LESahmedcarrodi PDF
LESahmedcarrodi PDF
LESahmedcarrodi PDF
Ahmed body
C. Hinterberger, M. Garca-Villalba, and W. Rodi
Institute for Hydromechanics, University of Karlsruhe, Kaiserstrasse 12,
76128 Karlsruhe, Germany
1 Introduction
The automotive industry has a high demand for reliable simulation methods capable of tackling the complex turbulent air flow around vehicles. The
Ahmed reference model is a generic car-type bluff body with a slant back. It
is frequently used as a benchmark test case for this kind of flow. In spite of
the relatively simple geometry of the Ahmed body, the flow around it retains
some main features of the flow around real cars.
The Ahmed body, Fig. 1, was first defined and its characteristics described
in the experimental work of Ahmed et al [1]. Therein, it is shown that most
of the drag of the body is due to pressure drag, which is generated at the rear
end. The structure of the wake is very complex, with a separation zone and
counter-rotating vortices coming off the slant side edges, whose strength is
mainly determined by the base slant angle. The maximum drag was found for
a critical slant angle of 30. Above this angle a sudden drop in drag occurs
which corresponds to a change in the wake structure. Below this angle, strong
counter-rotating vortices are present and the flow separates in the middle
region of the top edge and reattaches at the sloping surface. For angles above
the critical angle, the counter-rotating vortices are weaker, the separation
occurs along the entire top and the side edges and there is no reattachment.
More recently, Lienhart et al [2] performed more detailed experiments on
the same body, albeit at a somewhat lower velocity. They measured the mean
and fluctuating velocities by LDA and obtained surface oil-flow pictures for
two rear vehicle body slant angles ( = 25 and = 35), i.e. just below and
above the critical angle. Their results show the differences in flow behaviour
for the two angles considered.
In recent years, there have been several computational studies of this flow.
Han [3] and Gillieron and Chometon [4] using the RANS approach obtained
qualitatively good results in terms of flow structures but they do not show
velocity profile comparisons. In two recent workshops [5, 6] the flow around
the Ahmed body was a test case and the results presented there have shown
and streamwise direction, an outer O-grid structure was chosen in Grid2, Fig
2(b). In addition, Grid 2 includes more refinement in the region of the slant
back, especially close to the top and side edges, Fig 2(c-h). For both grids,
the near-wall cell centre has a wall distance on average of about 40 wall units
(yP + ), but it varies from approximately 10 in the separated regions along
the slant back to 150 close to the top rear edge. The spanwise and streamwise
extent of the grid cells is up to a factor of 10 larger for Grid 2, except in the
refinement regions close to the edges. For Grid 1 this factor is even larger. This
means that the boundary layer at the body surface is highly underresolved.
Fig. 2. Typical grids used for the LES. (a,c,e,g) Grid 1. (b,d,f,h) Grid 2. (a,b)
Sketches of the block structure. (c-h) Slices in different planes showing the grid
point distribution in the body region. (c,d) Cut in xz-plane . (e,f ) Cut in xy-plane.
(g,h) Cut in yz-plane.
3 Results
3.1 Time averaged profiles
For the front part, mean streamwise velocity profiles in the symmetry plane
are compared with the experiment in Fig. 3. It can be seen that the flow
upstream of the body and in the free stream above it is properly represented
in the simulation. This is because in this region the level of turbulence is so
low that the flow is nearly potential flow.
In Figures 4 and 5, the mean streamwise velocity profiles and root mean
square velocity fluctuations are compared with the experimental results in
the rear body part and the near wake (in the symmetry plane). The general
agreement with the experiment is reasonably good taking into account that
neither grid is fine enough to resolve adequately the boundary layer developing on the body up to the slant back. However there are some discrepancies
between the computations and the experiment mainly concerning the velocity
profiles on the slant back. In the experiment, the flow separates right at the
corner of the sloping surface and it reattaches in the middle of the surface.
In the simulation, the flow first stays attached before it separates somewhat
downstream of the corner and no reattachment occurs on the slant back. This
is most likely due to the poor resolution of both grids near the wall leading
to an incorrect prediction of the approaching boundary layer.
The main differences between the results of the two simulations are found
at the beginning of the sloping surface. There, Grid 2 is much finer than Grid
1, and it can be seen that the turbulence intensities are very close to the
Fig. 4. Mean streamwise velocity profiles in the rear body and near wake (symmetry
plane)
experimental ones for Grid 2 while they are too low for Grid 1, Fig. 5. In
addition, the prediction of the separation point is closer to the experiment
in the simulation performed with the finer grid, in which the flow separates
earlier, Fig. 4.
The complex unsteady flow on the slant back leads to high fluctuation
intensities which are very difficult to predict. In fact, no RANS method has
succeeded in predicting the high values observed in the experiments [5, 6]
while the present LES calculations (Fig. 5) have.
Figure 6 shows the mean velocity vectors in the symmetry plane, Fig.
