Sei sulla pagina 1di 9

919

Theoretical and experimental investigation of the


aerodynamic drag due to automotive cooling systems
R H Barnard
Department of Aeronautical, Civil and Mechanical Engineering, University of Hertfordshire, College Lane, Hateld,
Hertfordshire AL10 9AB, UK
Abstract: Theoretical expressions for the aerodynamic drag due to the internal cooling ow in road
vehicles are examined. The advantages of a ducted system and the sizing and location of inlet and
outlet apertures are discussed. Experimental tests to evaluate theoretical relationships in model scale
are described, and practical conclusions concerning drag minimization are drawn.
Keywords: Automotive cooling systems, vehicle cooling system design, vehicle cooling system drag,
automotive internal ows
NOTATION
A
AB
Ac
Af
Ai
AO
cp
CpO
CpOB
CD
CDc
Cp
D
Dc
kp
m_
M1
p1
pO
Q
r
S
T
_
Vc
Vi
VO

blocked area of wind tunnel test section


area of rear or `base' face of vehicle
radiator core area
model projected frontal area
duct inlet area
duct outlet area
specic heat of air at constant pressure
static pressure coecient at outlet
mean pressure coecient on rear `base' face
with cooling system banked o
vehicle drag coecient dened by
CD D=12 V 21 Af
drag coecient contribution due to the cooling
system
pressure coecient
aerodynamic drag
aerodynamic drag due to the cooling system
radiator core pressure loss coecient
cooling air mass owrate
free stream Mach number
free stream static pressure
static pressure at outlet
heat transfer rate
pressure ratio in Brayton or `ramjet' cycle
wind tunnel cross-sectional area
rise in air temperature through radiator core
cooling air volume owrate
radiator normalizing or `core' speed
inlet air speed
outlet air speed

The MS was received on 23 July 1999 and was accepted after revision for
publication on 19 January 2000.
D09099

IMechE 2000

V1

free stream air speed (synonymous with car


speed on road)






inclination of outlet plane to vertical


ratio of the principal specic heats of air
ideal thermodynamic eciency of Brayton cycle
air density

1 INTRODUCTION
1.1 The importance of cooling system drag
The internal cooling ow in a road vehicle can add
signicantly to the overall drag. In 1985, Williams [1]
found that, for production cars, the increment in drag
coecient due to the engine cooling airow was at that
time in the range from 0.042 to 0.06 with the fan turned
o. Similar gures for vehicles of that era may be found
in Hucho [2]. In the 1980s, this would have represented
around 10 per cent of the total vehicle drag.
Improvements in cooling system design have taken
place in recent years, as outlined below, and nowadays
car manufacturers aim for a CD increment of around
0.02. Low cooling system drag gures can, however, be
misleading, because they may merely indicate that the
cooling ow due to ram pressure alone is quite small,
and power-consuming fan assistance may be required
much of the time.
The major contribution to the cooling system drag
comes from the surface friction on the ns and tubes of
the radiator core, the surface area of which may be some
100 times the core frontal area. In addition, there may
be some drag from items such as a decorative intake
grille. If the ow is discharged vertically upwards or
Proc Instn Mech Engrs Vol 214 Part D

920

R H BARNARD

downwards, the axial momentum is lost completely,


resulting in a high drag penalty, an eect that often
seems to be overlooked. Apart from direct losses, there
may also be signicant detrimental interference between
the internal and external ows.
When fan assistance is required, the power absorbed
is equivalent to a further increase in drag. At 70 miles/h,
1 kW absorbed in fan power is equivalent to an increase
in drag coecient of 0.03 on a typical mid-size car.
Because the radiator ow represents a signicant
contribution to drag, it is now customary to model it in
wind tunnel testing.
1.2

