Elizabeth M. Lee-Rausch
John T. Batina
NASA Langley Research Center
Hampton, Virginia 23681–0001

AIAA Paper No. 93–3476

Presented at the
AIAA 11th Applied Aerodynamics Conference
Monterey, California
August 9–11, 1993


structural damping. Lee-Rausch* John T.2 With the advent of more powerful com- Branch. Structural Dynamics Division. indi- From the first calculations of Ballhaus and Goorjian1 cates that the fluid viscosity has a significant effect on the to more recent applications of the Euler and Navier-Stokes * Research Engineer. are studied using an unsteady Navier-Stokes algorithm in Nomenclature order to investigate a previously noted discrepancy between Euler flutter characteristics and the experimental data. aeroelastic analysis of wings using the unsteady Euler equations. The Aij generalized aerodynamic force resulting from algorithm.6.  nondimensional semispan location This stability analysis is used to determine the flutter char.141 using the Euler and Navier-Stokes equations to compare with the lin- ear V-g flutter analysis method. For the linear stabil- ity analysis. and number of modes in T dimensional time the structural model are investigated. which is Introduction used in conjunction with the time-marching analysis.  mass ratio acteristics of the wing at free-stream Mach numbers of 0. puters and more efficient algorithms.S. The paper presents a brief descrip- k U1 tion of the aeroelastic method and presents unsteady cal. ative computational efficiency. Virginia 23681–0001 Abstract supersonic flutter boundary for this wing while the structural damping and number of modes in the structural model have The flutter characteristics of the first AGARD standard a lesser effect. mode i ously modified for the time-marching. These modifi. the field of computational aeroelasticity has pro- ** Senior Research Scientist. aeroelastic configuration for dynamic response. Aeroelastic Analysis and Optimization Branch. b root semichord cations include the incorporation of a deforming mesh algo. No copyright is asserted in the United States under Title 17. Rec free-stream Reynolds number based on root acteristics of the isolated 45 swept-back wing. The U. 2!c tended and evaluated for applications that use the Navier- Stokes aerodynamics. Time-marching aeroelastic calculations are per.141 using the generalized aerodynamic forces gener. the transonic small distur- 1 . the aeroelastic method is ex.S.6 FLUTTER USING NAVIER-STOKES AERODYNAMICS Elizabeth M. Wing 445. researchers have in Copyright  1993 by the American Institute of Aeronautics and Astronautics. recent years computed more detailed and sophisticated sim- Government has a royalty-free license to exercise all rights under the copyright claimed ulations of aeroelastic phenomena. implicit. Batina** NASA Langley Research Center Hampton. was previ. U.96 and 1. Q free-stream dynamic pressure culations. equations for the aeroelastic analysis of three-dimensional Structural Dynamics Division. c root chordlength rithm and the addition of the structural equations of motion for their simultaneous time integration with the governing Cp pressure coefficient flow equations. wings. torsion mode formed at a free-stream Mach number of 1. the unsteady generalized aerodynamic forces of UF flutter speed the wing are computed for a range of reduced frequencies U1 streamwise free-stream speed using the pulse transfer-function approach. M1 free-stream Mach number culations which verify this method for Navier-Stokes cal. All other rights are reserved by the copyright owner. Code.1). Aeroelastic Analysis and Optimization gressed rapidly. CALCULATION OF AGARD WING 445. gi structural damping for mode i reduced frequency. upwind pressure induced by mode j acting through Euler/Navier-Stokes code (CFL3D Version 2. Inc. The V-g analysis. ! angular frequency ated by solving the Euler equations and the Navier-Stokes ! uncoupled natural frequency of the wing first equations. which is a three-dimensional. The flutter char- acteristics of the wing are determined using these unsteady angle of attack generalized aerodynamic forces in a traditional V-g analysis. In this paper. A linear stability analysis and a time-marching aeroelastic analysis are used to determine the flutter char. Because of their rel- herein for Governmental purposes. Member AIAA. Associate Fellow AIAA. Effects of chordlength fluid viscosity.

unsteady Euler/Navier-Stokes code (CFL3D tional stability analysis is used to determine flutter charac. in previous studies.6 using the system. which define the flutter boundary over a range of Mach number from 0. typi- cally have been applied to the analyses of flexible wings over a limited range of flow conditions. Calcu- lated flutter results are compared with experimental data for seven free-stream Mach numbers. of these modifications and presents unsteady calculations other advantage of the linear stability analysis is that it can which verify the modifications to the code for the solution of provide more information regarding the effects of individ. 17 for cases with low an. 13 and to compute the flutter character- istics of the AGARD Wing 445. One advan. This paper gives a brief description small and if there are no static aeroelastic deformations.12. Wing 445. A linear stability anal- of airloads may not be accurate in some cases such as those ysis and a time-marching aeroelastic analysis are used to involving large amplitudes of motion and separated flow. Lee-Rausch and Batina compute the complete flutter boundary for the same wing using the Euler equations on a structured grid. fluid viscosity. pitching motions are presented to verify the performance tant to note. pare favorably with those obtained from a harmonic analy- Many of the aeroelastic analyses performed with the sis. However. Effects of However. the Euler equations have been used to compute the flutter boundary for the AGARD standard aeroelastic configura- tion for dynamic response. the Navier-Stokes equations.3–11 Only recently. 13. structural damping. modifications are made to an existing three- namic forces are assumed to be locally linear and a tradi. the unsteady generalized aerodynamic forces 2 .6. In Ref. 1.14. In Ref.6 for free-stream Mach num- bers ranging from 0. 2 Similarly.bance (TSD) equation and the full potential (FP) equation have been applied to a wider variety of three-dimensional configurations than the Euler and Naiver-Stokes equations. the flutter in the structural model are investigated. 15. Other analyses have been performed using harmonic Navier-Stokes equations. Version 2. the unsteady aerody. the linear stability analysis is used in nonlinear flow equations have been obtained by calculating conjunction with time-marching analyses to study the aeroe- the transient response of the coupled aerodynamic/structural lastic characteristics of the AGARD Wing 445. For the linear sta- characteristics obtained from time-marching analyses com.499 to 1. dimensional. and the computed flutter speeds and frequencies are compared with the experimental values measured for this wing. 13 with experimental data for Wing 445. algorithm and the addition of the structural equations of mo- cant computational savings over the time-marching analysis tion for their simultaneous time integration with the govern- if the number of modes necessary to model the structure is ing flow equations. It is impor. 13. the computed aeroelastic results predict a premature rise in the flutter boundary as compared with the experimental bound- ary.6. The purpose of this paper is to extend the capability presented in Ref. characteristics at the supersonic free-stream Mach numbers can be investigated. 12. an unstructured-grid Euler code is used to compute the flutter characteristics of Wing 445. that the assumption of superposition of the deforming mesh algorithm. such as the Euler and Navier-Stokes equations. Results from calculations per- ual modes and the effects of structural damping from fewer formed about a rigid wing undergoing forced plunging and computations than the time-marching approach. the TSD and FP equations also have been utilized for more detailed analyses of the aeroelastic characteristics of these configurations. determine the flutter characteristics of the wing.14 (See Fig.) These comparisons show good agreement in flutter characteristics for free-stream Mach numbers below one.499 to 0.96.14–17 In the harmonic analysis. These teristics based on the calculated harmonic loads. In contrast.1) for the aeroelastic analysis of wings. for free-stream Mach numbers above one. bility analysis. the higher-order meth- ods. loads. For these reasons. however. modifications include the incorporation of a deforming mesh tage of the linear stability analysis is that it offers a signifi. 13 In Ref.6 using the Navier-Stokes Figure 1 Comparison of computed Euler flutter results equations so that the source of the discrepancy in the flutter from Ref. An. and number of modes gles of attack and small amplitudes of motion.

