Professional Documents
Culture Documents
Effective cooling system design is important in the endurance event of the FSAE-A
competition. Research into sidepod design and radiator fan optimisation requires further
attention. To provide a further understanding on the aerodynamic effects of a sidepod
two investigations were performed using CFD. The first investigation identified that high
turbulence intensity levels created more uncertainty in predicting the radiator pressure
drops. Velocity of the airflow vary around bends in the sidepod. The airflow is
accelerated around the convex corners and decelerated around the concave corners.
Turbulence levels dissipate slower around convex corners. The deviation of radiator
pressure drops obtained using CFD from MUR's experimental results was more
significant if inlet flows were more turbulent. The second investigation illustrated that
high degrees of curvature in a sidepod may lead to internal flow separation. The latter
will occur if the diameter of curvature is too small or the separation length is too large.
Validating using data obtained from MUR may not be comprehensive, hence, raising the
importance for experimental testing to validate CFD results. RNG k- model proved its
resilience in all cases simulated. This thesis forms the foundation for future research and
design of sidepods in FSAE cars.
Contents
I.
II.
Introduction
Background Information
A. Contribution to cooling system design
B. Sidepod and radiator flowfield discussion
C. Heat exchanger modelling in FLUENT
D. Turbulence modelling in FLUENT
III. Methodology
A. Effects of inlet turbulence
B. Effects of varying curvature
IV. Results
A. Effects of inlet turbulence
B. Effects of varying curvature
V. Modification & further results
A. Effects of inlet turbulence
B. Effects of varying curvature
VI. Verification & validation of results
A. Effects of inlet turbulence
B. Effects of varying curvature
VII. Extension to current work
VIII. Conclusion
Acknowledgements
References
Nomenclature
FSAE-A = Formula Society of Automotive Engineers-Australasia
1
1
A study on the pressure variation in a ducted heat exchanger using CFD, UNSW@ADFA
2
2
2
3
4
4
5
5
5
6
6
7
7
7
8
9
9
10
11
11
12
12
CFD
CAD
OEM
MUR
RNG
FVM
kL
1/
C2
=
=
=
=
=
=
=
=
=
I. Introduction
The FSAE-A competition is organised annually for all competing universities in the Australasia region. The
Academy Racing team is a regular participant of this competition. In this competition there are several events
that allow judges to assess and rank teams accordingly. The endurance event is assessed by pushing the car to its
limits by surviving 34 laps of the autocross track. This event forms 35% of the overall score and is of paramount
importance for all competitive teams [1]. In the 2010 FSAE-A competition the Academy Racing team placed
7th overall and 8th in the endurance event [2]. To improve this score, further design considerations can be
implemented into improving the cooling system of the car. In the FSAE-A competition in 2010, cost
optimization of the cooling system was part of the cost judging criteria. This was included in the combined score
awarded to the participating teams. In the 2010 FSAE-A competition, Academy Racing's WS10 car achieved a
score of 71.2 out of 100 [2]. This result can also be further improved upon. Thus, improving the cooling system
design will be necessary to perform better at subsequent FSAE-A competitions. This project aims to provide an
understanding into the aerodynamic considerations associated with sidepod flows, thus, contributing in part to
the overall cooling system design for the Academy Racing team's car.
failing to accelerate or decelerate in time. This leads to other problems associated with intermittent pipe flows
(e.g. cavitation of coolant, coolant backflow, etc.). Sidepods are designed to overcome these problems. The
design of sidepods will be able to provide a dedicated air intake for heat dissipation of the radiator [4].
Radiator efficiency is affected by the pressure variation across it. This is in turn determined by the velocity
of the airflow through it. The sidepod has an overall pressure difference between the ends (i.e. inlet and outlet).
This pressure difference varies during different external flow conditions. If there is a pressure drop large enough
across the sidepod, then there will be a flow velocity through the radiator that maximises its heat dissipation. If
the pressure difference across the sidepod is not large enough, there will be inadequate airflow through the
radiator. To overcome this, a radiator fan can be modelled to provide the necessary airflow through the radiator
in all stages of operation. The radiator fan provides the necessary driving force by creating a pressure gradient
that pushes or pulls airflow through the radiator even in adverse situations. [5] states that equation 1 can be used
to compute the pressure drop across the radiator solely by considering the pressure loss coefficient, k L, and
normal airflow velocity on radiator surface, v.