The use of ducted systems

Until the late 1970s, most cars used an unducted


radiator system where the radiator core was simply
placed in close proximity to the intake aperture. Since
the fan was usually driven directly from the engine, the
ow was eectively drawn through the core at all speeds.
Having negotiated the engine block, the air was discharged through the underside of the engine bay. Much
of the time, the forced air owrate was far greater than
necessary, and anything up to 5 kW of power could be
used. The introduction of thermostatically operated
electrical fans signicantly reduced these losses, but the
absence of continuous fan forcing meant that inlet
ducting was necessary in order to prevent much of the
ow from simply spilling around the outside of the core,
thereby contributing to drag without assisting in the
cooling. Many domestic cars do therefore now use some
form of intake ducting, and the advantages are evident
from the full-scale data presented by Renn and Gilhaus
[3]. However, apart from some early experimental
installations such as that reported by Paish and Stapleford [4], fully ducted systems are normally only found
on racing vehicles.
1.3

The essential features of a ducted ow system

A fully ducted system allows the ow speed through the


radiator to be controlled, and, as shown later, this
provides a means of minimizing the drag. A ducted
system comprises a heat exchanger or `radiator' core
sandwiched between inlet and outlet ducts. In practical
vehicle cooling systems the air ow speed through the
core is usually lower than the free stream approach
speed, and therefore in the intake section there will
normally be a decelerating air ow with an attendant
adverse pressure gradient.
For styling purposes, a small intake area is attractive,
but in a ducted system there is a limit to the amount of
diusion that can be achieved without ow separation
from the walls, and it is often advantageous to decelerate the ow externally by using a large area intake. In
model testing, the relatively low Reynolds numbers are
Proc Instn Mech Engrs Vol 214 Part D

more likely to induce separation in the diuser, and can


thus give a false indication of the ow characteristics.
In the outlet duct, making the exit area smaller than
the core area will provide a favourable pressure gradient
and the geometric shape of the outlet ow path will not
then be particularly critical. In many racing vehicles, the
ducted cooling system is designed to rely on ram air
pressure alone. In the following analysis, it is assumed
that that ow will normally be ram driven, and fan
assistance will only be needed in slow trac or abnormal
conditions.
2

OBJECTIVES OF THIS PROGRAMME

Although some well-established theoretical relationships


are available, there are insucient published experimental data to test fully their practical limitations. In
particular the eect of outlet conditions has not been
systematically investigated. Furthermore, most published data are from case studies of vehicles with partially ducted and non-idealized systems in which the ow
geometry is not fully dened. Even the theoretical
implications for low drag design seem to have received
only limited attention, and so further examination was
thought necessary.
Some preliminary results from experimental studies
were given in reference [5]. However, it was concluded
that, in these early experiments, the inlet diuser angle
was too severe, thus leading to ow separation. A
completely new set of experimental tests has therefore
been undertaken using a reworked model with a more
modest intake diuser and revised geometry, as described here.
The objectives of the programme were:
(a) to examine the practical implications of theoretical
relationships relating to the system drag and cooling
air volume owrate;
(b) to test experimentally the applicability of these
expressions, particularly at model scale;
(c) to investigate the controlling parameters for system
drag minimization.
3

THEORETICAL RELATIONSHIPS

3.1

Core pressure loss coecient

The heat exchanger or radiator core provides most of


the total pressure loss in a typical cooling system. The
losses are usually expressed by the simple relationship
ptotal 12 V 2c kp
where Vc is the nominal air ow speed in the core and kp
is known as the pressure loss coecient. This coecient
is dependent on the geometry of the core and is usually
D09099

IMechE 2000

INVESTIGATION OF THE AERODYNAMIC DRAG DUE TO AUTOMOTIVE COOLING SYSTEMS

treated as being a constant, although there is in fact a


small Reynolds number dependence.
3.2

Eect of the inlet and outlet aperture sizes

The relationship between the outlet area and the core


area is particularly important, as it eectively controls
the ow speed through the core. If it is assumed that the
air behaves as an incompressible uid at the low speeds
normally applicable to road vehicles, a useful relationship relating the ratio of this area to the core to free
stream speed ratio Vc =V1 may be derived quite simply
using Bernoulli's equation including losses; thus
p1 12 V 21

1
2
2 V c kp

12 V 2O pO

where the O subscript denotes outlet conditions.


