Sei sulla pagina 1di 11

AIAA-2001-0606 NUMERICAL SIMULATION OF VORTICAL FLOWS USING VORTICITY CONFINEMENT COUPLED WITH UNSTRUCTURED GRID

Mitsuhiro Murayama and Kazuhiro Nakahashi Department of Aeronautics and Space Engineering Tohoku University, Sendai, JAPAN Shigeru Obayashi Institute of Fluid Science Tohoku University, Sendai, JAPAN

AIAA-2001-0606 NUMERICAL SIMULATION OF VORTICAL FLOWS USING VORTICITY CONFINEMENT COUPLED WITH UNSTRUCTURED GRID
Mitsuhiro Murayama*, Kazuhiro Nakahashi and Shigeru Obayashi Tohoku University, Sendai 980-8579 Aoba-yama 01, JAPAN

ABSTRACT This paper discusses the use of the vorticity confinement method coupled with the unstructured grid approach to simulate vortical flows. The method is evaluated by several vortical flow computations of the leading-edge separation vortices on delta wings and the wing tip vortices of NACA0012 wing. It is shown that the vorticity confinement can keep the vorticity away from the numerical diffusion effectively. Although further study to reduce the dependency of the confinement coefficient on the grid density is required, the present results indicate the possibility of accurate vortical flow computations by the vorticity confinement method coupled with unstructured and adaptive refinement grids. Accuracy of the vortical flow simulations, however, is still not good enough. Generally, the computational grid becomes rapidly coarser as it becomes far away from the body surface. There, the vortices are highly diffused due to the numerical discretization error. By the use of a highly dense grid, numerical dissipation of the vortices may be minimized. However, flow computations around 3-D complex bodies with large-scale separations and vortices are difficult within the realistic number of grid points. One approach to crisply capture the vortex is to use the adaptive grid method. We proposed an effective and efficient adaptive grid refinement method for the improvement of grid resolution around a vortex center using the vortex center identification method. The resulting method was applied to flows around a delta wing and showed significant improvements in resolution of the leading-edge separation vortices [1]. For further improvements, however, not only the way to reduce dissipation by the refinement of the grid but also the model to keep the vortices from diffusion will be needed. Vorticity confinement method has been proposed to reduce the diffusive property of the vortical flow simulations [2-4]. In the method, the source term added to the Navier-Stokes equations works as it convects the discretization error back into the vortex center and thus confine the vortex. The method is applied to flows around helicopter rotors with fuselage, and shows reasonable improvements in the vortex resolution [3]. However, the method is still needed to be tuned for the confinement parameter at a particular grid. In addition, the vorticity confinement has been used on Cartesian grids or structured surface conforming grids, not on unstructured grids. Unstructured grids will be more suitable for the numerical simulation around 3-D highly complex bodies. The objective of this paper is to develop the vorticity confinement method coupled with the unstructured grid method. The capability of the resulting method is evaluated by the computations of the following four flow fields with vortices,

1.

INTRODUCTION

With the recent progress, the computational fluid dynamics (CFD) is very close to its matured stage in the computation of flows around airplanes at designed conditions. However, it is still difficult to deal with complex flows at off-design conditions where flow separates and vortices characterize the flow features. These vortical flows are often encountered at various important engineering problems. For example, flows around a delta wing at high incidence are characterized by the leading-edge separation. This separation vortex generates a large non-linear lift increment called vortex-induced lift at moderate angles of attack. At higher angles of attack, however, this vortex is to burst, resulting in a sudden decrease of the lift. The BVI (Blade-Vortex Interaction) problems and vibratory loading problems are also caused by impingements of vortices on helicopter rotor and aircraft fuselage. Wing-tip vortices of airplanes during the take-off and landing are another serious problem to deal with the congestion at airports accompanied with the rapid growth of the aviation.
* Graduate student Professor, Department of aeronautics and space engineering, Associate Fellow AIAA Associate professor, Institute of Fluid Science, Senior Member Copyright 2001 by American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

1 American Institute of Aeronautics and Astronautics

(1) a 2D single vortex in freestream, (2) vortical flows around a double delta wing, (3) vortical flows with vortex breakdown around a delta wing, (4) a wing tip vortex of a stationary NACA0012 wing.

