ÇUKUROVA UNIVERSITY
INSTITUTE OF NATURAL AND APPLIED SCIENCES
Ms. Sc. THESIS
Serkan MEZARCIÖZ
AERODYNAMICS OF A MODEL BUS
DEPARTMENT OF MECHANICAL ENGINEERING
ADANA, 2006
ÇUKUROVA ÜNİVERSİTESİ
FEN BİLİMLERİ ENSTİTÜSÜ
AERODYNAMICS OF A MODEL BUS
Serkan MEZARCIÖZ
MASTER TEZİ
MAKİNA MÜHENDİSLİĞİ ANABİLİM DALI
Bu Tez ..../..../2006 Tarihinde Aşağıdaki Jüri Üyeleri Tarafından
Oybirliği/Oyçokluğu İle Kabul Edilmiştir.
İmza: İmza: İmza:
Prof. Dr. Beşir ŞAHİN Doç. Dr. Hüseyin AKILLI Doç. Dr. Galip SEÇKİN DANIŞMAN ÜYE ÜYE
Bu Tez Enstitümüz Makina Mühendisliği Anabilim Dalında Hazırlanmıştır.
Kod No:
Prof. Dr. Aziz ERTUNÇ Enstitü Müdürü
Not: Bu tezde kullanılan özgün ve başka kaynaktan yapılan bildirişlerin, çizelge, şekil ve fotoğrafların kaynak gösterilmeden kullanımı, 5846 sayılı Fikir ve Sanat Eserleri Kanunundaki hükümlere tabidir.
ABSTRACT
M.Sc. THESIS
AERODYNAMICS OF A MODEL BUS
Serkan MEZARCIÖZ
DEPARTMENT OF MECHANICAL ENGINEERING INSTITUTE OF NATURAL AND APPLIED SCIENCES
UNIVERSITY OF ÇUKUROVA
Supervisor: Prof. Dr. Beşir ŞAHİN Year: 2006, Pages:67
Jury: Prof. Dr. Beşir ŞAHİN
:Doç. Dr. Hüseyin AKILLI : Doç. Dr. Galip SEÇKİN The ultimate goal of this work is to examine the turbulent, 3D flow around a
bus-shaped body by means of CFD. Another goal of this study is to show the accordance between the numerical and experimental results. In the present study, RANS k-ε turbulence model is employed to simulate the flow around the vehicle, and the results of the experiments performed by PIV are used to validate the corrections of the numerical simulation. A comparison of predicted results for time-averaged flow data, particularly, around the forward face of the model with the experimental results of Particle Image Velocimetry (PIV) showed that numerically predicted present results and experimental results agreed well. The flow is assumed to be incompressible, viscous, turbulent, and steady flow.
Key Words: Aerodynamics, Bus, Reynolds Average Navier Stokes (RANS), k-ε
Turbulence Model
I
ÖZ
YÜKSEK LİSANS TEZİ
BİR OTOBÜS MODELİNİN AERODİNAMİĞİ
Serkan MEZARCIÖZ
ÇUKUROVA ÜNİVERSİTESİ
FEN BİLİMLERİ ENSTİTÜSÜ
MAKİNA MÜHENDİSLİĞİ ANABİLİM DALI
Danışman: Prof. Dr. Beşir ŞAHİN
Yıl: 2006, Pages:67
Juri: Prof. Dr. Beşir ŞAHİN
:Doç. Dr. Hüseyin AKILLI
: Doç. Dr. Galip SEÇKİN
Bu çalışmanın ana amacı hesaplamalı akışkanlar mekaniği aracılığı ile otobüs
şeklindeki bir cismin etrafındaki 3 boyutlu, türbülanslı akışın incelenmesidir. Çalışmanın diğer bir amacı ise deneysel ve nümerik sonuçlar arasındaki benzerliklerin gösterilmesidir. Bu çalışmada araç etrafındaki akışın simülasyonu için k-ε türbülans modeli kullanılmış ve nümerik simülasyon sonuçlarının doğruluğunu teyit etmek için PIV deneylerinin sonuçları kullanılmıştır. Modelin ön yüzünün etrafındaki tahmin edilen akış özellikleri ile parçacık görüntülemeli hız ölçüm deneylerinin sonuçlarının karşılaştırması, nümerik olarak tahmin edilen mevcut sonuçların ve deneysel sonuçların birbirine uygun olduğunu göstermektedir. Akış sıkıştırılamaz, türbülanslı, sürekli kabul edilmiştir.
Anahtar Kelimeler: Aerodinamik, Otobüs, Reynolds Average Navier Stokes (RANS), k-ε Türbülans Modeli
II
ACKNOWLEDGMENT
I would like to express my gratitude to everyone who assisted in this study.
First and Foremost, I would like to thank my supervisor, Professor Beşir ŞAHİN,
who gave me the opportunity, support, and freedom that I needed to conduct this research.
I would like to thank Assist.Prof. Hüseyin AKILLI for his scientific and technical
guidance and his support during my graduate studies.
I would like to thank Tural TUNAY for transferring all his knowledge, for helping in
the study and for his friendship.
I would also like to thank Research Assistant Cahit GÜRLEK for his very useful
comments and experimental support.
I appreciate gratefully the help from my friends, İbrahim PEHLİVAN and
Serin MAVRUZ for their contributions to this project.
Finally, I am greatly indebted to my parents for nurturing and cherishing
my academic endeavors.
III
CONTENTS PAGE
ABSTRACT......................................................................................................... I
ÖZ......................................................................................................................... II
ACKNOWLEDGEMENT……………………………………………………… III
CONTENTS……………………………………………………………………. IV
LIST OF FIGURES ……………………………………………………………. VI
NOMENCLATURE……………………………………………………………. IX
1. INTRODUCTION…………………………………………………………… 1
1.1.Importance of Aerodynamics and Flow Around Road Vehicles………... 1
1.2. Importance of Aerodynamics for Buses………………………………... 10
2. PREVIOUS STUDIES AND SCOPE OF THE PRESENT STUDY……….. 12
2.1. Scope of the Present Study……………………………………………... 21
3. MATERIALS AND METHODS……………………………………………. 22
3.1. Vehicle Geometry and Dimensions…………………………………….. 23
3.2. Determination of Dimensions of the Flow Domain…………………..... 24
3.3. Grid Generation and Refinement Study………………………………... 26
3.4. Boundary Conditions…………………………………………………… 28
3.4.1. Boundary Conditions for Simulations……………………………. 29
3.5. Simulation of the Flow…………………………………………………. 29
3.5.1. Discretization……………………………………………………... 30
3.5.2. Convergence……………………………………………………… 32
3.6. Governing Equations for Fluid Flow…………………………………… 33
3.7. Reynolds Averaged Navier Stokes (RANS) Equations of Motion……... 34
3.8. Overview of Turbulence Modeling…………………………………….. 36
3.9. Turbulence Models……………………………………………………... 37
3.9.1. Time Averaged Turbulence Models……………………………… 37
3.9.1.1. Zero Equation Model……………………………………... 38
3.9.1.2. One Equation Model……………………………………… 38
3.9.1.3. Two Equation Models……………………………………. 39
3.10. Near Wall Treatment………………………………………………….. 44
IV
4. RESULTS AND DISCUSSION…………………………………………….. 46
5. CONCLUSIONS…………………………………………………………….. 60
5.1. Future Studies…………………………………………………………... 62
REFERENCES…………………………………………………………………. 63
CIRRICULUM VITAE………………………………………………………… 67
V
LIST OF FIGURES PAGE
Figure 1.1. An aerodynamically improved truck (MAN)…………………. 2
Figure 1.2. Induced air resistances in the case of a fully faired semi-trailer
outfit……………………………………………………………. 3
Figure 1.3. Induced air flow at the vehicle body…………………………... 4
Figure 1.4. Longitudinal swirls caused by induced draft…………………... 5
Figure 1.5. Air flow through a motor vehicle……………………………… 6
Figure 1.6. Application of Rear Flap to a Truck (MAN, 2004)……………. 7
Figure 1.7. Forces and moments in aerodynamic measurements…………... 8
Figure 1.8. Components of aerodynamic drag in an optimized bus body
(MAN, 2004)………………………………………………….…
9
Figure 1.9. Aerodynamic drag at oblique flow for different passenger cars
(BOSCH, 2002)…………………………………………………. 10
Figure 1.10. An application of rounded corner to a bus (TEMSA Diamond)... 11
Figure 1.11. A bus having sharp corners (TEMSA Tourmalin)……………… 11
Figure 2.1. Geometry of the vehicle body and computational domain used
by Krajnovic and Davidson (2001)……………………………... 15
Figure 2.2 Kinetic energy contours at the centerline (y=0) around the 25º
Ahmed body using the linear k-ε and cubic non-linear k-ε
models with specified Chieng and Launder wall functions Craft
et al (2001)……………………………………………………… 17
Figure 2.3. Geometry of the cube with its channel dimensions ( Krajnovic
and Davidson, 2002b)…………………………………………… 18
Figure 2.4. Schematic representation of the computational domain with
vehicle body (Krajnovic and Davidson, 2004a)……………..... 21
Figure 3.1. The dimensions of the vehicle model…………………………... 23
Figure 3.2. Schematic representation of the computational domain with
vehicle body (Krajnovic and Davidson, 2004a)………………… 24
Figure 3.3. Position of the bus in the channel (side view, dimensions in
