Dry wind tunnel system

(12) (lo) Patent

United States Patent No.: US 8,393,206 B1

Chen (45) Date of Patent: Mar. 12, 2013

(54) DRY RIND TUNNEL SYSTEM Primary Examiner Andre Allen (76) Inventor: Ping-Chih Chen, Scottsdale, AZ (US) (74) Attorney, Agent, or Firm Keith L. Jenkins, Registered Patent Attorney, LLC; Keith L. Jenkins (*) Notice: Subject to any disclaimer, the term of this patent is extended or adjusted under 35 U.S.C. 154(b) by 321 days.

(57) ABSTRACT (21) Appl. No.: 12/702,510 This invention is a ground flutter testing system without a (22) Filed: Feb. 9, 2010 wind tunnel, called Dry Wind Tunnel (DWT) System. The DWT system consists of a Ground Vibration Test (GVT) (51) Int. Cl.

hardware system, a multiple input multiple output (MIMO) G01M9100 (2006.01) force controller software, and a real-time unsteady aerody- (52) U.S. Cl . ......................................................... 73/147 namic force generation software, that is developed from an (58) Field of Classification Search ........................ None aerodynamic reduced order model (ROM). The ground flutter See application file for complete search history.

test using the DWT System operates on a real structural (56) References Cited model, therefore no scaled-down structural model, which is required by the conventional wind tunnel flutter test, is U.S. PATENT DOCUMENTS involved. Furthermore, the impact of the structural nonlin- 4,327,581 A * 5/1982 Jackson et at ................... 73/147 earities on the aeroelastic stability can be included automati- 4,372,157 A * 2/1983 Caruthers et at ................ 73/147 cally. Moreover, the aeroservoelastic characteristics of the 4,372,158 A * 2/1983 Doggett et at . ................. 73/147 4,372,159 A * 2/1983 Doggett et at . ................. 73/147 aircraft can be easily measured by simply including the flight 4,665,578 A * 5/1987 Kawada et at . ................ 14/74.5 control system in-the-loop. In addition, the unsteady aerody- 4,862,739 A * 9/1989 Dobbs ............................. 73/147 namics generated computationally is interference-free from 7,942,049 132 * 5/2011 Fritz et al ........................ 73/147 the wind tunnel walls. Finally, the DWT System can be con- FOREIGN PATENT DOCUMENTS veniently and inexpensively carried out as a post GVT test JP 04106445 A * 4/1992 with the same hardware, only with some possible rearrange- JP 05079944 A * 3/1993 ment of the shakers and the inclusion of additional sensors.

JP 2002168727 A * 6/2002 RU 1172362 C * 9/1994 1 Claim, 6 Drawing Sheets * cited by examiner

Gb'T Hardware

Data Acquisition 5ystern

T,_._,_,_,_._. _._,___._. ._._._._._

-----------

i

i Real-time unsteady i

mI MC7 Farce Controller ; i

aererlyna}r#ic farce

5 i3€t

I generation set-ware

►

} 400

} 3 } Computer ; # ;

U.S. Patent Mar. 12, 2013 Sheet I of 6 BI

US 8,393,206 Hardware Data Acquisition System ^ ~ ^ , ~ ` no/rvxCIrurwe Controller ~ , 5un/ ~ ~ ^ ~ ` ^Computer ^ " 300 Figure I

Mar. 12, 2013 Sheet 2 of 6 U.S. Patent US 8,393,206 B1 Figure 2

U.S. Patent Mar. 12, 2013 Sheet 3 of 6 US 8,393,206 BI GVT Hardware 2c Figure 3

U.S. Patent Mar. 12, 2013 Sheet 4 of 6 BI

US 8,393,206

102;3 60.1 t~ Ms .................

AGE

..............

Im

GVT Hard'ovare Data Sys mm Wzak-q;Wsition Sym 2 m NW40R m IM , Aemdynamir Force rp ~ ..............