(1)
It is found that a large kL value and a large increase in airflow velocity from the inlet of the sidepod to the
plane surface of the radiator produce the best cooling performance [6]. The latter can be achieved by intelligent
optimisation of the sidepod geometry by creating regions that speed up the airflow. [6] also states that the inlet
of the sidepod should not be made to be of a converging nature. This is to prevent the generation of backflow
and vortices upstream of the sidepod and flow separation on the exterior near the inlet of the sidepod. The speed
of natural airflow at the inlet of the sidepod is limited by the velocity of the race car along the autocross track. In
a typical track, the front velocity of the car is not expected to exceed 100 km/h, owing to the lack of straights in
the track [7]. As such the airflow at the inlet of the sidepod does not exceed 30 m/s, and hence, is
incompressible. This allows for commonly known equations (i.e. steady flows, Bernoullis equation &
continuity equation) to be used and model the airflow of a FSAE sidepod.
Sidepods often have varying degrees of curvature and cross-section areas to allow for optimised airflow
through the radiators. By varying the inlet, duct and outlet geometries of the sidepod, the internal airflow can be
studied and used to a design engineers advantage [5]. To achieve best cooling performance, a higher air mass
flow rate into the sidepod has to be obtained. However, from the perspective of vehicle aerodynamic drag, a
minimum amount of airflow should be diverted from the main flowfield around the car into the sidepod [8]. [8]
also mentions that considerations have to be made for outlet geometries of sidepods to prevent the loss of
downforce on the car. Outlet airflow design should also be considered as this airflow must not increase the
pressure of the low pressure region underneath the car to an unacceptable level.
The free-stream natural airflow entering a FSAE car sidepod is disturbed by several effects. One of them is
the turbulence generated from the rotation of the front wheels. The front wheels of the car of the Academy
Racing team are situated just to the front of the inlet of the sidepod. The effect of a rotating wheel on the inlet
turbulence of a side-pod varies with the velocity at which the free stream airflow is acting on the rotating wheel
[9]. This leads to a variety of inlet turbulence generation at the inlet. This can be modelled in ANSYS FLUENT
and further details can be found in the turbulence modelling section.
C. Heat exchanger modelling in FLUENT
The Academy Racing team uses a Borland Racing radiator used previously in the Formula Ford race cars.
No data for this radiator or other similar types of radiators within the same class could be obtained readily from
OEM. This means the only way to obtain heat transfer and pressure loss coefficients would be from the
experimental testing of the Borland radiator. From this testing, we can identify the pressure loss coefficient, k L,
and heat loss coefficient, h. Using equation 1, the pressure variation can then be estimated for a variety of
velocities. A quadratic curve of best fit can then be used to model the pressure variation distribution to be
analysed for porous media parameters. Section 7.2.3.6.11 of the ANSYS FLUENT user guide recommends the
use of two equations, 7-2 and 7-25 of the user guide (reflected here as equations 2 and 3 respectively) [10].
These two equations, when combined, culminate in equation 4. In these equations, S i represents the momentum
source term, n represents the radiator thickness,
represents the viscous inertial resistance factor and C2 is
the inertial resistance factor. The coefficients found from the quadratic curve of best fit can now be used to find
and C2, which are used as inputs for the porous media region.
| | )
(
(2)
(3)
3
A study on the pressure variation in a ducted heat exchanger using CFD, UNSW@ADFA
| | )
(
(4)
The FSAE team from MUR has performed experimental testing on their car radiator (similar class of
radiators) and have obtained a fourth-order polynomial fit for their results as shown in figure 1 [6]. Seeing that
the operating range of a FSAE race car exceeds a forward velocity of 5 m/s, This polynomial fit was then
compared for a larger velocity range to ensure that it would follow a similar trend to the experimental results.