mm)…………………………………………….……….………. 25
VI
Figure 3.4. Position of the bus in the channel (cross-sectional view)………. 25
Figure 3.5. 3D view of the computational domain………………………….. 25
Figure 3.6. Distribution of y* values for the grid having 442.496 nodes…... 27
Figure 3.7. Distribution of y* values for the grid having 786.586 nodes…... 27
Figure 3.8. Distribution of y* values for the grid having 1.186.819 nodes… 28
Figure 3.9. Convergence of the residuals to 10-5…………………………… 33
Figure 3.10. Subdivisions of near wall region (FLUENT, 1998)…………….. 44
Figure 4.1. The symmetry planes of the bus model…………………………. 46
Figure 4.2. Time-averaged patterns of experimental streamlines around
forward face of the bus model along the central axis in side view
plane(Gürlek,2006)…………………………………………….. 47
Figure 4.3. Time-averaged patterns of numerical streamlines around
forward face of the bus model in side view symmetry
plane……………………………………………………………. 48
Figure 4.4. Comparison of time–averaged experimental and numerical
patterns of streamlines in plan view planes around the forward
part of the bus model …………………….................................... 49
Figure 4.5. Comparison of time–averaged (a) experimental and
(b)numerical vorticity in plan view planes of forward portion of
the bus model.. ………………………………………………… 50
Figure 4.6. Comparison of time–averaged experimental and numerical
streamlines in plan view planes downstream of the bus model … 50
Figure 4.7. Time-averaged experimental velocity vector map downstream of
the bus model in horizontal symmetry plane……………………. 51
Figure 4.8. Time-averaged numerical velocity vector map downstream of
the bus model in horizontal symmetry plane……………………. 52
Figure 4.9. Time-averaged patterns of streamlines downstream of the bus
model, in side view plane. a: experimental data, b: numerical
data………………………………………………………………. 52
Figure 4.10. Time-averaged velocity vector map in side view symmetry
plane of the bus model……..…………………………………… 53
VII
Figure 4.11. Time-averaged velocity vectors in side view plane in the
forward face of the model………………………………………. 54
Figure 4.12. Time-averaged velocity vectors on the horizontal symmetry
plane near front of vehicle model………………………………. 54
Figure 4.13. Time-averaged velocity vector map in side view symmetry
plane downstream of the model…………………………………. 55
Figure 4.14. Contours of static pressure in the vertical symmetry plane…… 55
Figure 4.15. Contours of static pressure in the vertical symmetry plane near
the top of the model……………………………………………. 56
Figure 4.16. Velocity vectors on the vertical symmetry plane of the vehicle… 56
Figure 4.17. Time-averaged patterns of streamlines and corresponding
various plan views having different heights from the bottom of
the channel as a:Y/H=0.015, b:Y/H=0.045, c:Y/H=0.09,
d: horizontal symmetry plane of the vehicle, ………………….. 57
Figure 4.18. Time-averaged pattern of streamlines and corresponding
distribution of velocity contour in different planes having
different heights from the bottom of the channel as
a:Y/H=0.896, b:Y/H=1,045, c:top of the vehicle,
d:Y/H=1.433…………………………………………………….. 58
Figure 4.19. Planes mentioned in figure 4.17 and 4.18………………………. 58
Figure 4.20. Time-averaged patterns of streamlines and corresponding
velocity contour in side view plane……………………………... 59
Figure 4.21. Time-averaged patterns of streamlines and corresponding
velocity contour in plan view plane……………………………... 59
VIII
NOMENCLATURE
A surface area (m2)
B Width of the channel
vc specific heat at constant volume
F height of the channel
H Height of thevehicle
FD Drag force
I turbulence intensity (%)
k turbulent kinetic energy (m2/s2)
L length of the vehicle
l characteristic length scale of the turbulent flow
N frame number
p pressure (Pa)
np local pressure (Pa)
∞p free stream pressure (Pa)
q& heat flux
′′− ji uuρ Reynolds (turbulent) stress
jq− Reynolds (turbulent) heat flux
jq filtered heat flux
Pr Prandtl number
tPr turbulent Prandtl number
R specific gas constant
Re Reynolds number
S strain rate magnitude
T flow time (sec)
T temperature (K)
iu time averaged (mean) velocity components in the yx, and direction z
iu′ fluctuating velocity component in the yx, and direction z
IX
∞U free stream velocity (m/s)
iu velocity components ( for wvu ,, =i 1, 2 and 3 respectively)
ix independent space coordinates ( zyx ,, for =i 1, 2 and 3 respectively)
V volume of a computational cell (m3)
cw frequency (rad/s)
X non-dimensional x -distance ( DxX /= )
X1 Distance between inflow and model front face
X2 Distance between outflow and model back face
fX projected fin pattern length
Z non-dimensional −z coordinate ( sz /= )
Greek Symbols
ρ fluid density (kg/m3)
α thermal diffusivity (= )/( pCk ρ )
P∆ pressure drop
ν kinematic viscosity (m2/s2)
tν turbulent kinematic viscosity (m2/s2)
τ non-dimensional time (= ) sUt /
ε turbulent dissipation rate (m2/s3)
∆ averaged grid size
β frequency parameter
µ viscosity, kg/(ms)
tµ turbulent (eddy) viscosity
effµ effectivite viscosity ( teff µµµ += )
ijτ Reynolds stress tensor or SGS stress term or filtered stress
ϑ characteristic velocity of the turbulent flow
X
λ a second viscosity related to the dynamic viscosity )32( µλ −=
ϕ a flow property
Φ)
filtered variable
θ corrugation (wavy) angle (º)
κ von Karman constant
φ cell centered value of a flow variable (field variable)
ω time average square of the vorticity fluctuations
η ratio between the characteristic time scales of turbulence and mean flow field
subscripts
in at the inlet
w in the wall
exit at the exit
av average
crit critical
∞ free stream value
superscripts
* non-dimensional value
‘ fluctuating component
^ filtered variable −
time averaged (mean component)
XI
1.INTRODUCTION Serkan MEZARCIÖZ
1. INTRODUCTION
1.1. Importance of Aerodynamics and Flow Around Road Vehicles
Aerodynamic structure and the flow around the road vehicles have been
investigated for a long time. Significance of the aerodynamics is obvious, since it
affects the fuel consumption, wind noise, and vibration.
The complexity of the structures in this three dimensional flow makes
experimental studies very difficult. Furthermore experimental studies often provide
only information on some limited partition of the flow. (Point, line or a plain).
Computational Fluid Dynamics (CFD) gives a description of the flow in the entire
computational domain (Numerical Wind Tunnel) (Krajnovic and Davidson , 2002a).
The drag forces of importance to the vehicle designer are dominated primary
by the wake forces. Thus the prediction of the pressure coefficient at the rear of the
vehicle is of great importance. Although the RANS simulations have been successful
in predicting many parts of the flow around the vehicles, they have failed to predict
the effects of the unsteady wake on the body. It is believed that an unsteady
simulation such as Large Eddy Simulation (LES) will have greater success than
RANS in predicting the pressure at the rear of the vehicles and give better insight
into the flow around these bodies (Krajnovic and Davidson, 2001: Howard and
Pourquie, 2002).
Good aerodynamics is becoming more and more important, even for
commercial vehicles, as fuel prices rise. The characteristic value that describes the
air resistance is the drag coefficient (CD). Low CD values indicate low drag and allow
a higher terminal speed and lower fuel consumption. The air resistance and flow
characteristics around vehicle can be determined in a wind tunnel experiments or
Computational Fluid Dynamics (CFD) analysis. In this study a CFD analysis with a
Reynolds Average Navier Stokes (RANS) and k-ε turbulence model will be
employed in stead of experimental study.