2.

VORTICITY CONFINEMENT METHOD

The formulation of the vorticity confinement for compressible flows can be expressed as follows [4]:
QdV + (F(Q) G(Q)) ndS = SdV t

process [1]. The vortex-center identification method employed in this study [6] applies the critical point analysis locally. The vortex centerlines are defined locally one by one using assumption that the local velocity field can be linearly parameterized in tetrahedral computational cells. By using the topological feature, there is no need to decide an intuitive and subjective threshold for the refinement. The cells containing the vortex centerlines and its neighboring cells are flagged for the refinement as shown in Fig. 1. The grid refinement is started from these cells.

(1)

where Q = [ , u , v, w, e]T is the vector of conservative variables; is the density; u , v , w are the velocity components in the x , y , z directions; and e is the total energy. The vectors F (Q) and G (Q) represent the inviscid and viscous flux vectors and n is the outward normal of which is the boundary of the control volume . The term, S , is the confinement term and can be written as
0 fb i j S = fb fb k f V b

(a) (b) Delta Wing ( = 20.5  , M = 0.3, Re = 0.95 10 6 )

(2)

(a) (b) NACA0012 Wing ( = 10.0  , M = 0.12, Re = 0.9 10 6 ) Fig. 1. The refinement indicator using vortex center, (a) vortex centerlines, (b) cells containing the vortex centerlines and its neighboring cells for the refinement The refinement algorithm employed in this study is the Rivaras bisection algorithm [7,8]. A tetrahedron is bisected by a plane passing through the midpoint of its longest edge and through the nodes opposite to this edge. This bisection algorithm allows an improvement of the refined cell shape because of its property to minimize maximal angles. Another advantage of the bisection refinement compared to the commonly-used equi-dividing method is that the refined region is diffused naturally around the indicated cells. This diffusive property of the bisection algorithm is preferable for the present feature-adaptation method.

where f b is a body force per unit mass. Now, this body force for the confinement of the compressible flows is defined as  f b = n (3)
  n =

(4)

where is a confinement parameter which controls  the strength of the confinement and is the vorticity vector. Although the vorticity confinement method is extended to model a boundary layer, the confinement for a boundary layer is not used in this study. The hybrid unstructured grid comprised of tetrahedral, prisms, and pyramids generated by the method described in Ref. 5 is used to compute the boundary layer near the body surface accurately.

3.

ADAPTIVE REFINEMENT METHOD

4.

FLOW SOLVER

An adaptive grid refinement method is used to increase the mesh resolution in the vicinity of the vortex centers. We use the distinct topological flow feature, vortex centerlines, identified by the vortex-center identification method as the refinement indicator instead of the commonly-used variations of the flow variables, such as the density and total pressure which may be noisy and need trial-and-error

The equations (1) are solved by a finite volume cell-vertex scheme, where the control volume i at mesh i is a non-overlapped dual cell. With this control volume boundary, Eq. (1) can be written in an algebraic form as follows,
Q i 1 + = [ S ij h (Q ij , Q ij , n ij ) S ij G (Q ij , n ij )] + S i Vi j ( i ) t j (i )

(5)

