Computer Modeling in

Engineering & Sciences

Tech Science Press

DOI: 10.32604/cmes.2023.043322

ARTICLE

Parametric Optimization of Wheel Spoke Structure for Drag Reduction of an

Ahmed Body

Huihui Zhai1 , Dongqi Jiao2 and Haichao Zhou2, *

1
Institute of Electrical and Information Engineering, Zhenjiang College, Zhenjiang, 212028, China
2
School of Automotive and Traffic Engineering, Jiangsu University, Zhenjiang, 212013, China
*Corresponding Author: Haichao Zhou. Email: [email protected]
Received: 29 June 2023 Accepted: 20 September 2023 Published: 30 December 2023

ABSTRACT
The wheels have a considerable influence on the aerodynamic properties and can contribute up to 25% of the total
drag on modern vehicles. In this study, the effect of the wheel spoke structure on the aerodynamic performance of
the isolated wheel is investigated. Subsequently, the 35° Ahmed body with an optimized spoke structure is used
to analyze the flow behavior and the mechanism of drag reduction. The Fluent software is employed for this
investigation, with an inlet velocity of 40 m/s. The accuracy of the numerical study is validated by comparing it
with experimental results obtained from the classical Ahmed model. To gain a clearer understanding of the effects
of the wheel spoke parameters on the aerodynamics of both the wheel and Ahmed model, and five design variables
are proposed: the fillet angle α, the inside arc radius R1 , the outside radius R2 , and the same length of the chord
L1 and L2 . These variables characterize the wheel spoke structure. The Optimal Latin Hypercube design method
is utilized to conduct the experimental design. Based on the simulation results of various wheel spoke designs, the
Kriging model and the adaptive simulated annealing algorithm is selected to optimize the design parameters. The
objective is to achieve the best combination for maximum drag reduction. It is indicated that the optimized spoke
structure resulted in a maximum drag reduction of 5.7% and 4.7% for the drag coefficient of the isolated wheel and
Ahmed body, respectively. The drag reduction is primarily attributed to changes in the flow state around the wheel,
which suppressed separation bubbles. Additionally, it influenced the boundary layer thickness around the car body
and reduced the turbulent kinetic energy in the wake flow. These effects collectively contributed to the observed
drag reduction.

KEYWORDS
Ahmed body; wheel spoke design; parameter optimization; drag reduction; numerical simulation

1 Introduction
The automotive industry is facing rapid development and increasing customer diversification and
personalized needs, which puts immense pressure on automotive companies to optimize products
and produce high-caliber vehicles. Aerodynamics is a major concern in the automobile industry as
it directly influences safety, comfort, fuel consumption, and driving range. The need to minimize fuel
consumption in automobiles is driven by greenhouse gas CO2 emissions. One effective measure for

This work is licensed under a Creative Commons Attribution 4.0 International License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the
original work is properly cited.
956 CMES, 2024, vol.139, no.1

reducing CO2 emissions is to decrease the vehicle’s aerodynamic drag. It has been reported that the
aerodynamic drag of a mid-size car contributes up to 46% of fuel consumption at highway speeds, and
a 5% reduction in aerodynamic drag could result in a reduction of 1.5 g/km of CO2 [1]. Therefore,
engineers and scholars have made significant efforts to improve the aerodynamic performance of
vehicles using passive and active devices [2,3]. Kim et al. argued that the classical approach to reducing
drag, such as shape modification, is not feasible in the modern world due to customers’ aesthetic
demands [4]. Inspired by the secondary feathers of birds, Kim et al. proposed a new automatic moving
deflector to be applied to the Ahmed body, resulting in a maximum drag reduction of 19%. Therefore,
the urgency to develop new drag reduction methods without compromising aesthetics is a growing
area of research interest.
Numerous studies have pointed out that wheels have a considerable influence on the aerodynamic
properties and can contribute up to 25% of the total drag on modern vehicles. The aerodynamic
drag has a direct relationship with the flow field around the car, and the high contribution of
wheels to drag is due to the strong interactions between the wheel wake and the vehicle underbody
through pressure modifications [5]. Hence, when body shape optimization reaches its limits, wheel
aerodynamics becomes a prominent concern in order to reduce fuel consumption. Additionally,
reducing wheel drag benefits the range of electric vehicles. Consequently, over the past two decades,
wheel aerodynamics has been the focus of numerous studies and optimization efforts in academia and
industry [6–8].
Frackell [9] conducted the first experimental work on an isolated wheel under realistic conditions.
The results showed that the pressure distribution measured on the wheel surface highlighted two main
specificities of the flow: the wheel rotation moved the forward stagnation point approximately 15°
down towards the front contact patch on the ground, and jetting-vortex effects appeared on either
side of the contact patch. Brandt et al. [10] designed several different types of wheels and studied
the effects of certain rim parameters on the flow field and drag using computational fluid dynamics
(CFD). The study concluded that drag has a linear relationship with the coverage area of the wheel
spoke. Fang et al. [11] investigated the influence of rim cover area on drag using wind tunnel tests and
found that the full cover wheel has the lowest drag coefficient. Fu et al. [12] studied the aerodynamic
characteristics of wheels and cars using different wheel opening areas and design locations, and it is
shown that, under the same opening area condition, increasing the number of openings can improve
the aerodynamic performance.
Despite the known potential for wheel modification to reduce drag, there is still considerable
room for improvement in understanding flow phenomena and their interaction with surrounding
parts in wheel aerodynamics. Elofsson et al. [13] presented a schematic picture of near-wake structures,
including two jetting-vortices and two vortices downstream of the upper wheel boundary separation.
Saddingtion et al. [14] analyzed the flow field around the wheel using the Laser Doppler Anemometry
(LDV) system, and the results showed that the upper vortices appeared weaker and merged with the
jetting-vortices within one diameter downstream behind the wheel axis. Computational fluid dynamics
(CFD) has provided further insight into understanding the wheel flow field and aerodynamics.
Simulations with isolated wheels have demonstrated the ability of CFD to capture specific flow features
and improve understanding of vortex dynamics and their impact on aerodynamic drag. When the
air reaches the wheel, a stagnation point occurs, and the flow divides, resulting in three main vortex
structures around the front wheel [15]. However, the wheel coverage area affects the flow field and
leads to changes in drag coefficients. A fully covered wheel blocks the air entering the wheel cavity,
reducing the vortex size and minimizing flow separation around the wheel. Furthermore, it has been
found that drag reduction from covering the rim track is more efficient than covering the center [16].
CMES, 2024, vol.139, no.1 957

