pilot and systems inputs will still be present as uncorrelated signals. These will impact the quality of the plots and may lead to erroneous conclusions.
A method for extrapolation of flutter test data by employing a two degree of freedom math model to predict the flutter onset has shown promise. This has the advantage of using actual test data to provide a measure of the flutter margin from the last test condition flown. Although employing many limiting assumptions, these flutter margin calculations provide the flight test engineer with another tool for safe flutter clearance.
STORE FLUTTER TESTING
The addition of external stores on the aircraft and the many download variations have a tendency to reduce flutter speeds. The modes that participate in a store flutter event are generally low-order wing modes. Stores testing is generally restricted to the subsonic range due to aircraft limitations. Flutter speed tends to increase with asymmetric loads and modes become more difficult to excite in such configurations. When doing a symmetric stores loading test, it is important that the loads are as truly symmetric as possible, that is that everything on one side of the airplane weighs the same as everything on the other side. This should be verified by careful mass properties tracking of stores, racks, and pylons. Fuel distribution should also be watched carefully. The aircraft-to-pylon, pylon-to-rack/launcher, and rack/launcher-to-store interfaces greatly increase the nonlinearities which can create analysis difficulties. Care should be taken to reduce these "rattles" as much as possible. It is critical that such attachment fittings as rack sway brace bolts be torqued down similarly and to maintenance specifications. Since the flutter speed of external store loading configurations are occasionally influenced by acceleration loads, such maneuvers as "wind-up turns" may be used to produce these accelerations.
Exciters have been incorporated directly into the stores or pylon/racks/launchers. This helps to ensure that the store is adequately excited without having to be concerned with amplitude or frequency attenuation through the intervening structure between the exciter and the store. A vertical and lateral accelerometer at the tip of the store, perhaps at the front and rear, and maybe on the pylon, are very common instrumentation installations (see Figure 8 Example Store Flutter Instrumentation Installation).
Serious store oscillation or flutter problems can be remedied in whole or in part by the addition of isolators between the various components, or between the airplane and the store components. These usually take the form of dashpots to change critical frequencies, such as missile launcher pitch, and reduce the potential for adverse coupling. Slight changes in the store location can produce beneficial mass redistribution or move the store out of an adverse flow field.
REFERENCES
Jones, W.P., ed., Manual on Aeroelasticity, AGARD Vol. V, London, England, 1960.
Van Nunen, J.W.G. and Piazzoli, G., Aeroelastic Flight Test Techniques and Instrumentation, AGARD-AG-160, Vol. 9, AGARD, London, England, February 1979.
Zimmerman, Norman H., Prediction of Flutter Onset Speed Based On Flight Testing at Sub-Critical Speeds, in AIAA/AFFTC/NASA-FRC Conference on Testing of Manned Flight Systems, 4-6 December 1963.
Military Standard Airplane Strength and Rigidity, Flutter, Divergence and Other Aeroelastic Instabilities, MIL-A- 008870.
Aeroelasticity, Chapter 12, Flying Qualities Theory and Flight Test Techniques, USAF Test Pilot School, Edwards AFB, California.
Norton, William J., Captain, USAF, Random Air Turbulence as a Flutter Test Excitation Source, Society of Flight Test Engineers 20th Annual Symposium Proceedings, September 1989.
Norton, William J., Captain, USAF, Limit Cycle Oscillation and Flight Flutter Testing, Society of Flight Test Engineers 21st Annual Symposium Proceedings, August 1990.