The Blockage Effects on Vehicle Aerodynamics in Closed-Wall

Wind Tunnel – A CFD Study

Prasanjit Das
Candidate for the Degree of Master of Engineering
Supervisor: Professor Makoto Tsubokura
Division of Mechanical and Space Engineering

Abstract correction equations of closed-wall wind tunnel were

based on simplified bluff body geometry. Cooper et al.
An investigation into the total blockage effects on [1] compared a number of correction methods and
aerodynamic drag in closed-wall wind tunnel testing recommended not to use the classical formulae for
has been carried out by computational fluid dynamics vehicle aerodynamics. Wind tunnel corrections for
(CFD). The simulation of flow past two different separated flows were developed first by Maskell [2].
realistic, full-scale, heavy-duty truck models was His analysis was originally derived for circular plates
conducted. The models were placed at different yawing normal to the flow and it based on conservation of
angles in the virtual wind tunnels with blockage ratio of momentum. Recently, a semi-empirical correction was
about 10%; which is exactly the same as the DNW proposed by Mercker [3], which for the first time
experimental wind tunnel geometry. To compare the introduced additional blockage effects associated with
results with ideal (blockage free condition) case, other the pressure gradient along the test section at empty
simulations with larger cross section at the blockage condition. The purpose of the present study is to
ratio of 1.1% were also conducted. Comparisons of the investigate the closed wall wind tunnel blockage effect
ideal CFD results to the corrected Cd values obtained on full-scale heavy-duty truck by using CFD, to assess
by some existing blockage correction methods for the blockage correction equations for real vehicle
closed-wall wind tunnel are also examined. model, and to determine blockage free aerodynamic
drag.
Key words: Wind tunnel blockage Corrections,
Aerodynamic drag, Closed-wall wind tunnel, CFD Numerical Methods
Governing equations and discretization:
Introduction
The governing equations being solved in the LES are
At the development stage, it is important to assess the
spatially filtered and Navier-Stokes equations:
vehicle performance in account of engine power. In a
u i
typical class of heavy-duty truck, power required to 0, (1)
x i
overcome rolling resistance and accessories increases
linearly with vehicle speed, while energy losses due to u i  P 
 ui u j   2    SGS S ij , (2)
aerodynamic drag increase with the cube of the speed. t x j x i x j
At a typical highway speed of 70 mph, aerodynamic The bar over the physical quantity indicates the spatial
drag accounts for approximately 65% of the energy filtering operation for LES. The filtered strain rate
output of the engine. Even modest reductions in
tensor S ij and pressure P in Eq. (2) are expressed as
aerodynamic drag can significantly reduce fuel
consumption. Lower fuel consumption will result in a 1  u j u i 
,
S ij   (3)
reduction in pollution emissions, and, more 2  xi x j 

importantly, a reduce dependence on fossil fuels. In
automotive industry, to assess the performance of road  
P  p /   uiu j  ui u j / 3 , (4)
vehicle, aerodynamic testing is normally conducted in In Eq. (2), the last term on the right represents the
wind tunnel. For the constrained effects imposed by the effect of sub-grid-scale (SGS) turbulence, which was
walls of a wind tunnel around the vehicle, there is modeled under the eddy viscosity assumption. The
always a need to correct the aerodynamic quantities conventional Smagorinsky model (Smagorinsky. 1963)
such as aerodynamic forces and moments. The was used, and the eddy viscosity coefficient was
corrected results can then reflect the performance under modeled as
blockage free conditions. To date, most of the blockage
 SGS  C s f d  2 2 S ij S ij , (5)
where ∆ is the volume of numerical element, and
model coefficient C s is set to be 0.15, which is
generally suitable for external flows. The damping of
the turbulent effect near a wall boundary is explained
by the Van-Driest type damping function as follows:
 l
f d  1  exp , (6)
25
where l  is the distance from the wall in wall
coordinates.
The governing equations were discretized by using the
vertex-centered unstructured finite volume method. The Fig 2. Space elements on the vehicle components
second-order central differencing scheme was applied
for the spatial derivatives and blending of 5% first-
The triangle element on the surface of the vehicle is
order upwind scheme for the convection term was indicated in Fig 2. The fluid space was decomposed by
employed for numerical stability. The third-order
tetrahedral elements. To maintain finer resolution
upwind scheme was adopted for the spatial derivative
around the vehicle, hierarchical allocation is carried
far away from vehicle, where coarser grid was out, as shown in Fig 3.
allocated. For time marching, the third-order Adams-
Moulton semi-implicit scheme was used. Pressure-
velocity coupling was preserved by using the
Simplified Marker and Cell (SMAC) algorithm.