when using the Navier-Stokes equations.19 and the flutter dynamic pressure and frequency associated with this Mach number can be es. An analysis of each of these options for locity for a given Mach number does. In addition. Varying the free-stream ve- flow condition. The time-marching aeroelastic procedure for a problem Figure 2 Variation in free-stream Reynolds utilizing the Navier-Stokes equations is almost identical to number with variation in dynamic pressure the procedure described in more detail in Ref.6 at fied for the static and dynamic computations at each dy. Changing the density or velocity will. However. change the Reynolds number associated with that perimental dynamic pressures). aeroelastic transients are computed at several values of dynamic pressure which bracket the flutter point. timated by interpolation. perimental flutter dynamic pressure ( Q=Qexp ). This stability analysis is used to determine the flutter characteristics of the wing at free-stream Mach numbers of 0. Time-Marching Aeroelastic Analysis The time-marching aeroelastic procedure used in this study is typical of those currently in use. The frequency and damping characteristics of the transient responses at each dynamic pressure are determined from a least squares curve fit.11. 12.141 using the Euler and Navier-Stokes equations for comparison with the linear V-g flutter analysis results. M1 = 0.141 using the GAF’s generated with the Euler equations and with the Navier-Stokes equations. time-marching aeroe- lastic calculations are performed at a free-stream Mach num- ber of 1. either the and (b) indicate that for a constant free-stream density. varying the density (especially for conditions above the ex- however. Euler equations. To determine the flutter conditions at a given free-stream Mach number. re- namic pressure of interest. (b) M1 = 1:141.96 and 1. 18 In general. When comparing with experimental data. the experimental flutter conditions of the Wing 445. 2(a) and (b).96 and 1. 13 to demonstrate the aeroelastic capability for the Navier-Stokes equations and to evaluate the effects of viscosity on the flut. This requirement raises the issue spectively. The flut- ter characteristics of the wing are determined using these unsteady GAF’s in a traditional V-g analysis. Figures 2(a) ber. vary- free-stream density or the free-stream velocity can be devi. These equa- tions then are written in terms of a linear state-space equation such that a modified state-transition-matrix integrator can be used to march the coupled fluid-structural system forward in time. however. Figures 2(a) and (b) show the variation in the of how to vary the dynamic pressure in a computational free-stream Reynolds number (Rec ) with the variation in the fluid dynamics.(GAF’s) of the wing are computed for a range of reduced frequency using a pulse transfer-function analysis. a free-stream Reynolds number must be speci. ter characteristics at these flow conditions. The time-marching Euler flutter characteristics are recom- puted on a mesh that is similar to the Navier-Stokes mesh in order to more effectively isolate the effects of viscosity. 13 for the for time-marching aeroelastic calculations. time-marching aeroelastic analysis to obtain free-stream dynamic pressure nondimensionalized by the ex- the flutter conditions for a specific free-stream Mach num. Computed flutter results are compared with the experimental data and with other computational results from Ref. require a 3 . the aeroelastic equations of motion are formulated in terms of a finite modal series of free-vibration modes. (a) M1 = 0:96. The fluid forces are coupled with the structural equa- tions of motion through the generalized aerodynamic forces.141 is shown in Figs. ing the free-stream velocity results in a smaller variation in ated from the experimental value to obtain variation in the Reynolds number than maintaining a constant velocity while dynamic pressure.