From continuity,
Ac
VO Vc
AO
Substituting for VO in equation (1), and rearranging
gives
s
1 CpO
Vc
2

V1
Ac =AO 2 kp
where CpO is the pressure coecient at outlet:
pO p1
1
2
2 V 1
3.3

Theoretical estimates for the drag due to the cooling


system

As described by Soja and Wiedemann [6], the drag force


Dc due to the cooling system may be derived from the
equation of linear momentum for steady ow conditions. Taking a horizontal axis, and considering a control volume that extends from upstream free stream
conditions to the outlet aperture, this gives
_ 1
Dc mV

VO cos

pO

p1 AO cos

where represents the inclination of the ow to the


horizontal axis at exit and m_ Ac Vc is the mass
owrate through the duct.
Using the continuity relationship
Ac
VO Vc
AO
again, and non-dimensionalizing by the free stream
dynamic pressure 12 V 21 and the vehicle frontal area
Af , equation (3) becomes


Vc Ac
Ac Vc
CDc 2
1
cos
V1 Af
AO V1
AO
cos
4
Af
An essentially equivalent expression is given in reference
[6], and is quoted by Hucho [2], and a version for purely
CpO

D09099

IMechE 2000

921

axial ow is given by Hoerner [7]. In the above expression, the core speed Vc is by implication the ow
approach speed to the core, not the actual speed within
the core passages.
The last term in the equation represents the magnitude of the horizontal component of the force on the
control volume due to the pressure at outlet. However, it
is important to appreciate that the pressure force on this
area will already have been counted as part of the
external vehicle drag or lift and must not be counted
twice. Therefore, when superimposing the cooling ow
drag on the drag measured or predicted for the basic
vehicle shape (with no cooling ow), the last term
should not normally be included. In practice, the pressure in the region of the outlet may be modied by the
presence of the outlet ow. In this case, there will be a
change in drag, due to the dierence between the actual
pressure coecient at outlet and the pressure coecient
for the blocked duct case, as described later.
In cases where the ow is discharged vertically so that
is 90 , equation (4) reduces to the extremely simple
relationship
Vc Ac
5
CDc 2
V1 Af
This expression is also given in references [6] and [2].
Comparing equations (4) and (5) shows that the vertical discharge case will give a greater drag than rearwardly inclined or axial discharge. The reduction in drag
due to rearward discharge is given by the second term in
the parentheses in equation (4), and for 0 , this
reduces to
 
Ac Vc 2 Ac
6
2
AO V1 Af
4

OPTIMIZATION OF THE COOLING SYSTEM


FOR LOW DRAG

For a given rise in the temperature T of the cooling air


as it passes through the heat exchanger, the heat transfer
_ since the heat
should depend on the mass owrate m,
transfer rate is given by
_ p T
Q mc

where cp is in this case the specic heat of air at constant


pressure.
Optimization of the system therefore requires nding
the lowest drag that can be obtained for a given mass
owrate within the physical size constraints of the
vehicle. For a given air density, the mass owrate is
directly proportional to the volume ow rate _ , which is
equal to the product Ac Vc . Thus, ignoring the pressure
term, equation (4) may be rewritten as


_ 1
_ 1
CDc 2
1
cos
8
V1 Af
AO V1
Proc Instn Mech Engrs Vol 214 Part D

922

R H BARNARD

It will be seen, that for a given volume owrate _ , free


stream speed V1 and vehicle frontal area Af , the drag
coecient is dependent only on the outlet area AO and
the outlet angle . For low drag, both the outlet area
and the outlet angle should be small.
Figure 1 shows the eect on the cooling system drag
coecient of changes in the outlet area for a constant
outlet angle of 0 and a constant volume owrate. This
illustrative graph has been produced for an arbitrary,
but practically realistic, case of the cooling system
volume owrate being equal to 3 per cent of the product
of the free stream speed and the vehicle frontal area; that
is _ 0:03Af V1 .
Putting this condition and the condition 0 into
equation (8) gives