Target Vehicle Model

The configuration of the two different full-scale heavy-

duty trucks is shown in Fig 1.The surface of the vehicle Fig 3. Space elements around the vehicle
is reproduced by about 1.5 million triangle meshes. To
reproduce the fine structure, the surface resolution is Computational Domains and Boundary
around 5 to 10 mm around the side mirror, and Conditions
relatively fine elements are allocated around the cabin.
The engine and power train is reproduced by the To validate our numerical method based on the DNW-
moderate elements with the resolution of 20 to 50 mm. LLF wind tunnel data, we adopted the same wind
tunnel geometry in the simulation, as shown in Fig 4(a),
Larger elements are allocated to reproduce the cargo
in which the blockage ratio of the projected frontal
panel. area of the vehicle and the cross sectional area of the
test section is about 10%. To compare the corrected
drag with the ideally obtained value, we have also
conducted the case with the blockage ratio of around
1%, as shown in Fig 4(b). In both cases, a uniform
velocity distribution U 0 is defined at the inlet (about 22
m/s and 25 m/s in the DNW and ideal cases,
respectively) about 40 m upstream of the vehicle. All
velocity components were gradient-free for the
steamwise direction at the outlet. The log-law profile
was assumed on the velocity and surface friction on the
wall was estimated and directly imposed as Neumann
boundary condition. In DNW case, shown in Fig 4(a),
the exit of the wind tunnel nozzle is located 3.79 m
windward of the vehicle and diffuser type collected is
mounted after the test section.

Fig 1. Models with (Left) and without (Right) the air

deflector
(a) Mercker correction:

Mercker [3] provides a blockage correction equation

based on the work of Lock for solid blockage and on
the work of Maskell, Thom, and Glauert for wake
blockage. There is important thing of Mercker
correction is considered about yaw angle effect, vehicle
geometry as well as wind tunnel geometry. The
correction expressed by the ratio of dynamic pressure is
(b)  
2
qc  2 AM .2VM 2A 1 2 AM 
 1  K3   M . CDm (0)   

q  3/ 2 2 AN 4 2 AN 
L p .2VM .(2 AN )
 
(9)
where  s is solid blockage factor,  w is wake blockage
factor, AM is vehicle frontal area at  deg. yaw angle,
VM is vehicle volume, and CDm (0) is the measured
drag coefficient at 0 deg. yaw angle.
Mercker recently modified his procedure in order to
consider the static pressure change over the test section
Fig 4. Computational domains: DNW geometry, at empty state. The drag coefficient is fully corrected in
(b) Ideal
order to consider wake distortion and horizontal
pressure buoyancy by
Review of Wind-Tunnel Blockage C D T  C D uw  C pw  C DHB , (10)
Correction Equations
C D w  C DT / qc / q  , (11)
It is defined that the total blockage correction factor is
Wake distortion term:
the sum of velocity acceleration (blockage factor)
caused by solid and wake blockage; however these are