As the speed of a vehicle increases, so does its drag. The increase in drag is as
the square of the speed, as can be seen from the formula of the drag force FDRAG
1
1.INTRODUCTION Serkan MEZARCIÖZ
2...21. VACF DDRAG ρ= ( 1.1)
Where, CD :Coefficient of drag
ρ :Density of fluid
A :Cross-sectional area exposed to the flow
V : Velocity
Most of the vehicle manufacturers are focusing on the subject of
aerodynamics and they are making some researches to improve the aerodynamic
structure and gain lower CD values. (See Figure 1.1).
Figure 1.1. An aerodynamically improved truck (MAN)
Here are some examples of the researches;
Development of Truck and Bus Aerodynamics using Computational Fluid
Dynamics (Takeuchi and Kohri, 1997).
Aerodynamic Simulations by Using Discontinuous Interface Grid and
Solution Adaptive Grid Method (Uchida et al., 1998).
2
1.INTRODUCTION Serkan MEZARCIÖZ
Development of Rear Spoiler for Hatchback Vehicles using Concurrent CFD
Method (Yamane et al., 1999).
The total drag of a vehicle is made up of the following components:
Pressurized-air resistance and induced air resistance, Surface air resistance, and
internal air resistance.
The induced air resistance is caused by the differences in air pressure that
arise between the top and bottom of the vehicle as it moves along Figure1.2.
Together, the pressurized-air resistance and the induced air resistance make up the
majority of the total drag, accounting for some 50-90 %.
Figure 1.2. Induced air resistances in the case of a fully faired semi-trailer outfit
Pressurized-air resistance is determined by the size of the areas of separated
flow. The main factor here is the size of the rear separation zone. At points where the
flow separates, a partial vacuum is formed, giving rise to the pressurized-air
resistance. Fundamentally, the aim is to ensure that the separation areas and hence
the partial vacuum zones are small (MAN, 2004).
The target should be to keep the separation zones and hence vacuum zones on
the vehicle as small as possible. By specifically influencing the turbulence on the
3
1.INTRODUCTION Serkan MEZARCIÖZ
rear end separation, a smaller vacuum and hence a smaller pressure resistance can be
achieved. A “stretching” of the boundary layer in the rear-end area can also lead to a
significant reduction in resistance. The induced resistance is a part of the vehicle’s
pressure resistance. Air pressure differences between the vehicle’s upper and lower
side produce cross-flows (Figure 1.3) that form two large longitudinal swirls together
with the roof flow (Figure 1.4). (BOSCH, 2002).
Figure 1.3. Induced air flow at the vehicle body
In their immediate vicinity such swirls induce low pressures. The “dead water
area” at the rear end is extended and thus leads to an increased pressure resistance.
4
1.INTRODUCTION Serkan MEZARCIÖZ
Figure 1.4. Longitudinal swirls caused by induced draft
The term surface air resistance is used to describe the frictional resistance of
the “outer skin” of the vehicle to air flow. It is more pronounced on long vehicles
such as semi-trailer outfits and buses. The surface air resistance makes up 3-30 % of
the total drag.
The internal resistance is the proportion of the air resistance to which the
vehicle is subject by virtue of the through-flow of air for cooling and for interior or
cabin ventilation (MAN, 2004). Air not only flows around a vehicle, but also through
the vehicle in order to cool down the aggregates and to ventilate the passenger
compartment. When air flows through the cooler, engine compartment, wheelhouses,
and passenger compartment (Figure 1.5), losses of momentum arise from friction as
well as turbulences and flow separation in the vehicle’s interior. The resulting
internal resistance which constitutes about 3-11% of the total aerodynamic drag only
makes up a small part. Internal air resistance accounts for 3-11 % of total drag.
(BOSCH, 2002). In this study internal resistance will not be considered.
5
1.INTRODUCTION Serkan MEZARCIÖZ
Figure 1.5. Air flow through a motor vehicle
The drag coefficient of a vehicle can be significantly reduced by rounding the
front section and using a front apron, a roof spoiler and side skirts. Only a slight
improvement in the drag coefficient is possible by altering the external shape of a
commercial-vehicle body since rounding the corners and edges reduces the load
space and hence the payload.
An aerodynamically designed cab causes an increase in the flow impinging
on a non-optimized body. The CD value is then higher than that with a sharp-edged
cab and a non-optimized body. The reason is that the front section of the body in the
case of a sharp-edged cab lies in a separation zone and is thus subject to a lower air
resistance.
In combination with such air deflectors, an aerodynamically optimized cabin
provides significant reductions in the drag of the vehicle as a whole.
Another measure for reducing resistance is to enclose the exposed running
gear of the vehicle with fairings. This reduces the air resistance of the vehicle
especially in a cross-wind.
Turbulence and the associated separation of the air flow increase drag to a
considerable extent. The rear of a vehicle combination, e.g. a semi-trailer outfit, is a
particularly problematic zone.
In order to optimize this are too, vertical metal flaps hinged to the rear of the
vehicle were used, an application of this can be seen in Figure 1.6. These flaps open
6
1.INTRODUCTION Serkan MEZARCIÖZ
at a speed of 50-55 km/h and reduce the size of the separation edge at the rear of the
vehicle.
To ensure that these flaps open and close automatically, there is a funnel on
each of the top corners at the rear of the vehicle. The airflow when the vehicle is in
motion blows through these funnels and inflates two air sacks situated behind the
flap, thereby opening the latter. When the vehicle slows down, the flaps close again
and reduce the length of the vehicle to the legal limits. TÜV, the German technical
testing organization, allows the prescribed vehicle length to be exceeded while the
vehicle is in motion (MAN, 2004).
Figure 1.6. Application of Rear Flap to a Truck (MAN, 2004)
Since the main dimensions of a vehicle are to a large extent regulated by law,
a reduction of the aerodynamic drag is possible only by means of a reduced CD-
value. Drag coefficients of CD=0.15 achieved in prototypes, indicate the high
potential for future development compared to today’s mass-produced passenger cars
(CD=0.26- 0.45) and commercial vehicles (CD= 0.6-0.8) (BOSCH, 2002).
7
1.INTRODUCTION Serkan MEZARCIÖZ
In order to achieve an aerodynamic body design, it is still necessary to
undertake experimental tests in a wind tunnel despite the availability of computer-
aided computational methods.
Corresponding to the degrees of freedom in three-dimensional space, three
forces (force of aerodynamic drag, lateral force, and lift) and three moments
(pitching, rolling, and yawing moment) (Figure 1.7) act on the vehicle. They are
usually determined by measuring the three forces coordinates on each wheel of the
vehicle. By geometrically adding up the forces recorded in the wind tunnel, the
aerodynamic forces and moments can be determined. This will not be considered
here in more detail. Out of the measured forces, the aerodynamic coefficients, e.g.
CD-value, are computed by combining them with the remaining constants. In this
case the vehicle flow should not be changed and a nearly distance and friction-free
measurement should be made possible (BOSCH, 2002).
Figure 1.7. Forces and moments in aerodynamic measurements
8
1.INTRODUCTION Serkan MEZARCIÖZ
Modern wind tunnels are used not only to determine the aerodynamic drag
and their coefficients, but also for the acoustical investigation of vehicles. The
increasing demand for noise comfort in passenger cars, in this respect requires
optimization and tests. The challenge involved in these tests lies in lowering the
ambient noise level in so-called aero-acoustic wind tunnels, to such a level that the
actual noise measurement on the vehicle cannot be falsified very much by the
tunnel’s own operating noise. Operating noise has been lowered down to 60dB
(BOSCH, 2002).
The vehicle’s surface resistance, in theory called frictional resistance, is of
significance in long vehicles, e.g. busses. Figure 1.8 shows the cumulative resistance
in an aerodynamically-efficient bus body consisting of negligible resistance at the
vehicle front, relatively high resistance at the rear end, and body resistance consisting
basically of frictional resistance which steadily increases with the length of vehicle
(BOSCH, 2002).
Figure 1.8.Components of aerodynamic drag in an optimized bus body (MAN, 2004)
When air flow attacks a vehicle at an angle, the drag coefficient significantly
changes. Figure 1.9 shows the influence of the angle of approach on the CD-value for
different passenger car designs (BOSCH, 2002). In this study oblique flow condition
is neglected.