Compwo-t ' al tyl

P igu re, 4

Patent Mar. 12, 2013 U.S. Sheet 5 of 6 US 8,393,206 B1 Figure

Patent Mar. 12, 2013 U.S. Sheet 6 of 6 US 8,393,206 B1 H , Figure 6

US 8,393,206 B1

1 2

DRY WIND TUNNEL SYSTEM If the flutter wind-tunnel test was able to use the real- structure then the control laws would automatically be STATEMENT OF GOVERNMENT included, since the real structure has control laws included in INVOLVEMENT it to be able to fly. With the use of a real-structure there would 5 be no scaling discrepancies between the tested structure and the actual structure. The impact of the structural nonlineari- This application involves an invention made with United ties on aeroelastic stability can automatically be included. If States Government support under a Small business Technol- the flutter test could be conducted without a wind-tunnel then ogy Transfer (STTR) Program, Phase I entitled, "GVT-Based the unsteady aerodynamics generated computationally would Ground Flutter Test without Wind Tunnel," having Contract be interference-free from the wind-tunnel walls.

Number: NNXOSCD34P awarded by NASA Dryden Flight In order to be able to conduct a flutter test without a wind- Research Center (National Aeronautics and Space Adminis- tunnel, a ROM (Reduced Order Model) of the unsteady aero- tration's Dryden Flight Research Center, Edwards, Calif.

dynamic model is needed. This unsteady aerodynamic ROM 93523).

represents an aerodynamic transfer function that inputs the physical structural deformation and outputs the aerodynamic FIELD OF THE INVENTION forces. Thus, this requirement immediately rules out the Computational Fluid Dynamics (CFD)-based ROM's The present invention relates generally to a method by because all CFD-based ROM's involve some type of modal which software tools and hardware systems simulates flight approach that assumes the structural mode shapes are known.

tests/wind tunnel tests on an aircraft, part of an aircraft, or any other flexible structure to search the aeroelastic/aeroser- SUMMARY OF THE INVENTION voelastic (AE/ASE) instabilities of the test structure. More particularly, the present invention relates to a method which The object of this invention is to provide a method for uses software and hardware components for the simulation of accurate calculations of an aircraft structures' aeroelastic a flexible structure's AE/ASE characteristics at various flight 25 instabilities at a comparatively low testing outlay.

The term "Dry Wind Tunnel System" ("DWT System") conditions.

refers to a system that consists of a Ground Vibration Test BACKGROUND OF THE INVENTION ("GVT") hardware system, a data acquisition system, a real- time unsteady aerodynamic force generation software mod- Prior to flight testing an aircraft structure or any part of a 30 ule and a multiple input multiple output (MIMO) force con- flexible structure, flutter wind-tunnel tests are required to troller software module.

search for AE/ASE instabilities of the structure's configura- As used herein, the term "Ground Vibration Test hardware tions. Flutter wind-tunnel testing is an expensive and compli- system" ("GVT System" or "GVT Hardware" or "GVT cated process because it involves the design and fabrication of Hardware System") refers to a test that includes state-of-the- a scaled-down aeroelastic structural model, which is a scaled- 35 art equipment including devices such as aircraft soft-support, down version of a real structure, and also because wind- structural excitation, vibration sensors, and data acquisition tunnel time is costly. Flutter wind-tunnel tests are necessary and analysis tools. The GVT Hardware used by the DWT because if a flutter instability is not found prior to the flight System is essentially a ground vibration test system which test then major instabilities may occur during the flight tests consists of vibration sensors, shakers, and load cells. The which can make the structure extremely unstable, hence 40 vibration sensors are devises such as accelerometers, linear endangering the aircraft structure and passengers. Unfortu- variable differential transformer's (LVDT), and linear veloc- nately, creating a scaled-down version of a real structure may ity transducer (LVT). The accelerometers measure the accel- introduce discrepancies in the structural characteristics of the eration of the structure at each sensor location. The LVDT's scaled-down structure when compared to the real structure. measure the displacement of the structure at each sensor For example, the control surface actuators with correspond- 45 location. The LVT's measure the velocity of the structure at ing stiffness and damping, are difficult to represent accurately each sensor location. The shakers are a structural excitation in the scaled-down structure. Also, the incorporation of the device, in which they exert forces onto the structure at each accurate modal damping in the scaled-down structural model shaker location. Each shaker can exert forces different from is almost impossible. These discrepancies between the actual the other shakers. The load cells are devices which measure/ structure and the scaled-down versions can lead to uncertain- 50 monitorthe amount of forcebeing exertedby the shakers onto ties in the measured aeroelastic instability boundary. In addi- the structure.