This extrapolation of kL is found in figure 2.
From figure 2, it can be seen clearly that for
Figure 2. Extrapolation
Figure 1. Fourth order
velocities after 6 m/s the fourth-order
polynomial fit for kL from
of polynomial.
polynomial fit does not follow a similar trend
MUR.
as the experimental results. In order to obtain
a curve of best-fit that adheres to the trend of
the experimental data, a power-law model
was used. This model is found in equation 5.
This was substituted into equation 1 and a
series of loss coefficient values, kL, were
plotted against the normal velocity. These
values
(5)
(6)
Figure 4. Plot of
Figure 3. Power-law model
allowed
for
a
quadratic
curve-fit
to
be
used
quadratic curve of best-fit
for kL versus velocity
and this is expressed in equation 6. This
using cftool function in
showing conformal to trend
quadratic expression is modelled to a 95%
MATLAB.
obtained from experimental
confidence level using a MATLAB function
called cftool. The coefficients of equations 6
results.
were then compared with those obtained from
equation 4 to find
and C2, which will
then be used as the porous media inputs. This
comes from assuming that the radiator is
fitted tightly around the walls of the sidepod,
such that all airflow must flow through the
radiator, that the operating pressure was
assumed to be at 101 325Pa (1 standard
atmosphere) and that kL observed the same
trend for all velocities. Figure 3 shows the
power law approximation of kL versus velocity, while figure 4 shows how p is modelled via the cftool
function.
Figure 5. Coarse mesh showing
upstream (green), radiator (blue) and
downstream (pink) zones.
4
A study on the pressure variation in a ducted heat exchanger using CFD, UNSW@ADFA
III. Methodology
A. Effects of inlet turbulence
It is evident that sidepod inlet turbulence and velocities vary during different stages of the
autocross/endurance events. Hence, it was pertinent to conduct an investigation into the airflow effects of
varying sidepod inlet turbulence intensities and velocities. Pressure variations, velocities and turbulence
intensities were measured for various regions downstream of the sidepod. A general design of a sidepod was
created from CATIA and mesh in ANSYS. Adopting the FVM, the initial coarse mesh created had 36, 960
nodes and 34, 338 elements. Figures 5 and 6 show this mesh created and a cut section through the mesh.
Consideration was made to
Figure 7. CAD model (left) designed with S-bend and CAD sketch
ensure that the different zones
(right) showing how the S-bend was varied.
of the mesh were conformed to
having identical faces along
Diameter
zone interfaces. This promoted
better solution convergence
h
rates and more accurate results.
Diameter
After which, the mesh was
imported into FLUENT. The
inlet conditions of the sidepod were varied
Figure 8. Table of parameters used for varying cases.
according to four different velocities (5, 10, 20,
30 m/s) and turbulence intensities (1, 3, 6, 10
Diameter of
Separation
Inlet turbulence Inlet velocity No. of
%) respectively. Temperature variation across
curvature (mm)
length (mm)
intensity (%)
(m/s)
cases
the duct was not measured and as such no pre400, 500, 600
100, 200
1, 10
20, 30
24
formulation of temperature related coefficients
was necessary. The sidepod pressure variations were not measured, as the results obtained from the CFD
simulations will show that the conditions that were simulated were of an internal flow nature without any
external influences. External influences are crucial in dictating the pressure drop across the sidepod.
Lengt
5
A study on the pressure variation in a ducted heat exchanger using CFD, UNSW@ADFA
IV. Results
A. Effects of inlet turbulence.
Upon compiling the results of the simulations, the absolute pressure values were found to vary across the
entire cross-section. This comes about due to the presence of velocity changes throughout the interior of the
sidepod. A longitudinal section was taken along the length of the sidepod and along the length of the crosssection face. Figure 9 shows how this longitudinal section was created. Using this longitudinal section, contour
plots of velocity magnitudes and absolute
Figure 11. Turbulence intensity contour plots at different
pressures were found. These are
locations of the sidepod. Sidepod outlet (left-most), radiator
expressed in figure 10. It can be seen in
nd
rd
outlet (2 from the left), radiator inlet (3 ), cross-section plane
figure 10 that the velocity increases
nd
at bend (2 from the right) and sidepod inlet (right-most) are
around the convex (inner) corners and
reflected. The convex corner of the cross-section plane is
decreases around the concave (outer)
located at the bottom, while the concave corner is located on
corners. This can be likened to a
top.
meandering river where the flow of the
water is always fastest on the inside of a
bend owing to the path of least flow
resistance.
