1.Critically assess the aerodynamic characteristics of a vehicle.
2.Select and specify the most appropriate methods for wind tunnel testing of scale models and interpret the results of the test.
3.Analyse and critically evaluate the aerodynamic cooling systems.
1. Faculty of Science, Engineering and Computing
School of Mechanical and Aerospace Engineering
Assignment:
CFD analysis of commercial vehicle
Module:
ME7733 Automotive Aerodynamics and StructuralAnalysis
Setter: Dr K Volkov / Dr A Heidari
Deadline:28 February2017
Name: SHIH CHENG TUNG
KU number: K1617281
Course: Msc Automotive Engineering
2. Contents
1. Introduction
2. Methodology and methods
Design issues
The wall y plus (y+)
Inflation of boundary layer
Mesh quality
3. Initial analysis
Computational flow domain
Wind tunnel blockage ratio
Different distributions of velocity u
Different distributions of pressure
Evaluation and discussion
4. AERODYNAMIC MODIFICATIONS
The feature of design
Analysis and discussion
5. WIND TUNNEL TESTING OF A SCALE MODEL CAR
REYNOLDS NUMBERS
6. AERODYNAMIC COOLING SYSTEM
LMTD: Log-Mean Temperature Difference
Effectiveness-NTU: effectiveness Number of Transfer Units
Analysis and discussion
7. References
3. 1. Introduction
Aerodynamic technology plays an important role in automotive industry, which
effects not only the performance of vehicle but also contributes to the fuel
consumption (Volcov, 2017a).There are two main methods to conduct the
experiments which are wind tunnel analysis and computational analysis. Most of
aerodynamic simulations can be done and analysed in an efficient and convenient
way by computational fluid analysis. “CFD is powerful and spans a wide range of
industrial application areas for professional study (Versteeg et al., 2007). The
adoption of predicting internal and external flows through computational fluid
dynamics (CFD) increased significantly in the past decade. The widespread
availability and efficient solution algorithms enable the use of commercial CFD
codes by graduate engineers for research, development and design tasks in
industry. Moreover, CFD systems of analysis including fluid flow, heat transfer and
associated phenomena require appropriate software to conduct high quality
outcomes as well. ANSYS CFX software is a multi-function, high-efficiency fluid
dynamics program with an advanced solver technology. The reliable and accurate
solutions could be achieved quickly through features of geometry, mesh, setup,
solution and result. (ANSYS, Inc., 2011)
2. Methodologyand methods
Design issues
The vehicle aerodynamics has a strong relation with automotive design, which
contributes to most of issues including strategy of performance, cooling, comfort and
stability. It is essential to consider the dynamic of air flow surround the car body that
could be optimised in numerous ways. The underbody flows and flow around rotating
wheels are parts of external flow while cooling system and air conditioning system
are classified to internal flow. In addition, the coefficients of drag and lift are two
common values to be referred as aerodynamic functions. For example, the history
shows aerodynamic drag has been improved from 0.8 to 0.27 since 1920. (Volcov,
2017b)
The wall y plus (y+)
4. The wall y plus (y+) is a value of distance to deal with modelling of turbulent flow,
which cannot be measured. In CFD analysis, to describe how coarse or fine a mesh
is for a particular flow. It could provide the information that whether the mesh is
generated coarsely. The y+ could help estimating surface flow of wall which present
the turbulent and laminar influences ratio and the thickness of the first layer (Wall y+
Strategy for Dealing with Wall-bounded Turbulent Flows). In this case, in order to
produce a fine mesh the Shear Stress Transport (SST) model is applied with a tiny
number of y+. Generally, it is better to make y+ close to 1 so that flow separation
could be predict accurately in the unstable boundary layer with the -k based SST
model.
