Downloaded from SAE International by University of Wisconsin - Madison , Monday, September 10, 2018

Application of Computational Methods

to the Aerodynamic Development
of a Prototype Race Car
Joseph Katz
San Diego State Univ.

COPYRIGHT Lee Bykstra

SOCIETY
-- OFAUTOMClTlVS
-
Special Chassis Inc.

ENGINEERS, INC.

aerodynamics. Thus, the objective of the present study is

ABSTRACT to demonstrate that the combined use of. numerical and
experimental techniques can help to improve the under-
A three-dimensional computer simulation technique standing of the fluid dynamic flow over the vehicle and
was combined with wind-tunnel testing during the aero- save wind tunnel and model develo~menttime.
dynamic development of an enclosed-wheel prototype race
car. This approach proved that valuable time can be For the present study a three-dimensional computa-
tional fluid dynamic method was incorporated early in
saved by investigating some of the important design pa-
rameters before a vehicle is built. One of the major ad- the design cycle of the prototype race car, shown in the
upper part of Fig. 1. The surface grid of the computa-
vantages of a computational approach is that it contains tional model is shown at the lower part of Fig. 1. At this
information such as pressure or velocity distribution on phase numerous ideas were investigated numerically, and
and near the whole vehicle. This abundance of data is
compared with results of quarter-scale wind-tunnel test-
essential for understanding major design trends and sen- ing. Once the vehicle concept was finalized, wing airfoil
sitivities, and can steer the design toward fruitful mod- shape and its positioning was developed numerically and
ifications. Once the vehicle's body plan is finalized, the validated later during the following wind tunnel
method can be used to further modifv local details and to
.-- -

design and position a complicated rear wing cluster. At

this phase of wing design, the availability of the pressure
distribution on the entire wing surfaces is vital to a suc-
cessful design. Since such information is readily available
from computational tools, valuable wind-tunnel time is
saved, and vehicle development cycle is shortened.

INTRODUCTION

Computational aerodynamic tools are widely used in

the aerospace industry, but in general are considered to
be expensive and require large computational facilities.
Recently, however, this technology began to migrate to
the race-car industrv because of its maturitv and the es-
calating cost of win4 tunnels with proper rnbving ground
simulation. So, interestingly, the emerging reasons be-
hind this increase in the use of computational methods is
the need to save development time and cost (compared to
wind tunnel testing), resulting in an overall better prod-
uct. Another advantage of computations, when compared
with wind tunnel methods, lies in the fact that the results
can be used over and over to study new parts of the prob-
lem. This is not the case with wind tunnel experiments
where once the model was taken out of the test section,
new questions can not be answered. Of course, compu-
tational fluid dynamic methods are far from being the
perfect prediction tool, but also wind tunnel experiments Fig. 1 Rear view of the prototype race car and its nu-
are not quite flawless in simulating open-air r o a d - ~ ~ h i ~ l emerical grid (using 2558 ~ a n e l s ) .~ c t u a computational
l
grid is denser.
 Downloaded from SAE International by University of Wisconsin - Madison , Monday, September 10, 2018

