Aerodynamic Analysis

Armagan Altinisik1
of a Passenger Car at Yaw
Turkish Automotive Factory, Inc. Company
(TOFAS),
Yeni Yalova Yolu Cad. No: 574,
Angle and Two-Vehicle Platoon
Bursa 16369, Turkey Experimental and computational studies were performed to study the drag forces and the
e-mail: [email protected] pressure distributions of a one-fifth scale model FIAT Linea at increasing yaw angle and
two-vehicle platoon. Experiments were performed in the Uludag University Wind Tunnel
Onur Yemenici (UURT) only for the yaw angles of 0 deg, 5 deg, and 10 deg due to the test section dimen-
Assistant Professor sional restriction. Supplementary tests were performed in the Ankara Wind Tunnel (ART)
Mechanical Engineering Department, to evaluate the aerodynamic coefficients up to yaw angle of 40 deg. The test section
Faculty of Engineering, blockage ratios were 20% and 1%, respectively, in the UURT and ART tunnels. The
Uludag University, blockage effects for the yaw angles up to 10 deg were studied by the comparison of two
Gorukle Campus, wind tunnel results. The aerodynamic tests of two-vehicle platoon were performed in the
Bursa 16059, Turkey ART tunnel at spacings of “x=L” 0, 0.5, and 1. Static pressure distributions were
obtained from the model centerline and three vertical sections. In the numerical study,
Habib Umur three-dimensional, incompressible, and steady governing equations were solved by STAR-
Professor CCMþ code with realizable k-e two-layer turbulence model. Experimental and numerical
Mechanical Engineering Department, Cp distributions and Cd values were found in good agreement for considered yaw angles
Faculty of Engineering, and two-vehicle platoon. Maximum drag coefficient was obtained at yaw angle of 35 deg
Uludag University, for both experimental and numerical calculations. The two-vehicle platoon analysis
Gorukle Campus, resulted with the significant drag coefficient improvement for the leading car at spacings
Bursa 16059, Turkey of x=L ¼ 0 and 0.5, while for the tail car drag coefficient remained slightly above the
vehicle in isolation. [DOI: 10.1115/1.4030869]

Keywords: automotive aerodynamics, drag coefficient, pressure distribution, blockage

effect, yaw angle, platoon

1 Introduction several previous studies [11–17]. Aerodynamic parameters in

yawing angle change are strongly influenced by vehicle shape as
As a result of high oil prices, the automotive industry changed
discussed by Howell [18]. In the above literature studies, the most
their focus from maximizing the power of cars to maximizing fuel
critical yaw angle, where the drag force coefficient is maximum,
efficiency. Thus, the aerodynamic resistance of automobiles is
was found in the range of 20–35 deg. Moreover, after critical yaw
considered to be a crucial part of automotive design. Aerodynamic
angle, the drag coefficient is no longer sensitive to yawing angle
drag typically accounts for 65% of the total resistance for medium
change.
sized car at 100 km/hr. Hence, reducing drag contributes signifi-
Platoon aerodynamics for vehicle interactions is much more
cantly to the fuel economy of a car [1].
critical than to investigate isolated vehicle aerodynamics. The
There are various experimental and numerical studies con-
extremely small longitudinal spacings between adjacent vehicles
ducted on drag optimizations in the past literatures such as shape,
in a platoon make the flow field more complicated than that
style optimizations, and wake controls by vortex generator appli-
around a single vehicle. Vehicle interference has a remarkable
cations and active control methods [2–6]. The generic car models,
effect on the drag forces, especially on the following vehicles.
such as Ahmed’s model, SAE standard model, Motor Industry
Zabat et al. [19] carried out experiments on two 1/8 scale minivan
Research Association (MIRA) model, and DrivAer models, have
in tandem position and found that the overall drag was signifi-
been widely used in experimental and numerical analyses due to
cantly lower in a uniform platoon than that measured on a single
the restricted access and short life span of real car models [7–10].
vehicle. The effect of the vehicle platoon on the aerodynamic
However recently, aerodynamic studies on production cars can
coefficients of a passenger car has been an active study for
also be found in various engineering journals and papers.
researchers. These studies demonstrated that the drag coefficient
For passenger cars as well as for commercial vehicles, there are
of the leading car was improved significantly (20–50%) at spac-
considerable increases in the coefficients of resistance when they
ings of 0–1 car length. However, the tail car drag coefficient did
are approached by a side wind. Vehicle is more stable if the geo-
not show the same improvement in all studies [20–25].
metric center, center of gravity, and stagnation point should be
The aim of this study is to investigate the aerodynamic pressure
fall in line. But in case of crosswind, the flow becomes asymmet-
distributions on the model surface and drag coefficients at increas-
ric and stagnation point shifts toward direction of crosswind and
ing yaw angle and two-vehicle platoon at different vehicle spac-
this shift affects the stability of the vehicle. The drag coefficient “
ings. The one-fifth scale model FIAT Linea was tested for yaw
Cd ” and the lift coefficient “ Cl ” increase to a certain level as the
angles of 0–10 deg in the UURT tunnel. It was not possible to
yaw angle “ b ” reaches to a certain level. The aerodynamic char-
increase the test yaw angles more than 10 deg due to the test sec-
acteristics of the vehicle under crosswind have been the topic of
tion side walls. The same experiments were repeated up to yaw
angle of 40 deg in ART tunnel both to compare the blockage-free
1
Corresponding author. results with the UURT tests and to observe the aerodynamic coef-
Contributed by the Fluids Engineering Division of ASME for publication in the
JOURNAL OF FLUIDS ENGINEERING. Manuscript received March 29, 2015; final
ficients (Cp ; Cd Þ for higher yaw angles. Aerodynamic analysis of
manuscript received June 5, 2015; published online August 4, 2015. Assoc. Editor: the two-vehicle platoon at spacings of 0, 0.5, and 1.0 car length
Francine Battaglia. was carried out in the ART tunnel. Significant Cd improvements