Since
the
flow
is
incompressible,
simple
Bernoullis
principle can be applied. This leads to a
lower static pressure in the regions of
Figure 12. Velocity contour plot for the cases of
curvature diameter=500mm, inlet velocity=30m/s,
turbulence intensity=10%.Left plot shows the case for
separation length=100mm, right plot shows for
separation length=200mm.
6
A study on the pressure variation in a ducted heat exchanger using CFD, UNSW@ADFA
maximum of 8% and a minimum of 6.5%. These values continually decrease across the radiator and through the
aft section behind the radiator component. It is also observed that there is a higher value of turbulence intensity
closer to the concave corner (i.e. the region of lower velocity). This also means that the regions of higher
velocities have a lower value of turbulence intensity. In a sidepod, the radiator can be aligned to receive a higher
velocity flow rate and lower turbulence intensities. It has also become evident that the turbulence dissipation
rate nearer to the walls is much greater than
Figure 15. Absolute pressure (left) and velocity
further away from the walls as shown in figure
magnitude (right) contour plots along the
11.
longitudinal section.
Coarse Refined
Elements Elements
115412
233799
109244
343505
108934
354640
107037
211635
109306
213777
122460
220663
design.
Figure 16 shows
the
variation
of
turbulence
intensity
along the longitudinal
section and the crosssection at the corner
closest to the radiator.
This general variation
is similar to that
obtained for the coarse
mesh in figure 11.
However, it must be
noted that the turbulence intensities at the boundary layer of the regions varies slightly when compared to figure
11. Figure 16 depicts a more accurate solution where the turbulence region of influence affected by the
boundary layer is of a smaller proportion of the overall cross-section of the sidepod. Furthermore, from the
longitudinal section contour plot of turbulence intensity, it can be found that the convex corners of bends seem
to reduce the dissipation rate of turbulence as the turbulence intensities appear to be more resilient in the two
convex corners of the sidepod. This means that if the cooling system designer orientates the radiator to receive a
higher velocity flow rate in this specific sidepod geometry, then the radiator will also be subjected to more
inconsistent pressure drops across it due to the lower turbulence dissipation. Further optimisation studies can be
conducted to discover the right levels of turbulence that is acceptable for efficient radiator performance.
B. Effects of varying curvature.
The mesh created was inflated around the wall regions to ensure sufficient grid resolution to reflect these
modifications. Figure 17 shows the coarse and refined mesh grid elements. This shows how well resolved the
refined grid is.
The claim that flow separation occurs for longer separation lengths is again proven in figure 18 where the
velocity contour plots for the longitudinal sliced section is compared for the curvature diameter of 500mm.
From this figure, it is obvious that flow separation occurs at the claimed location of the second concave corner
in the S-bend. Figure19 shows the same situation for 600mm curvature diameter. It can be found here in figures
18 and 19 that the velocity increase is greater for smaller curvature diameter, thus increase the likelihood of flow
separation in the region of the second concave corner. This also suggests that the critical separation length of an
S-bend is reduced for a smaller curvature diameter in order to prevent flow separation. Figure 20 shows the
turbulence intensity contour plots for all four cases shown in figures 18 and 19. Figure 20 further proves the
above statement. In figure 20, the red region is larger for the case of the 500mm curvature diameter than that of
the 600mm curvature diameter. This means that the flow separation is greater in the S-bend of the smaller
curvature diameter. The inlet turbulence due to the slightly converging cross-section areas is also shown in
figure 20.