The next important phase of using computational THE COMPUTATIONAL METHOD AND THE PRO-
aerodynamic tools can be identified as the period after TOTYPE RACE-CAR
the actual vehicle was built. Even during this phase, nu-
merous aerodynamic aspects of the vehicle can be inves-
tigated numerically. For example, changes such as a new The fluid dynamic equations, solution of which con-
wing can be developed numerically with sufficient confi- tains the information on the aerodynamic loads on a mov-
dence, since the flow is attached on most of the body-work ing vehicle, can be found in most related text books (e.g.
of such streamlined vehicles.
The following data presented in this article, therefore,
6
Ref. 2). The solution of those equations called the
~avier-stokesequations) is quite complicate , even for
simple cases, and for practical engineering applications
serve to demonstrate the above mentioned applicability of numerous simplifications must be incorporated. One of
the numerical tool. As the numerical tool, the panel code the earliest models used in conjunction with subsonic
of Ref. 1 was selected, because of the previously gained airplane-wing design was based on the inviscid flow as-
experience in its application (Ref. 2 and Ref.3) and its sumption which, in attached flow conditions, is capable of
suitability to the smaller workstation environment. The calculating the pressure field over three dimensional bod-
experimental data for this vehicle was obtained with a ies. Such methods have matured in recent years, requir-
114-scale model which was tested in two different facili- ing less computational times for engineering applications,
ties. At first, a fixed ground plane method was used and and therefore are widely used in the aerospace industry.
the wind-tunnel test section speed was 160 Kmlh (100 Thus, for a highly streamlined vehicle with limited and
mph) with a 3.5% frontal-area blockage. Later, a rolling localized flow separation, such a method can estimate the
ground facility was used with a frontal-area blockage of pressure distribution and resulting aerodynamic loads.
1.7%, but free-stream speeds were reduced to 96 Km/h Since, within the formulation of these methods the viscos-
(60 rnph) because of the limit on the rolling belt. The ity terms are omitted from the governing equations, the
exper~mentaldata presented here in support of the nu- skin friction contribution to drag cannot be calculated
merical calculations is based on the latter data with the directly (though some codes have an interactive bound-
rolling ground simulation. ary layer calculation capability to overcome this setback).
Another drawback, resulting from the absence of viscous
terms, is the inability t o directly model regions of sepa-
rated flows. However, the objective of a good design is
to minimize these areas and then such codes can be used
to identify areas where the flow is likely to separate. So,
during a successful utilization of such a method, the body
shape can be modified until the flow becomes attached
over the surface; leading to a close agreement between
the calculated and the actual pressure field.
As mentioned earlier, for the present study a three-
dimensional inviscid solver (panel-code) was used and
more details on the numerical aspects can be found in
Ref. 2; user related details are provided in Ref. 1. The
applicability of this model to airplane wings and stream-
lined bodies is documented in the open literature (e.g.
in Ref. 2) and its validation for enclosed-wheel race cars
is provided in Ref. 3. Typical results of this validation
(taken from Ref. 3) are presented in Fig. 2. The lower
curve in this figure shows the computed and measured
pressure distribution along the upper centerline of the
streamlined body. Thc shape of the vehicle body and lo-
cation of the pressure ports are shown schematically on
the inset in this figure. In this case the flow was attached
on the entire upper surface and clearly, the computed
and measured data compare well. The upper curve in
Fig. 2 shows a similar comparison along the centerline of
the underbody tunnel (parallel to but not along the ve-
hicle's centerline). The discussion in Ref. 3 clarifies that
the good comparison is a result of a longitudinal vortex
originating along the tunnel side, which attaches the flow
inside the underbody tunnel (similar vortices were seen
on the present vehicle, as well, as indicated by Fig. 3).
Thus, in conclusion, in spite of the simplified mathemati-
cal formulation of the present numerical method, its ma-
tureness makes it practical for engineering applications.
Also, previous validations imply that the method can be
xi^ used to calculate the attached-flow pressure distribution,
lift, and induced drag, on complicated wing surfaces.
Fig. 2 Comparison between the computed and mea-
sured pressure distribution along the body and tunnel The computational grid and a photograph from a sim-
centerlines of a prototype race car (no wheels, after Ref. ilar angle of the prototype race car are shown in Fig. 1.
3). This type of race car utilizes underbody tunnels to create

162
 Downloaded from SAE International by University of Wisconsin - Madison , Monday, September 10, 2018

of Ref. 4, were aimed precisely a t this problem and it was

found that, in general, the downforce of the vehicle with-
out the wheels increased by close t o ACL = -2 whereas
the drag was reduced somewhat. The large difference in
the downforce reported in Ref. 4 is probably a product
of the very low ground clearance and the stationary wind
tunnel floor in those experiments. These two factors lim-
ited the flow under the vehicle with the wheels, causing
the underbody tunnel flow t o stall, which may explain
the large loss of downforce and increase in drag with the
wheels on. As mentioned earlier, the wind-tunnel results
presented in support of the following figures are based on
experiments with rolling ground simulation. In this case,
the effect of not modelling the wheels in the computa-
tions was found to be on the order of ACL = -1.5 (that is,
more downforce without the wheels), as will be explained
in the next section.