Although the effect of wheel design and the main flow features has been studied both exper-
imentally and numerically, few studies have focused on the optimization of wheel spoke design to
reduce aerodynamic drag. Therefore, the objective of this study is to investigate the influence of wheel
spoke structure parameters on the aerodynamic drag of an isolated wheel. Subsequently, the kriging
model and ASA method are employed to optimize the wheel spoke design and achieve maximum drag
reduction for the Ahmed body, as well as the drag reduction mechanism is also discussed. The paper is
organized as follows: Section 2 presents the geometry model and simulation details. Section 3 discusses
the investigation of parameters and the optimization process for the wheel spoke structure. Section 4
reports the results and discussions, including the differences in the flow field and the drag reduction
mechanism. Finally, Section 5 provides concluding remarks.

2 Geometry Model and Simulation Details

2.1 Geometry Model
In this current study, a generic car model with the wheels and wheelhouses is used, and the
reference model chosen is the Ahmed body. The selection of this model is based on its extensive use in
previous studies for validation and benchmarking purposes in the field of road vehicle aerodynamics.
The geometry model of Ahmed body, as depicted in Fig. 1, represents a 1:4 scale model of a lower-
medium size hatchback vehicle. It features a short curved front section, followed by a long straight
middle section, and an angled rear end. Varying the rear end angle between 0° and 40° enables the
investigation of different flow characteristics and significant changes in drag coefficients [17]. The
rear angle plays a crucial role in determining drag, as approximately 85% of the aerodynamic drag is
attributed to pressure drag. Rear angles larger than 30° correspond to a low drag configuration. Based
on the findings of previous research and Strachan’s test results [18], the Ahmed body with a rear angled
upper surface of 35° is selected for the present study. The detailed dimensions of the Ahmed body are
presented in Fig. 2.

Figure 1: Ahmed body of a simplified vehicle

The purpose of this investigation is to determine the sensitivity of the wheel spoke structure
parameters and identify an optimal aerodynamic shape that minimizes the drag coefficient of the
wheel. Additionally, the aim is to enhance the overall aerodynamic performance of vehicles equipped
with low drag wheels. To establish a universal design principle, an analysis is conducted on the opening
ratio of 15 popular car models currently available in the market. For instance, Fig. 3 illustrates the
investigation of a popular SUV.
The selected wheels in this study primarily exhibit an opening ratio concentrated in the range of
50%–55%, as demonstrated in Fig. 4. Therefore, the focus of this research is on a five-spoke wheel with
an opening ratio of 50%. The dimensions and features of the wheel used in this study closely resemble
those of a real car, maintaining the same proportions as the Ahmed body, as depicted in Fig. 5. To
simulate realistic conditions, four wheels are attached to the car body using cylindrical axles. The
958 CMES, 2024, vol.139, no.1

center of these cylindrical axles is positioned at a vertical height of 75 mm, as shown in Fig. 6. In order
to accurately capture the pressure peak and jetting effect at the wheel edges, following the simulation
results provided by Humimic [19], the contact patch region of the wheel is modeled as a convex plate
with a height of 0.2 mm and a length of 25 mm for computational analysis. This modeling approach
ensures the quality of the mesh and captures the airflow behavior around the wheel effectively.