Journal of Fluids Engineering Copyright V

C 2015 by ASME DECEMBER 2015, Vol. 137 / 121107-1

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 were observed for the leading car; however, the drag coefficient
remained same or slightly worsened for the tail car compared to
the vehicle in isolation. Cp distributions on the model centerline
and three vertical sections were measured and compared with the
numerical results. Considering that there are few studies of pres-
sure distributions obtained over the exterior surfaces of cars in the
literature, apart from very localized areas, in this study surface
pressure distributions were obtained both for increasing yaw angle
and two-vehicle platoon analysis. For all tests, the freestream ve-
locity was taken as U1 ¼ 30 m/s. In the numerical part of the
study, STAR-CCMþ solver with realizable k-e two-layer turbulence
model was used. Reynolds-Averaged Navier–Stokes (RANS)
equations were solved by using SIMPLE algorithm. The compari-
son of the calculated Cp distributions and Cd results were in good
agreement with the experimental results.

2 Experimental Study
The dimensions of one-fifth scale model of FIAT Linea and
schematic view of pressure tap positions on the model surface
were given in Fig. 1. A smooth underbody without detailed fea-
tures and a stationary floor was used. Cooling flow through the
model was not considered. Total number of the orifices drilled in
the model surface was 154. The aerodynamic analysis in the
Fig. 1 (a) Dimensions of one-fifth scale model and (b) posi- UURT tunnel was performed up to yaw angles of 10 deg. It was
tions and quantities of the pressure holes not possible to increase yaw angle due to UURT test section
dimensional restrictions, as shown in Fig. 2. The ART tunnel was

Fig. 2 Scaled view of one-fifth model in UURT test section at yaw angles of 0 deg, 5 deg, and
10 deg

Table 1 Aerodynamic wind tunnel test planning

Wind tunnels

Air velocity m/s Reynolds number (one-fifth scale) Tests UURT ART
6
30 1.87  10 Yaw angle 0–10 deg 
Yaw angle 0–40 deg 
Two-vehicle platoon analysis x/L:0, 0.5, and 1.0 

Table 2 Technical specifications of the UURT and the ART tunnels

UURT ART

Type Horizontal open-loop Closed-loop

Test sections L: 2 m, W: 0.7 m, H: 0.6 m L: 6.1 m, W: 3.05 m, H: 2.44 m
Maximum test velocity 30 m/s 90 m/s
Fan type Radial Axial
Fan power 22 kW 750 kW
Turbulence intensity level 0.05 0.005
(Cutoff frequency 1/3sw)a 60 6 7 kHz (U ¼ 30 m/s) —
Contraction ratio 2 7.5
Blockage ratio 20% (one-fifth scale) 1% (one-fifth scale)
a
Predicted from square wave test.

121107-2 / Vol. 137, DECEMBER 2015 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 3 Realizable k-e constants level of 5%. The closed-loop ART wind tunnel had a cross section
of 3.05 m  2.44 m. The freestream turbulence level was 0.5%
Ce1 Ce2 Ce3 rk re with a maximum freestream velocity of 90 m/s. Technical specifi-
cations of both tunnels were given in Table 2. Main differences of
Maxð0:43; g=ðs þ gÞÞ 1.9 1, Gb > 0 1 1.2 both tunnels were the turbulence intensity level and the test sec-
g ¼ S:k=e 0, Gb < 0
tion blockage level. The UURT tunnel has 20% blockage effect
for one-fifth scale model, while the ART tunnel has almost 1%
blockage effect that can be considered as blockage-free.
used to perform the aerodynamic analysis at higher yaw angles up The experimental tests were carried out at 30 m/s freestream
to 40 deg and to compare the blockage-free drag coefficients with velocity. Drag forces were measured by strain gauge balance sys-
the UURT results. The two-vehicle platoon tests were performed tem. The model was connected to the strain gauge balance system
in ART at spacings of 0, 0.5, and 1.0 car length. The detailed test through its four supports passing within holes in the floor and
plan was given in Table 1. calibrated with the standard loads applied on the model geometri-
The Reynolds dependence study for the present passenger car cal center. Calibration error was less than 1% according to the
model was performed by computational fluid dynamics (CFD) obtained calibration curve. The static mean pressures were
analysis [26]. Reynolds independency can be considered as obtained from the model centerline and the three vertical sections
reached for Re ¼ 1.88  106. At this point, the single vehicle drag by 64 channel pressure datalogger.
coefficient was calculated as 0.248, while the average drag coeffi- Uncertainty analysis was performed according to Ref. [27].
cient was 0.246 for 1.88  106  Re  12.5  106. Total uncertainty for Cp measurement was calculated as 1.13%.
The test section of the UURT wind tunnel was 0.7 m  0.6 m. Total uncertainty for Cd was calculated as 1.37% by considering
The maximum freestream velocity was 30 m/s with a turbulence strain gauge balance system accuracy.