9
1.INTRODUCTION Serkan MEZARCIÖZ
Figure 1.9. Aerodynamic drag at oblique flow for different passenger cars (BOSCH, 2002)
Consequently, for passenger cars as well as for commercial vehicles, there are
considerable increases in the coefficients of resistance when they are approached by
a side wind. In this case the maxima of the angle of approach β =25°-35° can be
achieved. When the vehicle is driven, angles of approach greater than 15° are
however exceptional.
1.2. Importance of Aerodynamics for Buses
Aerodynamic structure of the commercial vehicles especially buses are very
important, because aerodynamic forces and flow around the bus will cause a noise
called aero-acoustic noise, and vibration. Since the intercity buses transport
passengers, wind noise and vibration are much important than the trucks and trailers,
but the aerodynamic structure is also important for the trailers and trucks, from the
point of view of fuel consumption.
The experiments show that the fuel consumption can be lowered by
improving aerodynamic structure of the vehicle. The front apron and an optimally
positioned roof spoiler alone reduce fuel consumption by about 1.5 litres/100 km
(MAN, 2004).
10
1.INTRODUCTION Serkan MEZARCIÖZ
Also the flow characteristics around the elements of the bus such as mirrors,
wheels, air conditioning unit, escape hatches on the top of the bus, plays an important
role as a source of the noise and vibration. These parts must be designed and
positioned to the bus by considering the flow around the parts.
In order to improve the flow characteristics of the bus, the best way is to use
rounded corners in stead of sharp one. The flow characteristic of the bus, shown in
the Figure1.10, is better than the bus shown in Figure 1.11.
Figure 1.10. An application of rounded corner to a bus (TEMSA Diamond)
Figure 1.11. A bus having sharp corners (TEMSA Tourmalin)
11
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
2. PREVIOUS STUDIES AND SCOPE OF THE PRESENT STUDY
The drag forces of importance to the vehicle designer are dominated primary
by the wake forces, thus the prediction of the pressure coefficient at the rear of the
vehicle is of great importance. Also since the flow around a vehicle has an important
effect on the fuel consumption, noise, and vibration, lots of scientists and automotive
companies pay attention to this subject.
Takeuchi and Kohri (1997) has preformed a study for Mitsubishi Motors
Cooperation. The name of the study was “Development of Truck and Bus
Aerodynamics using Computational Fluid Dynamics“. This paper describes a
prediction method of aerodynamic drag and engine cooling performance for trucks
and buses using CFD. In particular, to obtain the accuracy of wake flow behind the
body, which is a dominant component of the total drag, they developed an adequate
method by comparing the experiment with calculation. Furthermore a practical
method for engine cooling air flow rate with regard to radiator and cooling fan
characteristics was shown.
“Aerodynamic Simulations by Using Discontinuous Interface Grid and
Solution Adaptive Grid Method” by Uchida et al. (1998), was carried out by the
financial support of Daihatsu Motor Cooperation Ltd. Aerodynamic simulations of
an automobile with the air-flow type spoiler using a discontinuous interface grid
method and flow simulation around the rear view door mirror using a solution
adaptive grid method were presented in their work. Consequently, it has become
possible to capture the detailed phenomena around these parts, such as the spoiler
and rear view door mirror.
Kuriyama (1998) conducted a study of “Transient Aerodynamic Simulation in
Crosswind” for Daihatsu Motor Cooperation Ltd. In this study, transient
aerodynamic simulation by using a sliding mesh of discontinuous interface and the
Arbitrary Lagrangian-Eulerian method was presented. This method uses the k-ε
turbulence model and the third-order upwind scheme, in order to introduce the
convective term of Navier-Stokes Equation to improve the data of flow field and
12
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
pressure distribution. Computed Yaw characteristics in crosswind are in good
agreement with the experiments.
Sumitani and Murayama (1999) reported that, airflow effect is one of the
important functions demanded of a rear spoiler. It helps prevent mud or dust from
swirling up behind the running vehicle, or, in the case of driving in rain or snow,
helps prevent rain or snow from adhering to the rear window. They often decide on
the shape of a spoiler in a relatively short time, focusing primarily on its appearance.
Therefore, they established a design method using recently developed computational
fluid dynamics to determine the central cross sectional shape of spoiler that realizes a
desired airflow effect and verified its effectiveness through testing.
In order to determine the flow around a road vehicle there are different
techniques, the most commonly used are wind tunnel experiments, and numerical
analysis of Computational Fluid Dynamics, in the case of CFD analysis two
challenging techniques Reynolds-Average Navier Stokes (RANS) and Large Eddy
Simulation (LES) are used.
Comparison of LES and RANS calculations of the flow around bluff bodies
was conducted by (Rodi 1997). His work compares LES and RANS calculations of
vortex-shedding flow past a square cylinder at Re = 22.000 and of the 3D flow past a
surface-mounted cube at Re= 40.000. The RANS calculations were obtained with
various versions of the k-ε model and in the square-cylinder case also with Reynold-
stress models, the various calculation results are compared with detailed
experimental data and an assessment is given of the performance, the cost and the
potential of the various methods. As a result of this study, Calculations obtained with
a variety of LES and RANS methods have been presented for two basic bluff body
flows with relatively simple body geometries albeit complex flow behavior. The
comparison with detailed measurements has shown that the main features of these
complex flows can be predicted reasonably well, at least with some of the methods.
The square cylinder results are not entirely satisfactory and do not provide a uniform
picture. It is clear, however, that for this flow the Standard k-ε model produces rather
poor results. This is to a large extent due to the excessive turbulence production in
the stagnation flow resulting from the use of this model. The Kato-Launder
13
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
modification removes this problem and yields improved results; a further
considerable improvement can be obtained when this model is combined with the
two-layer approach resolving the near-wall region. The excessive turbulence
production problem is also absent when a Reynolds-stress model is used which,
however, tends to over predict the periodic motion. In all RANS calculations, the
turbulence fluctuations are severely under predicted; the fairly high value of these
fluctuations in the experiment may stem from low frequency variations of the
shedding motion due to 3D effects which cannot be accounted for in 2D RANS
calculations. LES seems to pick up these motions and in general gives a better
simulation of the details of the flow (Rodi 1997).
The price to be paid for this is a large increase in computing time: the
UKAHY2 LES calculations took 73 h on a SNI S600/20 vector computer while the
RANS calculations using wall functions took 2 h and the ones using the two-layer
approach 8 h on the same computer. Further, it was found that none of the LES
results are uniformly good and entirely satisfactory and there were large differences
between the individual calculations which are difficult to explain. Reasons for the
lack of agreement with the experiments include insufficient resolution near the side
walls of the cylinder where the separated shear layer undergoes transition and a thin
reverse flow region is present, neglect of the turbulence in the incoming stream,
numerical diffusion and insufficient domain extent and number of grid points in the
span wise direction. Hence, this flow was selected once more as a test case for a LES
workshop held in Grenoble in September 1996 (Rodi 1997).
For the cube flow, the same problem with the excessive k-production in the
stagnation region exists when the standard form of the k-ε. Model is employed; and
this leads to poor predictions of the flow over the roof. These are significantly
improved by introducing the Kato Launder modification and also when the two layer
approach is used, and only with the latter can the complex structure of the near-wall
streamlines be simulated. However, both modifications increase even further the
length of the separation region behind the cube which is too large already for the
standard model.
14
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
RANS methods with statistical turbulence models will be needed and used for
many years to come in engineering calculations of the flow past bluff bodies.
However, inaccuracies must be accepted, and this comparative study has
demonstrated that LES is clearly more suited and has great potential for calculating
these complex flows. Further development and testing is certainly necessary, but
with the recent advances in computing power LES will soon be ready and feasible
for practical applications (Rodi 1997).
Krajnovic and Davidson (2001), investigated two large eddy simulation of the
flow past a bus-like vehicle body, and they compared the results with the
experimental data of E.G. Duell and A.R. George (1999). “Experimental study of a
ground vehicle body unsteady near wake The effect of the near wall resolution and
the modeling of the unresolved coherent structures in the near wall flow were
studied. The purpose of this work was to present LES of the flow around a bus-like
bluff body at Re=210000. They compared two LES in which the near wall region
was treated in different ways. It was determined that, although the wall functions are
inadequate to represent the thin vortices close to the wall, their use leads to results in
a near wake region that are similar to those in the simulation with a sufficient wall
normal resolution. That study indicated that the wall normal resolution has little
influence on the pressure coefficient at the rear face.
