Professional Documents
Culture Documents
Ryoji Kojima∗
University of Tokyo, Sagamihara, Kanagawa 229-8510, Japan
and
Taku Nonomura,† Akira Oyama,‡ and Kozo Fujii§
Japan Aerospace Exploration Agency, Sagamihara, Kanagawa 229-8510, Japan
DOI: 10.2514/1.C031849
In this study, the flowfields around NACA0012 and NACA0002 airfoils at Reynolds number of 23,000 and the
aerodynamic characteristics of these flowfields were analyzed using implicit large-eddy simulation and laminar-flow
simulation. Around this Reynolds number, the flow over an airfoil separates, transits, and reattaches, resulting in the
Downloaded by ROKETSAN MISSLES INC. on February 7, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C031849
generation of a laminar separation bubble at the angle of attack in a certain degree range. Over an NACA0012 airfoil,
the separation point moves toward its leading edge with an increasing angle of attack, and the separated flow may
transit to create a short bubble. On the other hand, over an NACA0002 airfoil, the separation point is kept at its
leading edge, and the separated flow may transit to create a long bubble. Moreover, nonlinearity appears in the lift
curve of the NACA0012 airfoil, but not in that of NACA0002, despite the existence of a laminar separation bubble.
set to 0.2, the value at which compressibility can be ignored and y,v
computational efficiency can be improved. The Reynolds number Re x,u
was set to 23,000, which is the same as that in the previous experi- z,w
mental studies. The specific heat ratio was set to 1.4. The angles of
attack were set to 3.0, 6.0, and 9.0 deg for LES computation, and 0.0,
1.5, 3.0, 4.5, 6.0, 7.5, and 9.0 deg for laminar computation. Note that
LES computation is approximately 300 times more expensive than
laminar computation. However, laminar computation is unable to
treat turbulent transitions.
B. Computational Methods
In this study, LANS3D [9] (developed in ISAS/JAXA) was Fig. 3 Computational grid around NACA0012 airfoil.
adopted and 3D implicit LES (iLES) and 2D laminar simulation was
Downloaded by ROKETSAN MISSLES INC. on February 7, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C031849
C. Computational Mesh
The computational meshes for the 3D iLES around NACA0012
and NACA0002 are shown in Figs. 3 and 4, respectively. For the span
direction, 20% chord length was computed using the periodic
α = 9 .0
c)
x/c
α = 3.0
a)
α = 3 .0
a)
α = 6.0
b)
Downloaded by ROKETSAN MISSLES INC. on February 7, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C031849
α = 9.0
c) α = 6 .0
b)
thin airfoils such as thin flat plates, where the separation point is fixed
at the leading edge.
First, the instantaneous flowfields around NACA0002 at α
3 deg are discussed. The isosurfaces shown in Figs. 9a and 10a
illustrate that the 2D coherent spanwise vortices are generated
periodically. These vortices break up and generate hairpin vortices, as
in the NACA0012 flowfields. Next, corresponding figures at α
6 deg are discussed. The isosurfaces shown in Figs. 9b and 10b
illustrate that the flow structure at α 6 deg similar to the α
α = 9 .0
3 deg condition. Comparing Figs. 9a and 9b, the 2D coherent
vortices at α 6 deg breaks up earlier than those at α 3 deg. The c)
vortex structures after the breakup are finer and more complex at
α 6 deg than at α 3 deg.
Finally, the similar figures at α 9 deg are discussed. The
isosurfaces shown in Figs. 9c and 10c illustrate that the 2D coherent
vortices at α 9 deg breaks up earlier than those at α 6 deg. The
hairpin vortices generated by the breakup of the 2D coherent vortices Fig. 9 Q-Isosurfaces of instantaneous flows around NACA0002 airfoil.
are stretched in a direction away from the wing surface.
Comparison of NACA0012 and NACA0002 The comparison of
6 deg as discussed later) are compared with those for which the flow
flowfields around NACA0012 at α 6 and 9 deg and those around
does not reattach after the separation (NACA0012 at α 9 deg and
NACA0002 at α 6 and 9 deg shows that the 2D coherent vortices
NACA0002 at α 9 deg as discussed later). This comparison
around NACA0002 are generated earlier and have spatially shorter
shows that the reattachment of the flow depends on whether the 3D
intervals than those around NACA0012. This is because the flow is
turbulent vortices generated by the breakup of the 2D vortices reach
further accelerated around the sharper leading edge of NACA0002.