2 American Institute of Aeronautics and Astronautics

where

Sij

is a segment area of the control volume

approximation
A = 0.5( A A I ) ,

boundary associated with the edge connecting points i and j . The term h is an inviscid numerical flux vector normal to the control volume boundary, and Q ij are values on both sides of the control volume boundary. The subscript of summation, j (i) , refers to all node points connected to node i . The Harten-Lax-van Leer-Einfeldt-Wada Riemann solver (HLLEW) [9] is used for the numerical flux computations. Second-order spatial accuracy is realized by a liner reconstruction of the primitive variables q = [ , u, v, w, p]T inside the control volume, viz., (6) q ( x, y, z ) = q i + i q i (r ri ) where r is a vector pointing to point ( x, y , z ) ; and i is the node number. The gradients associated with the control volume centroids are volume-averaged gradients computed using the value in the surrounding grid cells. A limiter, , is used to make the scheme monotone. Here Venkatakrishnans limiter [10] is used because of its superior convergence properties. To compute viscous stress and the heat flux terms in G (Q) , spatial derivatives of the primitive variables at each control volume face are evaluated directly at the edges. A one-equation turbulence model by GoldbergRamakrishnan (G-R)[11] was implemented to treat turbulent flows. This model does not involve wall distance explicitly so that it is of great benefit to unstructured grid method. The lower-upper-symmetric Gauss-Seidel (LU-SGS) implicit method originally developed for structured grid is applied in order to compute the high Reynolds number flows efficiently. The LU-SGS method on unstructured grid can be derived by splitting node points j (i) into two groups, j L(i ) , and j U (i) for the first summation in LHS of Eq. (5). With Q = Q n+1 Q n , the final form of the LU-SGS method for the unstructured grid becomes the following two sweeps: Forward sweep: Q = D 1 [R i 0.5 S ij (h j A Q j )] (7a)
jL (i )

[12] of the Jacobian as where A is a spectral radius of

Jacobian A . The lower/upper splitting of Eq. (7) for the unstructured grid is done by a grid reordering technique [13] that was developed to improve the convergence and the vectorization. For unsteady flow, the time accuracy of the LU-SGS solution algorithm is recovered by the Newton iteration using Crank-Nicolson method.

5.

RESULTS

5.1 CASE1: Single Vortex in Freestream For the validation of the present method, a 2D single vortex in freestream was computed first. The formulation of the initial vortex suggested in Ref. [4, 14] is written as follow,
U c r Rc , r < Rc u (r ) = Ar + B r , Rc r R0 U R A= 2c c2 R0 Rc
B=
2 U c R0 Rc 2 2 R0 Rc

(10) (11) (12)

(13) where R0 is an outer radius and Rc is a core radius of the vortex. U c is a maximum core velocity. This initial vortex conditions are added to freestream conditions. Solutions were obtained at a freestream Mach number of M = 0.5 . The Reynolds number is 1.2106. The initial vortex conditions were set to U c = U , Rc = 0.05 , R0 = 10 Rc . The outer boundary is a square whose edges have 1.0 length and periodic boundary conditions are applied. Two types of the unstructured computational grid were used as shown in Fig. 2. Grid 1 which has 14,094 node points has approximately uniform cells. Grid 2 which has 18,024 node points is constructed by the division of the Grid 1 and has relatively dense region in the middle of the grid and non-uniform grid density. The variations of the maximum vorticity magnitude around vortex center and vorticity contours with/without the vorticity confinement are shown in Figs. 3 and 4, respectively. The centered vortex moves downstream and passes through the downstream boundary and reappears from the other upstream side and moves to the center of the grid again. This process is defined as one cycle in these computations.

u (0 ) = 0 , u (Rc ) = U c , u (R0 ) = 0

Backward sweep:
Q i = Q 0.5D 1 i

jU ( i )

ij ( h j

A Q j )

(7b)

where h = h (Q + Q) h(Q) , and


D = ( jt + 0.5 Ri =
V

The term D is diagonalized by the Jameson-Turkel

j (i )

S ij A )I ,

(8) (9)

j (i )

S ij h + ij

j (i )

n S ij G ij + Vi S i

3 American Institute of Aeronautics and Astronautics

(a)

(b) Without confinement

(c)

(a) Grid 1 (14,094 nodes)

(a)

(b) (c) With confinement (=0.003)

Fig. 4. The vorticity contours on Grid 1 with/without the vorticity confinement, (a) after one cycle, (b) after ten cycles, (c) after fifteen cycles (b) Grid2 (18,024 nodes) Fig. 2. Computational grid In the case without the vorticity confinement, the vortex rapidly diffused. In the case with the vorticity confinement with moderate values of the confinement parameter , it can be seen that the degrees of the dissipation of the vortex are decreased by the confinement method and the strength of the vorticity is preserved even after 50 cycles. However, this effect is influenced by the values of the confinement parameter . Too large values of the parameter may lead to the non-physical results. Moreover, in the case using the non-uniform grid, Grid 2, the results are fluctuant, especially when the relatively large value of the parameter is employed. These results suggest that the effect of the confinement highly depends on the grid density and the confinement parameter.
30 40 50

