Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
International Journal of Automotive Technology, Vol. 20, No. 5, pp. 9971008 (2019)
DOI 10.1007/s1223901900948
Copyright © 2019 KSAE/ 11014
pISSN 12299138/ eISSN 19763832
997
EFFECT OF CROSSWINDS ON THE AERODYNAMICS OF TWO
PASSENGER CARS CROSSING EACH OTHER
Ahmad Hammad1), Tao Xing1)*, Ahmed Abdel-Rahim2), Vibhav Durgesh1) and John C. Crepeau1)
1)Department of Mechanical Engineering, University of Idaho, Moscow, ID 83844-0902, USA2)Department of Civil and Environmental Engineering, University of Idaho, Moscow, ID 83844-1022, USA
(Received 2 July 2018; Revised 9 February 2019; Accepted 25 February 2019)
ABSTRACTThe impact of aerodynamics on vehicle safety during crossing of passenger cars is investigated, in the absence
and presence of 30o crosswind. Three-dimensional, unsteady computational fluid dynamics (CFD) simulations were used to
simulate these maneuvers. The vortical structures surrounding one car in the case without crosswind were analyzed,
establishing the connection between force and moment fluctuations pre-interaction and the shedding frequency of these
vortices. The forces and moments acting on a passenger car during a crossing maneuver may change by up to 43 %, with the
maximum change associated with the windward car in the presence of crosswind. However, the duration of this increase in
forces is at most 0.01 s, which will not affect the stability of vehicles under normal conditions. The presence of crosswind
increased the rate of fluctuation of forces and moments. Wind tunnel experimental results are in good agreement with the
simulations, and the data available in literature. The analysis results do not show the necessity of enacting new safety policies
on highways, but future parametric studies are needed to fully investigate the impact of different crosswind speeds and
directions, the impact of discrepancy in vehicles sizes, and different vehicle lateral separating distances during crossing and
overtaking.
KEY WORDS : Car-car crossing, Vehicle aerodynamics, Computational fluid dynamics, Vortical structures
NOMENCLATURE
CFD : computational fluid dynamics
DES : detached-eddy simulation
DDES : delayed detached-eddy simulation
IDDES : improved delayed detached-eddy simulation
LES : large-eddy simulation
RANS : Reynolds-averaged Navier-Stokes equations
SUBSCRIPTS
x, y, z : coordinates
d : drag (in Cd)
1. INTRODUCTION
Rural roads in the US, compared to urban roads, see a
proportionately higher number of fatalities with respect to
traffic volume. This risk is most apparent on two-lane,
undivided rural highways shared by fast-moving cars and
trucks, especially in twisty sections of these roads where
winds are more variable in terms of speed and direction.
These vehicles impart aerodynamic forces and moments on
one another, and the significance of these forces and
moments in affecting vehicle stability has increased with
the trend towards lighter, more fuel-efficient vehicles. In an
overtaking or crossing maneuver on a highway involving
two vehicles, the flow fields around the two vehicles
interact generating transient aerodynamic forces that can
affect car handling and stability (Corin et al., 2008). When
the relative size difference between the two vehicles is
large (e.g., a car and a truck), these forces increase on the
smaller vehicle, and increase further under the influence of
crosswinds, (Howell et al., 2014). A vehicle is more stable
when its geometric center, center of gravity and stagnation
point are all in line. Under crosswind conditions, the air
flow around the vehicle becomes asymmetric and the
stagnation point shifts towards the direction of crosswind,
affecting the stability of the vehicle (Altinisik et al.,
2015b). As the computational power of commercially
available computers doubles approximately every two
years (famously observed and predicted by Gordon Moore
(Moore, 1998)), high quality transient computational fluid
dynamics (CFD) simulations of the complex interactions
between moving vehicles were not feasible until the late
2000s. In a transient CFD study on two vehicles crossing
each other using k-ε turbulence model (Zhang et al., 2010),
however, it is not clear if the geometries used are two-
dimensional (2D) or three-dimensional (3D). The study did
not include experimental validation of the results.
Instrumented experiments on a proving ground measured
the surface pressure distribution of a passenger car during*Corresponding author. e-mail: [email protected]
998 Ahmad Hammad et al.
both a passing and an overtaking (Kremheller, 2015).
These experiments showed that the asymmetric pressure
distribution induced by these maneuvers influenced lateral
acceleration and yawing rates, being pronounced in
vehicles with a larger frontal area. The change in surface
pressure increased when the lateral distance between
vehicles was decreased. A good correlation between CFD
and experimental data was found, however, only a brief
explanation of the comparison between CFD results and
experimental measurements was presented in this paper.
Previous studies in this area appear to have one or more
of the following limitations: (a) using 2D geometries; (b)
low grid resolutions; and (c) inappropriate turbulence
models that are too dissipative, destroying details of the
important vortical structures in the flow.
