Special Issue Article

Advances in Mechanical Engineering

2017, Vol. 9(7) 1–17
Ó The Author(s) 2017
Aerodynamic drag reduction and flow DOI: 10.1177/1687814017711390
journals.sagepub.com/home/ade
control of Ahmed body with flaps

Jie Tian, Yingchao Zhang, Hui Zhu and Hongwei Xiao

Abstract
In this study, we attempted a novel drag reduction technique for 25° and 35° Ahmed models by experimenting with two
types of flap structures, respectively, added to the slant edges of the two models. Different pairs of flaps were added at
various angles compared to the slant for the sake of comparison. The study comprehensively analyzed the effects of the
‘‘big-type’’ and ‘‘small-type’’ flaps on the aerodynamic drag and near wake of an Ahmed model in a greater range of flap
mounting angles. Parametric analysis results confirmed that large and small flaps are most efficient when configured on
the 25° Ahmed model at specific angles; up to 21% pressure coefficient reduction was achieved for the 25° Ahmed
model (flap configurations at slant side edge) and 6% for the 35° Ahmed model (flap configurations at both slant side and
top edges). The velocity and pressure contours indicated that the key to drag reduction is to weaken (if not eliminate)
the longitudinal vortex created at the side edges of the rear slant.

Keywords
Ahmed model, drag reduction, flow structure, flap structures, pressure coefficient

Date received: 7 September 2016; accepted: 27 April 2017

Academic Editor: Pietro Scandura

Introduction flow cannot separate and aerodynamic forces depend

instead largely on the flow structure created on the
Reducing the global consumption of vehicle fuel is a rear. The wake of the Ahmed model is a series of fully
crucial and urgent necessity if we are to respond swiftly 3D complex flows which are directly linked to the
and appropriately to climate change. Research on drag coherent structures in the vehicle as reported by Vino
reduction, which enhances fuel efficiency, thus repre- and Watkins3 and Beaudoin and Aider.4 The flow
sents a very significant environmental concern. To structures contain three major components: The recir-
reduce aerodynamic drag effectively, we surely need a culation bubble over the rear slant, the longitudinal
comprehensive understanding of the flow structure vortices created on the side edges or C-pillars of the
around the vehicle. The air drag surrounding an auto- slant, and the recirculation torus on the base of the
mobile, for example, is mainly constituted of pressure model.
drag and friction drag. For a basic bluff vehicle type of There is an interesting relationship (or competition)
body (i.e. the Ahmed model), up to 85% of the total between these three types of flow structures along with
drag is pressure drag and the remainder is friction
drag.1 Ahmed et al.1 also found that the rear end con-
tributes as much as 91% to the total pressure drag and State Key Laboratory of Automotive Simulation and Control, Jilin
is especially predominant at high speeds, as confirmed University, Changchun, China
by Hucho and Sovran2 in 1993.
Corresponding author:
There have been numerous previous studies on the Yingchao Zhang, State Key Laboratory of Automotive Simulation and
Ahmed body, which is a simplified car model of 3D Control, Jilin University, Changchun 130022, China.
bluff shape; its front is designed to be blunt so that the Email: yingchao@jlu.edu.cn

Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering

the slant angle of the upper rear surface of the model. aerodynamic drag by employing flow control tech-
Previous studies on the wake of the Ahmed model,1,5–7 niques. But their research variables were relatively sin-
via both experimentation and numerical simulation, gle and the experiments were conducted at the critical
have observed the critical angle to be around 30°—after flap mounting angle between a = 0° and a = 80°, the
this point, there is an abrupt decrease in drag along conclusions may not be applicable for other flap
with a full separation of the rear flow after increasing mounting angles. We comprehensively analyzed the
sharply to the maximum drag value.2 Wang’s previ- effects of the ‘‘big-type’’ and ‘‘small-type’’ flaps on the
ously published article8 provides a helpful schematic aerodynamic drag and near wake of an Ahmed model
diagram of the coherent structures for the 25° and 35° in a greater range of flap mounting angles. The primary
Ahmed models. With a slant angle of 25°, the flow goal of this study was to successfully utilize specific
separates at the lower edge of the slant and the two actuators to modify the flow structures of 25° and 35°
counter-rotating longitudinal vortices are dominant Ahmed models so as to reduce drag and select the opti-
throughout the entire flow structure; at the slant angle mal solution in different situations.
of 35°, conversely, it separates at the upper edge of the
slant along with a burst in the longitudinal vortices.9,10 Simulation details
Several previous researchers have explored flow con-
trol techniques both experimentally and by simulation 3D bluff body and actuators
over a 3D bluff body. Mixed results, some quite favor- The Ahmed model used here, as it was first described
able, have been achieved by applying active or passive in 1984, is the equivalent of one quarter the size of a
techniques. The active steady-state flow method, as first real automobile (Figure 1). To improve the simulation
introduced by Aubrun et al.11 in their research on wake efficiency, we used the half model in our steady formu-
flow reduction and control, can be used to cut the drag lation experiment. Like the rear window of a car, the
coefficient by 9%–12% and the lift by 42%. Khalighi slant surface takes an angle relative to the horizon.
et al.12 carried out wind tunnel tests and computational Although many of the classical steady-state Reynolds-
investigation on the transient wake flow structures of a averaged Navier–Stokes (RANS) approaches have some
bluff body equipped with a plate-shape drag reduction difficulties in capturing the main flow features, especially
device, they succeeded in modifying the flow field slightly below the critical slant angle, some other works,
behind the test model by disturbing the shear layer. As such as the one from Craft et al.,19 showed also that with
a consequence, the closure of the wake is altered and a proper combination of theoretical modeling and
reductions in aerodynamic drag of more than 20% are numerical setup, the steady-state RANS turbulence clo-
observed. Howell et al.13 conducted applied dents and sure strategies can lead to surprisingly good quantitative
through-holes to the zero-degree Ahmed model for predictions of the aerodynamic force coefficients. And
drag reduction. Beaudoin and Aider4 from Peugeot- our previous work has compared the Cd values obtained
Citroen mounted tail fins on all edges of both surfaces by our simulation with the experimental results of
at the tail of the 30° Ahmed model to secure a drag Ahmed et al.1 (Figure 2), the simulation results are closed
coefficient reduction of 25%. Castejon et al.14 achieved to the experimental results, and the error is very small.
automobile drag reduction for the SAE model (pro- Additionally, because of the limitation of time and com-
posed by Society of Automotive Engineers) without puting hardware resources, the steady RANS is used to
reducing the automobile aerodynamic stability. The simulate the time-averaged flow field in this study, so the
drag reduction of a pickup truck by a rear downward velocity got in the simulation is time-averaged velocity
flap was examined computationally and experimentally not instantaneous velocity. And as mentioned above, we
in the study of Ha et al.15 Fourrie et al.16 carried out examined the aerodynamic forces of 25°and 35° slant
experimental investigations to study the drag reduction angles with various flap configurations.
in a generic car model using a deflector. They achieved There are two types of flap shape used here: large (also
a drag reduction of up to 9%. Altaf et al.17 studied the called ‘‘big’’ in this article), which was 1 mm in thickness
shape, sizing, and design of flaps for drag reduction of and 20 mm in height, and small which was 1 mm in thick-
bluff bodies in heavy vehicle aerodynamics. Hanfeng ness and 8 mm in height; both were 4 mm in width with
et al.18 experimentally investigated the effects of deflec- an interval of 12 mm (Figure 3). The flap structures were
tors on the aerodynamic drag and near wake of an configured at the four edges of the slant as ‘‘top-type,’’
Ahmed model with a 25° slant angle. ‘‘side-type,’’ or ‘‘bottom-type’’ (Figure 4), a = 0° corre-
Insightful research into the flow structures and the sponds to the flaps parallel to the slant surface.
mechanism of controlling aerodynamic drag has shown
that the key to reducing drag is the control of the flow
separation and large coherent structures. The previous Parameters
studies improved the near wake of an Ahmed model Numerical simulation can achieve useful and accurate
with a 25° and 35° slant angle and then reduced the results with the notable advancements in computational
Tian et al. 3

Figure 1. Dimensions of the Ahmed vehicle model (length (mm); angle (°)).

Figure 2. Comparison between numerical and experimental

results.

Figure 3. Detailed characteristics of the two flap shapes: (a)

big and (b) small.
4 Advances in Mechanical Engineering

Figure 4. Configurations of flap structures at the four edges of slant surfaces: (a) top-type, (b) side-type, and (c) bottom-type.

Figure 5. Layout of model and mesh space.