C pw  C pwc  C pmb ,  (12)
more difficult factors to assess for unusual geometries where C pwc is the pressure coefficient in empty wind
such as the heavy-duty truck with including
tunnel at the location of wake closure and C pmb is the
complicated detail characteristics features and the
associated flow fields around them. pressure coefficient in empty wind tunnel at the
location of the base of the model.
Maskell correction: Horizontal pressure buoyancy term:
CDHB  1.75 / AM . VM / 2.G ,
Maskell [2] was the first to address the problems with (13)
non-streamline flow bodies, such as bluff-body testing
in closed-wall wind tunnel and that partially stalled Evaluation of the Correction Equation
shapes such as wings. Maskell’s theory holds for close
axis symmetric wake in three dimensional flows. The The time-averaged aerodynamic drag coefficient shown
correction of the dynamic pressure ratio is as follows: in Table 1 was obtained by DNW wind tunnel
geometry and ideal condition wind tunnel simulation
qc  5 A 
 1  C D uw  M  , (7) for the model with air deflector. All the aerodynamic
q  2  AN  coefficients are normalized for confidentiality reasons.
In Table 2, the aerodynamic drag is corrected by
C D w  C D uw / qc / q , (8) Maskell and Mercker methods. The corrected results
were obtained by using drag coefficient from DNW
where qc is corrected dynamic pressure, q is geometry simulation. The corrected value by Mercker
uncorrected dynamic pressure, AM the flat plate area, method shows good agreement with the ideal value,
when consider solid blockage, wake blockage and
AN the wind tunnel working section cross sectional
horizontal buoyancy term. Also Maskell’s correction
area, CDw is corrected drag coefficient and CDuw is equation has a similar trend in 0 degree yaw angle and
uncorrected wind-axis drag coefficient . it shows over correction in 10 degree yaw angle. On the
other hand, when taken into account all of the effects
including wake distortion term in the Mercker’s relevant length behind the vehicle over which the
method, drag is overcorrected. pressure gradient effect has to be calculated. This
location is measured up to a first order of
Table 1: Normalized drag coefficient for the model with air approximation, which occurs where C PT  0 , here
deflector (0 and 10 deg. yaw angle)
CPT is total static pressure coefficient. Recently,
Normalized drag coefficient (CFD)
Yaw angle(deg.) DNW geometry Ideal LES Mercker [4] has noticed that the true length (sensitivity
0 1.000 0.801 length) can be determined exactly by generating a
10 1.716 1.297 second gradient in the empty test section. It is turned
out that this length is usually shorter than the distance
Table 2: Normalized corrected drag coefficient for the model between vehicle base location and wake closure
with air deflector by correction Eq. (0 and 10 deg. yaw angle) location. For that reason, when using the wake closure
Normalized Corrected Drag coefficient based on location, the buoyancy effect on the wake
Mercker Equation
overcorrected the measured drag coefficient. We can
Factors considered 0 deg. 10 deg.
Solid + Wake blockage 0.847 1.329 say that at this moment we are yet in the position to
Solid + Wake blockage + HPB 0.789 1.278 determine the sensitivity length on a pure theoretical
Solid + Wake blockage + HPB 0.693 1.189 base other than on experimental grounds.
+ Wake distortion
Conclusions
Normalized corrected Drag coefficient based on
Maskell Equation Computational fluid dynamics (CFD) flow simulations
Factor considered 0 deg. 10 deg. have successfully been performed using large eddy
Solid blockage 0.787 1.174 simulation to assess the wind tunnel blockage effects
on real vehicle model. The Maskell’s correction
Table 3 and Table 4 present the result obtained with equation for aerodynamic drag is working well for the
model without the air deflector at 0 degree yaw. We model with the air deflector without considering the
can see that aerodynamic drag corrected by Mercker’s yawing effect. The drag coefficients under the effect of
method shows a better agreement when excluding the yaw angle obtained with the correction method
wake distortion effect. The corrected Maskell’s method proposed recently by Mercker had shown good
resulted in an over correction. From our CFD results, agreement with the ideally obtained values in both
we realize that Maskell’s method shows the worst models. But further investigation is needed for wake
result in non-zero yawing. induced drag increment for steep pressure gradient in
Mercker’s procedure.
Table 3: Normalized drag coefficient for the model without
air deflector (0 deg. yaw angle)
Normalized drag coefficient (CFD)
Acknowledge
Yaw angle(deg.) DNW geometry Ideal LES This work was supported by the Industrial Technology
0 1 0.812 Research Grant Program in 2007 from the New Energy
and Industrial Technology Development Organization
(NEDO) of Japan. This study was conducted in a
Table 4: Normalized corrected drag coefficient for the model
collaborative research project with Isuzu Advanced
without air deflector by correction Eq. (0 deg. yaw angle)
Engineering Center Ltd., and the geometry data and
Mercker Maskell experimental data received are greatly acknowledged.
Factors considered 0 deg. 0 deg.
Solid + Wake blockage 0.848
Solid + Wake blockage + HPB 0.764 References
0.802
1. K.R. Cooper et al.: "Closed-Test-Section Wind Tunnel
Solid + Wake blockage + HPB 0.762
+ Wake distortion Blockage Corrections for Road Vehicles."SAE
International. SP-1176.
2. Maskell, E.C. ‘A Theory of the Blockage Effects on
Concerning the relatively large discrepancies between
Bluff Bodies and Stalled Wings in a Closed Wind
the ideal drag coefficient and the corrected drag Tunnel’. ARC R&M 3400, Nov. 1963.
coefficient by Maskell’s method, one possible 3. E. Mercker et al.: The Influence of a Horizontal
explanation is that the Maskell’s correction equation is Pressure Distribution on Aerodynamic Drag in Open
based on simplified geometry (symmetric wake and Closed Wind Tunnels, SAE 2005-01-0867, 2005.
structure at rear section of a circular flat plate) and do 4. E. Mercker, BMW AG, Germany and Dr. Koro Kitoh,
not account for the effect of yawing. To consider the Kozo Kitoh Technology. Inc, 2011: Private
wake distortion in Mercker’s method, which depends communication.
on approximation location of the wake closure, the