AERODYNAMIC RESULTS

Aerodynamic considerations influence the shaping of

many vehicle components, and design tools such as com-
putations or wind tunnel data is needed from the prelim-
inary design stage (when a rolling vehicle does not ex-
ist) up to the development and refinement stages of the
existing vehicle. This is probably more important in the
preliminary design stage when a computational approach
can provide a reasonably quick avenue for checking out
numerous new ideas. For example, the small flap shown
in Fig. 4 (lower part) was designed t o increase the flow
emerging from the tunnels by creating a low pressure at
the exit. The upper part of the figure shows the car with
-- the flap and the full-span rear win^. The idea seemed
promisjng and this flap
gained size, until it developed
into a full-size wing, which was mounted lower than tra-
Fig. 3 Schematic description of the flow field below ditional rear wings. Note that the model at this stage
the vehicle (based on flow visualizations) and a similar had a blunt tail (Fig. 4), contrary to the streamlined tail
view of the computationa1 grid.

additional downforce by the body. A schematic descrip-

tion of these tunnels is shown in the upper part of Fig. 3,
along with a similar view of the surface grid used for the
computations. The flow visualizations, obtained with a
115-th scale model at 225 Km/h (140 mph), indicate that
in spite of the complex flow under the vehicle the flow in-
side the underbody tunnels was attached (upper part of
Fig. 3). This is a result of the two vortices created by the
side flow entering the tunnels, as described schematically
in the figure. The importance of this observation is that
in this case, as well as in the case reported in Ref. 3,
results of the computed pressure distribution (inside the
tunnels) fall close t o the experimental values.
The numerical grid of the race car, as shown in Fig.
1 and Fig. 3, does not have wheels. This simplification
in the model geometry results in a considerable reduction
in the computational effort, but several of the local un-
derbody flow details shown in Fig. 3 are compromised.
However, the primary focus of the aerodynamic improve-
ments was on the vehicle's upper surface and rear wing
cluster, where the flow was attached. It was felt that in
these areas the above simplifications were reasonable for
the purpose of engineering calculations. This rationale
was also reinforced by a similar study in Ref. 3 and by the Fig. 4 Geometry of the tunnel-exit flap proposed t o
experimental results of Ref. 4. The latter experiments, increase the flow under the car (lower figure and when
combined with a standard rear wing (upper Bgure).
 Downloaded from SAE International by University of Wisconsin - Madison , Monday, September 10, 2018

