n

Introduction of a New Realistic Generic Car Model 2012-01-0168

tio
Published
for Aerodynamic Investigations 04/16/2012

ibu
Angelina I. Heft, Thomas Indinger and Nikolaus A. Adams
Technische Universität München

s tr
Copyright © 2012 SAE International
doi:10.4271/2012-01-0168

Di
exist (see G.M. Le Good [2]): the use of strongly simplified
ABSTRACT car models and that of production vehicles.
State of the art aerodynamic research of vehicles often
employs strongly simplified car models, such as the Ahmed
and the SAE body, to gain general insights. As these models
for
Generic car models, such as the SAE model and the Ahmed
body, make it easy to relate the observed phenomena to
specific areas and thus help to understand basic flow
exhibit a high degree of abstraction, the obtained results can
only partly be used for the aerodynamic optimization of structures. At the same time, more complex flow phenomena,
ot
production vehicles. Aerodynamic research performed on e.g. at the underbody and the wheels/wheelhouses, cannot be
specific vehicles is on the other hand often limited due to reproduced due to the oversimplification of these geometries.
On the other hand, it is usually not feasible to investigate
-N

their short life span and restricted access. A new realistic

generic car model for aerodynamic research - the DrivAer these phenomena on a specific production vehicle, as, due to
body - is therefore proposed to close this gap. This paper its short life span and restricted access, typically little
focuses on the development of the model and the first validation data is available. Recognizing the need for a model
experimental results, namely force and pressure combining the strengths of both approaches, various more or
less generic models, such as the VW reference car and the
ly

measurements of the different configurations. The

experiments were performed in the recently updated Wind MIRA reference car, have been proposed (cf. G.M. Le Good
Tunnel A of the Institute of Aerodynamics and Fluid [2]). However, while these reference cars mark a step in the
On

Mechanics at the Technische Universität München. right direction, these models are still too generic to
completely understand the complex phenomena occurring at
realistic vehicles.
INTRODUCTION
Due to growing customer consciousness and various national To close this gap, the Institute of Aerodynamics and Fluid
iew

and international agreements, the reduction of CO2 emissions Mechanics of the Technische Universität München (TUM), in
has become increasingly important. As the car sector is one cooperation with two major car companies, the Audi AG and
of the big contributors to the overall CO2 emission, it is the BMW Group, therefore, proposes a new realistic generic
necessary to lower the fuel consumption of contemporary car model. The body is based on two typical medium-class
cars. Aerodynamic optimization of cars still offers big saving vehicles and includes three interchangeable tops and two
ev

opportunities, especially as its importance increases with the different underbody geometries to allow for a high
use of recuperation systems (see Hucho [1]). universality. To encourage the use of the DrivAer model in
independent research projects, the geometry and a
Pr

To further optimize the car geometry, it is important to comprehensive database with both numerical and
understand the occurring aerodynamic phenomena. So far experimental results will be published on the website of the
two basic approaches to investigate these aspects of the flow institute1.

1http://www.aer.mw.tum.de/en/research-groups/automotive/drivaer
The aim of this paper is to present this new geometry to a
broad audience and to provide first experimental results. In
the beginning the necessity of a new realistic generic car
geometry will be discussed, followed by a short summary of

n
the development of the DrivAer model. In the next section,

tio
the experimental wind tunnel setup will be presented and the
general approach and the data processing will be explained.
This passage will be followed by the discussion of the Figure 1. Dimensions of the SAE model.
experimental results categorized into force and pressure

ibu
measurements. Concluding the paper, the experimental
results will be summarized and a short outlook on future
investigations of the DrivAer model will be given.

DEVELOPMENT OF THE DRIVAER

s tr
MODEL
BACKGROUND

Di
The flow around a bluff body moving in proximity of a static
ground is governed by the interaction with the ground and Figure 2. Dimensions of the Ahmed body.
highly turbulent separation and reattachment. Both the
experimental reproduction of the relative movement between
vehicle and ground and unsteady investigations are highly
complex and, therefore, associated with high experimental
effort. Yet, to be able to further optimize road vehicles it is
for
On the other hand, as their shapes are very unlike actual car
geometries, these insights cannot be readily applied in the
development of production vehicles. Complex areas of the
necessary to completely understand these phenomena. car geometry, such as the A- and C-pillars, the highly curved
ot
rear end, and the wheelhouse region, are especially impaired.
The improvement of wind tunnel facilities, especially as to Therefore, during the actual optimization process, often real
the introduction of ground effect simulation through moving production car geometries are employed. As these are usually
-N

belts, allows a more precise experimental investigation of only accessible to a limited group of people, they are rarely
time-accurate flow content (see Janssen [3], Cogotti [4]). featured in more than one published work and, thus, cannot
Another important advance is the development of smaller and serve for validation purposes.
more accurate pressure transducers that can easily be fitted
into the surface of wind tunnel models. The gap between these two approaches makes a new realistic
ly

generic car model that combines the advantages of both

Categorizing the existing aerodynamic approaches in recent model types desirable. While several attempts to satisfy this
papers on the subject of vehicle aerodynamics, G.M. Le Good need have been made, for example by introducing the MIRA
On