fluid dynamics (CFD). Simulation facilitates the rapid

Table 1. The results of the grid independency study.
development of new devices while saving cost, but does
come with some amount of uncertainty in terms of accu- Mesh No. of cells Cd
racy. To improve the accuracy of our simulations which
we ran in STAR-CCM + software (CD-adapco), we Extra course 2,131,820 0.318
Course 3,357,874 0.302
researched the turbulent model, mesh style, and mesh
Medium 4,517,490 0.292
size on three levels: whole aerodynamic drag coeffi- Fine 5,089,280 0.289
cients, partial coefficients, and known statistics of the
flow field put forward by Makowski and Kim.20 To
determine the steady flow structures which can effi- To ensure a favorable simulation of the boundary
ciently reflect the main vortices,21 we utilized primarily layer flow near the model, we employed a prism layer
tetrahedral unstructured mesh of 5 mm in size in the mesh with five layers to control the y + value between
SST k–v turbulence model. 20 and 100 (Figure 6). We have done the grid indepen-
Using the 35° Ahmed model tested in 19841 as our dence research in the previous work to make sure the
standard case, we set the simulation parameters as simi- simulation results are irrelevant to the mesh model.
lar as possible to the following: The results of the grid independency study were shown
in Table 1. There is a maximum variation of only 9%
 Mesh domain: 15 3 3 3 1.5 m3; between the coarsest and finest meshing used, so the
 Velocity inlet: 60 m/s; results are assumed to be largely grid independent. So
 Pressure outlet: P = 0; we followed the previous mature simulation method in
 Reynolds number Re = 4.29 3 106, character this project. The total number of cells of our half model
length = 1.044 m; is about 5 million, which ensures simulation accuracy.
 Turbulence intensity: \0.5%.
Parametric study of various flap
The model is four times its own length away from configurations
the inlet and eight times its own length from the outlet,
which is sufficient to guarantee adequate air flow Simulation validation
growth around the vehicle. The distance (50 mm) Aerodynamic drag is mainly pressure drag, so we paid
between the model and the ground is the same as that only cursory attention to the friction coefficient and
in a standard wind tunnel test.22 Figure 5 shows the instead considered the pressure drag coefficient as our
layout of the model and mesh which involves two primary research object. In this article, Cp* is total pres-
blocks of grid refinement. sure drag coefficient, Ck* is the forebody pressure drag
Tian et al. 5

Figure 6. (a) Prism layer mesh around the model and (b) y + value contour.

Figure 7. Data comparison: x = 20.163 m; x = 20.123 m; x = 20.083 m; x = 20.043 m; x = 0.037 m at symmetry plane. Origins of
coordinate system: x = 0 (end of the car); y = 0 (symmetry plane); z = 0 (ground plane).

model. The experiment database was accessed online

Table 2. Comparison between the experiment data and through the ERCOFTAC homepage. We evaluated the
simulated 35° Ahmed model.
accuracy of our simulations with experimental velocity
Case Cp* Ck* Cs* Cb* at five lines including the date both at slant and the base
in the symmetry plane (Figure 7) and used that as the
Experiment 0.2013 0.0138 0.0983 0.0892 verification standard. As shown in Figure 8, the x com-
CFD 0.1985 0.0118 0.0979 0.0890 ponent velocity of the simulation matches closely with
CFD: computational fluid dynamics. the experimental velocity development at the z position.
But there is a deviation between CFD data and the
experiment, this may be because the flow separation at
coefficient, Cb* is the vertical base pressure drag coeffi-
the slant surface is difficult to predict and the k–v tur-
cient, and Cs* is the slant surface pressure drag coeffi-
bulence model cannot simulate the turbulence structure
cient. To verify the simulation accuracy, we compared
very accurately.
published 35° Ahmed model experiment results against
From the comparison between the simulation results
our simulation.
and the experimental results reported in the previous lit-
Table 2 shows a comparison between the experimen-
erature1,5 above, we can confirm that the simulation is
tal data1 and results based on the converged simulation
credible and the results obtained without flow control
of the 35° Ahmed model. Errors between the CFD
are accurate.
results and wind tunnel experiment results were small,
so the errors meet the engineering application require-
ment of 5%.
Drag reduction of 25° Ahmed model
We also compared the velocity at the x direction
between the experiment5 and simulation. Much of the Top-type flap configuration. The configuration of flaps at
data came from a paper by Lienhart and Becker5 which the junction between the roof and the slant has a com-
describes the velocity distribution in the wake of the mon feature of the roof spoiler of a real automobile. In
6 Advances in Mechanical Engineering

x=-0.163,y=0 x=-0.123,y=0
390 390
380
380
370
370
Posion[Z] (mm)

Posion[Z] (mm)
360
360 350
340
350
330
340 320
310
330
CFD Exp 300
CFD Exp
320 290
-20 -10 0 10 20 30 40 50 -20 -10 0 10 20 30 40 50
Velocity[Ux] (m/s) Velocity[Ux] (m/s)

x=-0.083,y=0 x=-0.043,y=0
390 400
380
370
360
Posion[Z] (mm)
Posion[Z] (mm)

350
340
330 320
300
310
280
290 260
CFD Exp CFD Exp
270 240
-20 -10 0 10 20 30 40 50 -20 -10 0 10 20 30 40 50

Velocity[Ux] (m/s) Velocity[Ux] (m/s)

x=0.037,y=0
230

200

170
Posion[Z] (mm)

140

110

80
CFD Exp
50
-20 -15 -10 -5 0 5 10
Velocity[Ux] (m/s)

Figure 8. X component velocity comparison between CFD and experiment in the shear layer.
Tian et al. 7

Figure 9. CFD drag evolution results of top-type flaps with different flap angles for 25° Ahmed model.