6(a-b), and in three transverse yz-planes, Fig. 6(c-h). The streamwise location of these transverse planes is indicated in Fig. 6(a) by dashed lines. The
colour of the vectors corresponds to turbulent kinetic energy. From the comparison of the experiment, Figs. 6(a,c,e,g) on the left, with the simulation,
Figs. 6(b,d,f,h) on the right, it is clear that the main flow structures are well
captured in the simulation. The size and the extent of the recirculation zone
behind the body are well predicted, Fig. 6(b). The development of the counterrotating vortices, which can be seen from the secondary flow vectors in the
Fig. 5. Root mean square streamwise velocity fluctuations in the rear body and
near wake (symmetry plane)
y z planes, is also in close agreement with the experiment. In Fig. 6(d), it can
be seen that these vortices develope half-way down the slant back, they grow
while they approach the end of the body, Fig. 6(f ), and they are strong and
fully developed in the near wake, Fig. 6(h). The level of turbulence obtained
in the simulation is also in good agreement with the experiment, as can be
seen from the coloured regions in Figure 6.
3.2 Flow visualisation and flow structures
The calculation results show complex time-dependent flow features in the wake
region. In Fig. 7, two typical instantaneous velocity fields in the symmetry
plane are shown. As discussed in section 3.1, the prediction of the flow on the
sloping surface is very difficult. Thus, in Fig. 7 on the left, an instantaneous
field is captured in which the flow tends to reattach on the slant back. However,
in Fig. 7 on the right, the flow in that zone is completely separated. Note that
in the experiment the flow reattaches (at least in the mean). The unsteadiness
of the near body-wake is clearly well captured. The flow between the body and
Fig. 6. Mean velocity vectors coloured by turbulent kinetic energy. (a,b) symmetry
plane y = 0 mm. (c,d) Close to the middle of the slant back x = 88 mm. (e,f ) End
of the body x = 0 mm. (g,h) Near-body wake x = 200 mm. (a,c,e,g) Experiment.
(b,d,f,h) Simulation performed with Grid 2.
the ground plate has a rather strong influence on the shape of the recirculation
zone.
4 Conclusions
A Large Eddy Simulation of the flow around the Ahmed body with a slantback angle of 25was performed. The flow around the Ahmed body is a very
10
11
challenging problem because of the complex geometry and the high Reynolds
number. The results obtained by the simulation are promising; the comparison
with the experiments shows that the flow structures are well captured. The
agreement of the time-averaged quantities is good although some discrepancies
are present, especially in the lower part of the slant back. The results also show
that this is a good test case for further development. The calculations for the
configuration with a slant angle of 35are in progress. Improvements of the
method (subgrid-scale model, discretisation, wall modelling, etc) are required
to obtain good results for the 35case.
Acknowledgments
This work was funded through the EU TMR-project LES of Complex Flows.
The calculations were carried out on the IBM RS/6000 SP-SMP high performance computer of the University of Karlsruhe, and the assistance of Mr.
Gernert is gratefully acknowledged. The authors are also grateful to Dr. J.
Frohlich for many helpful discussions.
References
1. S.R. Ahmed, G. Ramm, and G. Faltin. Some salient features of the time averaged
ground vehicle wake. SAE paper no. 840300, 1984.
2. H. Lienhart, C. Stoots, and S. Becker. Flow and turbulence structures in the
wake of a simplified car model (Ahmed model). In DGLR Fach Symp. der AG
STAB, 2000.
3. T. Han. Computational analysis of three-dimensional turbulent flow around a
bluff body in ground proximity. AIAA Journal, 27:12131219, 1989.
4. P. Gillieron and F. Chometon. Modelling of stationary three-dimensional separated air flows around an Ahmed reference model. In ESAIM proc., volume 7,
pages 173182, 1999.
5. S. Jakirlic, R. Jester-Z
urker, and C. Tropea, editors. Proceedings of 9th ERCOFTAC IAHR COST Workshop on refined turbulence modelling, 2001.
6. R. Manceau and J.P. Bonnet, editors. Proceedings of 10th ERCOFTAC IAHR
QNET-CFD Workshop on refined turbulence modelling, 2002.
7. S. Kapadia, S. Roy, and K. Wurtzler. Detached eddy simulation over a reference
Ahmed car model. AIAA paper no. 2003-0857, 2003.
8. R.J.A. Howard and M. Pourquie. Large eddy simulation of an Ahmed reference
model. Journal of Turbulence, 3, 2002.
9. M. Breuer and W. Rodi. Large eddy simulation of complex turbulent flows of
practical interest. In E.H. Hirschel, editor, Flow simulation with high performance computers II, volume 52 of Notes on Numerical Fluid Mechanics, pages
258274. Vieweg, Braunschweig, 1996.
10. C.M. Rhie and W.L. Chow. Numerical study of the turbulent flow past an airfoil
with trailing edge separation. AIAA Journal, 21(11):10611068, 1983.
11. H. Werner and H. Wengle. Large-eddy simulation of turbulent flow over and
around a cube in a plate channel. In 8th Symp. on Turb. Shear Flows, 1993.