Fig. 3 The y þ distributions on the model wall surfaces

Fig. 4 Midsection of the computational domains for b ¼ 30 deg and x/L 5 0 close spacing

Journal of Fluids Engineering DECEMBER 2015, Vol. 137 / 121107-3

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 4 Validation of the numerical method with the experi-
mental results

Scale (one-fifth) Blockage (%) CFD EXP D (%)

Cd 0 0.248 0.264 6.1

20 0.409 0.431 5.1

3 Numerical Study
A commercial CFD package STAR-CCMþ (version 6.06.011) was
used for the numerical analysis, supplementary to the experimen-
tal results. STAR-CCMþ is the standard CFD code of FIAT—
Chrysler Group and many major vehicle manufacturers. This code
was used by several authors in the past literatures [28–35]. Large
eddy simulation (LES) model would be an encouraging solution
to the problem due to its higher accuracy to reproduce unsteady
turbulence characteristics, but in turn it requires excessively large Fig. 5 UURT and ART model centerline pressure distributions
computational resources. Hanjalic [36] declared that RANS at yaw angles of 0 deg, 5 deg, and 10 deg
method will survive at least next few decades due to at least 108
cells were needed for LES method application and it requires 1
Cl ¼ (1)
higher computational loads compared to RANS method. It is real- U k
istic that the trend of RANS with k-e will continue, though hope- A0 þ As
e
fully we should expect to see more advanced RANS and URANS
(unsteady RANS) applications in the future. Roy et al. [37] dem- where A0 , As , and U* are the functions of velocity gradients.
onstrated that RANS method for simplified tractor and trailer geo- The transport equation for both the turbulence kinetic energy
metries was able to accurately predict the surface pressure “k” and the dissipation “ e ” are calculated as follows:
distributions and drag forces compared to experimental results.
Although time-dependent mean flows can be modeled using the  
@ðkuj Þ @ l @k
unsteady form of the RANS equations, a great majority of turbu- q ¼ lþ t þ Gk þ Gb  qe  YM þ Sk (2)
@xi @xi rk @xj
lent flows, including the flows around ground vehicles, can be  
economically and accurately modeled using steady RANS equa- @ðeuj Þ @ l @e e2
tions [38]. Jakirlic et al. [39] performed computational study on a q ¼ lþ t þ qCe1 Se  qCe2 pffiffiffiffiffi
@xj @xj re @xj k þ #e
1/2.5th scale realistic car model by means of RANS, URANS, and e
partially averaged Navier–Stokes (PANS) models and demon- þ Ce1 Ce3 Gb (3)
strated that the integral characteristics such as drag coefficients k
were very close to each other in RANS and URANS models but
slight differences were observed in the streamlines and lift coeffi-
cient originated from the underhood area. Cilies et al. [40] studied
RANS, URANS, and LES methods on formula 1 wheel aerody-
namics and demonstrated that RANS with realizable k-e turbu-
lence model showed good convergence as in URANS and LES
methods. Veluri et al. [41] performed joint computational and ex-
perimental studies on simplified tractor and trailer geometry by
steady RANS methods and the predicted overall drag was in good
agreement with experimental results. Considering that the lift
coefficient Cl is not the scope of this study and test model has no
engine cooling flow with smooth underbody, RANS equations
were solved incorporated the very well known SIMPLE procedure
[42]. The flow was assumed as incompressible (Mach <03) and
the second-order upwind scheme was used for steady-state calcu-
lations. The energy equation was excluded as there was no heat
transfer or temperature change. Water-tight model was obtained
by ANSA preprocessor.
The governing equations based on conservation of mass and
momentum were solved [26]. The realizable two-layer k-e model
was used for the turbulence model. Eddy viscosity model with
Boussinesq approximation was used in calculations. The realiz-
able k-e model is noticeably more accurate than the other k-e mod-
els in simulating the mean velocity components. Initial studies
have shown that the realizable model provides the best perform-
ance of all the k-e model versions for several validations of sepa-
rated flows and flows with complex secondary flow features [43].
Singh et al. [44] performed a validation of four turbulence models
(Spalart-Allmaras, standard k-e, RNG k-e, and realizable k-e) inte-
grated in FLUENT and proved that a realizable k-e model gives the
best match with the experimental results. Fig. 6 (a) CFD pressure distributions on the model centerline
In this model, Cl is not a constant any more, but rather it is at increasing yaw angles and (b) comparison of the CFD and
computed according to the experimental results at yaw angles of 10 deg and 30 deg