CDc 0:06 1

0:03

Af
AO


9

In Fig. 1, the values of CDc for these specied conditions


have been plotted for a range of values of the outlet
frontal area ratio AO =Af . It will be seen that, as long as
the required volume owrate can be maintained, reducing the outlet area lowers the drag. To maintain the
volume owrate it is necessary to increase the core area,
but there are obviously practical limitations to the
largest core that can be accommodated. Note that a
dierent arbitrary choice of _ would produce a similar
but displaced version of the curve.
From equation (2), it may be seen that changing the
loss coecient kp or the CpO value merely changes the
speed ratio Vc =V1 . This ratio, however, is important,
because, for a given volume owrate and free stream
speed, it controls the required size of the radiator core
Ac , as shown below.
Choosing again the arbitrary but realistic condition
given previously, of cooling air volume owrate
_ Vc Ac 0:03Af V1 , then
Ac
V1
0:03
Af
Vc

Fig. 1

10

Combining this with equation (2) gives


s
1 CpO
Ac
0:03
Af
Ac =AO 2 kp

11

From equation (11) we may calculate the corefrontal


area ratio Ac =Af for any combination of values of kp ,
CpO and coreoutlet area ratio Ac =AO .
The corresponding ratio of outlet to frontal area
AO =Af can be determined from
AO
Ac =Af

Af
Ac =AO

12

Figure 2 shows plots of Ac =Af against AO =Af calculated


using this procedure. This gure shows that, for a low
ratio of AO =Af (which, from Fig. 1, can be seen to
produce low drag), Ac =Af is large. This shows that, for
low drag, the radiator should be as large as can be
practically packaged.
It also appears from Fig. 2 that a low kp is preferable,
but this analysis assumes that all the cores can produce
the same heat transfer rate, whereas in reality, a high
density design is likely to be more eective. It should
also be remembered that the water pumping losses have
to be taken into consideration.
Figure 2 also shows the eect of changing the outlet
pressure coecient. A low outlet pressure coecient
reduces the size of radiator required or alternatively
permits a lower drag for a given radiator size.
Finally, for the same cases as illustrated in Figs 1 and
2, but with downward discharge 90 ) and zero
outlet pressure coecient CpO 0 the drag coecient
is always 0.06. This follows from equation (5), since
Vc Ac _ has been specied as constant. Downward
discharge therefore always produces high drag regardless of the details of the design.
An important observation is that improvements to the
internal design of the system or the selection of kp
merely dictate the size of the radiator required. The drag
is principally controlled by the outlet aperture size and
position.
It may also be concluded from this analysis that at
part load, where the cooling requirement is lower, drag

The variation of duct drag coecient CDc with the ratio of outlet area to frontal area AO =Af for a
constant volume owrate and outlet angle 0

Proc Instn Mech Engrs Vol 214 Part D

D09099

IMechE 2000

INVESTIGATION OF THE AERODYNAMIC DRAG DUE TO AUTOMOTIVE COOLING SYSTEMS

923

Fig. 2 Inuences of kp , CpO and AO =Af on the required core size for a given cooling air volume owrate and
free stream speed

reductions may be obtained if the outlet aperture is


reduced so as to restrict the volume ow to no more
than the required amount.
5

EXPERIMENTAL ARRANGEMENT FOR THE


VALIDATION OF THE THEORETICAL
RELATIONSHIPS

To test the validity of the above expressions, a model


radiator core (constructed as described below) was
mounted in a ducted arrangement inside an Ahmed
shape model, as shown in Fig. 3. The Ahmed shape,
which is described in reference [8] is commonly used as a
benchmark for computational uid dynamics (CFD)

evaluation and provides the basic aerodynamic characteristics of a road vehicle without the distractions of
styling detail. A considerable amount of experimental
and computational data exist for this shape.
The model was mounted on a conventional threecomponent force balance and placed 110 mm above a
ground board in a 1.5 m by 1.3 m open-return closedsection wind tunnel. The duct intake speed (and hence
the volume owrate of the internal ow) was determined
by means of an external pitot tube and static wall tappings in the intake aperture. The intake had wellrounded walls and could be considered as a simple
venturi meter. A discharge coecient of 0.98 was arbitrarily assumed. From the intake speed and the intake
core area ratio Ai =Ac , the core speed Vc can be inferred.