[2] identifies two main classes: the investigations performed reference car or the VW reference car (see G.M. Le Good
on strongly simplified generic models and those carried out [2]), the authors of this paper think that these models are still
on real production cars. too generic to allow for detailed investigations of complicated
flow phenomena. To this end, the DrivAer model is proposed
Commonly, especially time-accurate investigations resort to and will be - along with numerical and experimental results -
iew

the use of strongly simplified bodies, such as the SAE body, made available to the public. The computer-aided design
as presented by Cogotti [5], see Figure 1, and the Ahmed (CAD) geometry will be published on the homepage of the
body, as described by Ahmed [6], see Figure 2. These models Institute of Aerodynamics and Fluid Mechanics at the TUM
offer the advantage of a reasonable computational and and independent experimental and numerical studies using
ev

experimental effort compared to production vehicles and the the geometry are strongly encouraged.
possibility to examine the flow effects of different parts of the
vehicle with limited interference effects. Furthermore, in DEVELOPMENT
Pr

contrast to specific production vehicles, the simplified models

offer a broad spectrum of both numerical and experimental The geometry of the DrivAer model is based on the
validation data and are accordingly well suited for validation geometries of two medium sized cars, the Audi A4 and the
purposes. BMW 3 Series (see Heft et al. [7]). Audi AG and the BMW
Group generously provided the CAD data of the different
configurations of the original vehicles. The original CAD
surfaces were simplified and approximated by characteristic
curves. In the next step, the curves of both original cars were
merged to generate the new CAD geometry.

Figure 3 shows a sketch of the fastback configuration of the

n
1:2.5 DrivAer model measured in the Wind Tunnel A of the

tio
Institute of Aerodynamics and Fluid Mechanics.

ibu
s tr
Di
Figure 4. DrivAer body with different tops.

Figure 3. Main Dimensions of the 1:2.5 DrivAer Model.

for
ot
The flow phenomena at the backlight of contemporary
production vehicles can be divided into three main groups
-N

with distinct aerodynamic behavior (see Hucho [8]): estate

back vehicles, fastback vehicles, and notchback vehicles.
While the flow over an estate back vehicle detaches at the (a). Detailed Underbody
end of the roof thus creating a big wake region, the flow over
a fastback or a notchback vehicle is strongly influenced by
ly

the angle of inclination of the backlight. For fastback vehicles

the strong vortices emanating from the C-pillars typically
On

induce a downwash region on the backlight, thus, inhibiting

the detachment of the flow at the end of the roof or forcing
the flow to reattach on the backlight depending on the angle
of inclination. These C-pillar vortices are less pronounced for
notchback vehicles. Therefore, the flow detaches at the end of
iew

the roof at smaller angles of inclination, the reattachment

depends, amongst other things, on the length and height of (b). Smooth Underbody
the trunk. To allow for a thorough investigation of these Figure 5. Different underbody configurations: (a)
different behaviors, the DrivAer model was developed as a detailed underbody and (b) smooth underbody.
modular concept that comprises three interchangeable tops, as
ev

shown in Figure 4 - just as for example the SAE body and the
MIRA reference car (see G.M. Le Good [2]). Furthermore, as The results presented in this paper were obtained using the
shown in Figure 5, two different underbody geometries are mock-up configuration of a scaled 1:2.5 DrivAer model, i.e.
Pr

provided: a smooth underbody for symmetrical investigations without considering a cooling flow. However, future
and a detailed underbody based on the simplified underbody investigations will concentrate on cooling configurations for
geometry of the Audi A4. both conventional and electric cars.
EXPERIMENTAL SETUP into account. To obtain physically accurate results, the
blockage ratio
The DrivAer model was measured in the recently updated
Wind Tunnel A of the Institute of Aerodynamics and Fluid

n
Mechanics at the Technische Universität München, a (1)
horizontal Göttingen type wind tunnel. The open test section

tio
of the Wind Tunnel A has a length LT=4.8m, the nozzle has a should be as small as possible (see Hucho [8]). At the same
height of HN=1.8m and a width of WN=2.4m. Vortex time, it is desirable to satisfy the Reynolds number similarity
generators are installed at the nozzle exit to reduce the to ensure the physical similarity of the flow structures:

ibu
pressure fluctuations induced by the developing shear layers.
To allow for ground simulation (GS) with rotating wheels,
the wind tunnel has been equipped with a moving belt system
(see Mack et al. [9]). As the moving belt lies 60mm higher (2)
than the static ground configuration, the effective nozzle

s tr
height is reduced to HN,eff=1.74m. Especially for the If measuring a 1:2.5 model, the free stream wind speed
optimization of underbody geometries and the wheelhouse should be 2.5 times higher than for the 1:1 vehicle. In vehicle
region, it is essential to simulate the relative motion between aerodynamics it is common to perform the measurements at