uses upwind differencing based on GAF’s over a range of reduced frequency in a single time. each structural mode used in the flutter analysis. exponen. is discussed later in the paper. In contrast to the forced thin-layer Navier-Stokes equations. 11 and 23. This method in- will hold true. Although FLUTDET originally was designed to inter- calculations of the GAF’s and the V-g analysis.20 The capabili- the computed flutter condition. experimental free-stream density is specified. an additional term accounting ment between the pulse and harmonic methods. essary. assume that the unsteady aerodynamic forces are number due to changes in free-stream velocity are small. which calculates unsteady aerodynamic is not feasible for a Navier-Stokes analysis due to constraints forces from subsonic kernel matrices. eddy-viscosity model of Balwin-Lomax. these GAF’s are obtained from time. matched Reynolds number solution is to iterate between the 21. for the Navier-Stokes equations terpolates the given generalized aerodynamic forces (GAF’s) an additional parameter must be considered: the free-stream to compute the eigenvalues of the flutter determinant. types of upwind differencing account for the local wave- sis. Both marching calculation for each mode. The Reynolds number. a deforming mesh. forced harmonic oscillation and using the last cycle of os- cillation to determine the first harmonic component of the Upwind Euler/Navier-Stokes Algorithm GAF’s. then this functionality also computing flutter speeds and frequencies.28 monic method which tends to verify that the analysis is valid For unsteady cases. the pulse transfer- adiabatic wall temperature. For nonlinear number and the Reynolds number based on the computed aerodynamic methods. The only way to ensure a ties and techniques used in FLUTDET are described in Ref. The effect of Reynolds number function method is used in this study to compute the GAF’s variation on the time-marching cases considered in this study for input to the V-g analysis. the generalized mass and modal stiffnesses are re- requirement for the following reason. These differences are discussed later in the paper. tions. subsequently. However. because these schemes are naturally dis- tially shaped pulse. planform (ge- Linear Stability Analysis ometry). flutter characteristics are small for the cases considered in marching calculations by computing several cycles of a this study. This option face with GENFLU. Several types of flux limiting are available within placement to obtain the GAF. the unsteady forces are determined from the response propagation characteristics of the flow and sharply capture due to motion represented by a smoothly varying. nonequilibrium half-equation model of Johnson and King.change in the free-stream temperature and. the difference between the experimental Reynolds GAF’s as functions of reduced frequency. Multiple time-marching computations are required at various reduced frequencies to generate the GAF’s for The CFL3D24–26 code uses a three-factor. shock waves. which is harmonic method.27 and the in good agreement with the GAF’s computed using the har. additional artificial dissipation terms are not nec- force is divided by the fast Fourier transform of the dis. the original algorithm contains the nec- for predicting the linear small perturbation response about essary metric terms for a rigidly translating and rotating a nonlinear flowfield. the pulse analysis can determine the a cell-centered scheme. The V-g stability analysis does not ensure procedure used to obtain the flutter eigenvalues is contained that the Reynolds number associated with the computation of in the parametric flutter analysis program FLUTDET which the GAF’s will match the Reynolds number associated with is part of the FAST flutter analysis package. Euler mesh which moves without deforming. the GAF’s for a vibration mode are functions of free-stream Mach number. either flux-vector splitting or flux-difference splitting. the density. locally linear and utilize modal superposition of harmonic If the GAF’s are computed at the experimental Reynolds loads. The algorithm. two turbulence models are available: the equilibrium. compatible input from on computational resources. for the cases ana- other aerodynamic methods can be used. In addition to the lyzed in this study. finite-volume algorithm based on upwind-biased spatial dif- An alternate method for determining the GAF’s is the ferencing to solve the time-dependent Euler equations and pulse transfer-function analysis. the GAF’s of airfoils at tran. and reduced frequency. In the V-g analysis. quired input to FLUTDET. implicit. for the change in cell volume must be included in the time- 4 .22 Also in Ref. The program generates flutter the flutter speed is computed for a given free-stream den- characteristics and V-g plots for each specified free-stream sity. variations in free-stream Reynolds method. Therefore. plate are found to be in good agreement with linear theory For applications utilizing the thin-layer Navier-Stokes equa- calculations. such as the V-g in the previous section. if the assumptions A conventional V-g method is used in this study for of local linearity are maintained. For linear methods. 22. sonic speeds computed from a pulse analysis are shown to be algebraic. However. A fast Fourier transform of the unsteady sipative. Typically when comparing with experimental data. Similarly in Refs. As discussed Traditional flutter analysis methods. For nonlinear methods such as the TSD and Euler equations. Transonic small disturbance the code to prevent oscillations in the solution near shock results computed using the pulse analysis for a pitching flat waves which are typically found in higher-order schemes. the because of its computational efficiency. Linear aerodynamic methods typically compute the number. however. For cases involving results for a pitching and plunging airfoil show good agree. In the pulse analy. Also. it was not considered to be a necessary GAF’s.

a power of three laminated mahogany. the deforming mesh algorithm. For aeroe. as up to 8 iterations. the edge of each Research Center. As these C-type meshes transition from by out-of-plane (or vertical) displacements. 13 For the current Euler and of the wing are used to model the wing structure. Wing 445.17 Hz for the first torsion mode. It was found that specifying be updated at every time level so that it conforms to the the location of the leading-edge points for the tip wake alle- aeroelastically deformed shape of the wing.54 Hz for the second torsion mode as There is one case where this has been found to be a prob. previous Euler calculations. When using a C-H-type mesh. The Euler calculations required torsion. It was found that. Similarly.35 Hz for the second equations.6 This algorithm is extended in the current study for use on Navier-Stokes meshes. The points around the “leading-edge” of this tip wake sur- Deforming Mesh Algorithm face must negotiate a 360 turn. The semi-span model was at- are solved using a predictor-corrector method. determined by a finite element analysis. holes were drilled through the ma- lastic calculations. 11 was used by the current authors for time-marching aeroelastic calculations on Euler grids. wall-mounted model which was constructed of of the diagonal. the wing root was immersed in the wall the previous two time levels and then corrected using sev. second bending. Because the max- lem. Several mod- Diagonal springs are added along the faces of each cell in or.141. As suggested in Ref. bending mode. The structural equations modeling the mesh in this area lack stiffness in one direc- In the time-marching aeroelastic calculations and in the tion. this model must represent a realistic structure. The aerodynamic shape integration of the structural equations of motions. the C-type meshes are stacked along the span to form the three. In thickness of the wall boundary layer was 0. the 1. to obtain accurate equations for this network to determine the new locations unsteady solutions. The new of the original wing was preserved by filling these holes locations of the interior grid points then are determined by with rigid foam plastic. a panel aspect ratio of 1. first required to move the mesh. the first four natural vibration modes ficient to move the mesh. 11. a problem occurs for Navier-Stokes applications. The basic principles of the method The first AGARD standard aeroelastic configuration for do not change although some modifications to the deforming dynamic response. and a NACA 65A004 airfoil section. y and from 0.5 feet. 13 AGARD Wing 445.833 feet and the semi-span was 2. another aeroelastic motions of the wing are arbitrary in nature. In a C-H-type mesh for an isolated wing.65.8 inch or less. four Jacobi iterations were suf. 0.discretization of the governing equations. 30 indicates that the displacement eral Jacobi iterations of the static equilibrium equations. respectively.6 Hz for the first bending mode. and the Navier-Stokes calculations re. in turn.11. In this case.6. models the mesh between the trailing edge of the airfoil and the trailing-edge- as a network of springs and solves the static equilibrium wake line of points. This. causes poor convergence rates for the modal pulse transfer-function calculations. 48. The model used in this study is a springs is inversely proportional to a power of the length semi-span. els of the wing were tested in the TDT including full span der to prevent cell shearing. In order to ob- grid points on the outer boundary are held fixed. the displacement is determined from the hogany wing to reduce its stiffness.338 to 1. boundary layer. Because the viates this problem. the mesh must predictor-corrector procedure. Because the dynamic mesh is modeled using structural 38. is collapsed to zero thickness to form a “tip wake” surface.31 The wing had a quarter-chord sweep hexahedral cell is modeled as a spring with a stiffness that angle of 45 . represent first bending. therefore. described in Ref. Ref. 13. motion due to the first four vibration modes was dominated dimensional mesh. These slopes were matched by specifying by Robinson et al. This algorithm was originally match the slope of the instantaneous airfoil camber line at developed by Batina29 for tetrahedral cells and extended the trailing edge. The root chord of this model was was used in the present calculations. These static equilibrium equations allowed throughout the test.66. and the tain flutter data for a wide range of Mach number and density displacement of the wing surface is specified. These Navier-Stokes calculations.30 The modes have quired up to 12 iterations. The new grid tached directly to the wind tunnel wall (no splitter plate was point locations are first predicted by an extrapolation from used). At each time level. and 91. 31 over a range of Mach number a summation of forces at each grid point in the x. chordwise imum amplitudes of motion were small for these tests.30 was tested in the 16–foot mesh boundary conditions are required as described below. additional Jacobi iterations were modes. One such the flow solution is very sensitive to the angle of intersection method. and second torsion. The method the points in the wake using a quadratic function. Flutter data for this model tested in solving the static equilibrium equations which result from air are reported in Ref. a taper ratio of is inversely proportional to a power of the length of the edge. Natural boundary-layer transition was z coordinate directions. numbered 1 through 4. Therefore in the 5 . natural frequencies of 9. general mesh updating procedure is necessary. the stiffness of these and semi-span models. conditions in the TDT. Transonic Dynamics Tunnel (TDT) at the NASA Langley For the deforming mesh algorithm. As in Ref. determined from a ground vibration test. This modification the wing surface to the exterior flow field the airfoil section is implemented as described in Ref. 11. the slope of the trailing-edge wake must of the mesh grid points.11 for hexahedral cells.