121107-4 / Vol. 137, DECEMBER 2015 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 where Ce1 ; Ce2 ; Ce3 ; rk ; and re are the constants and defined as in separation areas and 10–15 on the rest of the model surfaces,
Table 3. Sk and Se are the user-defined source terms and S is the demonstrating good near-wall resolution especially in separation
modulus of the mean strain rate tensor. Gk and Gb are the turbu- zones. The Wolfstein shear-driven formulation was used as
lence production terms and defined as in k-e turbulence model. default in two-layer calculations.
The two-layer approach was suggested first by Rodi [45]. In CFD analysis was performed supplementary to the experimen-
this approach, the computation is divided into two layers. In the tal results at various yaw angles up to b ¼ 40 deg, and two-vehicle
layer adjacent to the wall, the turbulent dissipation rate e and the platoon simulations at spacings of x/L ¼ 0, 0.5, and 1.0. The fine
turbulent viscosity lt are specified as functions of wall distance. mesh containing between 16 and 18  106 polyhedral cells for
The values of e specified in the near-wall layer are blended various modelings was used in the model. Polyhedral mesh pro-
smoothly with the values computed from solving the transport vides better resolutions in the flow wake region, however more
equation far from the wall. The coefficients in the models are computational time and memory are needed [32]. High number of
identical, but model gains the added flexibility of an all yþ wall meshes were used in the solution domain considering the past
treatment with the two-layer approach. The two-layer formula- studies in the literature [26] and the hardware capacity. For the
tions make no assumption about how well the viscous sublayer is near-wall treatment, 16 prism layers were used starting from
resolved. By using a blended wall law to estimate shear stress, the 0.3 mm to 30 mm with the stretch factor of 1.2. Figure 4 shows
result will be similar to the low yþ wall treatment if the mesh is the sample midsections of the domains at b ¼ 30 deg and two-
fine enough. If the mesh is coarse enough (yþ > 30), the wall law vehicle platoon at spacing of x=L ¼ 0. The model was positioned
is equivalent to a logarithmic profile [46]. Figure 3 shows that the five model length (5 L) from the inlet. The distance from the end
yþ distributions on the model surfaces in this study are < 5 in the of the model to the downstream boundary was 10 L. Total length

Fig. 7 (a) CFD pressure contours at yaw angles up to 40 deg and (b) tuft test at yaw angle of
30 deg

Journal of Fluids Engineering DECEMBER 2015, Vol. 137 / 121107-5

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 8 UURT and ART pressure distributions on vertical sections at yaw angles of 0 deg,
5 deg, and 10 deg

Fig. 9 ART pressure distributions on vertical sections at yaw angles up to 30 deg

of the domain was 16 L.The model reference area was 0.08361278 0.5%. The total drag force was calculated by the integration of
m2 . The freestream velocity was 30 m/s with the Reynolds number pressure and friction over the car model and the validation of the
of 1.88  106 . Velocity inlet was used as uniform and outlet was numerical method was performed with the wind tunnel test
considered as atmospheric conditions by using pressure outlet. results, as shown in Table 4. Obtained results were in very good
Reference density was 1.225 kg/m3. The turbulence intensity was agreement within 5–6%.

121107-6 / Vol. 137, DECEMBER 2015 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 starting from the rear end of the roof (a) toward the deck lid (b) as
indicated with dark color. The trends were similar in both UURT
and ART tests but 20% blockage effect in the UURT tests showed
 40% suction pressure increase on model roof compared to the
ART tests.
The CFD centerline pressure distributions at yaw angles up to
40 deg are presented in Fig. 6(a). The pressure distributions were
almost the same for yaw angles of 0 deg and 10 deg but significant
changes were observed for yaw angles of b  20. The static pres-
sure distributions were totally changed toward the rear part of the
model at yaw angles between 30 deg and 40 deg. This will affect
the model drag coefficient in the direction of travel as will be
discussed in Sec. 4.3. Comparisons of the experimental and the
numerical centerline pressure distributions showed very good
agreement at yaw angles of 10 deg and 30 deg in Fig. 6(b).
Figure 7 shows the CFD pressure contours at yaw angles up to
40 deg and the tuft test result at yaw angle of 30 deg. The pressure
contours at yaw angles 0 deg; 5 deg, and 10 deg did not show
significant changes on the model surface except small variations
on the front windscreen and the rear part of the roof. However,
with the increasing yaw angle at b  20, the flow characteristics
Fig. 10 (a) Schematic view of the model at yaw angle and (b) over the model surface were totally changed starting from the
typical normalized drag coefficient at increasing yaw angle front windscreen toward the downstream of the model. The flow
separation area in the rear deck lid was decreasing with increasing
yaw angle. The same results were validated by the experimental
4 Results and Discussion tuft test at b ¼ 30 deg.

4.1 Centerline Static Pressure Distributions at Yaw

Angles. Figure 5 shows the comparisons of one-fifth scale model 4.2 Cp Distributions on the Vertical Axis at Yaw
centerline static pressure distributions starting from the rear end Angles. The comparison of the UURT and the ART Cp distribu-
of the roof at yaw angles of 0 deg, 5 deg, and 10 deg in the UURT tions on the vertical sections is shown in Fig. 8. The sections were
and the ART tunnels. There were no significant pressure changes denoted as Y1, Y2, and Y3, respectively. Static pressures
in the front part of the model at yaw angels up to 10 deg and not increased with increasing yaw angle in section Y1. The suction
presented in Fig. 5. Pressure distributions at yaw angle of 5 deg peak pressure on the upper edge of front door (UURT at b
showed almost same distribution with yaw angle of 0 deg How- ¼ 0 deg) disappeared at yaw angles of 5 deg and 10 deg. This was
ever, there were small changes observed with yaw angle of 10 deg due to the increased gap between section Y1 and the tunnel side