Inflation of boundary layer
It is necessary to model the boundary layer through layer inflation which generates a
valid mesh. The condition of inflation could be predicted by values of the non-
dimensional distance from the wall Y+, growth ratio r, moving speed u, vehicle length
L and air viscosity v. The following steps provide basic concept of calculating
inflation:
1. Calculate the Reynolds number Re at scaled-down length and identify flow
type
2. Choose suitable flow type equation to compute the total boundary layer
thickness
3. Select a proper value of desired Y+ for analysis
4. Determine the thickness of the first layer at scaled-down length
5. Carry out the number of layers by total boundary layer thickness, first layer
thickness and growth ratio
Step Formula Unit
1 𝑅𝑒0.1𝐿 =
𝑢×𝑥
𝑣
, if is turbulent 𝑅𝑒0.1𝐿 > 500000
if is laminar 𝑅𝑒0.1𝐿 < 500000
/
2 Laminar: 2/1
Re5
xx
Turbulent: 5/1
Re37.0
x
m
3 Wall function: Y+ < 300, if not Y+ < 2 /
4 m
5. 5
r)-(1
)r-(1
y
n
m
In this case, the kinematic viscosity of 20 degree Celsius smv /1051.1 25
and 50 mph = 22.35 m/s
that Reynolds number Re0.1L=
22.35×0.8
0.0000151
= 1184105.96 > 500000
Then choose turbulent as flow type and compute the total boundary layer thickness
5/1
)96.1184105(8.037.0
= 0.018 m
Proper value of Y+ is selected as 2 that
14/13
)96.1184105(7428.y
o = 0.00003156 m
Finally, to solve the number of layers n,
1.2)-(1
)1.2-(1
0.00003156018.0
n
, n
1.2-1
0.00003156
(-0.2)0.018
, 26
)2.1ln(
ln(115.07)
n
Mesh quality
It is worth spending most of time to set up the controlling parameters for creating a
proper mesh and to identify the important regions so that could create the high
resolution volume mesh on key surfaces. There are two ways could be applied easily
to identify approximate quality of mesh, skewness and orthogonal quality. The value
of skewness bases on the equilateral volume and the deviation from a normalised
equilateral angle. To use the method of skewness which is constructed on the
equilateral volume only applies to triangles and tetrahedral while another process
which based on the deviation could apply to all shapes of cell and face (lecture note:
domain-meshing). Commonly, the value of skewness should be kept under 0.95 in
order to manage the mesh qaulity (lecture note: domain-meshing).
Equilateral Volume deviation Normalised Angle deviation
Maximum quality 1 (Worst)
Skewness
6. Minimum quality 0 (Perfect)
Orthogonal quality is derived from fluent solver discretisatioin which present the
range between 0 (worst) to 1 (best). There are two equations for computing minimim
values of each face on a cell, which differ from the type of vector. To compute the
orthogonal quality for faces the minimum values of each edge are determined by the
edge normal vector and the vector from the face centroid to the centroid. According
to mesh metrics spectrum, it is acceptable to maintain value of orthogonal quality
over 0.15.
Cell Face
: The face normal
vector.
: Vector from the
centroid of the cell to the
centroid of that face.
: Vector from the
centroid of the cell to the
centroid of the adjacent
cell that shares the face.
In addition, the accuracy of simulation not only be effected by mesh quality but also
be contributed by the total count. Great number of cells could provide high resolution
to generate the exact results, however the processing time would be increased
significantly. In order to balance the time cost and performance of simulation, the
total cell count should be managed to the proper range. 1 million to 10 million cells
could offer a reasonable outcome, which take an acceptable time to process.
3. Analysis and discussion
7. Computational flow domain
The computational fluid dynamics analysis is conducted by ANSYS CFX software
which is a multi-function, high-performance program. A commercial box van in 8
metre length (L), 2.45 metre width (W) and 3.55 metre height (H) will be test in 20
degree Celsius air with 50 mile per hour speed.
After importing the geometry file which has been provided, a suitable flow domain
should be created for the car model. The domain size is determined for a cube
surround car model, that effects significantly to accuracy of results. In this case,
three different flow domain dimensions are created in order to investigate the
influence on the outcomes. The length in front of car body are placed at 4, 12 and 24
metres which represent 0.5L, 1.5L, and 3L while the length behind the car body are
20, 56 and 112 metres which represent 2.5L, 7L and 14L. The width of flow domain
is fixed in 10 metres as 4.08W whilst 10 and 12 metres height are determined as
2.82H and 3.36H. In order to simulate the actual road condition, the box van model is
placed as close as possible to lower ground of solid.