2. OBJECTIVE AND APPROACH
This study uses 3D CFD models to investigate variations in
forces and moments associated with car-car crossings, with
and without 30o crosswind (the drag coefficient is not
sensitive to yawing angle change after a critical yaw angle
of 25 ~ 30o (Altinisik et al. (2015b) influenced by vehicle
shape), influenced by vehicle shape, with complimentary
experiments on a single car model in the wind tunnel, to
validate the CFD results. The vortical structures in the
flowfield around the car were analyzed, and their
relationship with forces and moments were investigated.
To address the shortcomings mentioned in the literature
review section, the CFD simulations of vehicle interaction
used: (a) 3D, model scale geometries of vehicles; (b) fine
grids with up to 5.8 million grid points; and (c) Improved
Delayed Detached Eddy Simulation (IDDES) turbulence
model (Gritskevich et al., 2012). IDDES is a hybrid
RANS/LES (Reynolds-Averaged Navier-Stokes equations/
Large-Eddy Simulation) model that accurately resolves
small flow structures near the solid boundaries of the
vehicles and larger flow structures further away from the
boundaries, while avoiding grid induced separation of the
boundary layer.
3. METHODOLOGY
All simulations and experiments involving the passenger
car were run at 1 : 25 scale.
3.1. Car Geometry
A generic passenger car model (Figure 1) was designed
based on the average dimensions of mid-size family sedans
in the US market (e.g., Toyota Camry, Honda Accord, Ford
Fusion, etc.), in terms of overall length, height, width, track
width and wheelbase length. Other important proportions
(cabin length, beltline height, etc.) are also in line with
typical mid-size sedans. The vehicle lacks a wheel gap
between the wheels and the fenders, and wheel motion was
not considered in this study.
3.2. Experimental Methodology
The purpose of the experiments was to measure the drag
forces on a single car model. The drag force was used to
validate a 3D unsteady simulation of the same geometry.
The subsonic wind tunnel in the Department of Mechanical
Figure 1. True dimensions of car model (model dimensions
in parentheses). All dimensions in meters.
Figure 2. Car model setup in wind tunnel. Air is moving in
the negative X-direction. Schematic shows the orientation
of force (red) and moment coefficients (blue) obtained in
the simulations.
EFFECT OF CROSSWINDS ON THE AERODYNAMICS OF TWO PASSENGER CARS CROSSING EACH OTHER 999
Engineering at The University of Idaho was used to
conduct the experiment, and the vehicle model was 3D-
printed using the department’s facilities, with the necessary
attachments manufactured in the machine shop. The 3D-
printed car model is accurately sized to within 1 % in all
dimensions. Figure 2 shows the details of the experimental
set-up. The force balance uses strain gauges and the data is
fed to a Newport Electronics INFS4 Strain Gauge meter.
The balance was calibrated using standard weights.
In each experimental run, a data acquisition device
(National Instruments USB-6002) recorded the drag force
after the air velocity stabilized inside the wind tunnel. Each
data sample was 5 seconds long at 1,000 Hz. Force data
was averaged for each run, and the standard deviation
calculated. The wind tunnel came to a complete stop before
commencing the next run. Air velocity inside the wind
tunnel was changed by adjusting the variable-speed drive
frequency. The frontal area of the car (0.004408 m2) was
used to calculate the drag coefficient.
3.3. CFD Methodology
3.3.1. Simulation design and grid topology
Commercial mesh generating software, Pointwise
(V17.2R3), was used to generate the mesh. The dimensions
of the vehicles were scaled down to 4 % of their original
sizes. An unstructured grid was used on the surfaces of the
vehicles. The volume immediate to the vehicle surfaces
was filled with structured, triangular-aspect prisms to
provide enough grid points to resolve the boundary layer
(Figure 3). The first grid point away from the wall is at a
distance of 0.0001 m, for an approximate y+ value of 2.
The volume between these cells and the inner surfaces of
a cuboid containing each vehicle is populated with
unstructured, pyramid-shaped cells. The high-resolution
dynamic mesh zones behind and in front of each vehicle is
populated by structured cubes with the dimensions (X × Y
× Z) 0.004 × 0.004 × 0.004 m (see Figure 4 for a vertical
slice through the car mesh). The number of grid points in
each mesh is sufficient to resolve all small and large-scale
flow structures. A grid dependency study involving 3
single-car meshes (the fine grid resolution is identical to
the two-car simulations used in this study) showed
monotonic convergence on the grid triplet (refer to Section
4.1 for details).
A large, rectangular, low-resolution zone encloses the
dynamic mesh zones. The lateral distance separating the
two vehicles is approximately half-lane width (1.6 m). The
velocity of each vehicle is 30 m/s (67 mph), moving in
opposite directions parallel to the X-axis.
3.3.2. Numerical method
The commercial CFD software, ANSYS Fluent 17.1 &
17.2 were used in the simulations, using a least squares
cell-based discretization method, and a transient solver.
The SIMPLE algorithm was used for the pressure-velocity
coupling. A second order scheme was used for pressure
discretization. The momentum equations were discretized
using bounded central differencing. An implicit second
order scheme was used for temporal discretization. The
overall simulation design is shown in Table 1.