Fig. 3 Modied Ahmed model with two alternative internal ow ducting arrangements. Relevant data are:
core depth 55 mm, intake depth 52 mm, frontal area 0.0391 m2 , core area 0.006 16 m2
D09099

IMechE 2000

Proc Instn Mech Engrs Vol 214 Part D

924

R H BARNARD

Hence, knowing V1 , the speed ratio Vc =V1 can be


found.
The cooling drag increment CD was determined by
subtracting the drag values for the model with all
apertures blocked from those obtained with the cooling
system duct open. No blockage correction was normally
applied to the drag measurements, because the theoretical cooling drag increment values have been calculated
on the basis of actual measured values of free stream
stagnation and dynamic pressures and outlet static
pressure. For calibration purposes, however, the drag of
the basic model with inlet and outlet apertures blocked
was corrected using the simple continuity-based blockage
correction
factor
CD true CD (measured)
1 A=S2 , where A is the blocked area and S is the
tunnel cross-sectional area. This gave a corrected CD
value of 0.267, which is consistent with the experimental
results quoted by Ahmed et al. [8].
6

MODELLING THE RADIATOR CORE IN


SMALL SCALE

To model the pressure loss in small scale it would be


impractical to build an exact scale replica of the actual
core, and the loss eect is usually reproduced by using
some form of porous material such as a wire mesh grid.
The model core developed for these studies is described
in reference [5], and consisted of three wire mesh grids
separated by honeycomb spacers. The pressure loss is
produced mainly by the grids, the honeycomb being
intended to replicate the ow-guiding properties of the
core ns. It should be noted that, although modelling
the core in this way reproduces the required total pressure drop, the ow physics is rather dierent from that
in a real radiator. In the model, the drag mechanism is
primarily pressure drag on the grid elements, whereas in
a real radiator the drag is predominantly due to surface
friction. The use of grids does, however, have the
advantage that the kp value is more constant at the low
Reynolds numbers of model tests, as the loss mechanism
is essentially turbulent.
7

ASSESSMENT OF THE PRESSURE LOSS


COEFFICIENT kp

As described by Barnard and Ledakis [5], for the


experimental measurement of kp , the model radiator
was placed in a custom-built parallel-sided rectangular
duct which exactly enclosed it. The stagnation pressure
drop was measured by placing total pressure tubes
upstream and downstream of the core. The normalizing
speed Vc was taken as the duct ow approach speed and
was measured 5 cm upstream of the core by using a pitot
tube and static wall tappings. Note that this is not the
true speed within the core elements and is only used for
Proc Instn Mech Engrs Vol 214 Part D

normalization purposes. It is, however, consistent with


the nominal core speed value used in the experimental
expressions given earlier. The kp value measured by this
method was 10.07.
8

GEOMETRY OF THE EXPERIMENTAL


VEHICLE DUCT

As described in reference [5], preliminary testing of a


similar arrangement to that shown in Fig. 3 showed that
the ow was separating in the inlet diuser, thereby
invalidating the theoretical assumptions. Therefore the
coreintake area ratio Ac =Ai was reduced from 3:1 to
1.06:1 for the set of experiments described here. This
necessitated cutting down the size of the core kp was
unaltered at 10.07) and the use of a completely redesigned ducting system. The important dimensions of the
ducts are given in Fig. 3. The side walls of the ducts were
parallel. Two outlet locations and several outlet geometries were used, as shown in Figs 3 and 4. In all cases,
the outlet area was made smaller than the core area to
provide a favourable pressure gradient.
9