Figure 2: Detail sizes of Ahmed model

Figure 3: The different car wheel structure with five spoke

4
Count

0
20 30 40 50 60 70 80
Opening ratio %

Figure 4: The opening ration distribution Figure 5: Wheel model with five spoke
CMES, 2024, vol.139, no.1 959

Figure 6: Ahmed model with four wheels

2.2 Calculation Domain and Boundary Conditions

The shape and geometry of the Ahmed body are depicted in Fig. 7. The body has a length (L) of
1044 mm, a height (H) of 288 mm, and a width (W ) of 389 mm. The computational domain extends
approximately 12 times the body length (12 L), 5 times the width (5 W ), and 5 times the height (5 H).
The car body is positioned in the middle of the width direction, with a distance of 4 L from the inlet
and 7 L from the outlet. This arrangement corresponds to a blockage ratio of 4%. The inlet boundary
condition is set as a uniform inflow condition, while the outlet is configured as a pressure-outlet
condition with a pressure of 0 Pa. Frictionless boundary conditions are applied to the top and two
lateral sidewalls. The Ahmed body surface is modeled as a no-slip wall, and the floor is treated as
a zero-shear wall. As for the wheel surface, a rotating motion is considered with a specified angular
speed. To simulate the flow characteristics, the SST turbulence model is chosen, given its widespread
usage in the industry for modeling vehicle flows [20]. The SIMPLE algorithm and second-order
discretization method are employed to obtain the flow field and pressure distributions accurately.

Figure 7: Computing domain and boundary conditions

2.3 Model Validation

The numerical simulation method presented in this study is validated using the available data
of the Ahmed model in the published literature. The mesh generated around the Ahmed body is
shown in Fig. 8. The numerical study uses the same boundary conditions as the experiment, including
an incoming flow velocity of 40 m/s. The total number of elements generated using the Hypermesh
software is 11,326,856, with a minimum cell size of 0.02 m around the body surface. Two refined
regions are applied in the mesh generation, one with a cell size of 0.02 m and the other with a cell
size of 0.034 m. The maximum cell size in the outer surface of the domain is set to 0.5 m. By setting the
minimum cell size, the first layer is ensured to be at y+ < 10 in the present study, as shown in Fig. 9.
This choice is important for accurately capturing the near-wall flow.
960 CMES, 2024, vol.139, no.1

Figure 8: Mesh around the Ahmed body

Figure 9: The y+ values around the body surface

The simulation results are compared with both experimental data and simulation results from
previous studies [21,22]. Table 1 shows that the numerical values of the drag coefficient are in good
agreement with the experimental results. Fig. 10 presents a comparison of the velocity along vertical
lines in the symmetry plane, revealing a significant separation over the rear slant zone, which is
consistent with the experimental results [23]. These validation results confirm the accuracy of the
simulation model, including the mesh work and physics model used for simulating the Ahmed model
with wheels in this study. The validation outcome further strengthens the reliability of the results
presented in this research.

Table 1: Comparisons of drag coefficient with the published results

No. Reference Method Drag coefficient (C d ) Accuracy
1 Ahmed et al. [21] Experimental 0.26 –
2 Siddiqui et al. [22] CFD simulation 0.271 4.2%
3 Present study CFD simulation 0.261 0.38%
CMES, 2024, vol.139, no.1 961

0.4

0.3

0.2 CFD[Present work]

Rear body
Exp.[23]

0.1

0.0
-0.25 -0.20 -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15
X/m

Figure 10: Comparison of velocity distribution between simulation and experiment

2.4 Meshing
The Hypermesh software is utilized to generate a mixed mesh consisting of tetrahedral and
pentahedral cells for the Ahmed body with the wheel. Two mesh refinements are applied near the
wheel and car body surfaces using tetrahedral elements, while the pentahedral elements are generated
further away in the domain using solid mapping. Following the same mesh structure used for the
mesh validation case, three meshes, named Mesh 1, Mesh 2, and Mesh 3, are created, containing 15.02
million, 18.51 million, and 22.46 million cells, respectively. As indicated in Table 2, the drag coefficients
decrease with an increasing number of grid cells. The C d value error for Mesh 1, Mesh 2 and Mesh 3
is 0.002. It is worth noting that the calculation accuracy is positively correlated with the number of
cells; however, excessive cell count can reduce computational efficiency. Therefore, Mesh 2 is selected
for further research, as illustrated in Fig. 11.