The geometry used by Krajnovic and Davidson (2001), is shown in the
flowing figure.
Figure 2.1. Geometry of the vehicle body and computational domain used by
Krajnovic and Davidson (2001)
15
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
The relationship between the bus-like shape and the channel dimensions are;
A domain with an upstream length X1/H=8, a downstream length of X2/H=16, and a
span-wise width of B=5.92H, was used for the simulation. The values of the other
geometrical quantities are L=0.46m, H=0.125m, W=0.125m, S=0.3075m,
R=0.0.019m, r=0.0127m, c=0.01m, and C=0.5 m. The ground clearance of c/H=0.08
is similar to the clearance ratio of buses. The Reynolds number Re=U.H/υ was
210000 on the basis of the incoming mean velocity, U and the vehicle height, H. The
cross-section of the tunnel test section, the ground clearance and the position of the
model’s cross-section with respect to the tunnel were identical in LES and
experimental set-up (Krajnovic and Davidson 2001).
In the study “Computational Study of Flow around the AHMED Car Body”
carried out by Craft et al (2001). A number of RANS simulations of flow around the
Ahmed body have been undertaken on Refined Turbulence Modeling. The
simulations have involved two different turbulence methods: a linear and non-linear
k-ε model, and two different wall functions. In this study calculations were
converged until, velocity, mass and turbulence residuals were below 10-4. Grid used
is approximately 300.000 cells and the legs, or stilts, on which the model is
supported in the wind tunnel experiments, were not included in the computational
grid. Grid were adjusted to maintain y* values of as many as possible near-wall cells
around the body to within the limits 55
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
Figure 2.2. Kinetic energy contours at the centerline (y=0) around the 25º Ahmed body
using the linear k-ε and cubic non-linear k-ε models with specified Chieng and Launder wall functions Craft et al (2001)
The development of attached or separated flow over the rear slant is strongly
influenced by the presence of the side edge vortices which draw fluid out of the
boundary layer on the rear slant. The weaker, more defined vortices predicted by, the
non-linear model led to the separation of the boundary layer. The velocity field was
reasonable predicted by the linear k-ε model over the 25º Ahmed body. But the
predicted stream wise normal stress was an order-of-magnitude lower than the
experimental values. For the 35º Ahmed body, both linear and non-linear models
correctly predicted separated flow over the rear slant. The linear k-ε model gave
reasonable predictions for the wake dimensions and velocity, although the predicted
turbulent kinetic energy was too high. The wake predicted by the non-linear k-ε
model was both too high in the z-axis and too long in the x-axis. The relatively poor
prediction of the wake size with non-linear model is in agreement with an earlier
study of the flow over a square cylinder, Craft et al (2001).
The feasibility of use of large-eddy simulation (LES) in external vehicle
aerodynamics is investigated. The computational cost needed for LES of the full size
car at road conditions is beyond the capability of the computers in the near future
17
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
(Krajnovic, 2002). Since LES cannot be used for quantitative prediction of this flow,
i.e. obtaining the aerodynamic forces and moments, an alternative use of this
technique is suggested that can enhance the understanding of the flow around a car. It
is found that making LES of the flow around simplified car-like shapes at lower
Reynolds number can increase our knowledge of the flow around a car. Two
simulations are made, one of the flows around a cube and the other of the flow
around a simplified bus. The former simulation proved that LES with relatively
coarse resolution and simple inlet boundary condition can provide accurate results.
The latter simulation resulted in flow in agreement with experimental observations
and displayed some flow features that were not observed in experiments or steady
simulations of such flows. This simulation gave information to study the transient
mechanisms that are responsible for the aerodynamic properties of a car. The
knowledge gained from this simulation can be used by the stylist to tune the
aerodynamics of the car’s design but also by the CFD specialists to improve the
turbulence models (Krajnovic and Davidson, 2002b). Geometry of the cube with its
channel dimensions are shown in the Figure 2.3.
Figure 2.3. Geometry of the cube with its channel dimensions (Krajnovic and
Davidson, 2002b)
18
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
Turbulence modeling plays a vital role in any numerical simulation in terms
of accuracy and computational cost. Wurtzler (2003) presents a Spalart-Allamaras
based detached eddy simulation hybrid model and numerical results for the Ahmed
reference car model with 25º base slant angle. Highly three-dimensional and
unsteady wake flow behavior is documented by showing velocity vectors in the
trailing region. One-equation RANS model is also used for the same simulation.
Both techniques are compared by showing the capability of each technique in
capturing the minor flow details and in predicting the coefficient of drag (CD).
Finally, unsteady behavior of CD is studied in both cases. Average value of CD is
calculated and validated with the reported experimental data of Ahmed et al. and
numerical results of Gillieron and Chometon. Further, brief discussion about the
present day available turbulence modeling techniques including DNS, LES, RANS
and DES is also done in the study of Detached Eddy Simulation over a reference
Ahmed Car Model.
In the study of “a comparison of large eddy simulations with a standard k-ε
Reynolds-Averaged Navier Stokes model for the prediction of a fully developed
turbulent flow over a matrix of cubes” by Cheng et al (2003), a fully developed
turbulent flow over a matrix of cubes has been studied using the large Eddy
simulation (LES) and Reynolds-averaged Navier–Stokes (RANS) [more specifically,
the standard k–e model] approaches. A comparison of predicted model results for
mean flow and turbulence with the corresponding experimental data showed that
both the LES and RANS approaches were able to predict the main characteristics of
the mean flow in the array of cubes reasonably well. LES, particularly when used
with LDM, was found to perform much better than RANS in terms of its predictions
of the span-wise mean velocity and Reynolds stresses. Flow structures in the
proximity of a cube, such as separation at the sharp leading top and side edges of the
cube, recirculation in front of the cube, and the arch-type vortex in the wake are
captured by both the LES and RANS approaches. However, LES was found to give a
better overall quantitative agreement with the experimental data than RANS (Cheng
et al., 2003).
19
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
In journal of “LES as a powerful Engineering tool for predicting complex
turbulent flows and related phenomena” Inagaki (2004) reported the advantages of
LES over the k-ε model as follows; high prediction accuracy, capability of resolving
the unsteadiness of turbulent motion over a broad range of scales, and simplicity of
modeling the turbulent effects in the fluid phenomena containing multiphysics
(Inagaki, 2004).
In the present numerical study the dimensions of the channel was determined
by referencing the study carried out by Krajnovic and Davidson (2004a). The
relationship between the generic vehicle body and channel dimensions were chosen
in that study and the present numerical analysis to be the same as used in the
experiments by Ahmed et al (1984) and Lienhart and Becker (2003). The geometry
of the body and the computational domain used by Krajnovic and Davidson (2004a)
are given in Figure 2.4. The body is placed in the channel with cross section of
BxF=6.493Hx4.861H (Width x Height). The cross section of this channel is identical
with the open test section of the wind tunnel used in the experiments of Lienhart and
Becker (2003). The front face of the body is located at the distance of X1=7.3H from
the channel inlet and the downstream length between the rear face of the body and
the channel outlet is X2=21H. The body is lifted from the floor producing the ground
clearance of c/H=0.174, same as in the experiments. The Reynolds number, based on
the incoming velocity U and the car height H, of Re=7.68x10∞ 5 used in the
experiments Lienhart and Becker (2003) was reduced to Re=2x105.
Krajnovic and Davidson (2004a) have already demonstrated successful LES
of this lower Reynolds number case with no rear body slant angle, i.e. α=0º (generic
bus body). Krajnovic and Davidson expect that the slanted rear end will produce a
wider spectrum of turbulent scales that must be resolved in LES.
20
2.PREVIOUS STUDIES Serkan MEZARCIÖZ
Figure 2.4. Schematic representation of the computational domain with vehicle body
(Krajnovic and Davidson, 2004a)
2.1. Scope of the Present Study
The purpose of the present study is to show a numerical study of a flow
around a bus-like shape taking into account the 3D geometry, the effect on the flow
around the bus. This study has been done with the commercial software package
FLUENT. This code uses FVM and solves 3D flows.
In the present study the dimensions of the model was the same as the
dimensions of the model which was used by Gürlek (2006) in his experimental work.
For the channel dimensions of the present study, we referenced the ratios of
dimensions which were used by Krajnovic and Davidson (2004a).