An important aspect of the early aerodynamic inves-

tigation is the capability t o estimate the vehicle's aerody-
namic coefficients, such as downforce, drag, and front/rear
downforce distribution. This information can then be
used for performance prediction and dynamic simulation,
so t h a t t h e competitiveness of the vehicle can be esti-
mated, before major finances are committed. Typical
parameters affecting the aerodynamic performance are
vehicle ride height, attitude, and wing positioning and
setting. Because of the large number of test parameters a
comprehensive wind tunnel program is quite lengthy and
expensive. Therefore, computations can be used t o derive
information on the generic trends, and to sort out non-
promising modifications. In this case of a highly stream-
lined vehicle, wind-tunnel testing can focus later on refin-
ing the geometry and t o adjust for effects not accounted
for by t h e computations (e.g. effects of boundary layer
thickness or local flow separations).
As a n example, the effect of vehicle ride height and
pitch angle on the aerodynamic coefficients is presented
in Fig. 6 and Fig. 7. T h e upper curve in Fig. 6 shows
the calculated trend, indicating that the vehicle down-
force will increase with reduced ground clearance. This
is a result of the flow being accelerated between the car's
lower surface and the ground. As discussed earlier, the
computational model did not include the wheels, and
this resulted in an overprediction of the downforce by
Ac, = 1 . 5 . (This value of -1.5 was based simply on the
difference between the experimental and the numerical
results and an extensive study of the wheel effect, as in
Ref. 4, was not carried out in this case.) Thus, the open
diamond symbols in Fig. 6 indicate the corrected results
for the calculated lift (after subtracting -1.5 from the lift
Fig. 5 Numerical visualization of the cooling exit coefficient). Interestingly, the vehicle suspension travel
flow, used t o locate engine induction air inlet. was limited such that the experimental d a t a (solid sym-
bols) was in the range of h = lin t o 2in (2.5cm t o 5cm)
and beyond this height only the computed results were
of the actual car as shown in Fig. 1. So, in order t o
increase the wing/body interaction and to enable easier
numerical modelings of the geometry, the sharp trailing
edge (of the body and the beaver tail (behind the rear
A + Experiment
0 A 0 Computation
fenders) was developed. This modeling exercise started
late in 1990 and by mid 1991 the biplane rear wing shown
in Fig. 1 was developed. Of course, a t that time open-
wheel formula-one cars used such biplanes, but this study
clearly indicated the role of the lower wing in helping the
flow exiting t h e underbody tunnels.
Another example of using computations for less tra-
ditional application is described in the flow visualization
of Fig. 5. As noted earlier, the flow was attached on
the body's upper surface and computational results were
close t o those of the flow visualizations. In this partic-
ular case the numerical visualizations served to identify
the path of the warm cooling flow exiting ahead of the
windshield, so that a n engine induction air intake could
be laced in the cold free-stream. The surface-flow visu-
alization indicates that the cooling flow is concentrated
0 2 4 6 8
Front ride-height, hf, in
10
Y
-
above the cockpit and the intake was placed on top of
the right rear fender (see sketch in Fig. 5), where no low
momentum and warm airflow was e x ~ e c t e d .
* T h e sharp trailing edge of the body, spanning the Fig. 6 Variation of vehicle lift and drag with full-scale
whole width, allows definition of a wake line similar to ground clearance. T h e corrected lift c~.... takes into ac-
the one used t o specify the trailing edge wakes on airplane count the effect of not modeling t h e wheels. ~ x ~ e r i m e n t a l