70

Vorticity Magnitude

60 50 40 30 20 10 0 0

Original Result EPS0.001 EPS0.003 EPS0.005 EPS0.01

10

Cycle (a) Grid 1

20

70

5.2 CASE2: Vortical Flows Around a Double Delta Wing


Original Result EPS0.001 EPS0.003 EPS0.005 EPS0.01

Vorticity Magnitude

60 50 40 30 20 10 0 0 5 10 15 20

25

30

Cycle

(b) Grid 2 Fig. 3. The variations of the maximum vorticity magnitude around vortex center with/without the vorticity confinement (EPS: the value of the vorticity confinement parameter, )

The method was applied to vortical flows on delta wings with high incidence. The geometry used in the present study is a double delta wing shown in Fig. 5. The leading-edge sweep angle is 80 degrees at the strake and 60 degrees at the main wing. The thickness is 0.6% of the root chord length and the leading edge is rounded. The first point above the wing surface is located at a distance of 8.010-5 of the root chord length. The outer boundaries are located 10-root chord length away from the body surface. The total number of the initial grid points is 676,541. The computational grids in the cross flow plane at 62.5% chord length are shown in Fig. 6. Solutions were obtained at a freestream Mach number of M = 0.3 and angle of attack of 12. The Reynolds number based on the root chord is 1.3106. In this study, laminar flow was assumed.

4 American Institute of Aeronautics and Astronautics

Fig. 5. Surface grid of the double-delta wing (a) Initial Grid

Fig. 6. Cut view of the grid at x/c=0.625 First, the results without using vorticity confinement are discussed. Total pressure contours computed on the initial grid and the adaptive grid refined twice are shown in Fig. 7. The adaptive grid has 1,042,292 grid points. In this flow field, two leading-edge separation vortices originating from the strake and main wing interact with each other in the main wing region. The strength of the strake vortex becomes weaker at the downstream of the kink because the energy feeding to the vortex from the strake leading edge will decrease, while the strength of the main wing vortex increase as the distance from the kink increases downstream. The weak strake vortex is moved outward influenced by the velocity field induced by the relatively strong wing vortex and merges with the wing vortex eventually. Compared with the results on the initial grid, the primary and secondary vortices are more clearly captured by the adaptive grid refinement method. For the validation, the surface pressure coefficients at different axial locations were compared with experimental data by the Brennenstuhl and Hummel [15] in Fig. 8. Two pressure peaks by the inner strake vortex and outer wing vortex can be seen in this experimental data. In the computational results on the initial grid, however, the inner peak at x/c=0.625 by the strake vortex is much lower and it can be seen that the strake vortex has already been weaken by the numerical diffusion due to the lack of the grid density. By using the adaptive grid, the diffusion of the strake vortex is suppressed as shown in Fig. 7. However, the improvement of the surface pressure prediction is very small as shown in Fig. 8. The main reason of this discrepancy may be caused by the fact that laminar flow was assumed in the computation although the laminar/turbulent transition was probably observed in the experiment, as discussed in Ref. 16.

(b) Adaptive grid Fig. 7 The total pressure contours ( = 12.0  , M = 0.3, Re = 1.0 10 6 )
2 1.5 1 Experiment Initial grid Adaptive grid

-Cp

0.5 0 -0.5 0 0.2 0.4 0.6 0.8 1

Spanwise location (a) Surface pressure coefficients at x/c=0.625


2 1.5 1 0.5 0 -0.5 Experiment Initial grid Adaptive grid

-Cp

0.2

0.4

0.6

0.8

Spanwise location (b) Surface pressure coefficients at x/c=0.75

Fig. 8. Surface pressure coefficients at differential axial locations without the vorticity confinement ( = 12.0  , M = 0.3, Re = 1.0 10 6 )

5 American Institute of Aeronautics and Astronautics

The results using vorticity confinement are compared in Fig. 9. The confinement was applied to the regions without a boundary layer. Computations were performed with varying the value of the confinement parameter on the initial grid. With moderate values of the confinement parameter, the resolution of vorticity is effectively improved. However, the method creates extra correction in the case that the values of are too large. The total pressure contours with/without vorticity confinement at x/c=0.625 and 0.75 are shown in Fig. 10. From these results, it can be seen that the vortex core position is not affected by the confinement and the vorticity is concentrated to the vortex core reasonably. The confinement is more effective than the adaptive grid refinement method even with much less grid points.