Wind tunnel blockage ratio
The efficient estimation of flow domain could be calculated through blockage ration
which represent the ratio between frontal area of the car and the cross section area
of the flow domain. Blockage ratio is an important factor that influences the
performance of results and it should be as small as possible (Sahini, 2004).
However, the computational capabilities might be a limitation of minimising blockage
ratio because bigger flow domain equal more cells and longer processing time. In
this simulation, blockage ratio are indicated to 8.7% and 7.3% which are below 10%
8. properly. Moreover, according to the lecture note CFD Simulation.pdf of Automotive
Aerodynamics and Structural Analysis module from Dr. K Volkov, the dimensions of
flow domain exceeded the minimum values so that could maintain the sufficient
quality.
According to the three different domain volume simulation, the forces of lift and drag
are extremely similar in first and second results. The factors of first two simulation
are extending the frontal and back length of flow domain so that keeps blockage ratio
in the same value. It could be summarised that the results maintain the close values
with variation of domain length in proper flow domain volume. For the third
simulation, the blockage ratio is changed into 0.073 which influences the
performance of result as well. Due to extending the height of flow domain the forces
of lift and drag are increased at the same time. The study of Sahini (2004) states that
the drag variation with blockage ratio could be almost linear.
Different distributions of velocity u
Figure 1 and Figure 2 are the analysis of ANSYS CFX which present the contours of
velocity u and total pressure. The contributions of blockage ratio on the velocity and
total pressure contours are shown as figure 1 and figure 2.
Domain# H1 H2 L1 L2 Y+ Max Blockage
1 0.5L 3.5L
0.25m
2.82H 937.558 0.087
2 1.5L 8L 2.82H 3404.27 0.087
3 3L 15L 3,38H 830.694 0.073
H1 (m) H2 (m) L2 (m) Lift(N) Drag (N)
1 4 28 10 -429.844 1520.47
2 12 64 10 -439.203 1460.19
3 24 120 12 -343.887 1343.45
Domain 1
Domain 2
Domain 3
9. Figure 1, contours of velocity u on XY plane Figure 2, contours of total pressure on XY plane
Obviously, the differences are located behind the tail part of box van which can be
found in analysis of the Domain 3. The greater height of domain produces longer
pattern to both velocity and pressure while pattern of first two simulation are similar
in the same blockage ratio.
Different distributions of pressure
Another comparison of pressure are shown as following figures in similar unit of
matrix spectrums. For result of the Domain 3, the pressure distribution upon top of
box van occupies a smaller area whilst the top of box van in the Domain 1 is covered
by a larger area. However, more pressure are concentrated at the middle gap of box
van in Domain 3 instead of be dispersed averagely.
In the same time, the stronger force of lift presented to the Domain 1 which is related
to the difference pressure distribution upon the box van.
Evaluation and discussion
There are millions of adjustments of simulation that have to be set up properly. Not
only should the physics of the flow but also the appropriate allocation of boundary
conditions be considered in CFD analysis. For example, the mesh generation
includes plenty regulations such as relevance center, curvature normal angle,
smoothing and face sizing with elements size. Those factors could be utilised easily
to optimise quality of mesh which effects simulation accuracy. However, the amount
Figure 3, pressure contour of Domain 1 on XY
plane
Figure 4, pressure contour of Domain 3 on XY
plane
10. of elements is a certain consideration that indicates a long processing time.
Therefore, to balance the elements number and meshing quality could be a critical
part of CFD analysis and also worth spending time on.
The optimisation of Domain 1 with varying mesh average quality is presented by the
following table. The inflation condition of boundary layer is fixed with the same
number of layers, total boundary layer thickness and growth rate. There is an
approximate 24.4 million elements number which is conducted by 0.01 element size,
coarse relevance center and low smoothing. However, the greatest number of
elements might be not the highest quality mesh. As shown in the table, 0.05 element
size with fine relevance center and high smoothing takes a first place.
Certainly, the orthogonal quality method is not only way to identify the quality of
mesh. The results can be improved through assessing skewness and orthogonal
quality method at the same time. In addition, that could produce an optimised value
of y+ with exact flows of boundary layer.