wings. For more details on this topics see ~ ~ I andf .~ ~ . is based on 1 scale test with rolling ground at an
fdata
2. airspeed of 60 Mp
 Downloaded from SAE International by University of Wisconsin - Madison , Monday, September 10, 2018

available. A possible conclusion from this diagram is that tions. Furthermore, a computation includes information
the computed and measured data differ significantly, be- on the surface pressure and resulting loads on the vari-
cause of the opposing slopes. However, similar data on ous vehicle parts. This data reveals that the front/rear
other vehicles (e.g. Ref. 5 and Ref. 6) indicate that downforce ratio (not plotted) was not affected seriously
the correct trend is the computed one. This means that by varying the ride height. Another interesting aspect of
the downforce will increase as ride height is decreased, the downforce variation, as depicted by the combination
but when ground clearance reaches a certain minimum of the corrected and the experimental data in Fig. 6,
(Zin, according to this data), the trend is reversed. This is that at high speed the vehicle is protected from aero-
reversal at the lower ground clearance values is proba- dynamically overloading itself (since the downforce de-
bly caused by the viscous boundary layer on the vehicle's creases with very low ground clearance).
lower surface that reduces the flow between the car and
the road that feeds the large tunnels. So, the conclusion
from this diagram is that the flow under the car should The effect of pitch angle, a, on the vehicle's lift and
he increased a bit or else the size of the tunnels needs drag for two front ground-clearance values is presented
to be reduced, possibly by reducing their upward angle. in Fig. 7. For this case the vehicle was pitched about
An increase in the flow under the car can be achieved by its nose and the front ride height, h,, was held constant.
raising the panel between the front wheels or simply by The slopes of the calculated and measured data show
increasing the ride height (which is not a popular option). similar trends but, as noted earlier, the vehicle is placed
too close to the ground (which explains why there is more
The calculated drag coefficient data in Fig. 6 com- donwnforce with hf =2in than with hf =1.5in). Similarly,
pares quite favorably with measured values. This is a re- the downforce should increase with reduced pitch angle
sult of the larger computed downforce (yielding in larger (which it does), but the trend is reduced as the angle
induced drag) which, in a way, compensates for the fric- decreases because of the viscous effects limiting the flow
at the small front clearance. This trend reversal a t the
tion drag not accounted for by the numerical model. So, smaller ground clearance values is best demonstrated in
Fig. 6 may be viewed as an engineering calibration of
the computational tool and, after establishing the correc- Fig. 6, which indicates that at h, of less than lin (in full-
scale) the reduction in downforce is quite large. Because
tions aCL = -1.5 and ACD = 0.0, the effect of small model of this reason, only the calculated slopes of the lift and
changes can be estimated without any additional correc- drag (near a = 0") were plotted versus the pitch angle. As
discussed before, the slope of the experimental downforce
data is similar, except for the increase in the slope at
the positive angles of incidence (due to the very small
rear clearance h,). Another important parameter, when
pitchin the vehicle, is the frontlrear downforce ratio (not
plotted7 which did not change much with a, indicating
low level of pitch sensitivity. The calculated changes in
the drag coefficient in Fig. 7 are close to the measured
values since these changes are caused primarily by the
'lift induced' effect.