This leads to larger velocity differences across the shear layer and the wing surface, as observed in the experiments [17]. In the case of
stronger K-H instability around NACA0002 than those around the flowfields for which the flow reattaches after the separation, the
NACA0012. Flowfields around NACA0012 at α 3 deg have a turbulent vortex position seems to be lower than those for which the
separation point at the decelerating region. This leads to smaller flow does not reattach after the separation. Therefore, the flow seems
velocity differences across the shear layer, and the instability is to reattach in the former case due to the lower turbulent vortex
further suppressed, as compared to the case with leading-edge position. The distance from the shear layer to the wing surface or the
separation. 3D turbulent strength may relate to reattachment. This point should
In addition, the flowfields observed in Figs. 7–10 show that the be further investigated in the future research using both computation
appearance of how 2D vortices break up are approximately the same. and experiments.
The breakup of the 2D vortices is caused by the interaction of hairpin
vortices and 2D vortices, where the hairpin vortices are generated by B. Averaged Results
the breakup of the previously coherent vortex. The breakup of these In this section, the averaged flowfields of NACA0002 and
vortices resembles a chain reaction. NACA0012 are discussed using the results of the 3D iLES and the 2D
Finally, the flowfields for which the flow reattaches after the laminar simulation where the results of the 2D laminar simulation are
separation (NACA0012 at α 6 deg and NACA0002 at α 3 and used as supplements. Temporal and spanwise averaging are carried
KOJIMA ET AL. 191
x/c
α = 3.0
Separation
a)
α = 3.0
a)
α = 6.0
b)
Downloaded by ROKETSAN MISSLES INC. on February 7, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C031849
α = 9.0
α = 6.0
c)
b)
out for the results of the 3D iLES, whereas only temporal averaging is
performed for the results of the 2D laminar simulation.
1. Averaged Flowfields
NACA0012 Figure 11a shows that the flow around NACA0012 at
α 3 deg separates at 40% chord length, which is just downstream
of the maximum wing-thickness position (30% chord length). A low-
speed dead-water region is formed inside the separated shear layer. In α = 9.0
addition, the flowfields around NACA0012 at α 6 deg in Fig. 11b c)
show that the flow separates at 10% chord length, which is upstream
of the position of the maximum wing thickness, and reattaches to the
wing surface at 70% chord length, although the 3D vortices are
convected near the surface and clear reattachment is not observed in Fig. 11 x-direction velocity distributions and streamlines of time-
the instantaneous flowfields in Fig. 7b. Moreover, Fig. 11b shows a averaged flows around NACA0012 airfoil conducted by 3D iLES.
recirculation region surrounded by a separated shear layer and
reattached flow. This is the so-called separation bubble. Figure 11c
shows the averaged flow around NACA0012 at α 9 deg. At this simulation, and the experimental results [19]. Here the 2D laminar
angle, the flow separates around the leading edge and does not cases and the experimental results are plotted to support the 3D iLES
reattach. A large recirculation region is observed over the suction results for the separation and reattachment positions. Note that the 2D
surface. This flowfield is characterized as a leading-edge separation. laminar cases yield almost the same flowfields as the 3D iLES except
NACA0002 Figures 12a and 12b show that the flow around for the massively separated condition, although there is a slight
NACA0002 at α 3 and 6 deg exhibits separation and reattachment difference in the separation and reattachment positions. The separa-
similar to NACA0012 at α 6 deg, while the separation position of tion position in Fig. 13 shows that the trailing-edge separation is
NACA0002 is fixed at the leading edge. In these flowfields separation observed even at α 0 deg. The separation position moves linearly
bubbles are observed. Figure 13c shows that the flow around upstream with increasing angle of attack at α 1.5∼4.5 deg and
NACA0002 at α 9 deg does not reattach. Thus, it is characterized α 6∼9 deg, and the slope at α 1.5 to 4.5 deg is different from
as a leading-edge separation flowfield. The separation and re- that at α 6∼9 deg. At α 4.5∼6 deg, the separation position is at
attachment type of these computed flowfields are summarized in x∕c 0.3, which corresponds to the maximum wing-thickness
Table 2. Here the separation type is divided into leading-edge position, and the separation position moves suddenly upstream with
separation and trailing-edge separation. In this paper, leading-edge increasing angle of attack. This nonlinear behavior in the separation
separation denotes the separation upstream at the maximum wing- position is due to the change in separation characteristics at α
thickness position, whereas trailing-edge separation denotes the 4.5∼6 deg, whereas the separation position at α 1.5∼4.5 deg is
separation downstream at the maximum wing-thickness position. located downstream of the maximum wing-thickness position, and
