Sie sind auf Seite 1von 13

EVALUATION OF AERODYNAMIC DRAG OF GO KART BY MEANS OF

COAST DOWN TEST AND CFD ANALYSIS

Marco Evangelos Biancolini

Department of Mechanical Engineering Tor Vergata University, Rome Italy


e-mail: biancolini@ing.uniroma2.it

Keywords: CFD, Drag, Coast down, acceleration.

ABSTRACT
In this paper the problem of evaluation of aerodynamic drag forces in a competition go kart is
addressed. The study is collocated in a Land Speed Record Project and has the aim to evaluate the
maximum speed capability of the vehicle in its original configuration.
In order to investigate how vehicle components affect the overall drag a detailed numerical test has
been conducted by means of a CFD calculation software. To have a quantitative idea of actual track
performance several coast down tests have been conducted. The vehicle has been equipped with a
data logger for speed acquisition. An automatic clutch has been inserted in the transmission
allowing to disengage the engine when the coil ignition is turned off. Adopting a simple vehicle
acceleration model, several algorithms for data analysis has been tested to estimate actual drag
coefficients. Acceleration data are also examined to evaluate the actual driving capability of the
engine and to compare it with the engine torque measured in static conditions.

INTRODUCTION

Competition go kart optimization requires a deep knowledge of physical phenomena involved. In


the last decade several efforts were dedicated to establish advanced design and setup procedures
useful for the support of manufacturers and teams. The peculiarities of go karts (suspensions and
differential are missing) and the traditional design and optimization method that is based mainly on
empiricism, makes this task very difficult. In fact consolidated design procedures for racing cars are
difficult to apply to this kind of vehicle and requires to be extended. Furthermore is very difficult to
find useful data in the open literature and each parameter desired for the implementation of
advanced analysis tools has to be measured or estimated by simulation.
Top kart manufacturers are in Italy and so is quite obvious that the main contributions found in the
reviewed literature are given by Italian researchers. In (Vitale, 2001) a lumped parameters model
that represents the vehicle with three bodies connected by two torsional springs is presented, using
actual track data recorded to obtain the unknown parameters. A Multi-body modeling performed
with ADAMS has been presented in (Mirone, 2003) where the chassis is handled as a flexible
model. The author of the present paper started the studies about go kart dynamics in 2000 founding
the research group Tor Vergata Karting at the University of Rome Tor Vergata. The lumped
approach proposed in (Vitale, 2001) was also considered in (Ponzo, 2004) where the results of
simulations have been compared with on board logged data. The TVK research team has
subsequently developed a software tool in Fortran code to simulate the kart behavior in road
maneuvers, taking into account the effect of different frame geometry configurations (Baudille,
2002). Lap time simulation was investigated in (Baudille, 2004) where an integrated approach

1
based on a single mass vehicle model was used to predict a complete lap for several input
parameters such as path line, engine setup and grip. A general procedure for evaluating the vertical
load transfer was presented in (Baudille, 2006) in which a full detailed FEM model of the vehicle
with the actual mass distribution was used to evaluate vertical loads under elementary load
conditions. Vertical loads with generic conditions are then evaluated by superposition of collected
results. In (Biancolini, 2007) the optimization of a composite chassis is presented adopting a
detailed numerical model that allows to perform constrained tests, suitable for the extraction of
Moment Method diagrams (Milliken, 1995). Using load transfer data calculated by means of
structural FEM models is possible to optimize the track performances acting on the stiffness design
data.
A CFD study about go kart aerodynamics was presented by (Baudille, 2003) where a lattice
Boltzmann solver was adopted to investigate the drag generation and its effect on vehicle
performance (the effect on performance was further investigate in the aforementioned paper
(Baudille, 2004)). Several configurations were considered considering the driver position and the
interaction between two vehicles.
A preliminary design of a composite aerodynamic cover was presented by (Manieri, 2006) related
to the Project Land Speed Record promoted by TVK and the driver GianMaria Gabbiani..
The results presented here are related to a preliminary study in the project land speed record that
had the aim to quantify passive losses of the baseline vehicle in order to define an optimization
strategy for the record vehicle design.
The focus is the quantitative evaluation of passive losses of go kart considering simulation and
experiments. The numerical simulation has been conducted by means of a very detailed 3D model
implemented for the CFD solver Fluent with the aim to estimate the drag force and its distribution
on vehicle parts. The experimental estimation has been conducted by means of coast down test
(Hitoshi, 2005) equipping the vehicle with a clutch capable to disconnect the power transmission at
the desired speed and a data logger for the acquisition of speed.