tion, during flutter wind-tunnel tests the wind-tunnel walls The real-time unsteady aerodynamic force generation soft- interfere with the test results. ware is a reduced order model (ROM) of the unsteady aero- It is very costly to include structural nonlinearities such as dynamics that has a real-time computational efficiency for friction, free-play, etc. in the scaled-down structural model. 55 unsteady aerodynamic force generation at different flight And even when these structural nonlinearities are included in conditions. This computer software program reads the struc- the scaled-down version of the structure they will never be tural responses as its inputs, and the desired unsteady aero- identical to the real structure. These structural nonlinearities dynamic forces at each shaker's location are computed by can have a major impact on the aeroelastic characteristics of passing the structural response input to the reduced order the aircraft. The measurement of the coupling between the 60 model of the unsteady aerodynamics.

flight control system and the aeroelastic system to search for The ROM that the real-time unsteady aerodynamic force ASE instabilities of the aircraft is another critical design generation software uses is created by the GVTOPT Module requirement that can be accomplished by performing a flutter of the ZAERO Software System (ZAERO is ZONA Technol- wind-tunnel test in the presence of flight control laws. How- ogy, Inc.'s commercial software product for aeroelastic/aero- ever, performing this aeroservoelastic (ASE) measurement in 65 servoelastic analysis). The basic requirement in formulating the wind tunnel is also a very expensive process, thus theASE an unsteady aerodynamic ROM for the dry wind tunnel test is analysis largely relies on flight tests. that it contains only the aerodynamic characteristics of the

US 8,393,206 B1

3 4

configuration and does not involve any structural character- to a broad range of test structures, from components such as istics. This is because the coupling between the aerodynamics flexible wings to full aircrafts.

and structures is achieved only during the real-time test; thereby the structural characteristics cannot be known prior to BRIEF DESCRIPTION OF THE DRAWINGS the test.

Prior to the DWT System test, the ROM matrices, com- FIG. 1 demonstrates the interaction between the GVT puted by the GVTOPT Module at each flight condition, are Hardware System, the data acquisition system, and the com- saved in a look-up table database. During the DWT System puter system in which the MIMO force controller and the test, the ROM matrices saved in the look-up table can be real-time unsteady aerodynamic force generation software directly called by the real-time unsteady aerodynamic force 10 are uploaded.

generation software to generate the real-time unsteady aero- FIG. 2 illustrates a structure with sensors.

dynamic force for flutter testing.

FIG. 3 illustrates the GVT Hardware System.

The data acquisition system is such a device that connects FIG. 4 illustrates the Dry-Wind Tunnel System with all of hardware devices (such as sensors and actuators) and com- the component and how they interact each other.

puters. The data acquisition system collects the analog signals 15 FIG. 5 illustrates the real-time unsteady aerodynamic force measured from sensors, converts it into digital signals, and generation software process.

sends the converted digital signals to the computer for analy- FIG. 6 illustrates another real-time unsteady aerodynamic sis. Visa versa, the data acquisition system can also covert the force generation software process.

digital signals coming from the computer into analog signals, and send them back it to the hardware devices. 20 DETAILED DESCRIPTION OF THE INVENTION The MIMO force controller software is performed to ensure the forces exerted to the flexible aircraft structure by The present invention may be described herein in terms of the shakers are the desired unsteady aerodynamics forces, various functional components and various methods/steps. It which are computed by the real time unsteady aerodynamic should be appreciated that such functional components may force generation software at each shaker location. Further- 25 be realized by any number of structural components, hard- more, the MIMO force controller is applied to minimize the ware, and/or software configurations to perform the specified interference among the shakers and reduce the nonlinear functions. For example, the present invention may employ effects due to the nonlinear characteristics of the shaker various electronic components, hardware, and software, dynamics. The MIMO force controller software can be which can be suitably configured for various intended pur- designed using the standard robust control design technique 30 poses; such as testing a real aircraft structure or part of a real such as H_ and μ tools. To facilitate the MIMO force control- aircraft structure to predict the instabilities of said structure.

ler design process, a system identification technology is FIG.1 demonstrates the data flow between the GVT Hard- implemented to estimate a MIMO shaker dynamics, instead ware System 100, the data acquisition system 200, and the of using the analytical tool for a complicated mathematical computer 300, in which the real-time unsteady aerodynamic shaker model. Therefore, an experiment is performed by 35 force generation software 400, and the MIMO force control- exerting a set of sweep sinusoidal signals to the shakers to ler 500 are installed on.