By using GAMBIT program grid generation and identification of boundary
condition were performed and using FLUENT program finite volume analysis was
executed by k-ε turbulence model.
In order to get more precise results, in the present numerical simulation,
calculations will be converged until residuals were below 10-5 and the grid used has
935.004 cells. As in the study of Craft et al (2001), we also did not include the legs
of the model for the computation domain, and the present y* value was in range of
Craft et al (2001).
The analysis performed numerically was compared to the experimental results
of Gürlek (2006).
21
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
3. MATERIALS AND METHODS
Experimental fluid dynamics has played an important role in validating and
defining the limits of the various approximations to the governing equations. The
wind tunnel, for example, as a piece of experimental equipment, provides an
effective means of simulating real flows. Traditionally this gas provided a cost
effective alternative to full-scale measurement. However, in the design of equipment
that depends critically on the flow behavior, for example the aerodynamic design of
an aircraft or a bus, full-scale measurement, as part of the design process is
economically impractical. This situation has led to an increasing interest in the
development of numerical studies.
Computational Fluid Dynamics (CFD) has proven capability in predicting the
detailed flow behavior for wide-ranging engineering applications, leading typically
to an improved equipment or process design.
Most of the vehicle manufacturers are, now using CFD a simulation toll
originally developed for academic research and aerospace. CFD is used to solve the
RANS Equations that describe fluid flow and to generate a 3D model of that flow,
making possible more effective vehicles and minimizing the aero acoustic noise, and
vibration.
The object of CFD is to use computers to solve the previously intractable
conservation equations for fluids in order to accurately simulate flows. This typically
involves discretizing the problem in a finite set of elements, applying the
conservation of mass, momentum, energy, and chemical species (where necessary) to
these elements, placing additional boundary conditions at the edge of the
computational grids, and solving the resultant algebraic equations in an iterative
fashion (Öztürk, 2004).
Thus, CFD allows the analysis of fluid flow problems in detail, faster and
earlier in the design cycle than possible with experiments, costing less money and
lowering the risks involved in the design process. This trend is only likely to grow
more pronounced in the future as computers become increasingly cheaper and more
powerful while traditional forms of testing become increasingly expensive. For
22
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
improving aerodynamic structure of the ground vehicles, it is great importance to
carry out investigation on the flow behavior indeed. With the currently increasing
computer capability and CFD developed, numerical simulations of the 3D turbulent
flow around the road vehicles is becoming possible.
The purpose of the present study is to show a numerical study of a flow
around a bus-like shape taking into account the 3D geometry, the effect on the flow
around the bus. This study has been done with the commercial software package
FLUENT. This code uses FVM and solves 3D flows. For the solution, the grid was
generated by GAMBIT package program.
3.1. Vehicle Geometry and Dimensions
In the present study dimensions of the model geometry was the same as
dimensions which were used in the experiments of Gürlek (2006). The length of the
model L=175mm, the height of the model H is 67mm, and the width of the vehicle
W is 56 mm. The ground clearance of the vehicle c is taken as 9 mm. The geometry
of the model is shown in the Figure 3.1.
Figure 3.1. The dimensions of the vehicle model.
23
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
3.2. Determination of Dimensions of the Flow Domain
In order to perform a computer simulation of a physical flow problem, the
first step is to build a model of the flow domain. In the present study, to determine
the dimensions of the flow domain, the dimensions normalized by the vehicle height
H, is referenced as in Krajnovic and Davidson (2004a and 2004b). According to
Krajnovic and Davidson (2004a) the relationship between the channel dimensions
and vehicle height H, are as follows; The body was placed in the channel with cross
section of B x F=6.493H x 4.861H (Width x Height). The front face of the body is
located at the distance of X1=7.3H from the channel inlet and the downstream length
between the rear face of the body and the channel outlet is X2=21H and the distance
between the side walls of bus and channel S is 2.571H. Schematic representation of
the computational domain with vehicle body (Krajnovic and Davidson, (2004)) is
shown in Figure 3.2.
Figure 3.2. Schematic representation of the computational domain with vehicle body
(Krajnovic and Davidson, 2004a)
By referencing the ratios of Krajnovic and Davidson, (2004a), the present
flow geometry dimensions and the position of the vehicle model in the channel were
determined as follows;
Width of the channel (B) = 6.493xH = 6.493x 67=435 mm.
Height of the channel (F)= 4.861xH=4.861x67= 326 mm.
Distance between inflow and model front face (X1)= 7.3xH=7.3x67=489 mm
Distance between outflow and model back face (X2)= 21xH=21x67=1407mm
Overall length of the channel=X1+L+ X2= 489+ 175+1407=2071 mm.
24
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
With these results the dimensions of the computational domain and the
position of the model in the domain is determined and shown in the Figures 3.3 , 3.4
and 3.5.
Figure 3.3. Position of the bus in the channel (side view, dimensions in mm)
Figure 3.4. Position of the bus in the channel (cross-sectional view)
Figure 3.5. 3D view of the computational domain
25
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
3.3. Grid Generation and Refinement Study
A mesh is a set of points distributed over a calculation field for a numerical
solution of a set of partial differential equations. Essential CFD involves the setting
up partial differential equations and boundary conditions describing fluid flow over a
physical domain these equations are then solved numerically to generate the pressure
and velocity fields in the flow domain.
In order to perform the discretisation of the governing equations the physical
space must be subdivided into number of cells by means of a structured grid. The
grid can be constructed using a rectangular, a cylindrical or a body fitted coordinate
system.
In the present study the 3-D finite volume, rectangular mesh is built by using
GAMBIT computer code In order to build the models, firstly the geometry of the
flow domain is described by defining discrete set of points (nodes) and collection of
these points (elements). The points are joined to create lines, arcs, circles, and finally
the volumes are created.
To validate the numerical results, it is great importance to check whether the
grid that has been used to discretize the computational domain complies with the
requirements of the turbulence model used.
To check the grid used for the analysis, simulations were performed with
three different grids. The used firs second and third grids have 442.496, 786.586 and
1.186.819 nodes, respectively. The grids were compressed near the vehicle body
walls. The grids were adapted in order to further resolve the flow near the body wall
regions.
It is decided for the grid compatibility by checking the y* value, Craft et al
(2001) used a grid, which were adjusted to maintain y* values of as many as possible
near wall cells around the body within the limits 55
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
can not be used for this simulation. In the Figure 3.6, distribution of y* values for
442.496 nodes can be seen.
Figure 3.6. Distribution of y* values for the grid having 442.496 nodes
When the number of cells is 658.903 and nodes is 786.586, the value of y* is
in the range of 0
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
Lowering the y* value gives more precise results and provides us the grid
independency, but lowering the y* value, in other words refining the grid, is limited
by the computational cost of the simulation. In the present study y* value is lower to
the range of 0
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
3.4.1. Boundary Conditions for Simulations
1. Velocity Inlet boundary conditions are used at the inflow section of the
channel as approximately 90 km/h (25 m/s). it is assumed that only a constant and
uniform stream wise-direction velocity component exists and other velocity
components are assumed to be zero.
2. Free outflow boundary condition: The outflow boundary condition is used
to model flow exist where the details of the flow velocity and pressure are not known
prior to solution of the flow problem so that mass balance correction is applied at the
outflow boundary and other data at the exit plane is extrapolated from the interior.
Diffusion fluxes for all flow variables in the direction normal to the exit plane are to
be zero. Therefore the outflow velocity is consistent with a close to fully developed
flow assumption.
3. Wall boundary condition is used to bound fluid and solid regions at the
vehicle and channel surfaces. The boundary conditions are no- slip ( ) 0=== wvu
4. Fluid boundary condition at the fluid zone. Fluid zone is defined as group
of cells for which all active equations are solved. The fluid material is set as air.
3.5. Simulation of the Flow
In the present study, in order to simulate the flow around the bus-shape
model, a commercial package program FLUENT was employed.
For all flows, FLUENT solves conservation equations for mass and
momentum. For flows involving heat transfer or compressibility, an additional
equation for energy conservation is solved Additional transport equations are also
solved when the flow is turbulent. FLUENT employs the finite-volume-based
technique. In the finite volume method the solution domain is subdivided into a finite
number of discrete continuous control volumes (CVs) or cells, and the conservation
equations are applied to each CV. The governing conservation equations are
integrated on the individual control volumes to construct the algebraic equations for
the discrete dependent variables (unknowns) such as velocity, pressure, temperature
29
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
and scalars. The discretized equations are linearized and the resultant linear system
of equations is solved to yield updated values of the dependent variables (FLUENT,
1998).