Table 2: Grid independence results

Mesh Total cells numbers Cd
Mesh 1 15.02 million 0.415
Mesh 2 18.51 million 0.402
Mesh 3 22.46 million 0.404

3 Parameter Investigation and Optimization

3.1 Simulation of Isolated Wheel
While the wheel structure has a significant impact on the aerodynamic drag of the wheel, it is
important to note that wheel aerodynamics are independent of the overall vehicle aerodynamics. In
other words, reducing the aerodynamic drag of the wheel can have a positive effect on the vehicle drag.
This perspective has been supported by previous research [10].
962 CMES, 2024, vol.139, no.1

Figure 11: The mesh around the car body and the front wheel

To define the computational domain, the dimensions of the isolated wheel, including length
(LW = 150 mm), width (WW = 50 mm), and height (HW = 150 mm), are taken into consideration. The
boundaries of the domain are designed to be 20 times the length of the wheel (20 LW ) in length, 10
times the width of the wheel (10 WW ) in width, and 5 times the height of the wheel (5 HW ) in height,
as depicted in Fig. 12. The wheel is positioned at the center of the width direction, with a clearance
of 7 times the wheel length (7 LW ) from the inlet. The resulting blockage ratio is approximately 2%.
For mesh generation, the commercial software Hypermesh is employed to create the volume mesh
and boundary layer cells. A cylindrical-shaped volumetric control is applied around the wheel, while a
rectangular-shaped volumetric control is used in the wake region to accurately capture flow features in
these areas. Prism layers, triangular elements, and tetrahedral elements are utilized for meshing. The
prism layers, consisting of 5 inflation layers with a growth rate of 1.2, are employed to capture the
boundary layer flow. The wheel rotation is simulated using the moving reference frame method. To
test the grid sensitivity of the simulation, three different gird resolutions, named as the coarse, medium
and fine, are proposed, and the response Cd is shown in Table 3. The fine grid distribution around the
wheel are shown in Fig. 12b. Additionally, the fine grid ensured a y+ value of less than 2, as shown in
Fig. 13.

Table 3: Comparison of the different grid resolutions

Items Coarse Medium Fine
Total cells number 4.15 million 4.86 million 5.23 million
Cd 0.963 0.941 0.934
CMES, 2024, vol.139, no.1 963

3.2 Parameters Selection and DOE

The structure of the rim spoke plays a crucial role in aerodynamic drag as it affects the airflow.
Considering the same rim area coverage of 50% and a fixed wheel diameter of 75 mm, five design
parameters of the rim spoke are studied: the fillet angle α, inside arc radius R1 , outside arc radius R2 ,
and the arc tangent lines L1 and L2 , as depicted in Fig. 14.

Figure 12: Computational domain sizes and grids

Figure 13: Y + values distribution around wheel surface

964 CMES, 2024, vol.139, no.1

Figure 14: Schematic diagram of a single spoke

In the original spoke design, the fillet angle α is set at 0.384 rad, with inside arc radius R1 and
outside arc radius R2 of 8 and 55 mm, respectively. The arc tangent lines L1 and L2 are both 35.693 mm.
To examine the influence of different spoke design parameters on wheel aerodynamic drag while
maintaining the same opening area of 5S, the parameter variables are assigned the following values:
α ∈ (0.262, 0.523) rad, R2 ∈ (41.582, 58.679) mm while the remaining two parameters are determined
using Eqs. (1)–(3).
(π − α)
S = α · R2 2 − 20.912 · sin (α) · cos (α) + · 20.912 · (sin (α))
2
(1)
2
R1 = 20.91 · sin (α) (2)
L1 = L2 = R2 − 20.91 · cos (α) (3)

The Optimal Latin Hypercube design method [24,25], known for optimizing the combination
of factor levels, is employed to generate 20 sample points in this study. All 20 cases are simulated
with an inlet velocity of 40 m/s. The factors and corresponding levels, as well as the obtained results,
are summarized in Table 4. Fig. 15 showcases the designs of samples 7, 11, and 16. The impact of
parameter variables on the wheel’s aerodynamic drag coefficient is depicted in Fig. 16. It is evident
that the fillet angle α has the greatest influence on the drag coefficient, followed by the outside arc
radius R1 . Conversely, the arc tangent lines L1 and L2 exhibit minimal impact on the drag coefficient.

Table 4: Test schemes and calculation results

No. L1 (L2 )/mm R1 /mm R2 /mm α/Rad Cd
1 30.606 9.691 42.129 0.482 0.957
2 29.405 10.205 47.649 0.510 0.951
3 29.965 7.598 58.036 0.372 0.960
4 31.705 7.053 60.143 0.344 0.959
5 27.124 8.657 52.386 0.427 0.891
6 28.434 8.141 54.192 0.401 0.919
7 46.827 5.411 67.019 0.262 0.940
8 45.186 5.697 65.299 0.276 0.943
9 42.400 6.238 62.353 0.303 0.956
(Continued)
CMES, 2024, vol.139, no.1 965

Table 4 (continued)
No. L1 (L2 )/mm R1 /mm R2 /mm α/Rad Cd
10 31.886 9.187 50.664 0.455 0.951
11 34.289 7.871 55.193 0.386 0.950
12 41.103 6.516 60.966 0.317 0.954
13 34.155 8.390 53.303 0.413 0.885
14 32.602 8.923 51.507 0.441 0.946
15 37.772 7.325 57.352 0.358 0.951
16 28.863 10.452 46.967 0.524 0.968
17 39.893 6.794 59.663 0.331 0.950
18 43.794 5.958 63.832 0.289 0.938
19 29.991 9.949 48.376 0.496 0.950
20 31.254 9.430 49.911 0.468 0.957