8
A study on the pressure variation in a ducted heat exchanger using CFD, UNSW@ADFA
(8)
The maximum radiator pressure drops for 20 m/s
inlet velocity seem to show a 12Pa (2.5% of results)
fluctuation for varying inlet turbulence intensities,
while the minimum pressure drops hardly fluctuate. It
can also be seen that for the 10 m/s inlet velocity case,
that a 3% inlet turbulence intensity causes the largest
radiator pressure drop difference across the surface of
the radiator. To verify that these results are accurate
given, we can use the initial approximation (equation
6) obtained for the radiator pressure drops using the
cftool function in MATLAB. The radiator pressures
obtained from the Richardsons extrapolation were
plotted together with those obtained from figure 4.
This is represented in figure 23. In figure 23, only the
maximum pressure drops and maximum velocities across the radiator surfaces was taken into account as it was
difficult to obtain information on the minimum velocities due to the presence of boundary layers in the contour
plots.
Figure 23. Verification of results obtained from
CFD analysis with approximated quadratic
curve from equation 6.
From figure 23, it is obvious that the results obtained from CFD follow the same trend as the curve of best fit
from equation 6. However, it can be seen that the estimated results tend to deviate to a greater extent away from
this curve at higher intensities. This shows that at higher levels of sidepod inlet turbulence it may be more
difficult to accurately determine the radiator pressure drops. This may cause some concern for the cooling
system design as sidepod inlet flows are often subjected to large amounts of turbulence. Therefore, it is
necessary to account for this with a factor of safety when considering the external pressure drops across the
9
A study on the pressure variation in a ducted heat exchanger using CFD, UNSW@ADFA
Grid-independent
Radiator pressure drop (Pa)
Inlet Conditions Maximum
Minimum
1%, 20m/s
454.33
456.33
10%, 20m/s
468.50
433.79
1%, 30m/s
955.00
922.07
10%, 30m/s
909.65
912.79
1%, 20m/s
425.89
414.92
10%, 20m/s
447.71
413.93
1%, 30m/s
851.68
852.97
10%, 30m/s
918.41
858.65
1%, 20m/s
458.59
438.50
10%, 20m/s
459.80
449.10
1%, 30m/s
963.27
929.63
10%, 30m/s
963.22
906.06
1%, 20m/s
429.39
412.64
10%, 20m/s
451.98
431.29
1%, 30m/s
883.16
883.17
10%, 30m/s
949.53
910.48
1%, 20m/s
459.94
463.22
10%, 20m/s
464.61
448.95
1%, 30m/s
928.75
933.88
10%, 30m/s
965.00
930.09
1%, 20m/s
451.73
402.19
10%, 20m/s
454.39
454.39
1%, 30m/s
944.30
909.33
10%, 30m/s
903.05
880.19
B. Effects of varying
curvature.
Richardson's extrapolation
was conducted as in the first
investigation to obtain the
grid-independent solutions to
the radiator pressure drops.
Using equations 7 and 8, a
table of the grid-independent
radiator pressure drops was
compiled and is shown in
figure 24. From figure 24, the
effects of varying curvature
diameter on the values for
radiator pressure drops were
once again compared and are
shown in figure 25. Only the
maximum radiator pressure
drops were considered as it would be difficult to
approximate the lowest free-stream velocity normal
to the radiator cross-section due to the presence of
boundary layers and airflow separation. From
figure 25, it is also found that the radiator pressure
drops decrease for most cases with increasing Sbend separation length. This supports is similar to
the results discussed earlier.
all CFD analyses, real-life validation is essential and hence, a full experimental procedure of this investigation is
recommended in order to completely validate the obtained CFD results.
VIII. Conclusion
After performing these two investigations, several issues have been brought to light. The prediction of flows
in a sidepod via CFD simulations becomes more difficult with greater turbulence intensity levels at the inlet and
outlet of the sidepods. This will lead to additional flow velocity perturbations that will either mildly speed up or
slow down the airflow on the radiator surface. Such variation in airflow velocities is often coupled with bends in
a sidepod. These bends will allow designers to completely exploit the increase in the velocity of normal airflow
on the radiator. Studying these bends is likened to that studying water flowing in a meandering river, and using
this analogy, designers can create the ideal bend with appropriate degrees of curvature to speed up slower
velocities as much as possible without causing flow separation in slower regions.