that at α 4.5∼6 deg is located upstream of the maximum wing-
2. Separation and Reattachment Characteristics thickness position.
NACA0012 Figure 13 shows the separation and reattachment The comparison between the separation position of the 3D iLES
positions of NACA0012 computed by the 3D iLES and 2D laminar and that of the experimental data shows that the separation position of
192 KOJIMA ET AL.
Separated flow
α = 3.0
a)
Attached flow
Downloaded by ROKETSAN MISSLES INC. on February 7, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C031849
Table 2 Forms of separation, and presence or absence of the large airfoil thickness at a low angle of attack. Figure 13a also
of reattachment shows that the acceleration on the pressure surface near the leading
Airfoil NACA0012 edge is strong in this condition. The other is a trailing-edge
separation. The flow on the suction surface is separated around
Angle of attack (deg) 3 6 9
Separation type Trailing edge Leading edge Leading edge x∕c ∼ 0.4, and the flow from the pressure surface side goes to the
Reattachment No Yes No suction surface side around the trailing edge, leading to a negative Cp
Airfoil NACA0002 value. Due to the two factors previously discussed, the accelerated
Angle of attack (deg) 3 6 9 flow on the pressure surface does not decelerate well. As a result,
Separation type Leading edge Leading edge Leading edge attached flow with a negative Cp value is realized on the entire
Reattachment Yes Yes No pressure surface, where Cp on the suction surface slightly increases
near the trailing edge region because of trailing edge separation. This
is the reason of an inversion of pressure.
and 16c), where the flow does not reattach. With regard to The characteristics of the Cp distributions at α 6 deg in Fig. 17
NACA0002 at α 3 deg, reattachment is placed around the are different from those at α 3 deg. A strong negative pressure
position where the TKE is the highest. This is because the separated peak is observed near the leading edge. The flat pressure region is
shear layer is near the wing surface because of its low angle of attack, observed at x∕c 0.1 ∼ 0.5 deg, followed by the secondary
and the 3D structure of the vortex is not well formed before negative pressure peak at x∕c 0.55 and pressure recovery at
reattachment. The smaller TKE compared to the other cases supports the trailing edge. Compared with the averaged flowfield shown in
Downloaded by ROKETSAN MISSLES INC. on February 7, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C031849
this explanation, because the immediate reattachment prevent the Fig. 11b, the separation point corresponds to one at which the flat
growth of turbulent. Therefore, except for the thin airfoil with low pressure region starts, and the maximum TKE point corresponds to
angle of attack, the distance between the position where the 3D the secondary peak, which is at the end of the flat pressure region. The
structure is formed and the wing surface seems to be an important experimental data qualitatively agree with the computational data,
factor in reattachment. This is strongly related to whether the 3D
vortex reaches the wing surface.
C. Aerodynamic Coefficients
In this section, the aerodynamic coefficients are discussed on the
basis of the already analyzed flow characteristics.
1. Cp Distributions
The pressure coefficient (Cp ) distributions are discussed for Separation
clarifying the effects of the separation bubble on the airfoil surfaces.
Figures 17 and 18 show the Cp distributions on NACA0012 and
NACA0002 airfoils, respectively. Figure 17 includes the experimen-
tal data of the Cp distribution on NACA0012 airfoil at α 6 deg
[19]. Each Cp distribution is discussed separately below, whereas the a) α = 3 .0
Cp distribution at α 9 deg is not discussed in detail because it
massively separates. Here, the aspect ratio and Reynolds number of
experimental data compared are 2.78 and 2.3 × 104 .
NACA0012 The Cp distributions at α 3 deg in Fig. 17 show
that the Cp of the suction surface is smaller than the pressure surface
from x∕c ∼ 0.4 to the trailing edge. This inverse Cp distribution in the
wide region is due to the following two factors. One is the large
thickness of the airfoil. The Cp value on the pressure surface near the
leading edge is negative at α 3 deg, as shown in Fig. 17. This is Separation Reattachment
because the flow is well accelerated on the pressure surface because
α = 6 .0
b)
Separated flow
Reattached flow
Separation
Attached flow α = 9 .0
c)
Fig. 14 Separation and reattachment points of time-averaged flows Fig. 15 Turbulent kinetic energy distributions of flows around
around NACA0002 airfoil. NACA0012 airfoil.