Figure 15: Geometric models of some typical Schemes

98.85 100
100 95.876 97.7 100%

80 80%
Contributions for Cd

60 60%
49.732

40 40%

20 20%

0 0%
α R1 R2 L1 L2
Design variables

Figure 16: Contribution of different design variables for Cd

966 CMES, 2024, vol.139, no.1

3.3 Surrogate Model and Parameters Optimization

The Kriging surrogate model, known for its effectiveness in curve interpolation and response
surface approximation, is utilized in this study for the approximate evaluation of design schemes [26].
The correlation coefficient R2 is selected as a measure to evaluate the accuracy of the built Kriging
surrogate model. Fig. 17 illustrates the forecast accuracy between the simulation results and predicted
results, with an R2 value of approximately 0.98. This indicates that the established surrogate model
possesses sufficient accuracy for optimizing the parameters of the wheel spoke.
0.98

0.96
Simulation Cd

0.94

0.92

0.90

0.88
0.88 0.90 0.92 0.94 0.96 0.98
Predicting Cd

Figure 17: Kriging surrogate model accuracy

The adaptive simulated annealing (ASA) algorithm is chosen to obtain the global optimal
solution for the design parameters of the wheel spoke. In this algorithm, the objective function
aimed to minimize the wheel drag coefficient, while the constraint conditions are defined as α ∈
(0.262, 0.523) rad, R2 ∈ (41.582, 58.679) mm; R1 ∈ (5.411, 10.452) mm, L1 ∈ (27.125, 46.827) mm.
After 382 iterations, the optimal feasible solutions are determined and are listed in Table 5. Based on
these optimized parameters, a new wheel model is established and analyzed, resulting in a wheel drag
coefficient of 0.883. This represents a 5.7% decrease compared to the original drag coefficient of 0.934.

Table 5: Comparison of aerodynamic drag coefficients between original and optimized

L1 (L2 )/mm R1 /mm R2 /mm α/rad Cd
Original wheel 35.693 8 55 0.384 0.934
Optimized wheel 33.905 8.473 53.016 0.393 0.883

4 Results Analysis and Discussion

4.1 Flow Field around Wheel
To investigate the impact of the optimized wheel spoke on the flow field and aerodynamic drag,
several vertical sections and a longitudinal section are analyzed to compare the flow field information
before and after optimization, as depicted in Fig. 18. This analysis aims to reveal the differences
resulting from the optimized wheel spoke design.
CMES, 2024, vol.139, no.1 967

Figure 18: Four different sections

The pressure drag represents the primary component of the aerodynamic drag. Fig. 19 illustrates
the pressure distribution in the Y = 0 section of the wheel. It is evident that the air pressure around the
front of the wheel remains relatively unchanged, while the pressure at the rear undergoes noticeable
changes, with increased pressure values. The observed increase in pressure magnitude indicates a
decrease in pressure drag, resulting in a reduced overall wheel drag.

Figure 19: Pressure of longitudinal plane in wheel before and after optimization

Fig. 20 illustrates the streamlines and vorticity within the wheel cavity at different Z-direction
sections. Due to the obstructive effect of the wheel, the airflow initially separates at the front edge of the
wheel, forming two streams that propagate downstream. The presence of the wheel cavity introduces
unsteady flow characteristics and generates numerous flow vortices in the low-velocity airflow. When
968 CMES, 2024, vol.139, no.1

the airflow impacts the rear edges of the wheel cavity, the airflow velocity around the inner edge
decreases while vorticity increases. Under the driving effect of high-speed airflow from the front edge,
the vorticity within the inside zone is greater than that in the outside zone. With improved ventilation
capacity of the spoke, the vorticity distribution within the cavity of the optimized wheel is smaller
compared to the original wheel. Fig. 20a demonstrates minimal changes in the flow field around the
outside zone of the wheel, while significant differences are observed in the vicinity of the inside zone.
The optimized rim design displaces the vortex core backward and reduces the vorticity in the central
region. However, the vorticity intensity within the optimized wheel cavity is higher than that of the
original wheel. Fig. 20b compares the flow field at the Z = 0 mm section. Although the flow field
behind the wheel appears similar, the optimized wheel features a wider opening that more effectively
directs the airflow into the wheel cavity, resulting in a larger region of streamline vortices near the inner
edge of the rim. Fig. 20c reveals the presence of vortex pairs behind the optimized wheel, which are
absent in the original wheel. The vortices within the optimized wheel cavity exhibit relatively uniform
magnitudes, indicating a more stable airflow and reduced flow energy losses. Moreover, the regions of
higher vorticity in the optimized wheel are smaller compared to those in the original wheel.