Fig. 11 Comparison of the drag coefficients C d with increasing yaw angles b

Journal of Fluids Engineering DECEMBER 2015, Vol. 137 / 121107-7

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 12 CFD pressure contours at vehicle spacings of x=L 5 0,
0.5, and 1.0
Fig. 13 Comparison of the CFD and experimental centerline Cp
distributions at close spacings
wall by counterclockwise yawing angle. ART pressure distribu-
tions showed the same trends with UURT tests but with lower suc- 0 deg within 4.5% but the correction error slightly increased up to
tion pressures due to blockage-free effect. As the yaw angle 6.9% for yaw angle of 10 deg. ART test results showed very paral-
increases, a sharp increase in the suction pressure was observed lel results with CFD results with increasing yaw angle within 6%.
on the upper edge of the rear door frame in section Y2. In section Both experimental and numerical studies showed that the maxi-
Y3, strong suction peak was observed at yaw angle of 10 deg. mum drag coefficient was obtained at yaw angle of 35 deg. After
This was because the distance between section Y3 and tunnel side this angle, the drag coefficient started to decrease due to the total
wall was reduced resulting in an increase in local air velocity and drag force reduction caused by the significant static pressure
suction pressure. The flow structure was very complex along sec- change toward the rear part of the model (see Fig. 6(a)). Janssen
tion Y3 and strong pressure fluctuations were obtained at the bot- and Hucho [13] have obtained the same critical yaw angle for the
tom of the fender. maximum drag coefficient. Buchheim et al. [47] obtained the
The vertical section static pressure distributions for higher yaw same trend of drag increase versus yaw angle change up to 30 deg
angles (b > 10 deg) were obtained from the ART tests and given for VW 1600 notchback car.
in Fig. 9. Static pressures in section Y1 increased at increasing
yaw angles and a suction peak was observed on the upper edge of
the front door at yaw angle of 30 deg. In section Y2, the suction 4.4 Aerodynamic Analysis of Two-Vehicle Platoon
peak pressures were significantly increased after 20 deg and
became more evident with yaw angle of 30 deg. The strong nega- 4.4.1 Cp Distributions on the Model Centerlines at Various
tive static pressure increase was observed in section Y3. Pressure Spacings. Experiments were performed at spacings of 0, 0.5, and
fluctuations obtained for lower yaw angles were also observed for 1 vehicle length. Figure 12 shows CFD pressure contours on pla-
higher yaw angles at the bottom of the fender. toon members. Major changes in surface pressure were observed
on the tail car starting from the front end toward the middle of the
roof. Figure 13 shows the comparison of the numerical and the
4.3 Drag Coefficients (Cd Þ at Yaw Angles. The aerody- experimental centerline pressure distributions on platoon mem-
namic drag coefficient is defined as follows: bers versus the single car in isolation. The comparisons were in
very good agreement. Pressure distributions changed significantly
Fd starting from the rear half of the leading car (L0) toward the front
Cd ¼ (4) half of the tail car (T0) at spacing of 0 car length. As the spacing
1
qAðU1 CosbÞ2
2

where q is the reference air density, A is the model frontal area,

U1 is the freestream velocity, b is the yaw angle, and Fd is the
total drag force in the direction of travel (Fig. 10(a)) and calcu-
lated by the integration of pressure and friction over the car
model.
The drag coefficient was expected to increase with increasing
yaw angle due to the decreasing effective freestream velocity (see
Eq. (4)). However, the total drag force (Fd Þ in the model axis was
also decreasing significantly. As a result, the reduction in the
model drag coefficient was observed after the critical yaw angle
(bcrt Þ. The typical normalized drag coefficient with increasing
yaw angle can be shown in Fig. 10(b).
Figure 11 shows the comparison of the measured and the calcu-
lated drag force coefficients at increasing yaw angles. Drag coeffi-
cients of one-fifth scale model in the UURT tests were
significantly higher than the ART test results due to the higher
blockage effect. The continuity blockage correction [26] was
applied for the UURT results. The obtained corrected drag coeffi- Fig. 14 Drag coefficients of two-vehicle platoon at close
cients were in good agreement with the ART test results for b ¼ spacings