Table 3. Comparison of standard 25° Ahmed model with Top_big_20 and Top_small_40.

Case Cp* Ck* Cs* Cb*

Standard 0.2463 0.0055 0.1427 0.0981

Top_big_20 0.2021 0.0093 0.0813 0.1115
Top_small_40 0.2025 0.0085 0.0956 0.0984

this study, we explored two types of structures in con- about 60°, which is consistent with the experimental
tinuous and discontinuous shapes, that is, ‘‘big’’ and results presented in Beaudoin and Aider.4 As a contin-
‘‘small’’ types (Figure 3). Varying the configuration ued to increase, pressure drag coefficient increased due
alters the flow direction from the roof, as well as trans- to the drag reduction effect of the flap configuration
forms separation bubbles on the slant which may con- was lower than the drag produced from the flap
sequently change the pressure drag coefficient. By itself. The pressure drag coefficient reached its maxi-
varying the angle of the flap to the slant surface mum value (before beginning to decrease) when
(a = 0° corresponds to the flap aligned with the slant a = 115°, that is, when the flap was vertical to the
wall), we conducted a parametric analysis of changes in roof.
pressure drag coefficient. As shown in Figure 9, the Similar reduction was observed with the small-type
pressure drag coefficient changes with flap angle as the flap configurations. There was a reduction in the pres-
configuration of big and small flaps changes. For sure drag coefficient when the angle increased from 0
the large flap type, we found that as angle a increased, to 40° and reached its minimum (244 counts) when
the pressure drag coefficient first decreased and then a = 40°. Between a = 20° and a = 30°, there were also
increased (as expected). Between a = 10° and a = 20°, some sharp drop in the curve. The change of the pres-
there were some sharp drop in the curve. This is sure at the slant surface is similar to the big-type flap.
because the pressure changed a lot at the slant surface, The pressure drag coefficient then increased slightly
we can see that from Figure 10. When a was about 20°, and remained a relatively stable level (and still under
approximately parallel to the roof, the pressure drag the original value) until the end of the simulation.
coefficient reached its minimum value—44 counts (1 Table 3 shows a comparison between the original
drag count = 1023) less than the baseline model. and modified models with the top configuration of a
Beaudoin and Aider4 achieved 54 counts drag reduc- big-type flap at 20° (Top_big_20) and small-type flaps
tion for Uo = 40 m/s in their experimental study when at 40° (Top_small_40). The slant surface was mostly
a was about 25°, approximately parallel to the roof responsible for the reduction (261 counts), while there
(the slant angle of their Ahmed model is 30°). When a was an increase in pressure drag coefficient at the verti-
exceeded 20°, the pressure drag coefficient increased cal base (+ 13 counts) in Top_big_20. There was a 47-
until reaching the original level of the Ahmed model at count reduction of the slant surface pressure drag
8 Advances in Mechanical Engineering

Figure 10. Pressure distribution on the slant surface: (a) top_big_10 and (b) top_big_20.

coefficient in Top_small_40 and almost no change in the drag coefficient can be much more Reynolds depen-
the vertical base pressure drag coefficient compared to dant,4 so the small errors between the simulation
the baseline. results and the experimental results are understandable.
Figure 12 shows that the pressure at the slant surface
was reduced surprisingly, so there was some sharp drop
Side-type flap configuration. It is well known that the side in the curve. The pressure drag coefficient in present
edges of a slant surface are the positions at which longi- work then continued to increase and remained lower
tudinal vortices (C-pillar vortex) are generated, that is, than the original model throughout the simulation. In
why we set the flap on these positions. The longitudinal effect, the big-type flap varied the pressure distribution
vortices on the side edges interacting with the separa- on the slant surface by damaging or even preventing
tion bubbles on the slant surface4 are the predominant entirely the formation of longitudinal vortices. Certain
source of drag of the 25° Ahmed model. By configuring flap angles caused greater extent of damage.
big- and small-type flaps on the side edges (a = 0° cor- For the small-type flaps, there was nearly no reduc-
responds to the flaps laying over the rear slant), we are tion effect. Conversely, the pressure drag coefficient
able to destroy the longitudinal vortices effectually. exceeded the original model. This can be attributed to
As shown in Figure 11, for the big-type flap, the the fact that the discontinuous flaps are unable to stop
pressure drag coefficient was basically invariant until the crimping process of the longitudinal vortices, that
a = 40° at which point it began to decrease. There was is, the small flaps were unable to damage the vortices.
a critical moment when the flap approached a vertical This observation further confirms that preventing the
position to the slant surface—specifically, the pressure crimping process of the longitudinal vortices results in
drag coefficient dropped to its minimum at a = 80° drag reduction.
with a 52-count reduction. This trend is consistent with Table 4 shows a comparison between the original
the experimental results in Beaudoin and Aider;4 and modified models with the side configuration of the
Beaudoin and Aider first observed a plateau for 0° \ a big-type flap at 80° (Side_big_80). There was an overall
\ 67° before a sharp drag reduction (nearly 62 counts reduction of 52 counts: The pressure drag coefficient of
for Uo = 40 m/s) for aopt ’ 70°. It is well known that the slant surface (Cs*) was reduced by 48 counts and
the drag value is strongly dependent of the slant angle,1 the pressure drag coefficient of the vertical base (Cb*)
and for a more rounded bluff body, or even a real car, by 7 counts.
Tian et al. 9