Thus, after the initial calibration of the method, the

incremental predictions are quite good. For example, the
effect of wing angle change on the loads on various body
panels can be estimated. For this purpose, the loads on
a baseline configurations are calculated, and such an ex-
ample is presented in Table 1. This information is quite
useful for the structural design, a9 well, especially in view
of the large loads acting on the underbody tunnel (which
in fact was a stressed chassis part).
&
Front hood -0.06
Cockpit 0.44
Rest of Upper body 1.04
F. underbody -2.04
R. underbody (+ tunnels) -3.12
Low wing (main element) -0.99
Low wing flap -0.14
Upper wing -0.47
-
Total -5.34

Fig. 7 Variations in vehicle lift and drag versus body's

pitch angle. h, is full-scale ground clearance. The solid Table Typical results showing the calculated (un-
lines stand for calculated slopes and not for the absolute corrected) lift and drag contributions of various body
values of the computed data. Experimental data is based parts.
on 114-scale test with rolling ground at an airspeed of 60
Mph (96 Km/h).
 Downloaded from SAE International by University of Wisconsin - Madison , Monday, September 10, 2018

puted) lift coefficient increases as the wings are lowered

A +
Experiment
A 0 Computation
towards the body's upper surface. Therefore, the wind
tunnel experiments (only two points) focused a t the low
(corrected) body/wing clearance portion of this diagram, in order t o
determine the point where the computed trend is being
reversed by the boundary layer. So, when combining the
computed and experimental results, we can conclude that
the downforce will increase (along with the drag) down
t o a clearance close to 5in (12.7cm), at which clearance
this trend is reversed due t o the viscous flow effects. An-
other interesting aspect of the computations is that it
includes many details not included in this figure. For ex-
ample, it identifies t h a t the increase in downforce with
reduced clearance h (in the range of h >> sin) is mostly
generated a t the vehicle's lower surface (winglbody in-
teraction), while a t the same time, the lift of the upper
wing is somewhat reduced.
Previous reports on the effect of the horizontal trans-
lation Az of a rear wing relative to the vehicle body (e.g.
h, in Ref. 3 and Ref. 4), showed less sensitivity than the verti-
cal displacement h. This is also the case in Fig. 9 where
Fig. 8 Variation of vehicle lift and drag versus reaf- both experimental and computed d a t a show only a small
b'
wing cluster height h tndlcates full-scale values). Experl-
mental d a t a is base on 114-scale
' test with rolling ground
a t a n airspeed of 60 Mph (96 Km/h).
effect. Usually, too large of an overlap (minus x direction)
will reduce wing/body interaction but, as the experimen-
tal d a t a suggests, rear wing leading edge must be above
and slightly overlapping the trailing edge of the body.
Because of the unfavorable interaction between two
wings placed one above the other (see biplane in Ref.
8). the upper wing should be placed as high as possible
APPLICATION FOR WING DESIGN AND WING PLAC- within the maximum height set by regulations. There-
ING fore, the only other variable t o be investigated is the hor-
izontal placing of t h e upper wing. The d a t a in Fig. 10
shows very little sensitivity, although the computed d a t a
The shape of a race car body is often limited by reveals that the upper wing lift increases as it moves for-
certain regulations and engineering considerations other ward (but a t the same time the lift of the lower one is
than aerodynamics. Therefore, as cornering speeds in- reduced with almost no change in the total coefficients).
creased and inverted wings were added to increase the
aerodynamic downforce, the shaping and positioning of The intermediate conclusions from this brief inves-
those wings to fit within a particular set of rules be- tigation of rear wing cluster positioning are as follows.
came quite difficult. Past experience (e.g. Ref. 7) has First, the main wing (the lower one) should be placed
shown that race-car wing design is strongly affected by
the presence of other body components and the ground.
Therefore, standard airfoil shapes used in the aerospace
industry need t o be modified or completely redesigned for
race-car applications. Consequently, the first part of the
following discussion describes the use of computations in
selecting rear wing positioning relative t o the body, while
the second part of the discussion will focus on the airfoil
shape design. A+ Experiment
A O Computation
Numerous parameters affect the performance of the (corrected)
rear wing cluster. For instance, the angle of attack of
both wings and the flap angle of the lower wing can be
varied. This, when combined with the relative position-
ing of the two wings between each other and the body,
adds up t o a quite lengthy testing program. Again, the
use of computational methods can shorten this effort by
presenting the dominant trends versus the above vari-
ables. As an example for this type of evaluation Fig. 8
- 10 are presented. The first figure of this group (Fig.
8) depicts the effect of raising/lowering the wing cluster
(the geometry of which is depicted in the inset t o this fig-
ure). Previous data (as in Ref. 7) indicate t h a t lowering Ax, in
the rear wing of enclosed-wheel race cars will increase the
downforce as a result of the favorable interaction. This Fig. 9 Variation of vehicle lift and drag versus rear-
trend is then reversed when the wing is too close t o the wing cluster horizontal position (A= indicates full-scale
boundary layer formed above the vehicle's body. Similar values). Experimental d a t a is based on 114-scale test
behavior is shown in Fig. 8, where the corrected (com- with rolling ground a t a n &speed of 60 Mph (96 Km/h).
 Downloaded from SAE International by University of Wisconsin - Madison , Monday, September 10, 2018