excite the structure. The input sweep sinusoidal signals and FIG. 2 illustrates a structure 101 with sensors 102 on it. The the output forces signals on the structure measured by load sensors 102 are things like accelerometers, which measure cell sensors are used for model estimation of the shaker the oscillation at each sensor's location; linear velocity trans- dynamics. 4o ducer (LVT), which measure the velocity at each sensor's The DWT System concept would be particularly useful as location; and linear variable differential transformer's a pre-flight testing effort to identify any aeroelastic and aero- (LVDT), which measure the structure's displacement at each servoelastic instability on new and/or modified aircraft Inher- sensor's location.

ent structural nonlinearities such as friction and freeplay are FIG. 3 illustrates the GVT Hardware System 100. The notoriously difficult to model properly in linearized analyses, 45 GVT Hardware System includes sensors 102a, 102b, and but would naturally be present in the DWT System testing as 102c, placed on a structure 101. Shakers 103 exert forces onto it is carried out on the actual structure. The ground flutter test the structure. The load cells 104 measure the amount of force using the DWT System operates on the real structural model, the shakers are exerting onto the structure. The sensors 102a, therefore, unlike in the flutter wind-tunnel test no scaled- 102b, and 102c, measure the displacement, velocity and down structural model is involved. 5o acceleration at each sensor location.

Also, the impact of the structural nonlinearities on the FIG. 4 illustrates the Dry-Wind Tunnel System process.

aeroelastic stability can automatically be included. In addi- During the flutter test, one of the shakers 103 creates a small tion, the aeroservoelastic characteristics of the aircraft can be impulse as the initial disturbance to excite the structure 101.

easily measured by simply including the flight control system The sensors 102a, 102b, and 102c are used to measure the in the loop. The unsteady aerodynamics generated computa- 55 displacement, velocity and acceleration at each designed tionally would have no interference from the wind-tunnel locations of the test structure. The data acquisition system walls. The DWT System takes full advantage of the existing 200 transfers the data measured by the sensors 102a, 102b GVT Hardware setup in that the required DWT System's and 102c to the computer 300. Within the computer the infor- setup, for flutter testing, uses similar hardware as the GVT mation from the sensors 102a-102c, goes to the real-time System of the same aircraft structure, only with some possible 60 unsteady aerodynamic force generation software 400. The rearrangement of the shakers and the inclusion of additional real-time unsteady aerodynamic force generation software sensors. Once setup, the DWT System will perform the 400 uses the data from the sensor 102a-102c, to calculate the ground flutter test and requires only a link between the soft- desired forces the shakers 103 should exert at each shaker 103 ware programs and the GVT Hardware. Testing using the location onto the structure 101 to search AE/ASE instability DWT System would also be useful as a post-flight testing 65 of the structure 101. After the desired forces have been cal- procedure to resolve any discrepancies between the analysis culated by the real-time unsteady aerodynamic force genera- and flight test results. The DWT System concept is applicable tion software 400, the GVT Hardware System 100 starts

US 8,393,206 B1

5 6

exciting the structure to search for the potential AE/ASE field-panel method to solve the linearized transonic small instability of the structure 101. The load cells 104 measure the disturbance equation with the steady background flow being force exerted by the shakers 103. The data acquisition system imported from a high fidelity CFD code such as a Navier- 200 passes the load cell 104 measurements to the computer Stokes solver. The accuracy of the ZONA6 and ZONA7 300. Within the computer 300 the measurements are passed to 5 methods for subsonic and supersonic Mach numbers and the the MIMO force controller 500. The MIMO force controller ZTRAN method for transonic Mach number has been dem- 500 acts to enforce the forces exerted by the shakers 103 onto onstrated on many test cases.

the structure 101 are same as the desired forces generated by To obtain a time-domain aerodynamic transfer function, the real-time unsteady aerodynamic generation software.

requires transforming the frequency-domain AIC matrix into Specifically, the MIMO force controller 500 uses the load cell 10 time domain. This can be achieved using the minimum state 104 measurements to monitor if the forces exerted by the method that fits the AIC matrices at a set of frequencies in to shakers 103 onto the structure 101 are the same as the desired a rational function approximation which reads: force generated by the real-time unsteady aerodynamic force generation software 400. If any discrepancy between the measurement of the load cell 104 and the desired force gen- z 7f (r)7 = + [Ails+ [Az]sJXI+ ( )

erated by the real-time unsteady aerodynamic force genera- q— ~ [Aol

v vZ- tion software 400 occurs, the MIMO force controller 500 generates a correct excitation force signal to the shakers 103

q—[D] ~ [f]s— [ R]l 1 [EIJXI

L to excite the structure 101.