Figure 11. CFD drag evolution results of side-type flaps with different flap angles for 25° Ahmed model.

Figure 12. Pressure distribution on the slant surface: (a) Side_big_70 and (b) Side_big_80.

Table 4. Comparison of standard 25° Ahmed model with Side_big_80.

Case Cp* Ck* Cs* Cb*

Standard 0.2463 0.0055 0.1427 0.0981

Side_big_80 0.1942 0.0082 0.0943 0.0917
10 Advances in Mechanical Engineering

Figure 13. CFD drag evolution results of bottom-type flaps with different flap angles for 25° Ahmed model.

Table 5. Comparison of standard 25° Ahmed model with Bottom_big_130.

Case Cp* Ck* Cs* Cb*

Standard 0.2463 0.0055 0.1427 0.0981

Bottom_big_130 0.2187 0.0079 0.1071 0.1037

Bottom-type flap configuration. The flap configuration at a = 150°. So their effects to the pressure drag coeffi-
the junction between the rear slant and the vertical base cient are different.
is designed to modify the recirculation bubble on the For the small flaps configured on the bottom edge
slant surface and the torus on the vertical base of the of slant, there was no reduction in pressure drag coeffi-
model. Flaps placed here can stop the flow along cient regardless of the angle. In effect, the small flaps
the slant surface and raise it (a = 0° corresponds to the cannot prevent flow from the slant.
flap aligned with the slant wall). Although this may Table 5 shows a comparison between the original
lead to a stronger separation on the slant, the upward and modified models with the bottom configuration of
trend of separation flow, in turn, will constrain the the big-type flap at 130° (Bottom_big_130). The slant
longitudinal vortices and reduce the drag. surface accounted for most of the reduction (235
As shown in Figure 13, for the big-type flap, there counts), while there was an increase of pressure drag
was some reduction at small a angles (0°–40°) but not coefficient at the vertical base ( + 5 counts).
much. At 50°, the pressure drag coefficient suddenly
increased but then continuously decreased until reach-
ing its minimum at a = 130° (228 counts). The inverse Combination of different configurations. According to the
changing trend from a = 180° to a = 150° is similar to above results, there are four categories of schemes
the changing trend from a = 65° to a = 95° in which enhance drag reduction for the 25° Ahmed
Beaudoin and Aider.4 In Beaudoin and Aider,4 the model: Top_small_40, Top_big_20, Side_big_80, and
a = 0 case corresponds to a vertical flap along the ver- Bottom_big_130, and marked as ffi, ffl, , and Ð,
tical wall of the afterbody, from 65° to 95°, the drag respectively in Table 6. Clearly, the most efficient way
decreases nearly linearly toward a 24-count reduction to reduce the pressure drag coefficient is to modify the
for a = 95°. Between a = 150° and a = 160°, there flow circulation around the side edges of the slant (con-
were some sharp rise in the curve; this is because the figuration ). The configurations at the top side of the
flap in the upper part of the plane has positive effects slant surface both with large and small flaps are more
to the flow separation when a = 150°; however, when efficient than the bottom configuration. Interestingly,
a = 160°, the flap in the lower part of the plane and the slant surface pressure drag coefficient decreased
the flow field is very different from the one when markedly under all of these configurations.
Tian et al. 11

Table 6. Optimal drag reduction obtained by various configurations.