3.3.3. Dynamic mesh method
The sliding mesh model in Fluent was used to simulate the
relative motion between the vehicles involved. This model
is a special case of general dynamic mesh motion wherein
the nodes move rigidly in a given dynamic mesh zone, i.e.,
all the boundaries and the cells of a given mesh zone move
together in a rigid-body motion. In this situation, the nodes
of the mesh move in space (relative to the fixed, global
coordinates), but the cells defined by the nodes (the zones
behind and in front of the vehicles) do not deform.
Furthermore, mesh zones moving adjacent to one another
are linked across non-conformal interfaces (the zones
separating the imaginary “tunnels” in which the vehicles
move relative to one another, and the zones separating
these tunnels and the larger zone enclosing them,) thatFigure 3. Triangular prisms in the boundary layer close to
the car surface.
Figure 4. Slice through the mesh showing the unstructured
grid surrounding the car.
Table 1. Simulation matrix.
Simulation ReNo. of points
Initial separating distance (m)
No crosswind 4 × 105 5.2 × 106
2.787 (14.26 car lengths)Crosswind 4.5 × 105 5.8 × 106
1000 Ahmad Hammad et al.
allows for fluid flow from one mesh to the other (ANSYS,
2016).
3.3.4. Turbulence model
In all simulations, the flow around the vehicles involved
was modeled using Improved Delayed Detached-Eddy
Simulation “IDDES” (Shur et al., 2008). For 3D, high
Reynolds number turbulent flows, using RANS for
analysis results in the loss of vortical structure details,
especially the smaller, high-frequency eddies. DES
(Detached-Eddy Simulation, (Spalart et al., 1997)) and its
modification, DDES (Delayed DES (Spalart et al., 2006))
are models developed with the aim of accurately predicting
massively separated flows at a manageable computational
cost (Strelets, 2001). DES combines RANS in the attached
boundary layers, which are populated with small eddies
that have a length-scale much less than the boundary layer
thickness (δ), with LES in the complex, massively
separated regions away from the wall, where large,
unsteady turbulent scales play a dominant role. DES can
produce artificial separation of the boundary layer, in an
effect termed Grid Induced Separation, when the switch
from RANS to LES happens inside the boundary layer, due
to a refinement in the grid in that region (Menter et al.,
2003). This occurs when geometric features demand a fine
grid inside the boundary layer, resulting in wall-parallel
grid spacings in the upper regions of the boundary layer
that are smaller than the distance from the wall. This also
occurs when the boundary layer is thicker as it nears
separation. This motivated the development of DDES to
avoid the switch from RANS to LES in regions inside the
boundary layer which are not fine enough to resolve
velocity fluctuations (LES content) (Spalart et al., 2006).
Further improvement of DDES was proposed, the
IDDES model, combining DDES with an improved
RANS-LES hybrid model that enables Wall-Modeled LES
(WMLES) when unsteadiness is present in the boundary
layer (Shur et al., 2008). The presence of obstacles (for
example: sharp, backwards facing angles in car
geometries) can induce this unsteadiness. WMLES
resolves the inner-most part of the boundary layer using
RANS, switching over to LES formulation when the grid
spacing is sufficient to resolve the local scales. The IDDES
model surpasses both DES and DDES in simulating mixed
flows with both attached and separated regions.
3.3.4.1. Governing equations
The incompressible continuity and momentum equations
for Cartesian coordinates read as (Xing, 2014):
(1)
(2)
where V is the velocity vector, is the density, is the
dynamic viscosity, p is the pressure and g is gravitational
acceleration.
RANS equations in Cartesian tensor form are given by:
(3)
(4)
where and and are the mean and
fluctuating velocity components, respectively.
IDDES is a hybrid RANS-LES model, based on the
BSL/ST (baseline/standard) model (Gritskevich et al.,
2012).
SST IDDES formulation: Transport equations for k and
ω:
(5)
(6)
where k is the turbulent kinetic energy, ω is the specific
dissipation rate, Pk is the production term, and F1 is one of
the two SST blending functions.
3.3.5. Boundary and initial conditions
The faces of the large rectangular zone surrounding the
dynamic mesh zones have different boundary conditions
depending on the case (Figure 5, Table 2). Car 1 moves in
the positive X-direction.
• Constant pressure-outlet boundaries are defined by the
following conditions:
Gauge pressure = 0
Backflow turbulent intensity = 1 %
Backflow turbulent viscosity ratio = 0.1
( ) 0 V
2 T( ) ( ) ( )pt
V V V V g
i
i
( ) 0u
t x
i i j
j
ji
ij i j
j j j
k
ki j
( (
2(
)
3
)
)
u u u
t x
uu uu u
x x x x x x
i i iu u u
iu
iu
3
k t k
IDDES
( ) ( )k k
Uk k Pt l
2
ω t
2
1 k k
t
( ) (
2( )1
)Ut
kF P
Figure 5. Boundary conditions.