194 KOJIMA ET AL.
Separation Reattachment
α = 3.0
a)
2. CL − α
Next, CL − α is discussed on the basis of the flow and Cp
characteristics. The value of CL computed from the averaged results
of the 3D iLES and 2D laminar simulation are shown in Fig. 19.
CL − α of NACA0012 The CL − α of NACA0012, as shown in
α = 9.0 Fig. 19, is significantly different from the potential theory (2πα)
c) atα 1.5 ∼ 4.5 deg, and a strong nonlinearity appears, as reported
in the experimental study by Mueller and Batill [20]. However, the
value of CL increases suddenly at α 4.5 ∼ 6.0 deg and approaches
the theoretical value.
Fig. 16 Turbulent kinetic energy distributions of flows around The nonlinearity at α 1.5 ∼ 4.5 deg is caused by trailing-edge
NACA0002 airfoil. separation and the separation point moves toward the leading edge
with increasing angle of attack, as discussed when the separation and
reattachment points were considered. There are two opposite effects
Fig. 17 Cp distribution of averaged flows by 3D iLES around Fig. 19 CL − α of averaged flows by 3D iLES and 2D laminar
NACA0012 airfoil. simulation.
KOJIMA ET AL. 195
bubble can be characterized as a short bubble. At α 7.5 ∼9 deg, Directorate, United States Army Missile Command, Redstone Arsenal,
the flow completely separates. Alabama, Vol. 35809, 1967.
The aerodynamic characteristics are presented as follows. At [2] Schmitz, F. W., “The Aerodynamics of Small Reynolds Numbers,”
α 0 ∼4.5 deg, pressure does not recover at the trailing edge on the NASA Technical Memorandum, 1980.
[3] Lissaman, P. B. S., “Low-Reynolds-number Airfoils,” Annual Review of
suction side because of trailing-edge separation. In addition, the Fluid Mechanics, Vol. 15, 1983, pp. 223–239.
pressure-side flow is accelerated and the pressure at the pressure side doi:10.1146/annurev.fl.15.010183.001255
becomes very low. This effect and the movement of the separation [4] Shan, H., Jiang, L., and Liu, C., “Direct Numerical Simulation of Flow
point lead to the nonlinearity in CL − α. In addition, CD becomes Separation around a NACA 0012 Airfoil,” Computers and Fluids,
smaller because of trailing-edge separation. At α 4.5 ∼6 deg, CL Vol. 34, No. 9, 2005, pp. 1096–1114.
suddenly increases because of the suction side pressure recovery by doi:10.1016/j.compfluid.2004.09.003
reattachment. This is because the flowfield changes from trailing- [5] Galbraith, M. C., and Visbal, M. R., “Implicit Large Eddy Simulation of
edge separation to leading-edge separation and reattachment. The Low-Reynolds-Number Transitional Flow Past the SD7003 Airfoil,”
trend of an increasing CD is also changed. At α 6 ∼7.5 deg, AIAA Paper 2010-4737, June 2010.
[6] Abdo, M., and Mateescu, D., “Low-Reynolds Number Aerodynamics of
although the separation bubble becomes shorter, CL and CD become Airfoils at Incidence,” AIAA Paper 2005-1038, 2005.
larger. At α 7.5 ∼9 deg, CD becomes larger because of the [7] Oyama, A., Fujii, K., Shimoyama, K., and Liou, M.-S., “Pareto-
completely separated flows. With regard to L∕D, the L∕Dmax is Optimality-Based Constraint-Handling Technique and Its application to
observed approximately at α 6 deg, where the flowfield changes Compressor Design,” AIAA Paper 2005-4983, June 2005.
from a flow with trailing-edge separation to a flow with the separation [8] Oyama, A., and Fujii, K., “A Study on Airfoil Design for Future Mars
bubble. In addition, L∕D is affected by the nonlinearity of CL − α. Airplane,” AIAA Paper 2006-1484, Jan. 2006.
Downloaded by ROKETSAN MISSLES INC. on February 7, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C031849
The flow characteristics of NACA0002 are presented as follows. [9] Fujii, K., and Obayashi, S., “High-Resolution Upwind Scheme for
At α 0 ∼ 1.5 deg, the flow attaches. At α 1.5 ∼ 3 deg, the flow Vortical-Flow Simulations,” Journal of Aircraft, Vol. 26, No. 12, 1989,
separates and reattaches, resulting in a separation bubble. At pp. 1123–1129.
doi:10.2514/3.45890
α 3 ∼ 7.5 deg, the reattachment point moves downstream and the [10] Baldwin, B., and Lomax, H., “Thin Layer Approximation and
separation bubble becomes longer with increasing angle of attack. Algebraic Model for Separated Turbulent Flows,” AIAA Paper 1978-
This separation bubble can be characterized as a long bubble. At 257, Jan. 1978.