121107-8 / Vol. 137, DECEMBER 2015 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 increased toward 1 car length, the leading car roof pressure lower yaw angles were also observed for higher yaw angles at the
reached almost the same level as the vehicle in isolation and not bottom of the fender (section Y3).
plotted in the figure any more. However, there were still signifi- In the numerical part of the study, STARCCMþ code with realiz-
cant differences on the tail car (T1) engine hood surface pressure able k-e two-layer turbulence model was successfully used in the
while the rest of the model showed the same distribution with the analysis and obtained results were in very good agreement with
car in isolation. the experimental results. Both experimental and numerical studies
showed that the critical yaw angle for the maximum drag coeffi-
4.4.2 Drag Coefficients Cd of Two-Vehicle Platoon at Differ- cient was obtained at 35 deg. After this angle, the drag coefficient
ent Spacings. Figure 14 shows the experimental and the numerical started to decrease.
drag coefficient changes at various spacings of two-vehicle pla- The analysis of the two-vehicle platoon was performed in the
toon. Drag coefficient of each platoon member at various spacings ART tests at freestream velocity of 30 m/s. Cp distributions on the
was divided by the drag coefficient of the model in isolation “ model centerlines were measured at three different spacings of
Cd0 .” The leading car drag coefficient was improved almost 50% x=L ¼ 0, 0.5, and 1. The centerline pressure contours changed sig-
at spacing of 0 car length, while the drag coefficient of the tail car nificantly starting from the rear half of the leading car toward the
remained almost same. The tail car drag coefficient slightly front half of the tail car. The obtained Cp distributions were in
increased at spacing of 0.5 in both experimental and numerical good agreement with the CFD results. The leading car drag coeffi-
analyses. This was because the boundary layer on the tail car was cients were improved by 46%, 25%, and 12% at spacings of 0,
thicker than on the single car in isolation. The boundary layer sep- 0.5, and 1.0, respectively. However, the same drag improvement
aration from the leading car trunk influenced the tail car upstream was not observed with the tail car at spacings of 0 and 0.5 due to
with lower adverse pressure gradients and as a result, the separa- increasing separation area in the upstream of the tail car caused by
tion area increased resulting into a drag increase as discussed by the boundary layer separation from the leading car trunk. Tail car
Amromin [48]. Orselli [21] has observed the same drag increase showed 12% drag force improvement only at spacing of x=L ¼ 1.
for the tail car in the CFD analysis of two-MIRA platoon models The CFD results, on the other hand, showed 50%, 19%, and 7%
at spacing of 0.5 car length. The leading car drag coefficient drag improvements for the leading car at spacings of 0, 0.5, and 1,
slightly increased at spacing of 0.5, but it was still 25% under the respectively, while no drag improvement was observed for the tail
drag value of the vehicle in isolation. Finally, at spacing of 1 car car.
length, the experimental results showed that the drag coefficients Finally, authors have reached the conclusion that the STARCCMþ
of both leading and tail car were improved 12%. The CFD results, code with realizable k-e two-layer turbulence model was very suc-
on the other hand, showed less improvement compared to the cessful in modeling complex flows around a passenger car at
experimental results that were 7% for the leading car and 1% for higher yaw angles and two-vehicle platoon.
the tail car.

5 Conclusions Acknowledgment
Aerodynamic analysis has been performed on a one-fifth scale The authors wish to thank ART and TOFAS-FIAT due to their
passenger car model of FIAT Linea at various yaw angles and support for experimental tests, numerical analysis, and for the pro-
two-vehicle platoon. The aerodynamic analysis in the UURT tun- totype constructions.
nel was performed up to yaw angles of 10 deg. It was not possible
to increase yaw angle due to the UURT test section dimensional References