and 41 points distributed radially from the wing surface to the outer boundary. wrapped around the wing chord and 41 points distributed perimental flutter conditions. cling. the Reynolds number is specified as the exper. A planform view of the surface mesh for both the Euler grid and the Navier-Stokes grid is shown in Fig. a free-stream Reynolds number of 364.1 percent of the local chord. 4. Two cases of viscous time. shown in Figs. This mesh ber of 537. marching flutter calculations at M1 = 1. the GAF’s at each the 193 x 41 x 65 mesh. pute steady-state results in a grid density study. For time-marching calculations. the nondimensional 6 . The Euler and Navier-Stokes grids have identical chordwise and spanwise computational analysis. For all of the calculations. 65 points distributed from the wing root to the spanwise boundary (41 points on the wing surface).141 calculation. For the Euler grid. The Euler and Navier-Stokes computations are per- formed using a 193 x 41 x 65 C-H-type grid with 193 points wrapped around the wing and its wake (129 points on the wing surface). the same.141 are considered. The fine mesh has 177 points free-stream Mach number also are computed using the ex.141. Therefore. The steady-state viscous calculations are performed face is varied over the chord such that at least one point will at the experimental flutter conditions which corresponded to be in the laminar sublayer of the turbulent boundary layer. respectively. Both grids extend 6 root chord lengths from the wing to the upstream boundary. lengths from the wing to the upper and lower boundaries. 7 root chord lengths from the wing to the downstream boundary. For are computed at free-stream Mach numbers M1 = 0. This mesh topology is chosen rather than the O-O-type topology because the wind tunnel model has a sheared-off tip. 3(a) and (b). the modal deflections are defined distributions. the surface mesh for both grids is only by their out-of-plane deflections. 6 root chord (b) Navier-Stokes. the grid spacing normal to the sur- 1.700 per foot for the M1 = 1.96 calculation and a free-stream Reynolds num. (a) Euler. mesh sequencing and multi-grid cy- namic pressure. In flux-limiter. is a 265 x 81 x 65 C-H-type mesh identical in topology to As mentioned in a previous section. equations are solved using flux-vector splitting and a smooth imental value for calculations at each dynamic pressure. the grid spacing Steady and unsteady Euler and Navier-Stokes results normal to the surface is 0. Figure 4 Planform view of computational grids. the Euler and Navier-Stokes In one case. A partial view of the surface mesh on the wing and sym- Figure 3 Partial view of the 193 x 41 x 65 computational metry plane for the Euler and the Navier-Stokes grids is grids on the wing surface and symmetry plane. Convergence to steady-state is accelerated us- the other case.600 per foot for the An additional finer Navier-Stokes mesh is utilized to com- M1 = 0.96 and the Navier-Stokes grid. and 1 semi-span length from the tip to the side boundary. from the wing root to tip. Results and Discussion The distribution of points in the radial direction is not the same for the two grids. the Reynolds number is varied with the dy. ing local time-stepping.