4. AERODYNAMIC MODIFICATIONS
The feature of design
Commercial vehicles transport goods that develop the economic efficiently in
nowadays. However, people are concerned about the issues of environmental
pollution and carbon emission as well. According to the research, the aerodynamics
improvement can provide an effective solution. In this case, the design to reduce
drag of box van is came up with a roof spoiler which provides a simple and efficient
Face
sizing Mesh detail
Orthogonal
quality
Element
size 0.1 0.05 0.05 0.05 0.01
Relevance
center medium fine medium coarse coarse
Curvature
normal angle 9 9 9 9 9
Smoothing medium high medium low low
Elements 7000000 18407357 12778908 6996097 24375361
Average
quality 0.68 0.81238 0.71 0.709 0.80
11. application. Normally, the box vans generate a high coefficient of drag due to the
long cuboid shape. The sharp corners on vehicle cause turbulent flow above car
body. Therefore, a roof spoiler could be attached in order to improve the drag
coefficient.
A 1.2 metre length, 1.6 metre width and 1.0 metre height roof spoiler with varying
angle of edges is illustrated by Solidworks. The designed features not only guide air
flow to pass through the top of box van but also disperse air flow to right and let
sides. The strategy of distracting air flow to both sides of box van could conduct a
reasonable force of lift. Besides, it is not necessary to optimise force of lift for the
delivery truck yet the fuel consumption would be effected by friction between tyre
and road.
The modification for optimising aerodynamics is limited in size and shape in order to
ensure the quality of results. Firstly, the design should be generate without
increasing the contact area of flow so that it is better to attach on the corner of
vehicle. Secondly, the shape of modification should be devised within blunt turnings.
In other words, it is better to avoid all sharp angles because the low-quality mesh
would be conducted due to the extremely tiny volume of cells. Therefore, the proper
fillet could make the corners gentler when modelling the design in Solidworks.
Analysis and discussion
The coefficient of drag and lift formulas are determined as
d
d
d
Au
F
C 2
2
,
l
l
l
Au
F
C 2
2
According to the measurements on Solidworks, the frontal area Ad and top area Al
could be estimated to calculate both coefficients as following figures.
12. In the following table, there are various comparisons which include force of lift and
drag, coefficient of lift and drag and Y+ in the same dimension of flow domain.
Obviously, the first simulation without modification produces higher drag force and
coefficient which need to be improved in this case. In second simulation, the decent
decreases in the drag force and coefficient are contributed about 10 per cent by the
modification. Nevertheless, the lift force and coefficient are raised to the undesirable
values with 25 per cent.
Lift (N) Drag (N)
Frontal
area
Top
area
Drag
coefficient
Lift
coefficient
Y+
Original -439.203 1460.19
8.085 19.6
0,60 -0.074 3404.27
Modified -556.141 1321.16 0.54 -0.094 4353.76
The plots of streamline are provided by ANSYS CFX, which prove the optimisation
through observing the tail of box van. There are two clear eddies behind the tail of
box van in the first following figure while eddies are eliminated by attached
modification. It is expectable that the region of turbulence is lessened to effect drag
performance. However, the insufficient testing times and mesh adjustments both
influence the result of CFD analysis. The value of y+ need to be considered which
represent the quality of flow closed to the model surface.
13. 5. WIND TUNNEL TESTING OF A SCALE MODEL CAR
Reynolds numbers
Reynolds number (Re) Flow Field
0 ~ 1 Creeping flow
1 ~ 1e+3 Laminar flow, strong Re dependence
1e+3 ~ 1e+4 Transition to week turbulent flow
1e+4 ~ 1e+6 Turbulent flow, moderate Re dependence
> 1e+6 Strong turbulence, slight Re dependence
14. Osborne Reynold made the Reynold number become popular in 1883 so that it
provides an important contribution for professional study (Reynolds 1883). The flow
could be distinguished mainly by Reynolds number into two different types, laminar
flow and turbulent flow. For example, the viscosity causes an effect to reduce the
difference of velocity on the fluid between to the other fluid nearby. The following
table describes the range of Reynolds number that is related to various flow filed.