restrictions. The ART tunnel was used to perform the aerody- [1] Yang, Z., Schenkel, M., and Fadler, G. J., 2003, “Corrections for the Pressure
namic analysis at higher yaw angles up to 40 deg and to compare Gradient Effect on Vehicle Aerodynamic Drag,” SAE Technical Paper No.
the blockage-free drag coefficients with the UURT results. The 2003-01-0935.
[2] Aider, J., Franc, J., Beaudoin, O., and Wesfreid, J. E., 2010, “Drag and Lift
two-vehicle platoon tests were performed in ART at spacings of Reduction of a 3D Bluff-Body Using Active Vortex Generators,” Exp. Fluids,
0, 0.5, and 1.0 car length. 48(5), pp. 771–789.
Static pressure distributions on the model centerline showed [3] Gustavsson, T., and Melin, T., 2006, “Application of Vortex Generators to a
small deviations with increasing yaw angles up to 10 deg starting Blunt Body,” TRITA-AVE 13, KTH Engineering Sciences, Stockholm.
[4] Koike, M., Nagayoshi, T., and Hamamoto, N., 2004, “Research on Aerody-
from the rear part of the roof toward the end of the deck lid. Sig- namic Drag Reduction by Vortex Generators,” Mitsubishi Motors Research and
nificant pressure changes on the model centerline were observed Development Office, Technical Review No. 16.
with increasing yaw angles b  20 deg. The changes were very [5] Kourta, A., and Gillieron, P., 2009, “Impact of the Automotive Aerodynamic
evident not only in the rear part of the model but also throughout Control on the Economic Issues,” J. Appl. Fluid Mech., 2(2), pp. 69–75.
[6] Kumar, C. R., Chowdary, J. U., and Reddy, K. A., 2011, “Study of Aerody-
the centerline section. The static pressures along the model center- namic Drag Reduction Using Vortex Generators,” Int. J. Adv. Eng., Sci. Tech-
line were totally changed at yaw angles between 30 deg and nol., 7(10), pp. 181–183.
40 deg. The UURT tunnel suction pressures on the model roof [7] Ahmed, S. R., and Ramm, G., 1984, “Salient Features of the Time-Averaged
were almost 40% higher than the ART tunnel measurements due Ground Vehicle Wake,” SAE Technical Paper No. 840300.
[8] Cogotti, A., 1998, “A Parametric Study of the Ground Effect of a Simplified
to 20% blockage effect. In vertical section Y1, the suction peak Car Model,” SAE Technical Paper No. 980031.
pressure was observed on the upper edge of the front door in the [9] Heft, I. A., Indinger, T., and Adams, A. N., 2012, “Introduction of a New Real-
UURT test at yaw angle of 0 deg, while in the ART tests this suc- istic Generic Car Model for Aerodynamic Investigations,” SAE Technical Pa-
tion peak was not observed due to blockage-free effect. However, per No. 2012-01-0168.
[10] Strangfeld, C., Wieser, D., Schmidt, H.-J., Woszidlo, R., Nayeri, C., and
this suction peak disappeared with increasing yaw angle. This was Paschereit, C., 2013, “Experimental Study of Baseline Flow Characteristics for
due to the increased gap between section Y1 and the tunnel side the Realistic Car Model DrivAer,” SAE Technical Paper No. 2013-01-1251.
wall by counterclockwise yawing angle. Static pressures along [11] Docton, M. K. R., 1996, “The Simulation of Transient Cross Winds on Passen-
section Y1 have been increased with increasing yaw angle. In sec- ger Vehicles,” Ph.D. thesis, Durham University, Old Elvet, Durham, UK.
[12] Guilmineau, E., Chikhaoui, O., Deng, G., and Visonneau, M., 2013, “Cross
tion Y2, the strong suction peak pressure on the upper edge of the Wind Effects on a Simplified Car Model by a DES Approach,” Comput. Fluids,
rear door was developed at a yaw angle of 10 deg and became 78, pp. 29–40.
more evident for yaw angles 20 deg and 30 deg. In section Y3, the [13] Janssen, L. J., and Hucho, W.-H., 1973, “The Effect of Various Parameters on
suction pressures were increased as the yaw angle increased. This the Aerodynamic Drag of Passenger Cars,” Advances in Road Vehicle Aerody-
namics, British Hydromechanical Association, Cranfield.
was because of the fact that the local air velocity increased as the [14] Kohut, P., Sulitka, M., and Randa, Z., 2005, “Determination of Blockage Cor-
distance between sections Y2/Y3 and the wind tunnel side walls rection in Open-Jet Wind Tunnel,” 16th International Symposium on Transport
was reduced with yawing angle changing. Pressure fluctuations at Phenomena (ISTP-16), Prague.