Figure 20: Comparison of vorticity distribution in different sections

4.2 Aerodynamic Characteristics of Ahmed Body with Wheel

Fig. 21 illustrates the comparison of the time-average pressure coefficient distributions on the
body surfaces. It can be observed that compared to the original wheel, the pressure coefficient around
CMES, 2024, vol.139, no.1 969

the front wheel increases while it decreases around the rear wheel. This indicates that the rim spoke
structure has a direct influence on the airflow movement around the wheel. As for the back surface
of the car body, the pressure coefficient reduces on the slant rear surface, but due to changes in
airflow conditions in the under-body gap, there is a significant increase in the pressure coefficient
on the vertical rear surface. The drag coefficients are presented in Table 6. It can be seen that the
drag coefficient of the Ahmed body is reduced by 4.7%, while the front and rear wheels experience a
reduction of 16.7% and 17.1% in drag coefficient, respectively. From these findings, it can be concluded
that improving the geometric characteristics of the wheel’s aerodynamic shape is an effective measure
to enhance automotive aerodynamic performance and reduce drag.

Figure 21: Comparison of surface pressure coefficients around different Ahmed body

Table 6: Before and after optimization of vehicle aerodynamic drag coefficient

Cd of the car model Cd of the front wheel Cd of the rear wheel
Original model 0.402 0.0364 0.0246
Optimized model 0.384 0.0312 0.0210

The iso-surface of zero total pressure coefficient is selected to characterize the location, strength,
and size of flow separation, and the difference in the distribution of zero total pressure coefficient is
used to clarify the reasons for pressure drag [27]. Fig. 22 displays the iso-surface of zero total pressure
coefficient for the original wheel and optimized wheel. It is evident that the size of the zero total
pressure coefficient region is larger for the original front and rear wheels compared to the optimized
wheel. This indicates that the structural changes in the wheel spoke have a direct influence on the
airflow movements around the wheel, which also extends to the wake flow. Drag reduction relies on
pressure recovery around the car body. As depicted in Fig. 22, the separation bubbles around the front
wheel are larger than those around the rear wheel in both the original and optimized configurations.
This implies that the pressure recovery in the front wheel is slower, resulting in higher aerodynamic
drag compared to the rear wheel [5,28]. In comparison, the optimized wheel exhibits suppressed flow
separation and reduced separation bubble size based on the differences in the iso-surface of zero total
pressure coefficient. This delay in flow separation leads to a decrease in velocity fluctuations and
turbulent kinetic energy. Consequently, the aerodynamic performance of the Ahmed body with the
optimized wheel is improved.
970 CMES, 2024, vol.139, no.1

Figure 22: Iso-surface of zero total pressure coefficient for car body with different wheel model

The Q-criterion is based on the second invariant of the velocity gradient denoted as Q, which also
includes the condition that the ambient pressure should be higher than the pressure of the vortex to
separate it. The identification of vortex structures using the Q-criterion has been applied in vehicle
aerodynamics [29]. Fig. 23 illustrates the Q-criterion for the original and optimized wheel. In the
case of the front wheel, the large-scale vortical structure encompassing the five spoke appears similar
in both the original and optimized configurations. Additionally, there is a distinct horseshoe vortex
structure observed around the leeward side of the front wheel. However, in the optimized wheel, the
vortex development on the leeward side is reduced compared to the original wheel, indicating lower
fluctuations in the turbulent flow.

Figure 23: Vortex structure of front wheels before and after optimization (Q = 40000)

For the rear wheel, the original wheel model exhibits a prominent vortex structure on the windward
side, extending towards the spoke as depicted in Fig. 24. However, this large vortex is absent in the
optimized wheel, indicating effective control of flow separation on the windward side. In the optimized
wheel, the development of vortices on the leeward side of the rear wheel is reduced compared to the
original wheel. Additionally, the longitudinal vortex development is shorter and exhibits lower strength
and intensity compared to the original wheel. These differences in vortex structures imply that the
available kinetic energy for the eddies is reduced at an earlier stage.
CMES, 2024, vol.139, no.1 971

Figure 24: Vortex structure of rear wheel before and after optimization (Q = 40000)

The boundary layer distribution on four planes, defined by the range of velocities less than 0.99u,
is depicted in Fig. 25a. Fig. 25b presents a comparison of the boundary layer between the original
wheel and the optimized wheel. The values in the figure represent the normalized airflow velocity
using U/Uin , where Uin corresponds to the inlet velocity of 40 m/s. From Fig. 25b, it can be observed
that the boundary layer thickness around the wheel gradually increases along the car body, while the
velocity gradient decreases. However, there are differences in the boundary layer thickness around the
top of the car body. In plane P2, the boundary layer thickness is smaller for the original wheel, resulting
in higher viscous drag on the original car body. Conversely, in plane P4, the boundary layer thickness
is smaller for the optimized wheel, leading to higher viscous drag on the optimized car body. These
differences in the boundary layer affect the velocity field and result in variations in the iso-surface
of zero total pressure coefficient (Fig. 8). Therefore, it can be inferred that the pressure drag plays a
dominant role in the overall aerodynamic performance compared to the viscous drag.