Figure 1: Investigated vehicle running against its CFD model.

CFD MODELLING

Model description
Numerical modeling has been con ducted by means of the CFD solver Fluent. The reference
geometry was first presented by the TVK research team (Bertolino, 2003). Basic concepts for CFD
modeling of a go kart was described in the study published by (Baudille, 2003).
The detailed geometric model of the vehicle complete of the driver has been adopted for the
developing of a very fine surface mesh suitable to capture local flow field gradients. Fluid domain
includes the go kart, the ground and the external boundary of the virtual wind tunnel.

2
Boundary conditions of speed at the borders of the wind tunnel are imposed to assign the relative
motion of the vehicle. Furthermore tangential speed has been imposed at the wheels surfaces to
account for the spinning.
CFD analysis is iterated until a stable solution for flow fields and aerodynamic loads is met. At the
end of calculation all the fluid dynamic variables inside the domain and at the boundary, in
particular at the vehicle wetted surfaces, are available. A typical result is given in figure 2 where the
pressure on the surface is represented by a color map and the velocity field is represented by means
of stream lines. Animations obtained form this simulation are also available on the internet site
TVK (Biancolini, 2007).

Figure 2:CFD analysis results at a speed of 90 km/h.

CFD analysis results


Pressure distribution represented in figure 2 is related to a go kart running at a constant speed of 90
km/h. Positive pressure peaks are observed on the front wheels, on the helmet and on the body of
the driver and then on the front deflector and on the front spoiler. Integrating the pressure map
acting on each surface, taking also in account the shear loads, is possible to evaluate the resultant
load acting on each component of the vehicle. The plot of figure 3 represents the distributions
obtained subdividing the overall vehicle in the following parts: front spoiler, front deflector, driver,
engine, frame, radiator, tank, front tires and rear tires. Also total loads are given in the last columns.

3
Figure 3: drag force and down force distribution on the go kart components.

It’s interesting to notice that the overall downforce is negative (-30N) that means that at this speed
the vehicle is about 3 kg lighter. Considering the resultant drag, that is 173 N, is possible to
calculate the drag coefficient according to the equation (1).
Fdrag v   v 2 C x A front
1
2
A front  0.57484 m 2
kg
  1 .2 (1)
m3
km
v  90
hr
C x  0.804
Predicted value is quite high if compared with the literature value reported for a vehicle with
uncovered wheels and without wings (Milliken, 1995) that is 0.6. However the go kart is not
optimized for drag and the driver is almost completely exposed to the air flux.
In order to better understand how the drag losses are spent by the vehicle components, the relative
contribution of each parts, expressed in percent, is represented in the graph of figure 4.
About 33% of the power is dissipated by the driver. This is quite easy to understand because the
driver contributes for the main part of cross section and it’s well know that an hunched position can
be adopted to gain enough speed to undertake at the end of a fast straight.
The wheels contribution is about 20% and has to be accepted for this kind of vehicle with not
covered wheels, reducing the cross section of the wheels may lead an aerodynamic improvement
that will not be so benefic to compensate the loss in cornering capability of the go kart.
The 15% absorbed by the front spoiler at a first glance suggests that even a deep optimization of
such component can give a little improvement in performance. However a careful study of front
components has not only the aim of minimizing their drag loss, but also of improve the disturbed
flux pattern downstream the spoiler, in order to optimize the losses of all the components on the

4
vehicle. This is a common rule of aerodynamic optimization of the front part of a vehicle because
this part encounters the undisturbed flow ahead the vehicle (with the obvious exception of the
vehicle in the wake of the vehicle ahead) and produces the flow conditions encountered by the
components behind.
The last remark is about the radiator. In this case aerodynamic optimization has not only the aim to
minimize the loss but also to guarantee enough air flow to properly cool the liquid inside. Being
about 10% of the power loss spent at the radiator it seems better to change the angle of the radiator
when it’s oversized for the track condition instead to cover a part, as is usually done for go kart. In
this way only the flow requested for cooling is encountered by the surface.