(i) x/c=0.625

(ii) x/c=0.75 (a) Initial grid

(i) x/c=0.625

(ii) x/c=0.75 (b) Adaptive grid

(i) x/c=0.625 (ii) x/c=0.75 (c) Initial grid with the vorticity confinement, =0.01 Fig. 10. The total pressure contours with/without vorticity confinement

2 1.5 1
Experiment Initial grid EPS0.005 EPS0.01 EPS0.05

-Cp

5.3 CASE3: Vortical Flows Breakdown Around a Delta Wing

with

Vortex

0.5 0 -0.5 0 0.2 0.4 0.6 0.8 1

Spanwise location (a) Surface pressure coefficients at x/c=0.625


2 1.5 1
Experiment Initial grid EPS0.005 EPS0.01 EPS0.05

The confinement method was tested for the vortex breakdown simulations shown in Fig. 11. The geometry used in this computation is a slender delta wing of aspect ratio of unity and a sweep angle of 76 degrees. The freestream Mach number is 0.3, angle of attack is 32, and the Reynolds number based on the root chord is 1.0 10 6 . Laminar flow was assumed. This flow field was numerically simulated by the present unstructured grid method [1] and the predicted positions of the vortex breakdown showed good agreements with experiments of Hummel and Srinivasan [17]. In Fig. 11, the streamlines starting from the wing apex show the vortex breakdown pattern near the trailing edge of the delta wing. The results using the vorticity confinement computed on the same grid and conditions are shown in Fig. 12. With larger values of , the vortex breakdown does not appear.

-Cp

0.5 0 -0.5

0.2

0.4

0.6

0.8

Spanwise location (b) Surface pressure coefficients at x/c=0.75

Fig. 9. Surface pressure coefficients at differential axial locations with the vorticity confinement ( = 12.0  , M = 0.3, Re = 1.0 10 6 ); EPS is a value of the confinement parameter Fig. 11. The computed flow fields around a delta wing with vortex breakdown ( = 32.0  , M = 0.3, Re = 1.0 10 6 )

6 American Institute of Aeronautics and Astronautics

5.4 CASE4: Wing Tip Vortex of NACA0012 Wing Finally, the present method is applied to the computations of a wing tip vortex of a NACA0012 wing. The geometry used in the present study is a NACA0012 rectangular wing of aspect ratio 3 as shown in Fig. 15. The initial grid has nearly homogeneous grid density in the wake region and the total number of the grid points is 701,037. The axial direction coincides with the wing chord direction. The outer boundary is a sphere whose radius is 15-root chord length. Solutions were obtained at a freestream Mach number, M=0.12. The Reynolds number based on the root chord is 0.9106 and angle of attack is 10.0.

(a) =0.001

(b) =0.005

(c) =0.01 Fig. 12. The computed flow fields around a delta wing with vortex breakdown using the vorticity confinement ( = 32.0  , M = 0.3, Re = 1.0 10 6 ) Figures 13 and 14 show the results with large-scale vortex breakdown at higher angle of attack of 40. By the introduction of the vorticity confinement with large values of , the starting points of the vortex breakdown and its pattern are considerably influenced. The vorticity confinement gives rotation components to the vortex and confines the vortex. However, with larger values of confinement parameter, excessive rotational components are added and may destroy the flow physics.