Figure 25: (Continued)

972 CMES, 2024, vol.139, no.1

Figure 25: Locations and boundary layer distribution for the different four plane

The turbulent kinetic energy (TKE) is a crucial parameter that serves to connect the mechanisms of
drag reduction. It represents the mean kinetic energy available per unit mass of turbulent eddies. Fig. 26
illustrates the variation in TKE for the original and optimized wheel. The TKE is predominantly
concentrated on the upper slant surface. As the distance increases in the wake region, the TKE
decreases. However, when compared to the original wheel, the TKE in the surrounding flow area
exhibits a significant decrease. This indicates that the size of the low-vortex region is larger, and the
dissipation of flow energy can be effectively controlled. Typically, lower TKE signifies more stable
velocity pulsations, which may result in larger vortices and lower base pressure. This observation aligns
with the findings from the pressure coefficient results presented in Fig. 21.

Figure 26: (Continued)

CMES, 2024, vol.139, no.1 973

Figure 26: Turbulent kinetic energy at different planes in the original and optimized wheel

5 Conclusions
This study aimed to numerically investigate the influence of wheel spoke design on the aerody-
namics of an isolated wheel and the Ahmed model. The main conclusions are as follows:
(1) The aerodynamic characteristics of an isolated wheel are studied through numerical simulation,
focusing on four design parameters of wheel spoke: the fillet angle α, inside arc radius R1 , outside
radius R2 , and chord length L1 . The drag coefficient is selected as the objective. Using the Optimal
Latin Hypercube (OLH) design, 20 sample points are generated, and their corresponding drag
coefficients are calculated. Sensitivity analysis revealed that the angle α had the greatest impact on
the drag coefficient, followed by the outside arc radius R1 , while the arc tangent lines L1 and L2 had
minimal influence.
(2) By utilizing the integrated approach of the Kriging surrogate model and adaptive simulated
annealing algorithm, the optimal design parameters of the wheel spoke are obtained. Comparative
analysis of the flow fields between the original and optimized wheels demonstrated that the optimized
wheel influenced the airflow state inside the wheel cavity, reducing flow instability around the wheel.
As a result, the optimized wheel achieved a drag reduction of 5.7%.
(3) The aerodynamic drag of an Ahmed body equipped with the original and optimized wheel
is investigated, revealing a decrease of 4.7% in the drag coefficient for the Ahmed body with the
optimized wheel. Analysis of flow features indicated that the drag reduction primarily relied on
changes in the flow state around the wheel to suppress separation bubbles, as well as the influence on
boundary layer thickness around the car body and reduction of turbulent kinetic energy in the wake
flow. Further investigations can focus on developing new active or passive devices, for example the
active drag reduction of bionic non-smooth surface and the passive drag reduction of vortex generator,
to further reduce drag and enhance aerodynamic performance of the wheel and the vehicle.

Acknowledgement: The authors are grateful for the support by the National Natural Science
Foundation of China, and Postdoctoral Foundation of China.

Funding Statement: The authors are pleased to acknowledge funding of the National Natural Sci-
ence Foundation of China (Nos. 52072156, 51605198), and Postdoctoral Foundation of China
(2020M682269).

Author Contributions: The authors confirm contribution to the paper as follows: study conception
and design: Huihui Zhai; data collection: Dongqi Jiao; analysis and interpretation of results: Haichao
974 CMES, 2024, vol.139, no.1

Zhou; draft manuscript preparation: Huihui Zhai. All authors reviewed the results and approved the
final version of the manuscript.

Availability of Data and Materials: The data undering this article will be shared on reasonable request
to the corresponding author.

Conflicts of Interest: The authors declare that they have no conflicts of interest to report regarding the
present study.