Di
the vehicle and the ground [3]. The vehicle body is held from the wind speed u=140km/h which would correspond to a
above by a central strut while the wheels are supported necessary free stream velocity of almost u=100 m/s for a
separately by four horizontal struts from outside of the test 1:2.5 model. On the other hand, it has been observed that the
section. In the measured configuration, there is no physical drag coefficient reaches a relatively stable level for higher
connection between the body and the wheels. The model is
placed over a polyester-based belt of 1.39m width and a
distance of 4.53m. The basic configuration can be seen in
Figure 6.
for
Reynolds numbers. At the beginning of the DrivAer
measurement cycle, therefore, it has to be verified that the
drag coefficient reaches a constant level for the chosen
Reynolds number. To that end, the drag coefficient at
different Reynolds numbers will be examined.
ot
Taking the dimensions of the Wind Tunnel A and its capacity
into consideration, the chosen 1:2.5 model with a blockage
-N

ratio φ=8% seems to be an adequate compromise between the

two requirements.

To facilitate the instrumentation of the wind tunnel model

with pressure taps, the top part of the model was laminated,
ly

while the underbody geometry was cut out of high-density

foam. The force measurements were obtained using a main
On

internal 6-component force balance that was placed between

the top strut and the model and four separate 1-component
force balances attached to the wheels. The forces of all
balances were added up and averaged over three
measurement intervals of 10s each.
Figure 6. Picture of the experimental setup in the Wind
iew

Tunnel A. The time-averaged pressure measurements were conducted

using a multiport pressure measurement system. The system
has a full scale range FS=17kPa and an overall accuracy of
The moving belt system can operate up to a velocity of 50 ±0.15%FS and is connected to the surface pressure taps
ev

m/s. The boundary layer is reduced using a passive boundary through flexible tubing (see Vogel [10]). Up to 192 ports can
layer scoop. For a more detailed description of the wind be measured successively in one measurement cycle. For the
tunnel setup, its characteristics, such as the static pressure time-averaged measurements a sampling rate of 20 Hz and an
Pr

distribution and the boundary layer profile, be kindly referred averaging period of 10s were chosen.
to Mack et al. [9].
During one measurement cycle, 188 locations distributed
The decision on the size of the wind tunnel model is based on over the surface of the model were measured. In this paper,
various factors. On the one hand, it is important to take the only some of the most relevant regions will be discussed,
effects of wind tunnel blockage on the aerodynamic results namely the symmetry plane that was equipped with 61 probes
(in the configuration with detailed underbody), the z=60mm
plane which runs approximately through the stagnation point measurements were not conducted at fixed velocities, but
with 21 measurement locations (see Figure 7), the windshield rather at fixed Reynolds numbers. The Reynolds numbers
with 16 pressure taps (see Figure 21), the side window were determined as the equivalent of 10m/s, 20m/s, 30m/s
directly behind the A-pillar where 11 probes were located and 40 m/s in an air-conditioned dry air environment at 20°C

n
(see Figure 24), and finally the rear windows of the estate and at sea level (see Table 1). The chosen reference length

tio
back and the notchback configuration with 32 probes each lref=1.84m corresponds to the length of the scaled DrivAer
(see Figures 25 and 26). model.

Table 1. Correlation between the Reynolds number and

ibu
the velocity at standard conditions.

s tr
Di
When working with ground simulation, it is furthermore
Figure 7. Distribution of pressure taps in the z=60mm important to separate the aerodynamic from the frictional
plane of the 1:2.5 model. forces between the moving belt and the wheels. To isolate the
aerodynamic forces, a measurement series with operating

Additionally, time-accurate surface pressures were

investigated at the backlight of the fastback model. Areas of
for
moving belt and without wind was conducted and a quadratic
equation for the frictional forces was derived by curve fitting
(see Figure 9).
periodic detachment and reattachment can be identified with
spectral estimates. For this purpose 40 miniature pressure
ot
transducers of the type HCL12X5P were placed on the rear
slant of the fastback (see Figure 8). These pressure
-N

transducers have a pressure range of ±12.5mbar and feature a

typical error of ±0.05%FS and a maximum error of
±0.25%FS for combined non-linearity and hysteresis [11].
The instantaneous data was obtained during a sampling
interval of 120s at a sampling frequency of 2000Hz. The
ly

measurements were filtered with a low pass filter of 1000 Hz.

On

Figure 9. Rolling resistance over the moving belt

velocity.
iew

To obtain the correct aerodynamic forces, the rolling

resistance is calculated for the current moving belt velocity
and subtracted from the measured forces.
Figure 8. Distribution of pressure taps at the backlight of
ev

the fastback top. The measured forces are depicted as non-dimensional force
coefficients. The drag coefficient can be calculated as:
Pr

APPROACH AND DATA

PROCESSING
As the Wind Tunnel A of the Institute of Aerodynamics and (3)
Fluid Mechanics is not temperature and humidity controlled,
the measuring conditions can change quite drastically The pressure samples presented throughout this paper are also
between summer and winter. To maintain comparability, the non-dimensioned, i.e. calculated as the ratio of the