EFFECT OF CROSSWINDS ON THE AERODYNAMICS OF TWO PASSENGER CARS CROSSING EACH OTHER 1001
• In crosswind cases, two pressure-outlet boundaries are
replaced by velocity inlets, which have the following
conditions:
Turbulent intensity = 1 %
Turbulent viscosity ratio = 0.1
Velocity = 17.32 m/s, in the Z-direction
• All cases are initialized with the following conditions:
Time-step size: 0.0001 s
Initial separating distance between vehicles: 2 m
Initial air velocity inside domain: 0 m/s
3.3.6. Convergence criteria
All the simulations were determined to be converged when
the residuals of the continuity; x, y and z velocity; k; and ω
equations are lower than 1e-05.
3.3.7. CFD analysis method
In each simulation, forces and moments acting on the
vehicle surfaces were recorded at each time step. These
forces and moments are non-dimensionalized to form
coefficients using the following formulas:
(7)
, (8)
where F and M' are the force (N) and moment (Nm),
respectively, is the air density (1.1 kg/m3), u is the
relative velocity magnitude of air with respect to the
vehicle (m/s), A is the frontal area of the vehicle (m2), and L
is the total length of the vehicle (m). Fd is the magnitude of
the drag force, while Fy and Fz are the projected force
components in the y and z directions, respectively. Cd, Cy
and Cz are the drag, lift and side-force coefficients,
respectively. Mx, My and Mz
are the roll, yaw and pitching
moment coefficients, respectively.
Fast Fourier transform analysis was conducted on the
time histories of forces to identify their power spectrum.
The frequencies of different vortical structures can then be
associated with the FFT analysis. The flow around the
vehicles, and in particular the shedding vortices, where
visualized using the isosurfaces of Q-Criterion (Hunt et al.,
1988): in an incompressible flow, a vortex is a connected
fluid region with a positive second invariant of u, where
the vorticity magnitude is greater than the magnitude of
rate of strain:
(9)
where Q represents the local balance between shear strain
rate and vorticity magnitude. This criterion also requires
the pressure to be lower than the ambient pressure in the
vortex (Kolář, 2007; Holmén, 2012).
4. RESULTS AND DISCUSSION
4.1. Solution Verification and Experimental Validation
Wind tunnel experimental results for the drag coefficient
(Cd) of a single passenger car are clustered around an
average value of 0.349. The variability in each
experimental test is small, with a maximum standard
deviation of about 1 % from the average value. The
uncertainty in each drag force measurement after wind
tunnel calibration is around 12 %.
The mesh size (fine mesh) and time-step size used in the
simulations were determined using rigorous verification
(Xing and Stern, 2010, 2011) & validation (Xing et al.,
2008) for a single, stationary car (resembling conditions in
the wind tunnel). The results are summarized in Table 3.
Verification is used to estimate the numerical errors by
systematically refining the grid spacing and time-step sizes
on three meshes (coarse, medium and fine) using a
refinement ratio in three directions. As shown by
Figure 6, the grid triplet demonstrates monotonic
convergence and therefore the fine mesh (5.8 million) was
adopted for all the car-car crossing simulations. In Table 3,
RG defines the convergence ratio which is the ratio of the
solution differences for medium-fine and coarse-medium
solution pairs. PG is the observed order of accuracy, UG is
the numerical uncertainty estimate as a percentage of
experimental data, and |E| is the absolute relative error
between the fine mesh solution and experimental data. UV
is the validation uncertainty and UD is the experimental
uncertainty. UG is only 0.633 % based on the experimental
d,y,z
d,y,z 2
2 F
Cu A
x,y,z
x,y,z 2
2 M
Mu AL
2
i, j i, j j,i
2 2
i, j j,i
1(
2
1
)
)1( 0
2 2
Q u u u
u u S
‖ ‖ ‖ ‖
2
Table 2. Boundary conditions for simulation cases.
Faces Boundary conditions
ABCD All cases Slip wall
EFGH All cases Slip wall
ABFENo crosswind Constant pressure outlet
Crosswind Slip wall
DCGHNo crosswind Constant pressure outlet
Crosswind Slip wall
ADHE, BCGF
No crosswind Constant pressure outlet
CrosswindConstant velocity inlet
(17.32 m/s to simulate 30o crosswind)
Table 3. Verification & validation (V&V) study for Cd. All
percentages are based on experimental data D.
RG PG UG (%) |E| (%) UV
(%) UD
(%)
0.111 3.169 0.633 11.174 12.017 12
1002 Ahmad Hammad et al.
data. After solution verification is completed, validation is
performed by comparing |E| (11.174 %) with the validation
uncertainty UV (12.017 %). Since |E| < UV, the CFD model
was validated.
The aerodynamic blockage inside the wind tunnel was
corrected using the following proposed formula (Sykes,
1973):
(10)
where w is the blockage correction factor, m is an empirical
constant (m = 1.22 (Stafford, 1981)) B is the blockage ratio
and defined as:
(11)
where Am and Aw are the model frontal area and wind tunnel
cross-sectional area respectively. The corrected drag
coefficient is defined using the formula:
(12)
where Cdc and Cdm are the corrected and measured drag
coefficients respectively. The blockage correction method
used by Sykes is one of three methods that give the most
accurate results (Altinisik et al., 2015a) and applicable for
blockage ratios less than 5 %. The current blockage ratio,
B, of the passenger car inside the wind tunnel is 2 %.