For this reason. the 193 x 41 x 65 C-H mesh with 0.0 percent. turbulence is modeled over Steady-state flowfields are used as initial conditions for the entire wing surface. indicate that at M1 = 0. finer grid. Although Cp3 is derived from isentropic bench mark. After some investigation with different grid relations. Regions of supercrit- density study was performed for the steady-state viscous ical flow (indicated by the shaded areas) are determined in flow fields.4 percent. 50. (b) M1 = 1:141.96 and 1. for this study. Since no information is available on the cost of the aeroelastic computations.4 million grid points were used as a (Cp3 = –0. experimental transition location.960 η = 0.141 and = 0 .51 million dication of supersonic flow over the wing surface.504 η = 0. Results from the 265 x 81 x 65 C-H mesh Figs.0 percent of the semispan for both grids. coefficient distributions between the 265 x 81 x 65 C-H mesh and the 193 x 41 x 65 C-H mesh. and all of the subsequent results presented in ing the Baldwin-Lomax turbulence model with an adiabatic this paper were computed using this grid to minimize the wall temperature. a grid at M1 = 0.755 η = 0. the 193 x 41 x 65 grid was chosen Stokes calculations. the 0. shown in Figs.260 η = 0. 6 and 7 to illustrate the basic flow characteristics of the time-marching calculations Before beginning the flutter analyses of the wing.260 η = 0. Figure 5 Comparison of steady-state pressure Figure 5 Concluded.0697).755 η = 0.96 and flow at the edge of the boundary layer based on the assump- for M1 = 1. 6(a) and (b). 26. Comparisons of steady-state pres- transition locations near the leading edge of the wing.96 a 7 .141.504 η = 0. it is used in this case to give an approximate in- densities. For the points was found to predict very similar steady-state surface Navier-Stokes calculation.5 percent.960 (a) M1 = 0:96.000 η = 0. The pressure coefficient contours. 75. 6(a) and (b) by the critical pressure coefficient contour with approximately 1.0 percent. Viscous calculations are performed us. η = 0. Cp3 is an indication of supersonic pressure distributions as the finer mesh for M1 = 0.03273 for both the Euler and the Navier.000 η = 0. Figures 5(a) and (b) show the surface tion that for an attached flow the gradient of pressure through pressure coefficients at five span stations corresponding to the boundary layer is zero. and 96. This is a good approximation for the unsteady calculations. The surface pressure plots indicate that the coarser grid resolves the flow global time step (based on the root chord and the free-stream features across the span of the wing at least as well as the speed of sound) is 0. sure coefficient contours of these initial flowfields on the up- Steady-State Results per wing surface are shown in Figs.

the Eu. Although. the Eu. (b) Navier-Stokes.141 in Figs.2057). Pressure coefficient contours shown in Fig. 7(a) indicate that ler calculation predicts higher levels of acceleration over the for the inviscid computation. a weaker Regions of supercritical flow are determined for oblique shock forms on the outboard portion of the wing at M1 = 1. Fig. coefficient contour (Cp3 = 0. how. cept for a small area around the leading edge of the wing.6 at M1 = 1:141 and = 0 . 7(b) indicate that for the viscous computation. the su- large area of supercritical flow has formed on the forward percritical region encompasses the entire wing surface ex- portion of the wing. (a) Euler. At M1 = 1. the shock is located at approximately 75 percent ler and Navier-Stokes calculations predict similar regions of of the local chord. Figure 6 Comparison of steady-state pressure Figure 7 Comparison of steady-state pressure coefficient contours on the upper surface of coefficient contours on the upper surface of Wing 445. does not terminate with a shock. 7(a) and (b) by the critical pressure approximately 70 percent of the local chord. This area of supercritical flow.6 at M1 = 0:96 and = 0 . * Cp (a) Euler. On the outboard portion of recompression on the inboard portion of the wing. the wing. Wing 445. an oblique shock has formed wing (lower minimum pressure coefficient) and a more rapid on the aft portion of the wing. Pressure coefficient contours shown in supercritical flow. A comparison 8 . * Cp (b) Navier-Stokes.141. ever.

that the motion of the wing could be simulated not only by the deforming mesh algorithm but also by a rigid transla- tion and rotation of the grid.96 to 1.96. Ah is the lift coefficient due to pitching. 6 and 7 indicates that viscosity has a greater effect terms A11 and A22. which are defined as modes h and  respec.96. The results of the pulse analyses. Structural modes The generalized forces for the wing at M1 = 0. A33 . A comparison of inviscid and viscous results in Fig. The pulse calculations are of the wing. are plotted as real and imaginary components of the unsteady forces.1. 13 rapidly rise (See Fig.141 are computed for the inviscid and viscous cases using the pulse transfer-function analysis.01 root chord lengths.6 at M1 = 0:96 and = 0 computed using the Navier-Stokes equations. 8. indicate a more significant shift in the real parts Significant changes in the steady-state flow conditions from such that the Navier-Stokes forces are smaller in magnitude M1 = 0. are plotted for the real and imaginary components of the unsteady forces Aij as functions of the reduced frequency.005 for mode 4. 13 to verify the implementation and performance of the dynamic mesh algorithm for Euler computations. for the wing at M1 = 0. The first four structural modes are ana- lyzed. the maximum plunge amplitude is 0. indicate that the effect of viscosity is to delay the aft motion A plunging motion and a pitching motion about the root of the aerodynamic center. as a function of the reduced frequency k based on wing root semichord. A similar analysis is performed in Ref. The maximum pitch amplitude is 1 . This is consistent with the quarter-chord. is associated with an aft shift of the aerodynamic center scribed in a preceding section. shown in Figs.02 for mode 3.1. The maximum amplitude of the generalized displace- ment is 0. The results of the pulse analysis for the plunging and pitching motions.04. 9 . the forces are independent of the way in which the mesh was moved which verifies the implementation and performance of the dynamic mesh algorithm for the Navier-Stokes com- putations.02 for mode 1. 0. These generalized forces represent work divided by the dynamic pressure and c2 . 0. results shown in the steady-state surface pressure coefficient tively. flutter boundary in Ref. 8. are analyzed.) However. 1. The generalized force Ahh is the lift due to plunge. and 0.16 Therefore the results shown in Fig.141. in contrast to the results for M1 = 0.141 than at M1 = 0. In general. 9 and 10.96 indicates slight changes in the diagonal of Figs. and the frequency resolution of the pulse analysis is 0. Aij . This range of free-stream components of the GAF’s show a greater sensitivity to Mach number corresponds to the range of Mach numbers viscosity than do the imaginary components. 9 for M1 = 0. Ah is the pitching moment due to plunge.141. These simple modes were chosen so contours for M1 = 1. The remaining diagonal terms.96 are computed for the viscous A decrease in the magnitude of the real component of A22 case using the pulse transfer-function analysis method de. As shown in Fig. the real a shock at the tip of the wing. 10 indicates that the diagonal term A22 shows a more significant Pulse Transfer-Function Results difference in the real components of the viscous and inviscid Rigid pitching and plunging The generalized forces forces at the experimental flutter reduced frequency of 0. 10 restarted from a steady-state flow condition at = 0 . Fig. The diagonal terms A11. Figure 8 Comparison of generalized aerodynamic forces The frequency resolution of the pulse analyses is 0. and A is the pitching moment due to pitch.50 inch at the wing tip. shown in Fig. for the rigid plunge and pitch of Wing 445.141 are also indicated by the formation of in comparison with the Euler forces. The pulse calculations are restarted from a steady-state flow con- dition at = 0 .005 for mode 2. A33 on the surface pressures at M1 = 1. Figure 10 where the experimental flutter boundary and the computed indicates that the same is true for the GAF’s at M1 = 1. and A44. These displacements result in a maximum deflection of approximately 0.96 and 1.