α 7.5 ∼ 9 deg, the separated flow does not reattach. [11] Lele, S. K., “Compact Finite Difference Schemes with Spectral-like
The aerodynamic characteristics of NACA0002 are presented as Resolution,” Journal of Computational Physics, Vol. 103, No. 1, 1992,
follows. At α 0 ∼ 7.0 deg, CL − α is linear, although the flowfield pp. 16–42.
has changed from the attached flow to a flow with a long bubble. doi:10.1016/0021-9991(92)90324-R
CD − α is proportional to the square of α. At α 7.5 ∼ 9 deg, CD [12] Gaitonde, D. V., and Visbal, M. R., “Padé -Type Higher-Order Boundary
becomes much higher because of the completely separated flows. Filters for the Navier-Stokes Equations,” AIAA Journal, Vol. 38, No. 11,
2000, pp. 2103–2112.
With regard to L∕D, L∕Dmax is observed approximately at doi:10.2514/2.872
α 3 deg, where the flowfield changes from the attached flow to a [13] Nishida, H., and Nonomura, T., “ADI-SGS Scheme on Ideal
flow with a long bubble. Magnetohydrodynamics,” Journal of Computational Physics, Vol. 228,
For the low Reynolds number conditions examined in this study, No. 9, 2009, pp. 3182–3188.
L∕D of NACA0012 is strongly affected by the nonlinearity of CL , doi:10.1016/j.jcp.2009.01.032
and L∕Dmax is observed at the unstable angle of attack, where the [14] Boris, J. P., Grinstein, F. F., Oran, E., and Kolbe, R. J., “New Insights
flowfields changes from flows with trailing-edge separation to those Into Large Eddy Simulation,” Fluid Dynamics Research, Vol. 10,
with separation bubble. In this condition, thick airfoils such as Nos. 4–6, 1992, pp. 199–228.
doi:10.1016/0169-5983(92)90023-P
NACA0012 are not appropriate because the aerodynamic character-
[15] Shima, E., and Jounouchi, T., “Role of CFD in Aeronautical
istics possibly change drastically because of the change in flowfields. Engineering (No.14) -AUSM type Upwind Schemes-,” Proceedings of
On the other hand, thin airfoils such as NACA0002 are appropriate the 14th NAL Symposium on Aircraft Computational Aerodynamics,
because there is no sudden change in the aerodynamics character- National Aerospace Laboratory, Japan, 1997, pp. 7–12.
istics because of the fixed separation point owing to the sharp [16] Van Leer, B., “Towards the Ultimate Conservation Difference Scheme.
leading edge. V. A Second-Order Sequel to Godunov’s Method,” Journal of
Computational Physics, Vol. 32, No. 1, 1979, pp. 101–136.
doi:10.1016/0021-9991(79)90145-1
Acknowledgments [17] Gerakopulos, R., Boutilier, M., and Yarusevych, S., “Aerodynamic
The present research was supported by a Grant-in-Aid for Characterization of a NACA 0018 Airfoil at Low Reynolds Numbers,”
Scientific Research (20246122, 24246141, and 24246136), Strategic AIAA Paper 2010-4629, 2010.
Programs for Innovative Research of High Performance Computing [18] Anyoji, M., Nose, K., IDa, S., Numata, D., Nagai, H., and Asai, K.,
“Low Reynolds Number Airfoil Testing in a Mars Wind Tunnel,” AIAA
Initiative, and Mars Exploration Aircraft Working Group in Japan.
Paper 2010-4627, 2010.
The authors are grateful to the flutiful discussion with the members of [19] Kim, D. H., Yang, J. H., Chang, J. W., and Chung, J., “Boundary Layer
Mars Exploration Aircraft Working Group in Japan. and Near-wake Measurements of NACA0012 Airfoil at Low Reynolds
Numbers,” AIAA Paper 2009-1472, 2009.
References [20] Mueller, T., and Batill, S., “Experimental Studies of the :Laminar
Separation Bubble on a Two-dimensional Airfoil at Low Reynolds
[1] Schmitz, F., “Aerodynamics of the model airplane. Part 1. Airfoil Numbers,” AIAA Journal, Vol. 20, No. 4, 1980, pp. 457–463.
measurements. Translated from German by Translation Branch, doi:10.2514/3.51095
Redstone Scientific Information Center,” Research and Development