Figure 4: Contributions of components to aerodynamic loss.

EXPERIMENTS

The study exposed in the previous section is very useful to understand the aerodynamic loss
generation mechanism of a go kart and can be used to drive an optimization because it allows to
select the relative importance of the components. However the numerical value obtained for the
penetration coefficient is quite higher than the value expected for this kind of vehicles.
In order to clarify the quantitative value of aerodynamic losses of a go kart an experimental test is
mandatory. Two alternatives have been considered: wind tunnel testing adopting a scaled model (or
full scale testing if a wind tunnel big enough is available) or track testing on the actual go kart.
The second method has been selected because it is low expensive and can be conducted on an actual
go kart equipped with a speed logger. However a quite complex data analysis is required to extract
desired parameters.
Experiments have been conducted on the ISAM track of Anagni with a go-kart CRG 100, equipped
with a two stroke engine Maxter 100 cc, conducted by the driver Gianmaria Gabbiani with the
support of the team CIK and of the Doctor Chris Sewell of PI Research. The go-kart ready to go is
represented in Figure 5.

5
Figure5: go-kart ready for testing on the ISAM track of Anagni.

Speed data have been logged by means of the acquisition system Delta Clubman Kart Kit by Pi
Research. The installation of the system on the vehicle is illustrated in figure 6.

Figure 6: Acquisition system details: acquisition board PI “Delta Clubman”, steering wheel display PI “X-
Kart”, speed sensor mounted on the left front wheel , engine speed sensor mounted on the coil cable,
temperature sensor mounted on the cooling liquid pipe.

The vehicle belongs to 100cc category with fixed gear ratio. In order to allows the free deceleration
test a centrifugal clutch has been adapted between the engine power shaft and the pinion of the

6
transmission chain (see figure 7). The transmission is interrupted when the driver stops the rear axle
with a sudden brake action.

Figure 7:centrifugal clutch inserted in the transmission line.

On track losses measurement theory


The basic method for on track evaluation of losses is straightforward: the vehicle is accelerated until
the desired speed is reached and then the engine is turned off and disconnected to leave the vehicle
free to decelerate braked only by losses.
Vehicle motion is governed by the system of equations (2) that shows that total longitudinal force
applied to the system is given by the engine thrust, the tires roll resistance and the aerodynamic
drag. The engine thrust is calculated considering the available engine torque at the RPM imposed by
vehicle speed transformed by the transmission ratio and the rolling radius of the wheel. The rolling
resistance is proportional to vehicle weight and is supposed to be linearly dependent to the vehicle
speed. The drag is supposed to be proportional to flow field static pressure (that has a quadratic
dependence from speed and is proportional to air density), to the cross section and to a constant
penetration factor (the last approximation is true only in a range of speed that produces a fully
turbulent flow).

F.tot v   Ma
F.tot v   Fthrust (v)  Fdrag (v)  Froll (v)
 v  1
Fthrust (v)  T   (2)
 2Rrot Z tot  Rrot Z tot
Froll v   Mg  f 0  f1v 

Fdrag v  
1 2
v C x A front
2

7
When the desired speed is reached the driver performs a sudden braking action on the rear axle that
enable the clutch to disengage the transmission. The vehicle is then left free to decelerate until a
minimum speed is reached (or until a complete stop). Analyzing speed data logged during this
deceleration stage and performing a numerical differentiation of the acquired signal the deceleration
curve is obtained. Multiplying deceleration and vehicle mass the decelerating force is obtained that
consists in the sum of two contributions: rolling loss and drag loss. Interpolating the obtained curve
with a second order polynomial using a regression technique is possible to evaluate the unknown
coefficients that are directly related to loss model assumed.
Once the regression is completed the loss model that best fit deceleration data is obtained.
Equations 3 shows the polynomial adopted for regression and loss model coefficients matching.