Euler Navier-Stokes Real Real Im. Force Real Im. Im. a estimated value of 0. 10 . Reduced Frequency Reduced Frequency Reduced Frequency Reduced Frequency Figure 9 Comparison of generalized aerodynamic forces for the first four structural modes of Wing 445. this configuration has no static aeroelastic deflections.96 to 1. Since the wing A comparison of results in Figs. be considered. Im. The computed flutter character- M1 .03 was ap- Aeroelastic Results plied to each mode in order to evaluate the effects of struc- V-g analysis Flutter characteristics are determined in tural damping. the decrease in the real component of A22 indicates 2 and 3 (3 modes). attack. Im. The an aft shift in the aerodynamic center of the wing. Real Real Gen. Im.141 which is greater The linear stability analysis was performed for three for the GAF’s computed from the Euler calculation. 9 and 10. Real Real Real Im.141 at = 0 using shown in Fig. Im. the harmonic loads shown in Figs. 3 and 4 (4 modes).6 at M1 = 0:96 and = 0 . modes 1. 10 show similar effects due to viscosity as a V-g analysis for M1 = 0. However. 9 for these GAF’s. noted. Gen.960 shown in Fig. As modal configurations: modes 1 and 2 (2 modes). Im. The results for M1 = 0. Real Im. Force Real Real Im. The off-diagonal term A21 exhibits a decrease in the U istics in terms of flutter speed index b! fp and nondimen- magnitude of real component with an increase in M1 which sional flutter frequency ratio !! are shown in Figs. Gen. An additional case ter. Real Im. Real Real Gen. The real flutter characteristics determined from these V-g analyses are component of A44 exhibits a change in sign with increasing shown in Figs. 11 and 12. but since no structural damping was measured for the wind-tunnel model. 11 and is also an indication of an aft shift in the aerodynamic cen. Force Im. Other off-diagonal terms exhibit changes in sign with with structural damping (gi ) was computed for the 4–mode increase in M1 . the real component of A22 ex. The di. and A44 in Fig. Force Real Im. 12 along with the experimental results. Im. and modes 1. analyses.960 and 1. hibits a decrease from M1 = 0. at zero degrees angle of effects of increasing the free-stream Mach number. Im. 2. agonal terms A11 and A33 exhibit only slight changes with Therefore. static aeroelastic characteristics did not need to increase in M1 . 9 and 10 shows the has a symmetric airfoil section.

12 indicates that in- dicted in Ref.86–0. ing the effects of viscosity increases the flutter speed index by 9. 13. the significant decrease in Navier-Stokes GAF’s with structural damping included are the real component of A22 noted in the Euler pulse analyses within 5 percent of the experimental value of flutter speed from M1 = 0. Gen. Force Real Im. Based on the correlation.000 per foot. Im. Real Real Real Gen.000 Stokes GAF’s is lower than the flutter speed index predicted 11 . Real Real Im. The present 4–mode Euler flutter analysis predicts ison between the flutter characteristics predicted with the a lower flutter speed (approximately 7 percent) than that pre. Force Im. Force Im. The Reynolds numbers at flutter predicted in the V-g analysis. Reduced Frequency Reduced Frequency Reduced Frequency Reduced Frequency Figure 10 Comparison of generalized aerodynamic forces for the first four structural modes of Wing 445.141 provides the stabilizing aerody- index. Euler Navier-Stokes Real Im. the flutter speed index predicted with the Navier- analyses with the Navier-Stokes GAF’s range from 374. correlation with the experimental results. The 4–mode Euler and Navier-Stokes results (with namics that produced the rapid rise on the flutter boundary structural damping) are shown in Fig.96 to 1. Real Im. Force Real Real Im. The results shown in Fig. Im. Including modes 3 and 4 in perimental boundary and the flutter boundary predicted in the analysis tends to further stabilize the wing.6 percent for the 4–mode analysis which improves cor. 13 along with the ex. Im. in this Mach number range. 13. Im. In the 4–mode ses. 12 indicate that modes 3 and relation with the experimental data. Figure 11 also indicates that includ. 4 have a greater effect on the flutter characteristics for the tural damping stabilizes the wing which further improves Euler and Navier-Stokes results at M1 = 1. A compar- Ref. Im. The flutter speed index computed from the the results of the 2–mode analysis. Real Real Gen.5 percent dif- the flutter speed and frequency for either the Euler or the ference from the experimental free-stream Reynolds number Navier-Stokes results.6 at M1 = 1:141 and = 0 . (Q=Qexp = 0. a maximum of 4. Real Gen. The addition of struc. Real Im.92). 11 indicate that modes 3 and 4 do not significantly affect per foot to 382. This is thought to be due to the increase in cluding the effects of viscosity significantly improves the spatial resolution of the Euler grid used in the present analy. Euler and Navier-Stokes GAF’s in Fig. Real Im.141. Im. Real Im.