FIG. 5 demonstrates the real-time unsteady aerodynamic 20 force generation software's process for determining the cor- where s is the Laplace variable, L is the reference length, V_ rect force that should be applied by the shakers 103 (FIG. 4) is the free-stream velocity. [A 0], [A1 ], [Az], [D], and [E] are onto the structure 101 (FIG. 4) to search for the AE/ASE the resulting matrices from the minimum state method, and instabilities of the structure 101. The real-time unsteady aero- 25 [R] is a diagonal matrix with the diagonal coefficients being dynamic force generation software is a reduced order model the Ni assigned aerodynamic lags. Thus, the size of [R] is (ROM) of the unsteady aerodynamics that has a real-time NixNL . Usually, no more than ten aerodynamic lags are suf- computational efficiency for unsteady aerodynamic force ficient to obtain an accurate rational function approximation.

generation. All the variables presented in FIG. 5 are pre- The vectors {X} and {f (t)} in Eq. (2) are the structural computed via the GVTOPT module of the ZAERO software deformations and aerodynamic forces at the aerodynamic system at different flight conditions, and a lookup table for 30 panels, respectively. Therefore, for N, aerodynamic panels, these variables can be generated to incorporate variation of the size of the vectors {X} and {at)} is N PxI . However, in the the flight conditions.

GVT the structural deformations are measured at the sensor The basic requirement in formulating an unsteady aerody- locations and the aerodynamic forces are produced at the namic ROM for the dry wind tunnel test is that it contains only 35 shaker locations. To transfer these structural deformation and the aerodynamic characteristics of the configuration and does aerodynamic forces from the aerodynamic panels to the sen- not include any structural characteristics. This is because the sor and shaker locations requires an interpolation procedure coupling between aerodynamics and structures is achieved which can be accomplished by using the spline method such only during the real time test; thereby the structural charac- as the infinite plate spline method, the thin plate spline teristics cannot be known prior to the test.

method, and the beam spline method that jointly generate a In fact, this unsteady aerodynamic ROM represents an 40 spline matrices for such displacement and force transferal.

aerodynamic transfer function that inputs the physical struc- For Ns accelerometers, the displacement spline matrix reads: tural deformation and outputs the aerodynamic forces. Thus, this requirement immediately rules out the CFD-based ROM MNpx 1 — [Gs7NpxNJX _1Npxl (3) because all CFD-based ROM's involve some type of modal approach that assumes the structural mode shapes to be 45 where {X S.„} is the structural deformation at the N s sensors, known. On the other hand, the unsteady aerodynamic panel and [GS] is a displacement spline matrix that transfers the methods such as the Doublet Lattice method DLM, ZONA6, structural deformations from the sensor locations to the con- and ZONA7 do readily generate an aerodynamic influence trol points of the aerodynamic panels.

coefficient (AIC) matrix that relates the structural deforma- For NA shakers, the force spline matrix reads: tion (x) to the aerodynamic forces (f) which reads: 50 V,hakJ N,gx 1 - 1G} N,gxNpV,I Npx 1 (4) (w)J-q_[AIQo))] {X} (1) {f where f is the aerodynamic forces at the shaker locations, h,x where q- is the dynamic pressure, and w is the oscillating and [Ci ] is the force spline matrix that transfers the forces at frequency indicating that the AIC matrix is formulated in the the control points of the aerodynamic panels to the shaker frequency domain.

55 locations.

Closely examining the AIC matrix, one can see that each Substituting Egn.(3) and Egn.(4) into Egn.(2) yields a coefficient in the matrix represents the response on a receiv- reduced-order model of the unsteady aerodynamics shown in ing panel to a unit input on a sending panel. For a N P number the following equation: of aerodynamic panels in the aerodynamic model, the size of the AIC matrix is NpxN,. Thus, the AIC matrix truly repre- 60 sents an aerodynamic transfer function that contains only the z 5 ( ) [A1]s+ V1[A,ls'J{X,_)+

aerodynamic characteristics of the configuration. The AIC {fhak)=q- ~ [Aol+

matrix generated by the linear unsteady aerodynamic meth- ods can accurately produce aerodynamic forces at the sub-

q— [D] [lls— [R] ~ 1[El7X,e„J

L sonic and supersonic Mach numbers. For transonic flows, the 65 transonic AIC matrix can be generated by using an unsteady transonic method called ZTRAN. ZTRAN uses an overset