The CFD result on the fine mesh is 11 % lower than the
experimental data. The discrepancy between CFD
simulation result and experimental results can be explained
by: a) the effect of skin friction inside the wind tunnel, and
b) the nature of the simulation design. The two vehicles are
separated by a certain separating distance at the start of
each simulation. This distance is necessary to give the
airflow around the vehicle a chance to fully develop, and
the forces and moments to stabilize, before interaction
between the vehicles start. However, this distance is limited
by the computational affordability of CFD simulations.
Even though forces and moments in the simulation are
stable before interaction starts, it appears that there is still a
small aerodynamic effect of each vehicle on the other,
reflected in the lateral shifting of the stagnation point on
each vehicle by a small distance.
4.2. CFD Results
4.2.1. Two cars crossing without crosswind (Case 1)
4.2.1.1. Before interaction between the two vehicles starts
The vortical structures in the airflow surrounding Car 1 can
be seen in Figure 7. The main vortices and their
frequencies are defined in Table 4. The vortices are defined
such that “Right” and “Left” refer to the right and left sides
of the car along the direction of travel, respectively.
Vortices are abbreviated by T, B, W, and S, standing for
“Top”, “Bottom”, “Wake” and “Side”. Vortices T1, T2 and
T3 (Figure 7 (a)) are formed along the sharp roofline along
each side of the car. They are visually dominant, but they
are stable before the interaction. Along each side, these 3
vortices weaken towards the end of the roofline, forming 2
top vortices in the wake of the car, T4 and W (Figure 7 (b)).
Along the bottom surface of the car, a vortex, B, sheds off
the entire width of the rear bumper (Figure 7 (c)).
The asymmetry of the periods of similar vortices located
on the left and right sides (S, right and left) of the car is
explained by the location of the stagnation pressure point
(Figure 8). This point is slightly shifted away (on the
horizontal XZ-plane) from the symmetry plane of the car,
indicating that the incoming airflow impacts the car at an
angle that is not quite perpendicular. As shown in Figure 9,
spectral analysis of the drag coefficient during the pre-
interaction period (0.015 ~ 0.045 s in the simulation),
indicates dominant frequencies at 100, 170, 200 and 300
Hz, each corresponding to the shedding frequency ( f ) of
one or more vortices. Not all vortices influence the
1w mB
m
w
AB
A
dc dmC wC
Figure 6. Solution verification and comparison with
experimental data for a stationary, single car simulation
using simulations on 3 mesh sizes.
Table 4. Main vortices shedding frequency before
interaction between vehicles.
Vortex Frequency (Hz)
T4 163, 192
B 200
W (left) 159
W (right) 294
S (left) 435
S (right) 170
EFFECT OF CROSSWINDS ON THE AERODYNAMICS OF TWO PASSENGER CARS CROSSING EACH OTHER 1003
fluctuations in drag coefficient. Vortices T1, T2 and T3 are
longitudinal and continuously shedding, and the side
vortices (S) are too weak to influence the drag coefficient.
The rear top center vortex (T4) is influenced by the two
inner roofline vortices (T1 & T3). This results in a
shedding pattern dominated by two frequencies (163, 192
Hz). At 0.0276 s (Figure 10), a massive vortex spanning
the width of the car starts shedding, while further
downstream of it a small helical vortex is present. This
pattern is repeated starting at 0.0328 s.
The other two rear top vortices have different shedding
frequencies, (159 and 294 Hz for the left and right sided
vortices, respectively).
Figures 11 and 12 show the progression of each vortex,
shedding from the top edge of the trunk, then joined by the
Figure 7. Vortical structures using isosurfaces of Q-
criterion.
Figure 8. Horizontal slice through Car 1 highlighting the
shifted position of the stagnation point at the front of the
car.
Figure 9. Two cars crossing without crosswind: FFT
analysis of drag coefficient before vehicle interaction.
Figure 10. Rear center vortex (Figure 7 (b)), shown by
isosurfaces of Q-criterion = 5e6. The shedding frequency
(corresponding to fluctuations in drag coefficient)
alternates between ~ 170 and ~ 200 Hz.
1004 Ahmad Hammad et al.
longitudinal vortex T2 after it breaks down to form helical,
counter-rotating vortical structures.
Figures 13 and 14 show the asymmetrical side helical
vortices. The left-side vortex has a much higher frequency
(435 Hz) compared to the right-side vortex (~ 170 Hz).
This left-side vortex also contributes to the formation of the
left-side rear vortex.
4.2.1.2. During interaction between the two vehicles
From this point forward, the distance between the two cars
will be denoted by D, where D is the horizontal distance
along the X-axis between two vertical planes tangential to
the front of each car, measured in car lengths (the length of
each car is 0.192 m). Initially, D = 14.26 car lengths and
decreases when the two cars are approaching each other
and becomes negative when the two vertical planes pass
each other, i.e., the two cars leaving each other.