DRAG MEASUREMENTS WITH A LARGE


OUTLET APERTURE UNDER THE MODEL

For the rst set of tests, the model geometry was as


shown in the left-hand version in Fig. 3. The outlet was
on the underside and had an area only slightly smaller
than the core area.
In order to establish the correct value for the pressure
coecient at outlet, CpO , the model was placed in the
wind tunnel, and the static pressure at outlet was measured by means of a static probe traversed to various
points around the outlet aperture. An average coecient
was then calculated. This gave CpO 0:005.
If a vertical discharge of air is assumed, equations (2)
and (5), give the theoretical speed ratio and duct drag
coecient increment shown in Table 1. The experimental results were obtained at a test Reynolds number
based on a model length of 6:5  105 .
The experimentally determined value of the speed
ratio Vc =V1 is slightly higher than predicted, but this is
probably due to the assumed inlet discharge coecient
Table 1 The eect of cooling system drag: large aperture on
underside
Conguration
Experimental, all apertures
blocked
Experimenta, duct with
0.005 77 m2 aperture
underneath
Theoretical

CD (raw)

CD

Vc =V1

0.0983
0.0922

0.2988
0.2979

0.293
0.391
0.385

D09099

IMechE 2000

INVESTIGATION OF THE AERODYNAMIC DRAG DUE TO AUTOMOTIVE COOLING SYSTEMS

of 0.98 being too high. The drag increment is also


slightly higher than predicted, and this may be due to
interference eects between the internal and external
ows.
Note that the drag increment due to cooling ow is
not typical of a domestic car, because the radiator core
is unrealistically large in relationship to the vehicle
frontal area, although it could be applicable to a racing
vehicle where the cooling requirement is much greater,
owing to the high output power.
The theoretical drag penalty resulting from the
destruction of axial momentum due to downward discharge may be calculated from equation (6), and has a
value of 0.029, which represents 28 per cent of the total
measured drag increment. This shows the considerable
penalty associated with vertical discharge.

10

THE INFLUENCE OF THE OUTLET


APERTURE SIZE

As may be seen from equation (2), the outlet aperture


size is important, since it controls the ow speed through
the core. To test the theoretical prediction, the model
outlet aperture was reduced to half the original value by
covering the rear half of the outlet aperture. The results
are given in Table 2 and show that the ow is indeed
controlled by the outlet aperture, as predicted. The drag
is also lower than for the larger aperture, and this shows
Table 2 The eect of reducing the outlet aperture
area for a vertical outlet
Conguration

CD

Experimental, outlet aperture


reduced by 50%
Theoretical,
for same case

Fig. 4

D09099

Vc =V1

0.088

0.262

0.081

0.261

925

the advantages of controlling the outlet aperture to


match the cooling requirements at varying power outputs.

11

INFLUENCE OF OUTLET GEOMETRY

The geometry used for the above results was deliberately


arranged to give a near-vertical outlet ow in order to
test the theory. The vertical duct, is however, not typical
of the likely arrangement in a real vehicle, and the more
realistic conguration illustrated in Fig. 4 was tried. It
was expected that this arrangement would produce a
signicant rearward component. The results are shown
in Table 3.
From Table 3, it can be seen that the change in geometry has radically altered the drag and that the new
drag corresponds quite closely to the theoretical case of
an outlet angle of 30 . This angle appears to be reasonably consistent with the geometry.
Note that, although the drag values for the two
reduced outlet aperture cases are quite dierent, the
speed ratios Vc =V1 are nearly the same, as predicted by
the theory. These experiments show the importance of
the details of the outlet geometries and the dangers of
making crude assumptions about the outlet ow direction.
Table 3 The eect of venting the duct at less than
90
Conguration
Experimental, outlet aperture
geometry as in Fig. 3
Theoretical,
outlet angle 90
Theoretical,
outlet angle 30