Journal of Fluids Engineering DECEMBER 2015, Vol. 137 / 121107-9

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 [15] Mieller, C., 2002, “Lockheed Georgia Low Speed Wind Tunnel Honda Civic [32] Ahmad, N. E., Abo-Serie, E., and Gaylard, A., 2010, “Mesh Optimization for
Hatchback Airtab Modification Results,” Lockheed Georgia Company Wind Ground Vehicle Aerodynamics,” CFD Lett., 2(1), pp. 54–65.
Tunnel Test No. 561. [33] Heinzelmann, B., Indinger, T., Adams, N., and Blanke, R., 2012, “Experimental
[16] Tan, J., Chen, Z., Hu, Y., Parameswaran, S., Rahman, S., Gleason, M., and Sun, and Numerical Investigation of the Under Hood Flow With Heat Transfer for a
R., 2010, “Effects of Cross Wind on Sport Utility Vehicles (SUV): A Computa- Scaled Tractor-Trailer,” SAE Int. J. Commer. Veh., 5(1), pp. 42–56.
tional Study,” The Fifth International Symposium on Computational Wind [34] Regin, F. A., Manimanoharan, M., Reddy, A. B., and Nigam, P., 2013,
Engineering (CWE2010), Chapel Hill, NC, May 23–27. “Aerodynamic Analysis of Cabriolet Passenger Car: A Design Approach,” SAE
[17] Tsubokura, M., Nakashima, T., Kitayama, M., Ikawa, Y., Doh, D. H., and Paper No. 2013-01-0037.
Kobayashi, T., 2010, “Large Eddy Simulation on the Unsteady Aerodynamic [35] Zhang, Y., Ding, W., and Zhang, Y., 2014, “Aerodynamic Shape Optimization
Response of a Road Vehicle in Transient Crosswinds,” Int. J. Heat Fluid Flow, Based on the MIRA Reference Car Model,” SAE Technical Paper No. 2014-01-
31(6), pp. 1075–1086. 0603.
[18] Howell, J. P., 1993, “Shape Features Which Influence Crosswind Sensitivity,” [36] Hanjalic, K., 2005, “Will RANS Survive LES? A View of Perspectives,”
Proc. Inst. Mech. Eng., Part E, 9, pp. 43–52. ASME J. Fluids Eng., 127(5), pp. 831–839.
[19] Zabat, M., Stabile, N., Frascaroli, S., and Browand, F., 1995, “Drag Forces [37] Roy, J. C., Payne, J., and Payne, M. M., 2006, “RANS Simulations of a
Experienced by 2, 3 & 4-Vehicle Platoons at Close Spacings,” SAE Technical Simplified Tractor/Trailer Geometry,” ASME J. Fluids Eng., 128(5), pp.
Paper No. 940421. 1083–1089.
[20] Hong, P., Marcu, B., Browand, F., and Tucker, A., 1998, “Drag Forces Experi- [38] Makowski, F. T., and Kim, S.-U., 2000, “Advances in External-Aero Simula-
enced by Two, Full-Scale Vehicles at Close Spacing,” University of Southern tion of Ground Vehicles Using the Steady RANS Equations,” SAE Technical
California, California Path Research Report No. UCB-ITS-PRR-98-5. Paper No. 2000-01-0484.
[21] Orselli, E., 2006, “Computation of Drag Force on Single and Close-Following [39] Jakirlic, S., Kutej, L., Basara, B., and Tropea, C., 2014, “Computational Study
Vehicles,” M.S. thesis, Middle East Technical University, Mechanical of the Aerodynamics of a Realistic Car Model by Means of RANS and Hybrid
Engineering Department, Ankara, Turkey. RANS/LES Approaches,” SAE Int. J. Passenger Cars Mech. Syst., 7(2), pp.
[22] Rajamani, G. K., 2006, “CFD Analysis of Air Flow Interactions in Vehicle 559–574.
Platoons,” M.S. thesis, School of Aerospace, Mechanical and Manufacturing [40] Cilies, J. A., Issakhanian, E., Jimenez, J., and Iaccarino, G., 2012, “An Aerody-
Engineering, RMIT University, Melbourne, Australia. namic Investigation of an Isolated Stationary Formula 1 Wheel Assembly,”
[23] Schito, P., and Braghin, F., 2012, “Numerical and Experimental Investigation ASME J. Fluids Eng., 134(12), p. 021101.
on Vehicles in Platoon,” SAE Int. J. Commer. Veh., 5(1), pp. 63–71. [41] Veluri, P. S., Roy, C. J., Ahmed, A., Rifki, R., Worley, J. C., and Rectenwald,
[24] Tsuei, L., and Savas, O., 2001, “Transient Aerodynamics of Vehicle Platoons B., 2009, “Joint Computational/Experimental Aerodynamic Study of a Simpli-
During In-Line Oscillations,” J. Wind Eng. Ind. Aerodyn., 89(13), pp. fied Tractor/Trailer Geometry,” ASME J. Fluids Eng., 131(8), p. 081201.
1085–1111. [42] Patankar, S. V., 1980, Numerical Heat Transfer and Fluid Flow, McGraw-Hill,
[25] Watkins, S., and Vino, G., 2008, “The Effect of Vehicle Spacing on the Aero- New York.
dynamics of a Representative Car Shape,” J. Wind Eng. Ind. Aerodyn., [43] Shih, T., Liou, W. W., Shabbir, A., Yang, Z., and Zhu, J., 1995, “A New k-e
96(6–7), pp. 1232–1239. Eddy Viscosity Model for High Reynolds Number Turbulent Flows,” Comput.
[26] Altinisik, A., Kutukceken, E., and Umur, H., 2015, “Experimental and Numeri- Fluids, 24(3), pp. 227–238.
cal Aerodynamic Analysis of a Passenger Car: Influence of the Blockage Ratio [44] Singh, S. N., Rai, L., Puri, P., and Bhatnagar, A., 2005, “Effect of Moving Sur-
on Drag Coefficient,” ASME J. Fluid Eng., 137(8), p. 081104. face on the Aerodynamic Drag of Road Vehicles,” Proc. Inst. Mech. Eng., Part
[27] Moffat, R. J., 1988, “Describing the Uncertainty in Experimental Results,” Exp. D, 219(2), pp. 127–134.
Therm. Fluid Sci., 1(1), pp. 3–17. [45] Rodi, W., 1991, “Experience With Two-Layer Models Combining the k-E
[28] Williams, J., Quinlan, W. J., Hacket, J. E., Thompson, S. A., Marinaccio, T., Model With a One-Equation Model Near the Wall,” AIAA Paper No. 91-
and Robertson, A., 1994, “A Calibration Study of CFD for Automotive Shapes 0216.
and CD,” SAE Technical Paper No. 940323. [46] Suria, O. V., Testa, E., Repici, G., Peraudo, P., and Maggiore, P., 2011, “A
[29] Gaylard, A. P., Baxendale, A. J., and Howell, J. P., 1998, “The Use of CFD to PEM Fuel Cell Laminar and Turbulent Models Comparison, Aiming at Identi-
Predict the Aerodynamic Characteristics of Simple Automotive Shapes,” SAE fying Small-Scale Plate Channel Phenomena: A Mesh Independent Configu-
Technical Paper No. 980036. ration,” SAE Technical Paper No. 2011-01-1177.
[30] Connor, C., Kharazi, A., Walter, J., and Martindale, B., 2006, “Comparison of [47] Buchheim, R., Unger, R., Carr, G. W., Cogotti, A., Carrone, A., Kuhn, A., and
Wind Tunnel Configurations for Testing Closed-Wheel Race Cars: A CFD Nilsson, L. U., 1980, “Comparison Tests Between Major European Automotive
Study,” SAE Technical Paper No. 2006-01-3620. Wind Tunnels,” SAE Technical Paper No. 800140.
[31] Mokhtar, W. A., 2008, “Aerodynamics of High-Lift Wings With Ground Effect [48] Amromin, E. L., 2013, “Vehicles Drag Reduction With Control of Critical
for Racecars,” SAE Technical Paper No. 2008-01-0656. Reynolds Number,” ASME J. Fluids Eng., 135(10), p. 101105.

121107-10 / Vol. 137, DECEMBER 2015 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use