The interaction doesn’t start until D ~ 1.8 car lengths (t ~
0.04 s) and continues after crossing until D ~ − 4.5 (t = 0.06
s). This interaction duration is evident in the fluctuation of
the drag force (Figure 15 (a)) and side force coefficients for
both cars (Figure 15 (b)), before returning to the pre-
interaction average values.
Figure 11. Rear left vortex, shown by isosurfaces of Q-
criterion = 5e6. The shedding frequency of this vortex
(~ 150 Hz) is half that of the equivalent vortex at the other
side of the car.
Figure 12. Rear right vortex, shown by isosurfaces of Q-
criterion = 5e6.
Figure 13. Helical right side vortex, shown by isosurfaces
of Q-criterion = 3e5. The vortex starts forming at the edge
of the front bumper.
Figure 14. Helical left-side vortex.
Figure 15. Two cars crossing without crosswind: time
histories of (a) Drag force coefficient; (b) Side force
coefficient.
EFFECT OF CROSSWINDS ON THE AERODYNAMICS OF TWO PASSENGER CARS CROSSING EACH OTHER 1005
One noticeable observation is the alternation of these
force fluctuations from one set of force coefficients to the
other: at D = − 0.59 (t = 0.0475s (P1)), the drag coefficient
of both cars decreases to its minimum value (almost 20 %
less than average), at the same moment, the side-force
coefficient is at its average value. This coefficient climbs
up to its maximum value at D = − 1.21 (t = 0.0495 s (P2))
when the drag coefficient is at its average value (Cd = 0.32).
Another peak for both drag and side force coefficients is
observed at D = − 4.34 (t = 0.0595 s (P3)). The pressure
distribution on the sides of each car during the crossing
contributes to the fluctuations in the side-force coefficient.
At the maximum side force value (P2), the low-pressure
region between the two cars reaches its maximum extent
and strength (Figure 16). This low-pressure region grows
as the two cars approach each other, pulling the two cars
towards each other and contributing to the fluctuations in
side force (Figure 15 (b)). The main vortical structures
around the two vehicles do not change in shape or direction
during the interaction, but they lose their periodicity
(Figure 17).
The lift coefficient (Figure 18 (a)) hovers close to zero
for both cars and does not show significant variation during
interaction compared to its average fluctuation frequency.
The roll coefficient for both cars (Figure 18 (b)) is small,
which is expected, as the car geometry is symmetrical
around the vertical plane passing through its centerline,
consequently, the streamwise vortical structures are
symmetrical around that same plane. The vortices are
oriented along the longitudinal axis, not the vertical axis:
there is little variation between the vertical forces acting on
each side of the car to cause a significant roll moment. Still,
within that small magnitude of roll moment (less than
Figure 16. Horizontal slice at y = 0.02 showing the
pressure distribution around the cars at maximum side
force (P2 in Figure 15).
Figure 17. Overall flow pattern surrounding the cars and
pressure distribution on car surfaces during interaction
without crosswind.
Figure 18. Two cars crossing without crosswind: time
histories of (a) Lift; (b) Roll; (c) Yaw; (d) Pitch
coefficients.
1006 Ahmad Hammad et al.
0.0075 for either car), the interaction causes a spike at D =
− 1.21 (t = 0.0495 s) before returning to normal.
The yaw moment coefficient (Figure 18 (c)) does not
show a periodic pattern before the interaction starts,
varying between 0 and 0.075 for either car, which can be
attributed to the large number of symmetrical, streamwise
vortical structures of varying shedding frequencies on each
side of the car, with no particular vortex having a dominant
effect over the others. Again, there is a small spike at D =
− 0.12 (t = 0.0460 s), but this peak is in the same order of
magnitude of the largest fluctuation of the coefficient
before the interaction starts. There is no effect on the pitch
moment coefficient (Figure 18 (d)).
4.2.2. Two cars crossing with 30-degree crosswind (Case 2)
4.2.2.1. Before interaction between the two vehicles starts
The vortical structures in the airflow surrounding Car 1 can
be seen in Figure 19. Compared to the case without
crosswinds, the top vortices T1 (left & right) and the
downwind side vortex T3 (right) disappear. Side vortex S
only appears on the downwind (right) side of the vehicle.
All the other vortices keep their shape and structure, tilted
30o to the X-axis.
The mean drag coefficient of both cars before the
crossing is 0.41, a 34 % increase over the same car’s drag
coefficient without crosswind (Figure 20 (a)). The car is
less streamlined when the airflow is impacting it at an
angle, and an increase in drag force is expected. The lift
coefficient shows two stages, the first with a mean value of
0.14 between D = 9.57 and D = 4.88 (0.15 s to 0.03 s in the
simulation), and a second phase with a mean value of
− 0.10 between D = 4.88 and D = 1.76 (0.03 s to 0.04 s in
the simulation) (Figure 20 (b)). The side force coefficient
stabilizes later than the other two force coefficients. The
mean value for both cars is 1.47. This is an indication of the
stronger side forces acting on the vehicles by the crosswind
(Figure 20 (c)), when compared to the side forces
experienced in the similar case without crosswind.