The Reynolds number could be derived to a value of Inertial Stress Fi over Viscous
Stress Fv. For calculating Inertial Stress Fi, to apply the Newton’s second law
whereas the force on object F equal to object mass m multiply by acceleration a. The
viscous force could be computed by multiplying dynamic viscosity of the fluid μ,
shear velocity
y
u
and area A of plate
(Force)StressViscous
(Force)StressInertial
Re
L
t
L
L
L
v
t
v
L
A
y
u
ma
Fv
Fi
)(
)(
))((
Re
2
3
,
vL
Re
formula variable unit
Inertial
force
maFi m object mass kg
a acceleration m/s2
Viscous
force y
u
AFv AFv ,
y
u
v velocity of the
object relative to
the fluid
m/s
μ dynamic viscosity of
the fluid
kg/(m.s)
Reynolds
number
vL
Re
(Force)StressViscous
(Force)StressInertial
Re
ρ density of the fluid kg/m3
L travelled distance of
the fluid
m
15. 6. AERODYNAMIC COOLING SYSTEM
The heat exchanger is part of essential system in vehicle, which is designed in
various types such as front end stationary head, shell and rear end head types. In
order to cool down or heat up the required temperature, these heat exchangers
should be managed in an appropriate location, size and type. The reason is that heat
which is usually generated by vehicle operations effects working conditions
seriously. For example, the internal combustion engine of vehicle causes high heat
that has to be cooled down by cooling system to maintain the durability. The thermal
capacity which indicates ability of transferring heat between two fluids at different
temperatures. Thus, there are numerous methods need to be focused when carrying
out a solution of analysing the thermal performance or selecting a heat exchanger
type.
LMTD: Log-Mean Temperature Difference
The Log-Mean Temperature Difference (LMTD) is used to define a temperature
difference logarithmic average between the cold and hot flows at each end of
the system (LMTD and NTU methods, no date). Certainly, the LMTD method is
able to determine the size of heat exchanger to realise prescribed outlet
temperatures with specified mass flow rates and the inlet and outlet temperatures.
The heat exchanger which suffices the prescribed heat transfer conditions will be
chose to utilise the LMTD method (Zohuri, 2016). The following steps show the brief
process of the LMTD method (Talukdar, 2012b).
Select the suitable application of heat exchanger.
Determine inlet/outlet temperature, heat transfer rate using energy balance.
Compute the correction factor F and the LMTD ΔTlm optionally.
Calculate or choose the overall heat transfer coefficient U and compute the
total contact area of heat transfer As
16. Single-pass cross-flow with both fluids unmixed
Single-pass cross-flow with one fluid mixed and the other unmixed
According to the figures, the required area of the heat exchanger could be generated
through locating the point by two finding parameters.
17. Effectiveness-NTU: effectiveness Number of Transfer Units
Some of the other issue usually comes across the defined size and type of the heat
exchanger and specified inlet temperatures and fluid mass flow rates that the
analysis is conducted by the heat transfer rate and the outlet temperatures of the
cold and hot fluids. (Unit Operations Lab, 2013). The effectiveness–NTU method
which is came up with in 1995 by Kays and London could simplify analysis of heat
exchanger without the iteration (Talukdar, 2012b). The heat transfer outlet
temperatures of the heat exchanger in the case are unknown, but the surface area A
is not. Thus, the LMTD method is not practical to be used for this kind of problem
due to tedious iterations. For the effectiveness–NTU method, to conduct the heat
transfer performance could be easier then LMTD method in situation of specified
heat exchanger. To calculate the effectiveness the following procedure should be
executed (Navarro and Cabezas-Gómez, 2005).
Identify the parameter of geometric for a precise heat exchanger.
Define the number of transfer units NTU, heat capacity rate ratio C* and
minimum heat capacity rate Cmin = (Cc or Ch).
Chose the hot/cold inlet temperature T c,I, T h,i and UA values as effectiveness
relation ε depends only on NTU, C*, and flow arrangement.