however. The flutter speed index analyses with the Navier-Stokes GAF’s range from 486. Stokes time-marching aeroelastic analyses are included with ses. the flutter speed in.000 and flutter frequency from the Euler time-marching analysis per foot to 502. for Wing 445. As mentioned previously. However.6 at M1 = 1:141 and = 0 .96.46). Stokes time-marching analysis are high in comparison with the Navier-Stokes V-g (4–mode) results by approximately 5 percent. 13. Figure 12 also shows that the inclusion of structural Time-marching analysis Time-marching calculations damping destabilizes the wing at this Mach number which are performed at M1 = 1. This is also thought to be due to the increase in 1 percent difference). pressures ranging from Q=Qexp = 1. The flutter higher speed (approximately 3 percent) than that predicted results are. 13 along with the experimen. compare closely with the results of the Euler V-g (4–mode) ference from the experimental free-stream Reynolds number analysis. The Reynolds numbers at flutter predicted in the V-g the V-g results shown in Fig.85 (flutter dy- The present Euler flutter analysis predicts flutter at a slightly namic pressure predicted at Q=Qexp = 1. at this free-stream Mach number than at M1 = 0. (The Stokes equations because more significant discrepancies fact that the addition of structural damping destabilizes the were noted between the computed and experimental results wing is somewhat unusual). An additional Navier-Stokes time-marching calcu- 12 . variation in the free-stream Reynolds number at dynamic tal boundary and the flutter boundary predicted in Ref. The results of the Euler and Navier- spatial resolution of the Euler grid used in the present analy. The flutter characteristics predicted by the Navier- (Q=Qexp = 1. 12.28–1.4 to 1.141 using the Euler and Navier- further improves the correlation with the experiment.000 per foot.5 percent higher than the experimental value. In dex computed using the Navier-Stokes GAF’s with structural the time-marching analyses. 13.6 at M1 = 0:96 and = 0 . a maximum of 9. the viscous The 4–mode Euler and Navier-Stokes results (with structural time-marching calculations are performed with and without damping) are shown in Fig. the first four modes are used to damping is 17. independent of this change (less than in Ref.61). model the structure. with the Euler GAF’s by 46 percent of the experimental value.Figure 11 Comparison of computed flutter characteristics Figure 12 Comparison of computed flutter characteristics for Wing 445.6 percent dif.

structural damping and number of modes in the higher structural modes in the V-g analysis was less signif- structural model were investigated. A linear stability analysis mental data. The paper boundary as well. the results for the free-stream Mach number of and a time-marching aeroelastic analysis were used to deter. several issues concerning the modeling of this aeroelastic problem still need to be ad- dressed. The results of the V-g analysis indicated for the time-marching analysis were within 5 percent of the that the effect of viscosity on the flutter characteristics was flutter speed predicted by the V-g analysis. Although the flutter characteristics pre- presented a brief description of the aeroelastic method and dicted by the Navier-Stokes V-g analysis for the free-stream presented unsteady calculations which verified the method Mach number of 0.96 compared very well with the experi- for Navier-Stokes calculations. free-stream Mach number of 1. the aeroelastic method was extended and evaluated for the effect of structural damping was to delay the rise in the applications using Navier-Stokes aerodynamics. 1. was previously modified for the V-g results with Euler flutter results from Ref.6. at a free-stream Mach number of 1.141. Conclusions The flutter characteristics of an isolated 45  swept-back wing.96 and 1. marching analysis were within 2 percent of the flutter speed ditional V-g flutter analysis at free-stream Mach numbers of predicted by the V-g analysis. aeroelastic analysis of wings using the un. g flutter analysis method. ized aerodynamic forces of the wing were computed using Time-marching aeroelastic calculations were performed the Euler and Navier-Stokes equations for a range of re. information concerning transition over the wing model is unknown. Wing 445.141 included viscous structural damping added to all four modes.96 than at the In the linear stability analysis.141 using the Euler duced frequency using the pulse transfer-function method. icant at the free-stream Mach number of 0. which is a three-dimensional. The effects of in-plane deflections are an additional unknown. The Euler results for the time- ing these unsteady generalized aerodynamic forces in a tra. and Navier-Stokes equations to compare with the linear V- The flutter characteristics of the wing were determined us. Although the additional modeling of the viscous bound- ary layer did improve the correlation with the experiment for the cases considered in this study. The magnitude of the vis- cous damping corresponded to the amount of gi added in the V-g analysis. The addition of the viscous type of structural damping to the time-marching analysis predicted a drop in the flutter speed index which is the same percentage drop as that predicted from the V-g analysis.141. These modifications included the in- corporation of a deforming mesh algorithm and the addition to delay the rise in the flutter boundary on the supersonic of the structural equations of motion for their simultaneous side which significantly improved correlation with the ex- time integration with the governing flow equations. the unsteady general. The results of the V-g analysis also indicate that higher modes might be needed in the computational model at the higher M1 . accurate modeling of the tip geometry (the wind tunnel model has a cut-off tip) is difficult for body-fitted meshes. In addition. were studied using an unsteady Euler and Navier-Stokes algorithm in order to investigate a pre- viously noted discrepancy between Euler flutter character- istics and the experimental data. and the effect of this loss in geometric definition on the tip aerodynamics is unknown. upwind Euler/Navier-Stokes Figure 13 Comparison of Euler and Navier-Stokes code (CFL3D Version 2. Although the 13 . and the Navier-Stokes results 0.141 were still high in comparison with the experimental mine the flutter characteristics of the wing. time-marching. Since the wind tunnel test was performed under free-transition conditions. In this perimental boundary. The algorithm. and therefore. implicit. The V-g analysis also indicated that paper. it is difficult to evaluate the accuracy of the modeling of the boundary layer. The effect on the flutter characteristics of including viscosity.lation performed at M1 = 1.6. steady Euler equations. 13 and experimental data for Wing 445.1). Effects of fluid data.