3.5.1. Discretization
FLUENT uses a control volume based technique to convert the governing
equations to algebraic equations that can be solved numerically. This control volume
technique consist of integrating the governing equations about each control volume,
yielding discrete equations that conserve each quantity on a control-volume basis.
Discretization of the governing equations can be illustrated most easily by
considering the steady-state conservation equation for transport of a scalar quantity
φ . This is demonstrated by the following equation written in integral form for an
arbitrary control volume V as follows:
∫ Adrr
.νρ = ∫∫ +∇ΓV
dVSAd φφ φr
. (3.1)
where ρ is the density, νr is te velocity vector ( ), kwjviu ˆˆˆ ++= Ar
is the surface
area vector, is the diffusion coefficient for φΓ φ , φ∇ is the gradient of
,ˆˆˆ ⎟⎟⎠
⎞⎜⎜⎝
⎛⎟⎠⎞
⎜⎝⎛∂∂
+⎟⎟⎠
⎞⎜⎜⎝
⎛∂∂
+⎟⎠⎞
⎜⎝⎛∂∂
= kz
jy
ix
φφφφ and is the source of φS φ per unit volume.
Equation (3.6) is applied to each control volume, or cell, in the computational
domain. Discretization of Equation (3.2) on a given cell yields
(3.2) =∑ ffN
fff A
faces rr .φνρ VSAfN
fn
faces
φφ φ +∇Γ∑r
.)(
30
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
where is the number of faces enclosing cell, facesN fφ is the value of φ convected
through face fff Afrr ., νρ is the mass flux through the face, fA
r is the area of , f
nzyx kAjAiAA )(,ˆˆˆ φ∇++= is the magnitude of φ∇ normal to face , and V is the
cell volume.
f
Fluent stores discrete values of φ at the cell centers. However, face values of
fφ are required for the convection terms in Equation (3.1) and must be interpolated
from the cell center values. This is accomplished using an upwind scheme.
Then the discretized equations are linearized and resultant linear equation
system are solved to yield updated variables of the dependent variables.
In the segregated solution method, the governing equations are solved
sequentially. Because the governing equations are non-linear (and coupled), several
iterations of the solution loop must be performed before a converged solution is
obtained. Each iteration consists of the steps shown below:
1. Fluid properties are updated based on the current solution.
2. The and momentum equations are each solved in turn using
current values for pressure and face mass fluxes, in order to update the
velocity field.
vu, w
3. Since the velocities obtained in Step 2 may not satisfy the continuity
equation locally, a “Poisson-type” equation for the pressure correction is
derived from the continuity equation and the linearized momentum
equations. This pressure correction equation is then solved to obtain the
necessary corrections to the pressure and velocity fields and the face mass
fluxes such that continuity is satisfied.
4. Where appropriate, equations for scalars such as turbulence, energy,
species, and radiation are solved using the previously updated values of
the other variables.
5. A check for convergence of the equation set is made.
These steps are continued until the convergence criteria are met.
31
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
In the segregated solution method, the discrete, non-linear governing
equations are linearized to produce a system of equations for the dependent variables
in every computational cell. The resultant linear system is then solved to yield an
updated flow-field solution. Implicit means: For a given variable, the unknown value
in each cell is computed using a relation that includes both existing and unknown
values from neighboring cells. Therefore each unknown will appear in more than one
equation in the system, and these equations must be solved simultaneously to give
the unknown quantities.
In the segregated solution method, each discrete governing equation is
linearized implicitly with respect to that equation’s dependent variable. This will
result in a system of linear equations with one equation for each cell in the domain.
Because there is only one equation per cell, this is sometimes called a “scalar”
system of equations. A point implicit (Gauss-Seidel) linear equation solver is used in
conjunction with an algebraic multigrid (AMG) method to solve the resultant scalar
system of equation is linearized to produce a system of equations in which u
velocity is the unknown. Simultaneous solution of this equation system (using the
scalar AMG solver) yields an updated velocity field. u
In summary, the segregated approach solves for a single variable field
(e.g., p ) by considering all cells at the same time. It then solves for the next variable
field by again considering all cells at the same time, and so on (FLUENT, 1998).
3.5.2. Convergence
The convergence criterion for the continuity and momentum equations is set
as All the simulations are performed until the solution is converged with the
specified convergence criterions. The convergence of the residuals of the last 229
iterations can be seen in Figure 3.9.
510−
32
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
Figure 3.9. Convergence of the residuals to 10-5
3.6. Governing Equations for Fluid Flow
All physical systems obey three fundamental conservation laws: Mass is
conserved, Momentum is conserved, and Energy is conserved. The governing
equations of fluid flow and heat transfer represent mathematical statements of the
conservation laws of physics. The conservation equations are derived by considering
the mass, momentum and energy balances for an infinitesimal control volume. In the
absence of an external body force, differential forms of the governing continuity,
momentum and energy equations for the laminar case used in the present study can
be given as follows:
* Continuity Equation. The mass conservation or the continuity equation for
a time-dependent, three-dimensional, incompressible and Newtonian fluid is given in
Eqn. (3.3).
0=∂∂
i
i
xu
(3.3)
33
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
* Momentum Equation. The most useful form of the conservation of the
momentum equation is obtained by applying Newton’s Second Law of motion to an
infinitesimal fluid element and re-arranging the viscous stress terms yields the so
called Navier-Stokes Equation.
( )
⎥⎥⎦
⎤
⎢⎢⎣
⎡⎟⎟⎠
⎞⎜⎜⎝
⎛
∂
∂+
∂∂
∂∂
+∂∂
−=∂
∂+
∂∂
i
j
j
i
jij
iji
xu
xu
xxp
xuu
tu
νρ1 (3.4)
* Energy Equation. The energy of a fluid is defined as the sum of internal
(thermal) energy, kinetic energy and gravitational potential energy. The internal
energy equation for a time-dependent, three-dimensional, incompressible and
Newtonian fluid is given in Eqn. (3.5).
( )jj
ji x
Tx
Txt
T∂∂
∂∂
=∂∂∂
+∂∂ α (3.5)
3.7. Reynolds Averaged Navier Stokes (RANS) Equations of Motion
The Navier-Stokes Equations are a set of non-linear partial differential
equations and they can only be able to describe continuous and homogeneous fluid
flows approximately. However, at very small scales or under extreme conditions, real
fluids made out of mixtures of discrete molecules and other materials, such as
suspended particles and dissolved gases, produce different results form the
continuous and homogeneous fluids modeled by the Navier- Stokes Equations.
Although the full, unsteady Navier-Stokes Equations correctly describe nearly all
flows of practical interest, they are too complex for practical solution in many cases
and a special “reduced” form of the full equations is often used instead-these are the
Reynolds Averaged Navier Stokes (RANS) equations.
In Reynolds Averaging, the solution variables in the instantaneous (exact)
Navier-Stokes equations are decomposed into the mean (ensemble-averaged or time-
averaged) and fluctuating components. Now all flow variables in the equations can
34
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
be displaced by sum of a mean (ensemble) and a fluctuating component: Reynolds
averaging means replacing the time-varying quantities with time-averaged (mean)
and fluctuating components as
u=u + ; v = u′ vv ′+ ;w= ww ′+ ; p= pp ′+
Whereu v w , are mean velocity components, andu′ , v′ , w′ are fluctuating velocity
components.
The solution of the full steady Navier-Stokes Equations is sufficiently
accurate only for laminar cases of the fluid flow. For turbulent flows the Reynolds
averaged form of the equations are most commonly used. The RANS form of the
equations introduce new terms that reflect the additional modeling of the small
turbulent motions.