(a) Surface grid

(b) Close-up view

Fig. 13. The computed flow fields around a delta wing with vortex breakdown ( = 40.0  , M = 0.3, Re = 1.0 10 6 )

(c) Cut view of the fine grid at 95% semi-span Fig. 15. Computational grid of NACA0012 The vortex centerlines and vorticity contours obtained on the initial grid is shown in Figs. 16 and 17, respectively. The wing tip vortex far away from the trailing edge is highly diffused although the vortex centerlines are identified clearly.

a) =0.001

(b) =0.005

(c) =0.01 Fig. 14. The computed flow fields around a delta wing with vortex breakdown using the vorticity confinement ( = 40.0  , M = 0.3, Re = 1.0 10 6 )

Fig. 16. The vortex centerlines at the initial grid

7 American Institute of Aeronautics and Astronautics

14 12 10
Initial Grid Adaptive Grid 1 Adaptve Grid 2 Adaptive Grid 3 Adaptive Grid 4

Vorticity

8 6 4 2 0 2 4 6

10

Fig. 17. Contours of the vorticity magnitude at the initial grid Therefore, adaptive grid refinement method using the vortex center for the concentration of the grid points around the vortex core to decrease the diffusion of the vorticity is applied to the initial grid. The grid is refined twice and the resulting number of grid points is 815,262. The results are shown in Figs. 18-20. These results show the improvement about the vorticity magnitude by the adaptive grid refinement. More grid refinement may improve the results, but from the limitation of the computational resources, the physical model to keep the vortices from diffusion may be required.

Distance From Trailing Edge (x/c)

Fig. 20. The variation of the maximum vorticity magnitude around vortex center The results using vorticity confinement are shown in Fig. 21. The confinement was applied to the wake regions. The vorticity contours with vorticity confinement on the initial grid are shown in Fig. 22. In the case of the initial grid, the results show some improvements although the overall level of the vorticity magnitude is still low due to the poor grid density of the grid in the wake region near the trailing edge. However, larger values of the confinement parameter lead to extra effects as shown in Fig. 22(b). The results on the adapted grid refined twice with the vorticity confinement are shown in Figs. 21(b) and 23. It is apparent that the vorticity was confined and the method successfully decreased the numerical diffusion of the vorticity by the coupling of the adaptive grid refinement method and the vorticity confinement. By the use of the confinement, the diffusion of the vortex is suppressed at the station about ten chord length away from the trailing edge. The vorticity contours on a cut view at x/c=5.0 are shown in Fig. 24. In these figures, it can be seen that the vortex core position is not affected by the confinement and the vorticity is concentrated to the vortex core. However, the effect highly depends on the value of the confinement constant coefficient, and grid density. Moderate improves the results, while the large works extra correction. And in Fig. 21(b), the computed vorticity magnitude using the vorticity confinement is fluctuant and not smooth at both value of . This is because the grid size on unstructured grid is not regular although the adaptive grid refinement was applied and the grid density becomes approximately homogenous around vortex center. The confinement effects may depend on the grid. From these results, it can be seen that optimal which is different by the place and grid may be required again.

(a) Initial Grid

(b) Adapted grid refined twice

Fig. 18. Cut views of the grid at x/c=5.0

Fig. 19. Contours of the vorticity magnitude at the adaptive grid refined twice

8 American Institute of Aeronautics and Astronautics

14 12 10
Initial Grid EPS0.005 EPS0.01

Vorticity

8 6 4 2 0 2 4 6 8 10

Distance From Trailing Edge (x/c) (a) Initial grid


14 12 10
Adaptive Grid 2 EPS0.01 EPS0.02

=0.02 Fig. 23. Contours of the vorticity magnitude at the adaptive grid refined twice with vorticity confinement

Vorticity

8 6 4 2 0 2 4 6 8 10

Distance From Trailing Edge (x/c) (b) Adaptive grid refined twice

Fig. 21. The variation of the maximum vorticity magnitude around vortex center with vorticity confinement (EPS: the value of the vorticity confinement parameter )

(b) With confinement =0.02 Fig. 24. Vorticity contours in the cross flow plane at x/c=5.0 of the adaptive grid refined twice

(a) Without confinement

6.

CONCLUSION

(a) =0.01

The vorticity confinement method coupled with unstructured grid has been applied to the numerical simulations of four vortical flows. In the case of a single vortex in freestream, the effect of the vorticity confinement on unstructured was validated and problems about the grid dependency were pointed out. In the case of vortical flows around delta wings, the confinement was found more effective than the adaptive grid refinement method. In the vortex breakdown case, it was demonstrated that the use of excessive confinement parameters could suppress the breakdown and thus destroy the flow physics. Finally, in the case of a wing tip vortex of NACA0012 wing, it was shown that the confinement method could suppress the numerical diffusion of the vortex far away from the trailing edge by the coupling of the adaptive grid refinement method. The results obtained in this study show that, although further study to reduce the dependency of the confinement coefficient, on the grid density is required, the vortex confinement method coupled with unstructured and adaptive refinement grids has the possibility of accurate simulations of vortical flows.