Krist of ViGYAN Inc. “Current Status of Flutter Characteristics.. “Application of a Streamwise Upwind Algorithm for [23] Rausch. Including Flexibility. W.. [1] Ballhaus. and Yang. B. 91-1106.” AIAA [8] Obayashi. B. Rumsey [13] Lee-Rausch. References [15] Yang. “Transonic Time- Response Analysis of 3-Degree-of-Freedom Conven- tional and Supercritical Airfoils. G.. Y. April 1992. T. namic and Aeroelastic Calculations of Wings Using Eu.. 1988. W. Y. Batina J. No. 16.. and Sherrie L. T. August 1990.. 1038–1046.. Dynamic Meshes. Swept-Tapered Wings. St. [11] Robinson.. P. T. R. Jr. Number Effects on Transonic Aeroelastic Forces and [2] Edwards. “An Automated No. 91-1109. May 1982.” Journal of Aircraft.” AIAA Paper No. and Bennett. J. [10] Obayashi. Louis. [17] Rausch. Solutions Including Angle-of-Attack Effects. and Bennett.. T. M.” AIAA Paper of CFL3D. “User’s Guide [6] Schuster. D. and Guruswamy. and Ricketts. April 1991.. pp. Y. the V-g analysis did.” Pre. pp.. 11. for their many helpful discussions con- cerning CFL3D. “Static Aeroe. 88-2281. J. R. no. no. P. H. Thomas.. R. “Mach AIAA Journal. November 1991. Whitlow.. “Aeroelastic Analysis of Wings Using the Euler Acknowledgments Equations with a Deforming Mesh. Y. and Seidel. “Computation vol. Bennett. and Batina.. 93-1422.. H. 10. November 1989. Batina. “Three- of McDonnell Aircraft Company. K. “Modern Wing Flutter Analysis by Computational ler Equations. Y. R. J. indicate Vortex Interaction on a Flexible Delta Wing. January 1990. vol. G. E. D.. pp. 83-1811. 5. H..” Journal of Aircraft. 1978. for a Modular Flutter Analysis Software System (FAST lastic Analysis of Fighter Aircraft Using a Three. “Navier-Stokes Computations on of Aircraft. 1974. P. 89-1183. Version 1. T. J.. [20] Desmarais.. [19] Bennett. and Goorjian. D. 75–80.” May 1975. Virginia.. Vadyak. [18] Cunningham. “Spa- sented at the AGARD Structures and Materials Panel tial Adaption Procedures on Unstructured Meshes for Specialist’s Meeting on Transonic Unsteady Aerody. Paper No. [4] Guruswamy. Batina. “Vortical Flow Computations on pp. 28. Paper No. P.” AIAA the significant effect of viscosity on the supersonic flutter Paper No. 2. H. T. Guruswamy.. vol. for Dimensional Time-Marching Aeroelastic Analyses Us- sharing his experiences concerning the aeroelastic version ing An Unstructured-Grid Euler Method . and Yang. “Numerical Simulation of Vortical no. E. vol. J. G. Batina. R. 117–124. 92-2506.0). of Unsteady Transonic Flows by the Indicial Method.” [16] Mohr R.. “Vortical Flow Computations on Functions. J. the authors would like to thank [14] Edwards. vol. M. D. R... T. A. P.” namics and Aeroelasticity. S.” Journal of Aircraft. T. Hampton. M. R. J.” gley Research Center. 2. [3] Guruswamy.. H. 20. “Unsteady Shock- marching results. 962–968. J.. Y. 25... L.” NASA TM 100492. July 1983. a Flexible Blended Wing-Body Configuration. and Desmarais. S. and Yang.. J. J.. “Curve Fitting of January 1989. pp. M.” Journal of Aircraft..” Journal of Aircraft. H. W. G. their colleague Robert M... 90-0435. 1983. Paper No. N. T.. AIAA Paper No. and Malone. N. no. “Time-Marching Transonic Flutter sights into computational aeroelasticity and flutter. M. “Some Recent Applications of XTRAN3S... N. W. 82-3685.. Swept Flexible Wings Using Navier-Stokes Equations. and Goorjian. M. 27. W. The authors would also like to thank James L. pp. Kyle Anderson and Christopher L. Unsteady Transonic Computations Over Oscillating “Euler Flutter Analysis of Airfoils Using Unstructured Wings. R. April 1991. Batina.. 26.” AIAA Paper [21] Desmarais. Fluid Dynamics Methods. April 1993. R. Batina. April 1990. W. 90-3103. Dimensional Navier-Stokes Algorithm. Paper No. T.” AIAA [22] Seidel. G. 91-1013-CP.. 8. Robinson [12] Rausch.. Computational Methods for Unsteady Transonic Un. T. A.” Flutter Testing Techniques NASA SP-415. and Atta. and Bennett. M. 1.” Journal [7] Guruswamy.. however. G. T. and Yang.. Flows on Flexible Wings.... 89-0537. P. P. H.” AIAA Paper No... D. Prodedure for Computing Flutter Eigenvalues. 90-1152. vol.. T. AIAA Paper No. Finally. pp. The authors would like to thank Brian A.. April 1989. P. 778–788. R. G. R. “Wing Flutter of the Computational Aerodynamics Branch. Thomas.. and Batina. “Time-Accurate Unsteady Aerody. R. AIAA Paper No. Aeroelastic Transient Response Data with Exponential [5] Guruswamy. steady Aerodynamics and Aeroelastic Analysis.” AIAA Paper No. Bennett for sharing his many in. D. “Comparison of Finite Volume Flux Vector Splitting 14 .. B. and van Leer. A. Accurate Unsteady Aerodynamic Flow Computation. 1990. no. J. T.. Missouri. October 1991.” AIAA [24] Anderson.. Bennett.” AIAA Paper No. NASA Lan- Boundary Prediction Using Unsteady Euler Method. [9] Guruswamy. P. May 1978. W. April 1988. boundary for this wing. and Yang. 703–710. M. 436–443. R. F. vol. J..linear stability analyisis flutter results did not match the time.. April 1991. J.

vol. 89-1189. Flutter Characteristics of a 45 deg Sweptback Wing mation and Algebraic Model for Separated Turbulent Planform in Air and in Freon-12 in the Langley Tran- Flows. J.” AIAA Paper No..6. [25] Anderson.. [29] Batina. 9. “Thin Layer Approxi.. N. and Turbulence Mod. T. Land. S. E. 1985. L. for the Euler Equations. April 1989. AGARD CP-437. 24. Paper No.. . Jr.. March [28] Johnson.. with an Eddy-Viscosity/Reynolds-Shear-Stress Closure pp.” [31] Yates. August 1987. “Predictions of Transonic Separated Flow 1963. W. “AGARD Standard Aeroelastic Con- [26] Rumsey.: and Anderson. “Unsteady Euler Algorithm With Unstruc- “Extension and Applications of Flux-Vector Splitting tured Dynamic Mesh for Complex-Aircraft Aeroelastic to Unsteady Calculations on Dynamic Meshes.-Wing 445. and Foughner. Time Step. Jr.” NASA TM 100492. E. Model. T. Jr. and Rumsey.. C. Candidate Configura- Study of Grid Size. H. 1453–1460. May 1988. C. “Measured and Calculated Subsonic and Transonic [27] Baldwin. L.. Thomas. B. 78-257. 87-1152.. 1986. eling on Navier-Stokes Computations Over Airfoils. “Parametric figuration for Dynamic Response.. and Lomax. 5. Paper No. tion I. J.” AIAA Analysis. June 1987.” NASA TN D-1616.” AIAA Journal. [30] Yates.” AIAA Paper No. D.” AIAA Paper No. sonic Dynamics Tunnel.. C.. K. 15 . 85-1683.. K. J. January 1978. W. L.. no. C.