5.0 to turbulent flow is expected within the boundary layer

1 (usually near the point of maximum suction). The last
part oithis diagram consists of a recovery region, where
the pressure gradient is unfavorable, and this part is the
most important in the effort to avoid flow separations.
Liebeck, in Ref. 9, proposed a family of such curves based
on the Stratford separation criteria, so that if the pres-
sure distribution in the suction side is contained within
A* Experiment this diagram (as in Fig. ll),then the flow over this airfoil
A 0 Computation will stay attached (at a particular Reynolds number). Of
(corrected) course, in practice, the sharp corners in this schematic
target distribution must be avoided and the dashed line
shows a more practical pressure distribution, not exceed-
ing the boundaries set by the solid line. Following this
approach, which was developed for twc-dimensional air-
foils. the present computations can be used to develop
0.0
I I I I I , I I I I I thre&dim;nsional wing shapes. Thus, the pressure dis-
tribution on the race-car wine, at any spanwise section.
-5 -3 -1 1 3 5 should fit within a desirable Gessure distribution env;
in
AX. lope as demonstrated by the following two examples.
Fig. 10 Effect of upper wing horizontal position on Fig. 12 demonstrates such a design procedure for a
vehicle's lift and drag (A* indicates full-scale values). Ex- a single element wing and the boundaries for the target
perimental data is based on 114-scale test with rolling pressure distribution are shown by the solid line. In prac-
ground at an airspeed of 60 Mph (96 Kmlh). tice, the airfoil surface must be smooth (free of humps)
and an exact match of such a simplistic target pressure
is not necessarv. In this oarticular wine section. close to
one-quarter o i t h e chord is laminar, & reduce'the vis-
as low as possible, until the increase in wing/body in- cous drag. Also, the trailing edge pressure recovery is
teraction is reversed by the boundary layer effects. The less steep than allowed by the solid line, so flow separa-
leading edge of this wing should somewhat overlap the tion is unlikely. Another example is presented in Fig. 13.
car's trailing edge, resembling the overlap of the trailing Here the approach is applied to a two element airfoil and
edge flap in a rnultielement airfoil. Also, the second wing a target pressure distribution with a higher suction peak
should be placed as high as possible, and as forward as (near c,=-6) is selected from the family of curves of Ref.
possible (based on the effect of stagger suggested in Ref. 9. In the case of a two element airfoil. lareer levels of suc-
tion can be maintained near the trailing a g e of the main
element because of the interaction with the flap's leading
WINGIAIRFOIL DESIGN edge. Also, the solid line (in Fig. 13) describing the tar-
get pressure-distribution boundaries was developed for a
During the design or modification of race-car wings, single element airfoil (without a flap). Therefore, even
the distorted flow field caused by other body components
must be taken into account. For this task, information
about the pressure distribution over the whole wing sur-
face is needed, so that features such as maximum lift over
drag can be achieved. Experimental evaluation of this Laminarlturbulent rooftop
pressure field is virtually impossible within the time lim-
its required for the development of such vehicles. There-
fore, this aspect makes the use of computational methods
very attractive, and their application is demonstrated in
the following examples.
A typical wing design process can start with setting
up a target pressure distribution (on the suction side
of an airfoil) to meet tasks such as maximum lift over
drae. Fieure 11 describes schematicallv the main reeions
in sich a-target pressure distribution and Ref. 9 pro;ides
information about how to establish those diagrams. Sim-
ilarly, target pressure distributions can be dGveloped by
using more recent airfoil design methods, such as those
described in Ref. 10 and Ref. 11. Typically, near the air-
foil's leading edge, the suction builds up and within this
region the pressure decreases along the chord. This r e
gion, with a favorable pressure gradient, allows the devel-
opment of a laminar boundary layer as long as the pres-
sure decreases. This, of course, can result in a noticeable
reduction of the viscous friction drag. The next region,
shown in Fig. 11, is the top (called "rooftop") where Fig. 11 Schematic description of a desirable pressure
the pressure variations are small and also a transition distribution on the suction-side of an airfoil.
 Downloaded from SAE International by University of Wisconsin - Madison , Monday, September 10, 2018

-1 500
CONCLUDING REMARKS
Targel

Aclual The ability to generate information from two inde-

pressure pendent sources such as wind-tunnel testing and compu-
tations provides a more complete data base on a vehi-
cle's aerodynamics. Wind-tunnel data may be effective
-0M O - in measuring the important coefficients, whereas compu-
tations are more flexible when investigating numerous ge-
ometrical changes (since no models and parts need to be
fabricated). Also, computational methods often provide a
more complete explanation on the flow phenomena, which
can steer the designer into the direction of the desirable
modification. Such is the case when developing the shape
of a wing cluster with several elements, where the use of
computations can considerably shorten and improve the
design process. After calibrating the numerical model
(based on the experiments), incremental effects due to
minor modification in the attached flow regions can be
predicted quite satisfactorily.
To summarize this discussion, Fig. 14 presents the
evolution of the aerodynamic coefficients and the time
Fig. 12 Geometry of the upper airfoil, calculated cen- invested in wind tunnel testing versus the elapsed vehicle
terline pressure distribution, and schematics of the target development time. The first numerical model was con-
pressure distribution boundaries. structed at month 3 of this diagram, by which time, 60
hours of wind tunnel testing had been concluded. B e
if the pressure curve extends outside the boundaries of tween the third and tenth months, the combined experi-
the target-line near the trailing edge, as shown in this mental/numerical effort considerably increased the down-
figure, the flow stays attached on the wing (and this was force, as indicated on the figure. This can be contributed
to the rapid wing development and to the optimization of
demonstrated by flow visualizations). wing/body interaction. After the tenth month, the down-
The pressure distributions of Fig. 12 and Fig. 13, force increased only by a small increment but the drag
represent the conditions along the wing longitudinal cen- was gradually reduced by streamlining the rear trailing
terline. However, this pressure distribution may vary con- edge of the vehicle. The actual car first raced in the
siderably along the span due to the distorted flow caused 18-th month of this diagram (April 1992) and in its la&
by the wheels, underbody tunnels, and other body com- est version it had a downforce/drag ratio of 5.399 (and
ponents. Therefore, a single, ideal airfoil section usually downforce coefficient of 3.8). The relatively low usage
does not exist. Consequently, for maximum performance, of wind-tunnel time for this project and the rapid and
quite complex wing shapes (with twist, taper, etc.) must significant improvements in the aerodynamic coefficients
be developed and a t this stage computations can be very clearly demonstrates the positive effect of using the com-
helpful. bined wind-tunnel/numerical experimentation approach
during the vehicle aerodynamic development program.