Case Cp* DCp* (%) Cs* DCs* (%)

ffiTop_small_40 0.2025 20.0438 (217.9%) 0.0956 20.0471 (233.0%)

ffl Top_big_20 0.2021 20.0442 (217.9%) 0.0813 20.0614 (243.0%)
 Side_big_80 0.1942 20.0521 (221.2%) 0.0943 20.0484 (233.9%)
Ð Bottom_big_130 0.2187 20.0276 (211.2%) 0.1071 20.0356 (224.9%)

Table 7. Results of combination schemes for 25° Ahmed model.

Case ffi+ ffi+Ð ffl+ ffl+Ð +Ð ffi++Ð ffl++Ð

Cp* 0.2062 0.2400 0.1952 0.2279 0.2173 0.2343 0.2128

Table 8. Comparison of standard 35° Ahmed model with Table 9. Results of combination schemes for 35° Ahmed.
Top_big_30.
Case ffi+ffl ffi+ ffl+ ffi+ffl+
Case Cp* Ck* Cs* Cb*
Cp* 0.1868 0.1914 0.1922 0.1885
Standard 0.1985 0.0118 0.0979 0.0890
Top_big_30 0.1898 0.0096 0.0896 0.0906

with any of the combinations. The optimal combina-

tion scheme was ffi + ffl, at only a 12-count reduction.
We created several combinations of the configura-
tions in order to test another seven schemes: ffi + ,
Drag reduction analysis and discussion
ffi + Ð, ffl + , ffl + Ð,  + Ð, ffi +  + Ð, and
ffl +  + Ð. The results are shown in Table 7. No one Drag reduction analysis of 25° Ahmed model
combination was any more efficient than the single To delve further into the mechanism of drag reduction,
configuration of , that is, there was no accumulative we further compared the original 25° Ahmed model
effect to the configurations. against the optimal corresponding individual configura-
tion of Side_big_80 and Top_small_40. The velocity
contours on the symmetry plane (Figure 14) indicate the
Drag reduction of 35° Ahmed model flow separates after the configuration of the side big flap
Optimal 35° Ahmed model configuration. Because the 35° or top small flaps, a phenomenon which should be have
Ahmed model itself is in low drag regime, most of our increased the drag—but in fact, the total pressure drag
configurations for the 35° Ahmed were not as remark- coefficient significantly decreased as explained by the
able as the configurations in the 25° Ahmed model. The transversal velocity contours (Figure 15) on the cross-
optimal scheme of the 35° Ahmed model, Top_big_30 sectional plane (10 mm away from the tail) as well as the
(Top_big_30 for a 35° slant angle and Top_big_20 for a pressure distribution (Figure 16) on the slant surface.
25° slant angle correspond to the same location of the Although the separation region expanded on the
flap), had a 9-count reduction in total pressure drag slant surface during the simulation, the longitudinal
coefficient mostly due to the contribution of the slant vortices raised and were weakened in intensity as
surface (Table 8). evidenced by the transversal velocity contours in
Figure 15. The expanded lower velocity region in the
tail resulted from the enlarged separation region. Due
Combinations of different configurations. Similarly to the to the strong interaction between the flow from the
25° Ahmed model analysis, we further tested roof and the streamwise vortices created along the side
Top_big_30, Side_big_80, and Bottom_small_30 edges, the strong downforce generated from the crimp-
(marked as ffi, ffl, and , respectively) by testing four ing process of the longitudinal vortice suppressed the
combination schemes: ffi + ffl, ffi + , ffl + , and separation vortex. Placing flaps on the side edges of the
ffi + ffl + . The results are shown in Table 9. The slant to constrain the longitudinal vortices or configur-
reduction of pressure drag coefficient was very limited ing flaps on the top of slant to raise the flow intensified
12 Advances in Mechanical Engineering

Figure 14. Velocity contours on symmetry plane of (a) standard 25° Ahmed model, (b) Side_big_80, and (c) Top_small_40.

the flow separation and, in turn, weakened the longitu- longitudinal vortices stamped on the standard Ahmed
dinal vortices. Weaker longitudinal vortices caused the body, suggesting that longitudinal vortices were well
downforce decreased, the separation vortex strength- suppressed.
ened, and the drag generated by flow separation is As shown in Figure 17, we compared the velocity at
small, so the total drag decreased. This is why the different positions on the 25° Ahmed with Side_big_80
separation phenomenon shown in Figure 14 did not among experiment5 and simulation results CFD_ with
necessarily increase the drag. and without control. Because the experiment was con-
Beaudoin et al.23 showed that longitudinal vortices ducted at various velocities, we applied non-
are regions of low pressure which contribute to over- dimensional methodology24 to unify the velocity value
all pressure drag. In the pressure contour shown in with u=u‘ . The Reynolds numbers are 2.86 3 106 and
Figure 16, the negative pressure on the slant surface 4.29 3 106 in the experiment and simulations, respec-
of the reduction scheme models substantially tively (with character length based on 1.044 m) which
increased (i.e. the absolute pressure value decreased.). meet Reynolds number similarity requirements. All
It was diluted considerably compared to the three images above show similar overall development
Tian et al. 13

Figure 15. Transversal velocity contours on cross-sectional plane (10 mm from tail): (a) standard 25° Ahmed model,
(b) Side_big_80, and (c) Top_small_40.