4.2.2.2. During interaction between the two vehicles
As the vehicles approach each other, one car experiences a
jump in its drag coefficient to 0.54, a 35 % increase (Car 2)
(Figure 20 (a)). The other car (Car 1) is momentarily
shielded from the effect of crosswind, and the drag forces
acting on it drop by 15 % compared to the pre-interaction
mean value. The windward car (Car 2) also experiences the
largest deviation of side forces. Its side force coefficient
changes from a mean value of ~ 1.47 to 0.84 (Figure 20
(c)). The side force coefficient of the car on the leeward
side drops to 1.23. This pattern is repeated with the lift
Figure 19. Vortical structures using isosurfaces of Q-
criterion.
Figure 20. Two cars crossing with 30-degree crosswind:
time histories of (a) Drag; (b) Lift; (c) Side force; (d) Yaw;
(e) Roll; (f) Pitch coefficients.
EFFECT OF CROSSWINDS ON THE AERODYNAMICS OF TWO PASSENGER CARS CROSSING EACH OTHER 1007
coefficient: the windward car’s coefficient changes to
− 0.45, with no appreciable change for the leeward car.
The yaw moment coefficients (Figure 20 (d)) show the
same trend as the force coefficients: the windward car
shows a greater deviation from average (− 0.17 to − 0.10)
compared to the leeward car (0.20 to 0.18). Both roll and
pitch moment coefficients show no appreciable change
from the pre-interaction average (Figure 20 (e) and (f)).
5. CONCLUSION AND FUTURE WORK
Unsteady, three-dimensional CFD simulations were
performed to investigate the aerodynamic effects of vehicle
interaction in two cases: (1) two passenger cars crossing
each other without crosswind, and (2) two passenger cars
crossing each other with 30-degree crosswind. The
simulations investigated the change in forces and moments
associated with these vehicle encounters. Wind tunnel
experimental measurements of the drag coefficient of a
single car model without crosswind were used to validate
the simulation. The vortical structures surrounding one car
in case (1) were analyzed, establishing the connection
between force and moment fluctuations pre-interaction and
the shedding frequency of these vortices. A total of twelve
vortices were identified and analyzed. Ten vortices are
pairs on each side of the car: two vortices along the car
doors, six on the top, and two in the wake. The other two
vortices are central on the top and bottom of the car. The
six vortices on the top of the vehicle are stable and have no
effect on the unsteadiness of the forces and moments. The
two vortices along the doors are unstable but have minimal
effect on the forces and moments. The two vortices in the
wake and the two central vortices are the ones that have the
most significant impact on the forces and moments.
Two cars crossing without crosswind (Case 1): During
interaction, the most significant effect of one car on the
other is on the side force and the yaw moment. While the
side force is minimal before interaction, its magnitude
increases gradually, and its direction continuously changes
during interaction. The side force coefficient increases in
magnitude from an initial value of 0.01 to a maximum
value of 0.09. The yaw moment coefficient does not show a
periodic pattern before interaction. There is a peak in its
value when the cars are side by side, but this peak is in the
same order of magnitude of the largest fluctuation of the
coefficient before interaction starts. Under the conditions
analyzed, these effects last for ~ 0.02 s, and no fluctuation
lasts more than 0.005 s, so the impulse of any change in
side force is small. The two cars affect each other similarly,
and the interaction does not change the shape or direction
of the vortical structures surrounding each car.
Two cars crossing with 30-degree crosswind (Case 2):
As in the case without crosswind, the most significant
effect of one car on the other is on the side force and the
yaw moment. However, unlike Case 1 where each car
experiences roughly the same changes in force and
moment magnitudes, in Case 2 the windward car
experiences the largest changes in these magnitudes,
especially in the side force, which drops by 43 % during
interaction. The yaw moment coefficient also drops by 42
%. The leeward car experiences a smaller drop in the side
force of 16 % because it is shielded from the effects of
crosswind. However, the duration of these changes is
small, similar to the overall trends seen in Case 1.
5.1. Future work
This study involves two identical passenger cars interacting
with each other. The analysis of forces and moments, as
well as the identification and definition of major vortical
structures serves as a baseline for further studies involving
vehicles of disparate sizes, e.g., a car and truck, under a
range of scenarios and minimum lateral distance between
the two vehicles for safety concerns.
To fully investigate the effects of crosswinds on vehicle
interactions, more CFD simulations need to be run with
incrementally larger crosswind angles. Additionally, the
current V&V methodology and procedures were developed
for RANS models. In the IDDES model, RANS is used
inside boundary layer but large eddy simulation (LES) is
used in the separation region. It would be interesting to use
the recently developed general framework of LES (Xing,
2015; Dutta and Xing, 2017), and the five-equation and
robust three-equation methods for LES (Dutta and Xing,
2018) to not only evaluate the numerical errors but also the
modeling and total errors.
ACKNOWLEDGEMENTThis work has been funded by the
US Department of Transportation's University Transportation
Center program, Grant #DTRT13-G-UTC4O through the Pacific
Northwest Regional University Transportation Center (PacTrans).
The authors would like to thank PacTrans for their support.
REFERENCES
Altinisik, A., Kutukceken, E. and Umur, H. (2015a).