CD

Vc =V1

0.040

0.258

0.081

0.261

0.041

0.261

Model with revised underside outlet geometry giving a rearward component of velocity. The outlet
area was 0.0029 m2

IMechE 2000

Proc Instn Mech Engrs Vol 214 Part D

926

12

R H BARNARD

THE EFFECT OF PLACING THE OUTLET IN


A LOW PRESSURE REGION IN THE REAR
FACE

From equations (2) and (4) the eect of exhausting the


cooling air into a rearward-facing low pressure area at
the back can be predicted. Because of the lower axial
momentum change, the losses should be lower. The
outlet duct was therefore recongured to provide an
outlet aperture in the rear face as shown in Fig. 3.
Equation 2 shows that, by an appropriate reduction
of the outlet aperture, the speed ratio Vc =V1 and hence
mass owrate may be kept to the same values as for the
case of the large underside aperture of Table 1. The
mean pressure coecient on the rear face of the model
was measured as 0:289 when the duct apertures were
blocked and 0:264 when open. The latter value was
used as the outlet pressure coecient CpO . The theoretically predicted required duct outlet area was
0.0028 m2 . The model duct was constructed accordingly,
and the results are shown in Table 4.
Because of the rise in rear face (base) pressure when
the duct was open, the vehicle drag coecient should fall
by the amount
CpO

CpOB

13

AB
Af

where
CpOB mean pressure coecient on the rear face
when the duct is blocked o
CpO pressure coecient when it is open
AB base area, the area of the rear face of the model
(including the duct aperture) to which this
pressure applies (0.028 06 m2 )
As shown in Table 4, applying this small pressure correction appears to account for most of the discrepancy
between the measured and theoretical results, but it
cannot of course be made predictively, since the change
of pressure coecient will not be known.
Table 4 The eect of repositioning the outlet in
the rear face with appropriately reduced
aperture to maintain the same mass
owrate
Conguration
Experimental, duct with
0.005 77 m2 outlet
underneath
Experimental, duct with
0.0028 m2 outlet at rear
and CpO 0:27
Theoretical, for above duct
with outlet at rear
Theoretical, corrected for
base pressure reduction

Proc Instn Mech Engrs Vol 214 Part D

The reduction in drag compared with the underside


aperture case is signicant, and for a more realistic ratio
of radiator core to car frontal area the cooling system
drag would probably equate to less than 0.015, which is
low by current standards.
In a production car it is unlikely to be practical to
extend the outlet ducting to the rear. However, it is clear
that there are advantages in ensuring that the outlet is
directed as near axially rearwards as possible.
As noted previously, vertical exhausting will produce
a lift change due to momentum of a similar magnitude
to the drag increment, so venting the air downwards will
produce a signicant lift contribution, and venting it
upwards will produce a corresponding downforce contribution. The location of the outlet on the top of the
bonnet as on the Lotus Elise would thus have advantages for a high performance or racing vehicle. As usual,
however, a balance has to be struck between the relative
advantages of downforce production and drag minimization. In addition, consideration has to be given to
the design constraints that may be imposed by this
location of the outlet. A fully ducted design will also
carry a certain weight penalty.

CD

Vc =V1

0.0983

0.2988

0.0187

0.309

0.0323

0.292

0.019 57

THE INFLUENCE OF INLET APERTURE


AREA

Equation 2 shows that the core speed depends only on


the loss coecient, the outlet pressure coecient and the
outlet area. The inlet area is thus theoretically not
important. In practice, making the inlet aperture small
tends to produce an adverse pressure gradient and hence
to induce separation in the intake duct.
To test this prediction experimentally, the intake
aperture was reduced by 10 per cent by inserting a
contoured lining on the lower duct wall. The model was
re-tested, giving an overall vehicle raw CD value of 0.292
instead of 0.293. This represents a change of less than
0.4 per cent, which was within the range of experimental
repeatability. It is thus concluded that the inlet aperture
area does indeed not directly aect the ow losses to a
signicant degree, unless the resulting geometry produces a suciently adverse pressure gradient in the
intake diuser to degrade its function or eciency.
Theoretically, the location of the inlet, or the pressure
coecient in its vicinity, should not be important as long
as there has been no loss in stagnation pressure, but no
experiments were conducted to verify this.
14