Figure 16. Pressure distribution on the slant surface: (a) standard 25° Ahmed model, (b) Side_big_80, and (c) Top_small_40.

trend between the experiment and CFD_ without con- Figure 14(b) as well as the shear layer described by Lyn
trol, suggesting that our simulation technique is reli- and Rodi.25 There was no separation at x = 0.200 m
able. In the symmetry plane, at position x = 20.062 m for the standard 25° Ahmed due to the positive velocity
and x = 0.200 m, the velocities in the x direction both in the x direction. The negative values of the CFD_
have negative values. At x = 20.062 m, negative value control indicate that the separation region behind the
means the flow on the slant surface moves back, which body expanded and the vortex moved away from
is consistent with the phenomenon shown in the body, as shown in Figure 14(b). To further explore
14 Advances in Mechanical Engineering

Figure 17. Velocity comparison among experiment, CFD_ control, and CFD_ without control.
Tian et al. 15

Figure 18. Transversal velocity contours on cross-sectional plane (10 mm from the tail): (a) standard 35° Ahmed model and
(b) Top_big_30.

Figure 19. Pressure distribution on slant surface: (a) standard 35° Ahmed model and (b) Top_big_30.

the longitudinal vortices, we compared the velocity in the standard Ahmed model, which implies that the
the z direction at the position x = 20.088 m, y = strength of longitudinal vortices is poorer than the
20.170 m (near the side of the slant surface.) The velo- baseline; the velocity contour shown in Figure 15(b)
city magnitude in this direction was lower than that of supports this observation as well.
16 Advances in Mechanical Engineering