References
1. Huminic, A., Huminic, G. (2017). Aerodynamic study of a generic car model with wheels and underbody
diffuser. International Journal of Automotive Technology, 18, 397–404.
2. Altaf, A., Omar, A. A., Asrar, W. (2014). Passive drag reduction of square back road vehicles. Journal of
Wind Engineering and Industrial Aerodynamics, 134, 30–43.
3. Yang, X., Hu, Y., Gong, Z., Jian, J., Liu, Z. (2021). Numerical study of combined drag reduction bases
on vortex generators and riblets for the ahmed body using IDDES methodology. Journal of Applied Fluid
Mechanics, 15(1), 193–207.
4. Kim, D., Lee, H., Yi, W., Choi, H. (2016). A bio-inspired device for drag reduction on a three-dimensional
model vehicle. Bioinspiration & Biomimetics, 11(2), 026004.
5. Yu, X., Jia, Q., Yang, Z. (2022). Comprehensive study of the aerodynamic influence of ground and wheel
states on the notchback DrivAer. JEnergies, 15(3), 1–33.
6. Zhou, H., Jiang, Z., Wang, G., Zhang, S. (2021). Aerodynamic characteristics of isolated loaded tires with
different tread patterns: Experiment and simulation. Chinese Journal of Mechanical Engineering, 34, 1–16.
7. Josefsson, E., Hobeika, T., Sebben, S. (2022). Evaluation of wind tunnel interference on numerical prediction
of wheel aerodynamics. Journal of Wind Engineering and Industrial Aerodynamics, 224, 104945.
8. Jiang, F., Niu, J., Li, R., Wang, Y., Zhang, Y. (2023). Computational fluid dynamics analysis of effect
of braking plate configurations on the aerodynamic behaviors of an ahmed car. Flow, Turbulence and
Combustion, 110(2), 301–323.
9. Fackrell, J. E. (1974). The aerodynamic of an isolated tire rotating in contact with ground (Ph.D. Thesis).
University of London, London.
10. Brandt, A., Berg, H., Bolzon, M., Josefsson, L. (2019). The effects of wheel design on the aerodynamic
drag of passenger vehicles. SAE International Journal of Advances and Current Practices in Mobility, 1(3),
1279–1299.
11. Fang, J., Wang, F. L. (2019). A study on CFD simulation and wind tunnel test on aerodynamic effects of
wheel rotation and drag reduction optimization. Automotive Engineering, 41(9), 1006–1012.
12. Fu, L., Hu, X., Zhang, S. (2006). Numerical simulation of influence of holes in wheel spokes on automotive
external flow-field. Nongye Jixie Xuebao (Transactions of the Chinese Society of Agricultural Machinery),
37(1), 8–11 (In Chinese).
13. Elofsson, P., Bannister, M. (2002). Drag reduction mechanisms due to moving ground and wheel rotation
in passenger cars. SAE Technical Paper, 2002-01-0531.
14. Saddington, A. J., Knowles, R. D., Knowles, K. (2007). Laser Doppler anemometry measurements in the
near-wake of an isolated formula one wheel. Experiments in Fluids, 42(5), 671–681.
15. Croner, E., Bézard, H., Sicot, C., Mothay, G. (2013). Aerodynamic characterization of the wake of an
isolated rolling wheel. International Journal of Heat and Fluid Flow, 43, 233–243.
16. Qiu, Z. (2009). Wheel aerodynamic developments by module-based prototype rims and stationary rim shields
(Ph.D. Thesis). Chalmers University of Technology, Gothenburg.
CMES, 2024, vol.139, no.1 975

17. Strachan, R. K., Knowles, K., Lawson, N. J. (2007). The vortex structure behind an ahmed reference model
in the presence of a moving ground plane. Experiments in Fluids, 42(5), 659–669.
18. Corallo, M., Sheridan, J., Thompson, M. C. (2015). Effect of aspect ratio on the near-wake flow structure
of an ahmed body. Journal of Wind Engineering and Industrial Aerodynamics, 147, 95–103.
19. Diasinos, S., Barber, T. J., Doig, G. (2015). The effects of simplifications on isolated wheel aerodynamics.
Journal of Wind Engineering and Industrial Aerodynamics, 146, 90–101.
20. Liang, H., Sun, Y., Li, T., Zhang, J. (2022). Influence of marshalling length on aerodynamic characteristics
of urban EMUs under crosswind. Journal of Applied Fluid Mechanics, 16(1), 9–20.
21. Ahmed, S. (1984). Some salient features of the time-averaged ground vehicle wake. SAE Paper, 840300.
22. Siddiqui, N. A., Chaab, M. A. (2020). A simple passive device for the drag reduction of an ahmed body.
Journal of Applied Fluid Mechanics, 14(1), 147–164.
23. Meile, W., Brenn, G., Reppenhagen, A., Lechner, B., Fuchs, A. (2011). Experiments and numerical
simulations on the aerodynamics of the ahmed body. CFD Letters, 3(1), 32–39.
24. Montazeri, F., Tavakoli, M. R., Salimpour, M. R. (2023). Improving thermal performance of rectangular
microchannel heat sinks using porous layer: CFD simulation and optimization. Journal of Applied Fluid
Mechanics, 16(8), 1574–1586.
25. Park, J. S. (1994). Optimal latin-hypercube designs for computer experiments. Journal of Statistical Planning
and Inference, 39(1), 95–111.
26. Zhang, L., Zhou, H., Wang, G., Li, H., Wang, Q. (2022). Investigation of effects of tire contour on
aerodynamic characteristics and its optimization. Proceedings of the Institution of Mechanical Engineers,
Part D: Journal of Automobile Engineering, 236(12), 2756–2772.
27. McArthur, D., Burton, D., Thompson, M., Sheridan, J. (2018). An experimental characterisation of the
wake of a detailed heavy vehicle in cross-wind. Journal of Wind Engineering and Industrial Aerodynamics,
175, 364–375.
28. Wang, Y., Sicot, C., Boree, J., Grandemange, M. (2020). Experimental study of wheel-vehicle aerodynamic
interactions. Journal of Wind Engineering and Industrial Aerodynamics, 198, 104062.
29. Zhou, H., Chen, Q., Qin, R., Zhang, L., Li, H. (2021). Investigation of wheelhouse shapes on the
aerodynamic characteristics of a generic car model. Advances in Mechanical Engineering, 13(12),
16878140211066842.