THE EFFECT OF HEAT ADDITION

In these experiments, no account has been taken of the


eect of heat addition. In practice it would be dicult to
model the eect in small scale, as the number of
dimensionless groups to be matched increases. Simply
D09099

IMechE 2000

INVESTIGATION OF THE AERODYNAMIC DRAG DUE TO AUTOMOTIVE COOLING SYSTEMS

heating the model radiator to the same temperature as in


full scale would not produce dynamically similar conditions.
Some idea of the order of eects due to heating can be
gained by considering the heated duct as a low pressure
ramjet. The ideal cycle eciency depends on the pressure ratio r, the maximum possible ratio being given by
free stream stagnation pressure
free stream static pressure


1 2 = 1
1
M1
2

where M1 is the Mach number of the free stream


approach ow.
Even at racing speeds of 200 miles/h, M1 will only be
0.26 (for standard sea level conditions). For this condition, the value of r given by the above expression will
only be 1.049.
The ideal cycle eciency  is given by


 
1
r

1=

For the pressure ratio r of 1.049 calculated above, the


cycle eciency  would be 1.53 per cent. On top of this
low theoretical maximum cycle eciency, the jet eciency must be considered, so even under the most ideal
conditions only a tiny proportion of the waste heat
would be converted into work; the rest will just make the
air hot.
15

CONCLUSIONS

The theoretical expressions used in this paper give a


reasonably good prediction of the system performance
even in small model scale.
The outlet aperture area location and orientation are
critical controlling factors; a truly vertical outlet ow
will cause a high drag penalty.

D09099

IMechE 2000

927

For low drag coecients, either a rearward-facing


outlet is required or the ow should be directed so as to
attach to the bodywork. The drag may be minimized by
using a small outlet aperture, the limiting factor being
the maximum permissible size of the radiator core.
From this, it follows that signicant improvements in
vehicle eciency could be obtained if the outlet aperture
were controlled automatically to produce the correct
owrate for the engine power setting.
Exhausting to a rearward-facing low pressure area
can have the added advantage of reducing the vehicle
exterior pressure drag contribution in the outlet region.
The inlet aperture size is unimportant, as long as the
resulting inlet diuser geometry does not produce
separation. This will be more apparent in model scale
owing to the lower Reynolds numbers.
REFERENCES
1 Williams, J. An automotive front-end design approach for
improved aerodynamics and cooling. SAE paper 850281,
1985.
2 Hucho, W.-H. Aerodynamics of Road Vehicles, 4th edition,
1998 (SAE International, Warrendale, Pennsylvania).
3 Renn, V. and Gilhaus, A. Aerodynamics of vehicle cooling
systems. In Proceedings of Sixth Colloquium on Industrial
Aerodynamics and Road Vehicle Aerodynamics, Fachhochschule, Aachen, 1985, pp. 303311.
4 Paish, M. G. and Stapleford, W. R. A study to improve the
aerodynamics of vehicle cooling systems. MIRA Reports
1966/15, 1966 and 1968/4, 1968.
5 Barnard, R. H. and Ledakis, N. Physical modelling and
optimization of radiator cooling ow systems. In Proceedings of Second MIRA Conference on Vehicle Aerodynamics,
October 1998, session 5.
6 Soja, H. and Wiedemann, J. The interference between
internal and external ow on road vehicles. Ing. Auto.,
September 1987, 101105.
7 Hoerner, S. F. Fluid Dynamic Drag, 1965 (Hoerner Fluid
Dynamics, Brick Town, New Jersey).
8 Ahmed, S. R., Ramm, G. and Faltin, G. Some salient
features of the time averaged ground vehicle wake. SAE
paper 840300, 1984.

Proc Instn Mech Engrs Vol 214 Part D

Potrebbero piacerti anche