Turbulence intensities were identified to dissipate at a lower rate around convex corners and much faster
around concave corners. It was also observed that with increasing turbulence intensities at the sidepod inlet, the
deviation from experimentally derived data is more significant. For sidepods with an S-bend, it was found that
flow separation may occur if the curvature diameter becomes too small or the separation length is too large. If
the latter condition occurs, the pressure drop across the radiator will decrease. The resilience in the RNG k-
model was also seen in the two investigations where the realizable k- model was not required to be used in any
of the cases modelled. Experimental validation of the obtained results is crucial in assuring the designer that the
simulations had solved the right equations in the correct manner. Hitherto, there has been a lack of research into
sidepods for FSAE cars. Thus, this thesis will provide the groundwork for future research into sidepods and car
cooling system design for FSAE.
11
A study on the pressure variation in a ducted heat exchanger using CFD, UNSW@ADFA
Acknowledgements
The author of this thesis would like to thank several people who throughout the course of this project who
provided great support and advice, without which the author would not have been able to successfully complete
this thesis. The author would like to thank the thesis supervisor, Mr Alan Fien, for his undying support and
efforts in assisting the former to overcome his obstacles faced during the various phases of the project. Alan
had, on several occasions, sacrificed his spare time to assist the author with his problems. The author would also
like to thank independent advisor, Dr John Young, who provided great levels of guidance with the CFD
component of this project.
The author would also like to thank Dr Warren Smith, PLTOFFs Paul Gardner and Alistair Weir for their
insights into the requirements of the Academy Racing team. The author also wishes to thank his undergraduate
friends from Singapore who assisted in running simulations and extracting useful data for this project. Lastly,
the author wishes to thank the Academy Racing team members for providing a great working environment and a
lively atmosphere whilst working with them.
References
[1] About FSAE, Academy Racing website, http://www.fsae.unsw.adfa.edu.au/about.php, accessed 19 Oct 11.
[2] 2010 FSAE-A results, http://www.saea.com.au/wp-content/uploads/2011/09/2010-FSAE-A-Results.pdf, accessed 19 Oct
11.
[3] Milliken, W. F., Milliken, D. L., Race Car Vehicle Dynamics, SAE International Publishing, Twelfth edition, 1995,
p556.
[4] Hucho, W., Aerodynamics of Road Vehicles-From Fluid Mechanics to Vehicle Engineering, Butterworth & Co.
Publishers, p5, 1987.
[5] De Silva, C. M., Nor Azmi, M., Christie, T., Abou-Saba, E., Ooi, A., Computational Flow Modelling of Formula-SAE
Sidepods for Optimum Radiator Heat Management, Journal of Engineering Science and Technology, Vol 6, No. 1 (2011)
94-108 @ School of Engineering, Taylors University.
[6] Milliken, W. F., Milliken, D. L., Race Car Vehicle Dynamics, SAE International Publishing, Twelfth edition, 1995,
p558.
[7] Gardner, P. F., How fast does a FSAE race car travel in an autocross/endurance track?, Pers Comms, 7 Aug 2011.
[8] Adams, H., Chassis Engineering, HPBooks, published by the Berkley Publishing Group, 1993, pp102-103.
[9] Diasinos, S. The Aerodynamic Interaction of a Rotating Wheel and a Downforce Producing Wing in Ground Effect. PhD
thesis, School of Mechanical & Manufacturing Engineering, UNSW, March 2009.
[10] Section 7.2.3.6.11, Deriving the Porous Coefficients Based on Experimental Pressure and Velocity Data, Chapter 7:
Cell Zone and Boundary Conditions, ANSYS FLUENT user guide, Release 13.0.
[11] Chapter 6.2.2, Determining turbulence
help/html/ug/node178.htm, accessed 21 Oct 2011.
parameters,
Fluent6.1
online
help,
http://jullio.pe.kr/fluent6.1/
12
A study on the pressure variation in a ducted heat exchanger using CFD, UNSW@ADFA