Target pressure distribution

Actual pressure distribution

/ 160

F / 120
L
'
- /
0
Wind tunnel time
Acc. wind
tunnel time.
80 hours

---_-_- 40

I I I I I I Elapsed time, months

0.00 0 25 0.50 0.75 100 1.25
-
C

Fig. 13 Geometry of the lower airfoil, calculated cen- Fig. 1 4 Evolution of the aerodynamic coefficients
terline pressure distribution, and schematics of the target and invested wind-tunnel time (in hours) versus project-
pressure distribution on the main element. elapsed time (in months).
 Downloaded from SAE International by University of Wisconsin - Madison , Monday, September 10, 2018

REFERENCES

1. Ashby, L. D., Dudley, M. D., Iguchi, S. K., Browne,

L., and Katz, J., Potential Flow Theory and Operation
Guide for the Panel Code PMARC," NASA T M 102851,
March 1990.
2. Katz J., and Plotkin A., "Low-Speed Aerodynam-
ics: From Wing Theory t o Panel Methods," McGraw-Hill
Book Co., 1991.
3. Katz J. "Integration of Computational Methods
into Automotive Wind Tunnel Testing", SAE Paper 890601,
Detroit, Feb. 1989.
4. Katz J. and Largman R. "Experimental Study
of the Aerodynamic Interaction Between an Enclosed-
Wheel Racing-Car and its Rear Wingn, ASME J. of Flu-
ids Engineering, Vol. 111, No. 2, 1989, pp.154-159.
5. Wright, P. G., "The Influence of Aerodynamics on
the Design of Formula One Racing Cars," from "Impact
of Aerodynamics on Vehicle Design," Int. J. of Vehicle
Design, Technological Advances in Vehicle Design Series,
SP3, 1983, pp. 158-172.
6. Mercker, E., Breuer, N., Berneburg, H., and Em-
melmann, H. J., "On the Aerodynamic Interference Due
t o the Rolling Wheels of Passenger Cars," SAE Paper
910311, Detroit, Feb. 1991.
7. Katz J., and Dykstra L. "Effect of Wing/Body
Interaction on the Aerodynamics of Two Generic Racing
Cars," SAE 920349, Detroit, Feb. 1992.
8. Robinson, A., and Laurmann, J. A,, "Wing The-
ory," Cambridge University Press, 1956.
9. Liebeck, R. H . , "Design of Subsonic Airfoils for
High Lift," J . Aircraft, Val. 15, No. 9, 1978, pp. 547-
561.
10. Eppler, R, and Sommers D. M. "A Computer
Program for the Design and Analysis of Low-Speed Air-
foils," NASA T M 80210, Aug. 1980.
11. Drela, M., and Giles M. B.,"ISES: A Two Di-
mensional Viscous Aerodynamic Design Analysis Code,"
AIAA Paper No. 87-0424, Jan. 1987.