(b) =0.05 Fig. 22. Contours of the vorticity magnitude at the initial grid with vorticity confinement

9 American Institute of Aeronautics and Astronautics

REFERENCES [1] Murayama, M., Nakahashi, K., and Sawada, K., Numerical Simulation of Vortex Breakdown Using Adaptive Grid Refinement with Vortex-Center Identification, AIAA Paper 2000-0806, 2000. [2] Steinhoff, J. and Underhill, D., Modification of the Euler Equations for Vorticity Confinement: Application to the computation of interacting vortex rings, Physics of Fluids, Vol. 6, 1994, pp.2738-2744. [3] Steinhoff, J., Yonghu, W., and Lesong, W., Efficient Computation of Separating High Reynolds Number Incompressible Flows Using Vorticity Confinement, AIAA Paper 99-3316, 1999. [4] Hu, G., Grossman, B., and Steinhoff, J., A Numerical Method for Vortex Confinement in Compressible Flow, AIAA Paper 2000-0281, 2000. [5] Sharov, D. and Nakahashi, K., Hybrid Prismatic/Tetrahedral Grid Generation for Viscous Flow Applications, AIAA J., Vol. 36, No.2, pp.157-162, 1998. [6] Sawada, K., A Convenient Visualization Method for Identifying Vortex Centers, Transactions of the Japan Society for Aeronautical and Space Sciences, Vol. 38, No. 120, 1995. [7] Rivara, M. C., Selective Refinement/ Derefinement Algorithms for Sequences of Nested Triangulations, Int. Journal of Numerical Methods in Engineering, Vol. 28,1989, pp. 2889-2906. [8] Sharov, D. and Fujii, K., Three-Dimensional Adaptive Bisection of Unstructured Grids for Transient Compressible Flow Computations, AIAA Paper 95-1708, 1995. [9] Obayashi, S. and Guruswamy, G. P., Convergence Acceleration of an Aeroelastic Navier-Stokes Solver, AIAA Paper 94-2268, 1994. [10] Venkatakrishnan, V., On the Accuracy of Limiters and Convergence to Steady State Solutions, AIAA Paper 93-0880, 1993. [11] Goldberg, U. C. and Ramakrishnan, S. V., A Pointwise Version of Baldwin-Barth Turbulence Model, Computational Fluid Dynamics, Vol. 1, 1993, pp.321-338. [12] Jameson, A. and Turkel, E., Implicit Schemes and LU Decompositions, Mathematics of Computation, Vol. 37, No. 156, 1981, pp.385-397. [13] Sharov, D. and Nakahashi, K., Reordering of 3-D Hybrid Unstructured Grids for Vectorized LU-SGS Navier-Stokes Computations, AIAA Paper 97-2102, 1997. [14] Povitsky, A. and Ofengeim, D., Numerical Study of Interaction of a Vortical Density Inhomogeneity with Shock and Expansion

Waves, ICASE Report No. 98-10, 1998. [15] Brennenstuhl, U. and Hummel, D., Vortex Formation Over Double-Delta Wings, Proceedings of the 13th Congress of the International Council of the Aeronautical Science, ICAS Paper 82-6.6.3, 1982, pp. 1133-1146. [16] Fujii, K. and Schiff, L. B., Numerical Simulation of Vortical Flows Over a Strake-Delta Wing, AIAA Journal, Vol. 27, No. 9, 1989, pp.1153-1162. [17] Hummel, D. and Srinivasan, P. S., Vortex Breakdown Effects on the Low-Speed Aerodynamic Characteristics of Slender Delta Wings in Symmetrical Flow, Royal Aeronautical Society Journal, Vol. 71, 1967, pp.319-322.

10 American Institute of Aeronautics and Astronautics

Potrebbero piacerti anche