Effectiveness relations for heat exchangers:
Where ,
Cross-flow heat exchanger (sigle-
pass)
Effectiveness relation
Bothe fluids unmixed
Cmax mixed, Cmin unmixed
Cmax unmixed, Cmin mixed
18. Analysis and discussion
Heat exchangers could be analysed not only using Log-Mean Temperature
Difference (LMTD) but also applying the Effectiveness – Number of Transfer Units
(ε-NTU) methods. Both analysis of methods should be presented in below assumed
conditions
Consider no energy loss in the system
Be at a steady-state
Premeditate no phase changes in the fluids
The fluids’ heat capacities are not related to temperature
The overall heat transfer coefficient is not related to the position and fluid
temperature within the heat exchanger.
The LMTD and the ε-NTU method are two common solutions which conduct analysis
of heat exchangers from different viewpoints. The conceptions and parameters are
shared by both methods that approach the same results to thermal capacity of heat
exchanger (Sines, 2016). In this case, literature research are carried out into
analysis of designed heat exchanger. Furthermore, the discussion focuses on
analysing cross-flow heat exchanger compares the different characteristics through
both LMTD and ε-NTU method.
On one hand, the overall heat transfer coefficient with measured values of inlet and
outlet fluid temperatures could be computed efficiently by LMTD method.
19. Nevertheless, if it is used in calculating performance for the known inlet
temperatures and U the appliances of LMTD may be not suitable. The solution
needs additional two unknown values, hot and cold outlet temperature T c,o, T h,o,
which requires an iterative approach (Unit Operations Lab, 2013). On the other hand,
the analysis could be easily came up with by the ε-NTU method for predicting both
outlet fluid temperatures in situation of the known heat transfer coefficient and inlet
temperatures (Zohuri, 2016). It could be an advantage of ε-NTU method without
solving an arithmetic iterative solution of nonlinear equations.
7. Bibliography
ANSYS, Inc. (2011) ‘ANSYS Fluid dynamics’, .
LMTD and NTU methods (no date) Available at: https://gradeup.co/lmtd-and-ntu-
methods-i-3ac9d2bd-c289-11e5-a70e-bc0cab3fc1a6 (Accessed: 20 February 2017).
Navarro, H.A. and Cabezas-Gómez, L. (2005) ‘A new approach for thermal
performance calculation of cross-flow heat exchangers’, International Journal of Heat
and Mass Transfer, 48(18), pp. 3880–3888. doi:
10.1016/j.ijheatmasstransfer.2005.03.027.
Reynold, O. (1883) Philosophical Transactions of the Royal Society of London. (174
Vols). Philosophical Transactions of the Royal Society.
Sahini, D. (2004) ‘WIND TUNNEL BLOCKAGE CORRECTIONS: A
COMPUTATIONAL STUDY’, .
Sines, J. (2016) Difference between the Effectiveness-NTU and LMTD methods.
Available at: http://kb.eng-
software.com/display/ESKB/Difference+Between+the+Effectiveness-
NTU+and+LMTD+Methods (Accessed: 20 February 2017).
Talukdar, P. (2012a) HEAT EXCHANGERS. Available at:
http://web.iitd.ac.in/~prabal/MEL242/(30-31)-Heat-exchanger-part-2.pdf (Accessed:
27 February 2017).
Talukdar, P. (2012b) HEAT EXCHANGERS-2. Available at:
http://web.iitd.ac.in/~prabal/MEL242/(30-31)-Heat-exchanger-part-2.pdf (Accessed:
20 February 2017).
20. Unit Operations Lab (2013) HEAT EXCHANGER. Available at:
http://www.che.ufl.edu/unit-ops-lab/experiments/HE/HE-theory.pdf (Accessed: 20
February 2017).
Versteeg, H.K., Malalasekera, W., Versteeg, H. and Malalasekra, W. (2007) An
introduction to computational fluid dynamics: The finite volume methode. Harlow,
England: Prentice Hall.
Volcov, K. (2017a). lecture notes distributed in the topic ME7733, Kingston
University London, Roehampton Vale in February.
Volcov, K. (2017b). lecture notes distributed in the topic ME7733, Kingston
University London, Roehampton Vale in February.
Zohuri, B. (2016) ‘Effective Design of Compact Heat Exchangers for NGNP’, in
Zohuri, B. (ed.) Application of Compact Heat Exchangers For Combined Cycle
Driven Efficiency In Next Generation Nuclear Power Plants. Springer Nature, pp.
161–227.