In the absence of an external body forces, the governing time averaged
conservation equations for an incompressible, Newtonian fluid of constant density
and constant viscosity take the following forms:
1)Time averaged continuity equation:
0=∂∂
i
i
xu
(3.6)
2) Time averaged momentum equation:
( )⎥⎥⎦
⎤
⎢⎢⎣
⎡ ′′−⎟⎟⎠
⎞⎜⎜⎝
⎛
∂
∂+
∂∂
∂∂
+∂∂
−=∂
∂+
∂∂
jii
j
j
i
jij
iji uuxu
xu
xxp
xuu
tu
νρ1 (3.7)
3) Time averaged energy equation:
( ) jjj
ji
qxT
xTu
xtT
−∂∂
∂∂
=∂∂
+∂∂ α (3.8)
35
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
Reynolds averaging introduces additional terms in the momentum and energy
equations which are known as “Reynolds (turbulent) stress”, - and the
“Reynolds (turbulent) heat flux”, - , which appear on the right-hand side of the
momentum and the energy equations, respectively. The task of turbulence modeling
is to model these unknown, higher-order, extra terms in the mean flow equations and
thus close the system of equations. The need to model these correlations is the
‘closure problem’. It is also possible to model a transport equation for the heat flux,
but this is not a common practice. Instead, a turbulent thermal diffusivity is defined
proportional to the turbulent viscosity. The constant of proportionality is called the
turbulent Prandtl number, Pr
,′′ ji uuρ
jq
t (0.85
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
1.Models that use the Boussinesq Approximation: These are the eddy
viscosity models.
2.Models that solve directly for the Reynolds Stresses. These become
complicated fast by introducing further terms requiring modeling.
3.Models not based on time-averaging: These are Large Eddy Simulation
(LES) and Direct Numerical Simulation (DNS) methods.
3.9. Turbulence Models
A turbulence model is a computational procedure to close the system of mean
flow equations so that a more or less wide variety of flow problems can be
calculated. For most engineering purposes it is unnecessary to resolve the details of
the turbulent fluctuations. Only the effects of the turbulence on the mean flow are
usually sought. The most common turbulence models are classified as follows:
3.9.1. Time Averaged Turbulence Models
Time Averaged Turbulence Models are also referred as RANS Equations
Based Turbulence Models or Classical Models. These are:
1. Zero Equation Model (Mixing Length Model)
2. One Equation Model
3. Two Equation Model (k-ε Model, k-w Model)
4. Reynolds Stress Model
5. Algebraic Stress Model
These models model all scales of eddies present in the flow. Zero, One and
Two Equation Models use Boussinesq Hypothesis to relate the Reynolds stresses to
the time averaged velocity gradients as shown below:
- =′′vuρ ⎟⎟⎠
⎞⎜⎜⎝
⎛∂∂
+∂∂
xv
yu
tµ (3.9)
37
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
Where tµ is the turbulent (eddy) viscosity. The advantage of this approach is
that relatively low computational cost is required to compute the turbulent viscosity.
3.9.1.1. Zero Equation Model
The Zero Equation Model was the first turbulence model proposed by Prandtl
in 1925 and is also known as Prandtl Mixing Length Model. In zero-equation model,
the flow is characterized by one velocity scale and one length scale and the turbulent
viscosity is assumed to be constant. Therefore no transport of turbulence is resolved
in the zero equation models. The eddy viscosity expressed as
yuLC mt ∂∂
= 2µρµ (3.10)
Where Lm is the mixing length and is a constant. The model can be
successively applied to the simple flows and the mixing length can be specified by an
empirical formula in most situations. However, the model fail for separating flows
and cannot be applied to the rapidly developing and recalculating flows where
convective or diffusive transport of turbulence are important.
µC
3.9.1.2. One Equation Model
In one- equation model a transport equation turbulent kinetic energy is solved
and some physically based arguments for the mixing length is solved. This means
that the turbulent eddy viscosity now becomes a dependent variable that varies
spatially and with time for unsteady flows. The turbulent kinetic energy is considered
to be the velocity scale it is contained in the large scale fluctuations and is expressed
as follows:
k = ( )22221 wvu ′+′+′ (3.11)
38
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
The turbulent eddy viscosity is modeled by
st LkCµρµ ′= (3.12)
where and represent empirical constant and length scale, respectively. The
length scale distribution cannot be specified empirically for complex flows. Although
theoretical formulae for determining the length scale are proposed based on the mean
flow, they are very dependent on the problem type, they have not been tested
sufficiently and they require rather expensive computing time.
µC ′ sL
3.9.1.3. Two Equation Models
In two equation models, the RANS equations are closed by assuming the
turbulent stresses are proportional to the mean velocity gradients and the constant of
proportionality is the turbulent viscosity (Boussinesq Hypothesis). The constant of
proportionality is allowed to vary through out of the flow field and has been
correlated in terms of the turbulent kinetic energy and dissipation rate. Various
transport equations are developed for the turbulent kinetic energy and dissipation
rate.
* Standard Linear Turbulence k-ε Model
The standard linear k-ε turbulence model is well known and the model has
been tested for vortex shedding flow by Majumdar and Rodi (1985) previously.
Turbulent kinematic viscosity, vt, is related to kinetic energy k and turbulent
dissipation ε by a following formula:
εµ2kCvt = (3.13)
39
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
The forms of the k and ε equation of the Standard Linear k-ε model are then
gives as follows:
ε−∂∂
⎟⎟⎠
⎞⎜⎜⎝
⎛
∂
∂+
∂∂
+⎟⎟⎠
⎞⎜⎜⎝
⎛
∂∂
∂∂∂
=j
i
i
j
j
it
j
k
k
t
j xu
xu
xu
vx
vxDt
Dk (3.14)
kC
xu
xu
xu
vk
Cx
vxDt
D
j
i
i
j
j
it
j
t
j
2
21εεε ε
ε
−∂∂
⎟⎟⎠
⎞⎜⎜⎝
⎛
∂
∂+
∂∂
+⎟⎟⎠
⎞⎜⎜⎝
⎛
∂∂
∂∂∂
= (3.15)
The coefficients εµ σσ ,,,, 21 kCCC are constants in the sense that they are not
changed in any calculation. However, these constants need to be changed in order to
accommodate the effects such as low Reynolds number, near wall, etc. these five
coefficients have a value as shown below for the Standard k -ε model:
09.0=µC
C1 = 1.44
C2 = 1.92
00.1=kσ
30.1=εσ
The turbulent heat transport on the other hand is modeled by using the
concept of Reynolds’ analogy to turbulent momentum transfer by the use of Eqn.
(3.6).
The standard linear k -ε turbulence model is extended by Launder and
Spalding (1972) and named by ω−k and k – L models where Ls is the length
characterizing the macro scale of turbulence and ω is the time average square of the
vorticity fluctuations as defined by Wilcox and Rubesin (1980):
( )22
kC f
εω = (3.16)
40
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
ii
jj
xxuu∂∂
′∂′∂=νε (3.17)
ε
2/3kCL fs = (3.18)
The standard linear and extended k -ε turbulence models are not sufficient
for the predictions of near wall flow at high Reynolds numbers.
* Non-Linear RNG k -ε Turbulence Model
The Renormalization Group (RNG) k -ε model is proposed by Speziale
(1987) and has been validated against internal flow and impinging jet problem by
Rabbit (1997) and later is developed by Yakhot et al. (1992). This model is very
similar in form to the standard and extended k -ε turbulence models, however the
RNG k -ε model differs from the standard model by the inclusion of an additional
sink term in the turbulence dissipation equation to account for non-equilibrium strain
rates and employs different values for the various model coefficients. The form of
the k equation remains same. The turbulence dissipation, ε equation of the RNG k -
ε model includes the following sink term
k
C 23
0
3
1
1 εβη
ηηηµ
+
⎟⎠⎞⎜
⎝⎛ −
(3.19)
In the above term employs the parameterη , which is the ratio between the
characteristic time scales of turbulence and mean flow field as follows:
εη kS=
41
3.MATERIALS AND METHODS Serkan MEZARCIÖZ
where
S = t
ijijGSS µ=2
⎟⎟⎠
⎞⎜⎜⎝
⎛
∂
∂+
∂∂
=i
j
j
iij x
uxu
S21 (3.20)
The primary model coefficients of the RNG k -ε turbulence model are
εµ σσ ,,,, 21 kCCC and Von Karman constant κ . Recommended values of this model
coefficients are as follows:
085.0=µC
41.11 =C
68.12 =C
7179.0=kσ
7179.0=εσ
κ = 0.3875
* Realizable k -ε Turbulence Model
The term “realizable” means that the model satisfies certain mathematical
constrains on the normal stress, consistent with the physics of turbulent flows. The
realizable k -ε model proposed by Shih et al. (1995) intends to address the
shortcomings of traditional k -ε turbulence models by adopting a new eddy-viscosity
formula involving a variable originally proposed by Reynolds (1987) and a new
model equation for dissipation based on the dynamic equation of the mean-square
vorticity fluctuation. The kinetic energy equation is the same as that in the standard