US 8,393,206 B1

7 8

where [Ao],[A,],[A 2 ]=[Gf][[Ao],[A,],[A 2 ]][GS] with size of one velocity sensor, said at least one acceleration N A xN,, D=[Ci ] {D1 with size of N A xN L , E=[E] [GS ] with size sensor, said at least one load cell, and said at least one of N L xN,. computer; In a typical GVT set-up, the numbers of sensors and shak- (b) loading a real-time unsteady aerodynamic force gen- ers are typically small. 5 eration software stored in and executable on said at least Accordingly, the size of the matrices in Egn.(5) is small, one computer that is able to compute initial real-time implying that Egn.(5) can be computed very efficiently to unsteady aerodynamic force on said structure at said at generate aerodynamic forces at the shaker locations in real- least one shaker location; time. (c) controlling, using said at least one computer, said at The real-time unsteady aerodynamic force generation soft- 10 least one shaker to exert an initial impulse on such struc- ware 400 reads the structural responses 601 (FIG. 4) from the ture; sensors 102a-102c (FIG. 4) as its inputs. The inputs 102a- (d) receiving, in said at least one computer, initial data from 102c shown in FIG. 5 correspond to the output of the sensor said at least one displacement sensor, said at least one 102a-102c in FIG. 4, and which are LVT (displacement), velocity sensor, and said at least one acceleration sensor LVDT (velocity) and accelerometer sensors. The desired 15 via said data acquisition system, responsive to said ini- force 603 computed by the real-time unsteady aerodynamic tial impulse; force generation software 400 is passed to the shakers 103 (e) computing test inputs for controlling said at least one (FIG. 4) through the data acquisition system 200 (FIG. 4). shaker, responsive to said initial data and using said FIG. 6 demonstrates another process of the real-time real-time unsteady aerodynamic force generation soft- unsteady aerodynamic force generation software 400. In this 20 ware; process it uses the displacement 102a and acceleration 102c (f) controlling said at least one shaker, responsive to said as inputs, and outputs the desired force 603 to the shakers 103 test inputs and using said force controller software, to (FIG. 4) through the data acquisition system 200. The veloc- shake such structure; ity 102b is calculated by the integration of the acceleration (g) measuring, via said at least one computer: 102c. 25 i) structural response of said structure using said real- Both processes demonstrated in FIGS. 5 and 6 can be used time unsteady aerodynamic force generation software to determine the desired force. The selection of which process and structural response data from said at least one to use is based on what information and which sensors are displacement sensor, said at least one velocity sensor, available. and said at least one acceleration sensor; and 30 ii) said force exerted on said structure, using said force What is claimed is: controller software and said at least one computer, 1. A Dry Wind Tunnel (DWT) method that is used to responsive to data from said at least one load cell; predict the flutter boundaries of a structure without using a (h) computing, using said at least one computer running wind tunnel, said DWT method comprising the following said force generation software and responsive to said steps: 35 measured structural response as input, an unsteady aero- (a) providing: dynamic force at said at least one shaker location; (i) at least one shaker, for shaking such structure, (i) applying said computed unsteady aerodynamic force to coupled to such structure at at least one shaker loca- said structure through said at least one shaker; tion;

0) controlling, using said force controller software, said

(ii) at least one computer communicatively coupled to 40 applied forces to be the same as said real-time unsteady said at least one shaker and storing and operable to aerodynamic forces computed using said real-time execute a force controller software for generating unsteady aerodynamic force generation software; force control signals for controlling said at least one (k) monitoring said structural response data in real time; shaker; (1) if said structural response data indicates a decay motion, (iii) at least one displacement sensor mounted on such 45 incrementally changing said inputs to create incre- structure; mented test inputs to said real-time unsteady aerody- (iv) at least one velocity sensor mounted on such struc- namic force generation software; ture; (m) repeating steps (f) to (1) until said structural response (v) at least one acceleration sensor mounted on such data indicates a divergent motion; structure; 50 (n) designating a condition between said decay motion and (vi) at least one load cell mounted between said at least said divergent motion of such structure as a predicted one shaker and such structure; and flutter boundary of such structure.

(vii) a data acquisition system communicatively coupled to said at least one displacement sensor, said at least