Drag reduction analysis of 35° Ahmed model Funding

For the standard 35° Ahmed model, the flow separates The author(s) disclosed receipt of the following financial sup-
on the top of the slant and the longitudinal vortices on port for the research, authorship, and/or publication of this
the slant are very thin. The model is inherently in a low article: This project was supported by the State Key
Laboratory of Automotive Simulation and Control in Jilin
drag regime, accordingly. We compared Top_big_30
University, the Hong Kong Scholar Program, and
against the standard 35° Ahmed model to find out that
Postdoctoral Science Foundation of China (2012M510874,
the overall pressure drag coefficient for our scheme 2012T50314, and 2014M551180).
dropped by 9 counts (mostly contributed by the tail
slant). The change from the standard model was fairly
negligible apart from the fact that the separation area References
on the slant grew somewhat transversely and mixed bet- 1. Ahmed SR, Ramm R and Faltin G. Some salient fea-
ter with the longitudinal vortices, which were weaker tures of the time-averaged ground vehicle wake. SAE
for this configuration (Figure 18(b)). As shown in paper 840300, 1984.
Figure 19(b), the pressure on the slant of Top_big_30 2. Hucho WH and Sovran G. Aerodynamics of road vehi-
increased to a certain extent compared to the baseline. cles. Ann Rev Fluid Mech 1993; 25: 485–537.
3. Vino G and Watkins S. Flow structures in the near-wake
Furthermore, the longitudinal vortices disappeared
of the Ahmed model. J Fluid Struct 2005; 20: 673–695.
completely when this scheme was applied.
4. Beaudoin J-F and Aider J-L. Drag and lift reduction of a
3D bluff body using flaps. Exp Fluid 2008; 44: 491–501.
Conclusion 5. Lienhart H and Becker S. Flow and turbulence structures
In this study, we demonstrated the efficiency of adding in the wake of a simplified car model. SAE paper 2003-
01-0656, 2003.
specially designed flaps on the four edges of a standard
6. Drouin V, Giovannini A and Gilliéron P. Topology and
Ahmed model’s slant surface to break down the longitu- characterization of the vortical near-wake flow over a
dinal vortices and reduce the aerodynamic drag. For the simplified car model. In: Proceedings of the bluff body
25°Ahmed model, adding flaps on any possible edge can wakes and vortex induced vibrations (BBVIV3) confer-
weaken the longitudinal vortices. Vortex breakdown ence, Port Douglas, QLD, Australia, 17–20 December
modifies the wall pressure distribution of the slant. The 2002, pp.1–4.
optimal pressure drag coefficient reduction, 21.2%, was 7. Strachan RK, Knowles K and Lawson NJ. The vortex
obtained with a large flap placed on the side at 80° structure behind an Ahmed reference model in the pres-
angle. There was a reduction of 17.9% for the top con- ence of a moving ground plane. Exp Fluids 2007; 42:
figuration with both large and small flaps at 40° and 20° 659–669.
angles, as well. Flaps placed around the slant bottom 8. Wang XW, Zhou Y, Pin YF, et al. Turbulent near wake
were shown to be less efficient but to allow for a reason- of an Ahmed vehicle model. Exp Fluids 2013; 54: 1490
9. Gilliéron P, Leroy A, Aubrun S, et al. Influence of the
able pressure drag coefficient reduction, particularly a
slant angle of 3D bluff bodies on longitudinal vortex for-
large flap at 130° (up to 11.2% drag reduction). For the
mation. J Fluid Eng: T ASME 2010; 132: 051104.
35°Ahmed model, adding flaps in any position had prac- 10. Sims-Williams DB and Duncan BD. The Ahmed model
tically negligible effect. The optimal configuration (the unsteady wake: experimental and computational analy-
combination of Top_big_30 and Side_big_80) for the ses. SAE paper 2003-01-1315, 2003.
35° Ahmed model achieved only a 12-count reduction 11. Aubrun S, McNally J, Alvi F, et al. Separation flow con-
(6% drag reduction) in total pressure drag coefficient. trol on a generic ground vehicle using steady microjet
The flow of the original 25° Ahmed model does not arrays. Exp Fluids 2011; 51: 1177–1187.
separate on the slant, but creates strong longitudinal 12. Khalighi B, Zhang S, Koromilas C, et al. Experimental
vortices along the side. The different flap configura- and computational study of unsteady wake flow behind a
tions investigated in this study were observed to trans- bluff body with a drag reduction device. SAE paper 2001-
form the flow structures into the formation of the 35° 01-1042, 2001.
13. Howell J, Sims-Williams D, Sprot A, et al. Bluff body
Ahmed model (low drag regime), the flow which sepa-
drag reduction with ventilated base cavities. SAE paper
rates at the top of the slant consists of no or little longi-
2012-01-0171, 2012.
tudinal vortices. The results of this study demonstrate a 14. Castejon D, Catalano F and Severi G. Methods of drag
potential approach to flow control supported by the reducing without losing aerodynamic stability applied on
substantial drag reduction observed with various flap SAE reference model. SAE paper 2011-36-0079, 2001.
configurations. 15. Ha J, Jeong S and Obayashi S. Drag reduction of a
pickup truck by a rear downward flap. Int J Autom Tech
Declaration of conflicting interests 2011; 12: 369–374.
The author(s) declared no potential conflicts of interest with 16. Fourrie G, Keirsbulck L, Labraga L, et al. Bluff-body
respect to the research, authorship, and/or publication of this drag reduction using a deflector. Exp Fluids 2011; 50:
article. 385–395.
Tian et al. 17

17. Altaf A, Omar AA and Asrar W. Passive drag reduction 21. Rouméas M, Gilliéron P and Kourta A. Separated flows
of square back road vehicles. J Wind Eng Ind Aerodyn around the rear window of a simplified car geometry.
2014; 134: 30–43. J Fluid Eng: T ASME 2008; 130: 021101.
18. Hanfeng W, Yu Z, Chao Z, et al. Aerodynamic drag 22. Zhang Y. Research on automobile wind tunnel correction
reduction of an Ahmed body based on deflectors. J Wind based on simulation and test. Doctoral Dissertation, Col-
Eng Ind Aerodyn 2016; 148: 34–44. lege of Automotive Engineering, Jilin University, Chang-
19. Craft TJ, Gant SE, Iacovides H, et al. Computational chun, China, 2010.
study of flow around the Ahmed car body (case 9.4). In: 23. Beaudoin JF, Cadot O, Aider JL, et al. Characterization
Basara B (eds) Proceedings of the 9th Joint ERCOFTAC/ of longitudinal vortices on a 3D bluff body using cavita-
IAHR/QNET-CFD workshop on refined turbulence model- tion. Exp Fluids 2004; 37: 763–776.
ling, Darmstadt University of Technology, Germany, 4–5 24. Joseph P, Amandolese X and Aider J. Drag reduction on
October 2001. the 25° slant angle Ahmed reference body using pulsed
20. Makowski F and Kim S. Advances in external-aero simu- jets. Exp Fluids 2012; 52: 1169–1185.
lation of ground vehicles using the steady RANS equa- 25. Lyn DA and Rodi W. The flapping shear layer formed
tions. SAE paper 2000-01-0484, 2000. by flow separation from the forward corner of a square
cylinder. J Fluid Mech 1994; 261: 353–376.