Experimental and numerical aerodynamic analysis of a
passenger car: Influence of the blockage ratio on drag
coefficient. J. Fluids Engineering 137, 8, 081104-1–
081104-14.
Altinisik, A., Yemenici, O. and Umur, H. (2015b).
Aerodynamic analysis of a passenger car at yaw angle
and two-vehicle platoon. J. Fluids Engineering 137, 12,
121107.
ANSYS (2016). ANSYS FLUENT 17.2 Theory Guide.
Corin, R. J., He, L. and Dominy, R. G. (2008). A CFD
investigation into the transient aerodynamic forces on
overtaking road vehicle models. J. Wind Engineering
and Industrial Aerodynamics 96, 8-9, 13901411.
Dutta, R. and Xing, T. (2018). Five-equation and robust
three-equation method for solution verification of large
eddy simulations. J. Hydrodynamics 30, 1, 2233.
Dutta, R. and Xing, T. (2017). Quantitative solution
verification of large eddy simulation of channel flow.
1008 Ahmad Hammad et al.
Proc. 2nd Thermal and Fluid Engineering Conf.,
TFEC2017 & 4th Int. Workshop on Heat Transfer,
IWH201, Las Vegas, Nevada, USA.
Gritskevich, M. S., Garbaruk, A. V., Schütze, J. and
Menter, F. R. (2012). Development of DDES and
IDDES formulations for the k-ω shear stress transport
model. Flow, Turbulence and Combustion 88, 3, 431
449.
Holmén, V. (2012). Methods for Vortex Identification. M.
S. Thesis. Lund University Libraries. Lund, Sweden.
Howell, J., Garry, K. and Holt, J. (2014). The
aerodynamics of a small car overtaking a truck. SAE
Paper No. 2014-01-0604.
Hunt, J. C., Wray, A. A. and Moin, P. (1988). Eddies,
streams, and convergence zones in turbulent flows.
Center for Turbulence Research, Proc. Summer
Program, 193208.
Kolář, V. (2007). Vortex identification: New requirements
and limitations. Int. J. Heat and Fluid Flow 28, 4, 638
652.
Kremheller, A. (2015). Aerodynamic interaction effects
and surface pressure distribution during on-road driving
events. SAE Paper No. 2015-01-1527.
Menter, F., Kuntz, M. and Langtry, R. (2003). Ten years of
industrial experience with the SST turbulence model.
Turbulence, Heat and Mass Transfer 4, 1, 625632.
Moore, G. E. (1998). Cramming more components onto
integrated circuits. Proc. IEEE 86, 1, 8285.
Shur, M. L., Spalart, P. R., Strelets, M. K. and Travin, A. K.
(2008). A hybrid RANS-LES approach with delayed-
DES and wall-modelled LES capabilities. Int. J. Heat
and Fluid Flow 29, 6, 16381649.
Spalart, P. R., Deck, S., Shur, M., Squires, K., Strelets, M.
K. and Travin, A. (2006). A new version of detached-
eddy simulation, resistant to ambiguous grid densities.
Theoretical and Computational Fluid Dynamics 20, 3,
181195.
Spalart, P. R., Jou, W.-H., Strelets, M. and Allmaras, S. R.
(1997). Comments on the Feasibility of LES for Wings,
and on a Hybrid RANS/LES Approach. Greyden Press.
Ruston, Louisiana, USA.
Stafford, L. G. (1981). A streamline wind-tunnel working
section for testing at high blockage ratios. J. Wind
Engineering and Industrial Aerodynamics 9, 1-2, 2331.
Strelets, M. (2001). Detached eddy simulation of massively
separated flows. Proc. 29th Aerospace Sciences Meeting
and Exhibit, 879.
Sykes, D. M. (1973). Advances in Road Vehicle
Aerdodynamics. BHRA Fluid Engineering. Cranfield,
UK, 311321.
Xing, T. (2014). Direct numerical simulation of open von
kármán swirling flow. J. Hydrodynamics, Ser. B 26, 2,
165177.
Xing, T. (2015). A general framework for verification and
validation of large eddy simulations. J. Hydrodynamics,
Ser. B 27, 2, 163175.
Xing, T., Carrica, P. and Stern, F. (2008). Computational
towing tank procedures for single run curves of
resistance and propulsion. J. Fluids Engineering 130, 10,
101102-1–101102-14.
Xing, T. and Stern, F. (2010). Factors of safety for
richardson extrapolation. J. Fluids Engineering 132, 6,
061403-1–061403-13.
Xing, T. and Stern, F. (2011). Closure to “Discussion of
‘Factors of safety for richardson extrapolation’” (2011,
ASME J. Fluids Eng., 133, p. 115501). J. Fluids
Engineering 133, 11, 115502-1–115502-6.
Zhang, Z., Zhang, Y. C. and Li, J. (2010). Vehicles
aerodynamics while crossing each other on road based
on computational fluid dynamics. Applied Mechanics
and Materials, 29, 13441349.
Publisher’s Note Springer Nature remains neutral with regardto jurisdictional claims in published maps and institutional affiliations.