 Ma  Mg  f 0  f 1v   v C x A front
1 2
 2
 Ma  av 2  bv  c

1
 2 C x A front  a
(3)

Mgf1  b
Mgf  c
 0


The estimation of engine torque curve is now possible because the estimated loss model is
determined by deceleration test analysis. If the speed range during acceleration and deceleration test
is the same, the same loss coefficients can be assumed to be valid. Performing a similar data
analysis during acceleration test, the speed is differentiated to obtain acceleration, multiplied by the
mass to have the accelerating longitudinal force. Subtracting the losses using the loss model
coefficients is then possible to evaluate the thrust that can be transformed to obtain the torque curve
of the engine (see equation (3)).

Analyzed case and results


A detailed application of the exposed method is presented in this section. Logged speed data during
the test of free deceleration are represented in figure 8 where the portion of curved interesting for
data analysis is highlighted by a box. In figure 9 the data of the acceleration test are represented and
the portion of pure longitudinal acceleration along the straight that brings the go kart at its
maximum is highlighted by a box.
At this point the first part of data treatment, that is extraction of the portion of data to analyze, is
accomplished.

8
Figure 8: Speed data logged during the free deceleration test.

Figure 9: Speed data logged during the acceleration test.

Extracted data need to be filtered before analysis. After some trials a median smoothing with 5
points window has been chosen for the deceleration data. In figure 10 the comparison between
polynomial interpolation and force data obtained after smoothing and differentiating deceleration
data is presented. As can be noticed, being the sampling time fixed, force data are denser at low
speed values and more dispersed at high speed values.

9
0
Poly nomial
50
Experimental

Force (N)
100

150

200

250
0 20 40 60 80 100

Speed (kph)
Figure 10:Comparison between longitudinal force and polynomial correlation.

The result of figure 10 is obtained by means of a two steps procedure. The first step consists in
simple polynomial regression of experimental data based on least squares algorithm. However such
method handles acceleration data that suffers the numerical differentiation. For this reason obtained
results are used as guess values for a more robust algorithm that is based on the identification of the
differential system that govern vehicle motion. In this case speed data are compared with results
obtained integrating the equation of motion. A constrained optimization is then performed
minimizing the error between numerical prediction and experimental data. A reduction of the error
of a factor two is observed comparing the results before and after the minimization algorithm. The
results of identification allows to estimate the loss coefficients reported in equation 4.

C x  0.899

 f1  5.114  10 (4)
-4

 f  0.027
 0

Speed data of acceleration test have been processed introducing first a stronger smoothing with a
median filter that use a 15 points window, and then the numerical differentiation. The longitudinal
force of figure 11 is obtained. This curve is comprehensive of all the longitudinal loads including
both engine thrust, that is variable with speed according to the engine torque curve, and losses.

10
600

500
Longitudinal Force (N)

400

300

200

100

70 80 90 100 110 120 130

Speed (kph)
Figure 11:Longitudinal force obtained during acceleration test..

Subtracting the losses estimated according the laws of equation 2 and adopting the coefficients of
equation 4, the thrust produced by the engine can be estimated. Figure 12 shows a comparison
between a torque index obtained by acceleration data analysis and the same torque index provided
by the engine manufacturer.

2
Estimated from acceleration
Nominal

1.5
Torque index

0.5
70 80 90 100 110 120 130

Speed (kph)
Figure 12: Torque curve obtained by the presented method compared with manufacturer torque data.

11
SENSITIVITY ANALYSIS
Max speed is determined imposing the equivalence between engine maximum power and power
dissipated by losses:

 
Pmax   Mg  f 0  f1v   v 2 C x A front v
1
 2 
Transmission ratio for maximum speed can then be evaluated accordingly:

RPM P max 2Rrot


Z V max 
vmax

Parameter variation -40% -20% 20% 40%


Cx 18.9 7.8 -5.9 -10.7
f0 0.6 -0.6 -1.27
f1 -0.92 -0.46 0.46 0.93
Power -16.3 -7.4 6.5 12.3
Peso 0.3 0.17 -0.17 -0.35
Cross section 18.9 7.8 -5.9 -17

CONCLUSIONS

In this paper the problem of evaluate aerodynamic drag performance of a go kart has been
addressed. CFD analysis has been first adopted to understand how drag losses are distributed on the
components of the vehicle. Such results are very useful to drive drag optimization because drag
generation regions can be highlighted considering how they contribute to the overall result.
Furthermore CFD analysis allows to estimate the drag coefficient that is found to be about 0.8.
In order to have also an experimental estimation of the same parameter a on track test campaign has
been planned and executed. First, speed data acquired during coast down test have been analyzed to
identify unknown loss coefficients, then speed data acquired during acceleration test are analyzed
to estimate the engine thrust. A procedure for data manipulation including smoothing, numerical
differentiation and differential system identification techniques, has been adopted to careful
estimate loss coefficients and torque performance of the engine.
The drag coefficient is estimated to be about 0.9. A worst performance if compared with the CFD
result. However CFD analysis has been conducted on a different kart with a driver 1.70 m tall while
the measurements have been conducted with a driver about 1.85 m tall.
The rolling loss coefficient obtained is in line with what expected for the wheels mounted. The
torque data estimated is very close to the nominal one (15% of maximum error) but not in perfect
agreement. Torque value is overestimated at low and high speeds and underestimated at medium
speeds. This is probably due to the loss model that doesn’t account for drag coefficient variation
with speed.

ACKNOWLEDGMENTS

A special thanks to Dr. Baudille for his suggestions and to a Swedish friend for his help in CFD
analysis. For track measurements the author would like to express his acknowledgments to all the
sponsors for providing materials and support: ISAM for the track, CIK for vehicle setup, PI
Research for the measurement system, CRG and Maxter for the vehicle and the engine. A further
thanks to TVK Land Speed record team and to Gian Maria Gabbiani.

12
REFERENCES

Baudille, R., Biancolini, M. E., Reccia, L. (2002), “Integrated multi-body/FEM analysis of vehicle
dynamic behaviour”, The 29th FISITA World Automotive Congress, Helsinki, Finland, June 2002.
Baudille, R. (2003), “Ottimizzazione delle prestazioni aerodinamiche di kart da competizione”,
http://www.torvergata-karting.it/article/articleview/22/1/7/
Baudille, R., Biancolini, M. E., Reccia, L. (2004), “An Integrated Tool For Competition Go-Kart
Track Analysis”, The 30th FISITA World Congress 2004, Barcelona, Spain, May 2004
2002-01-1121
Baudille, R., Biancolini, M. E., Reccia, L. (2006), “Load transfers evaluation in competition go-
kart”, Int. J. Vehicle Systems Modelling and Testing, in press.
Bertolino, M. (2003), “Realizzazione di modelli tridimensionali”, http://www.torvergata-
karting.it/article/articleview/17/1/11/
Biancolini, M. E., Cerullo, A., Reccia, L. (2007),“Design of a tuned sandwich chassis for
competition go kart”, IJVD 2007 Volume 44 Nos 3/4.
Biancolini, M.E. (2007), “Il CX del go kart”, http://www.torvergata-
karting.it/article/articleview/57/1/7/
Hitoshi, T. (2005), Real-Life Coefficient of Drag - a Simple Extraction Method, AutoTechnology,
No.:2005-04.
Manieri, G., Urbinati, M. (2006), “Design of a land speed record go kart”, 8th FISITA Student
Congress, 22-27 October, Yokohama, Japan.
Milliken, W. F., Milliken, D. L. (1995), “Race car vehicle dynamics”, SAE.
Mirone, G. (2003), “Multibody modelisation of a go-kart with flexible elements frame: simulation
of the dynamic behaviour and experimental validation”, SAE 2003-01-2764.
Ponzo, C., Renzi, F. (2004), “Parametric multi-body analysis of kart dynamics”, The 30th FISITA
World Congress 2004, Barcelona, Spain, May 2004
Vitale, E., Frendo, F., Gheraldi, E. , Leoncini, A. (2001), “A lumped parameters model for the
analysis of kart dynamics”, 7th International Conference ATA, Florence, Italy.

INTERNET SITES

Top Speed Record Team Site, http://www.topspeedkart.com/


Tor Vergata Karting Research group, www.torvergata-karting.it
Gabbiani G.M., http://www.gabbianijr.com/
PI Research, http://www.piresearch.com/
Racing Team CIK, http://www.centroitaliakart.it/
ISAM proving ground and track, http://www.isam-spa.it/
CRG go-kart, http://www.kartcrg.com/
MAXTER engines, http://www.maxterengines.com/

13

Das könnte Ihnen auch gefallen