Professional Documents
Culture Documents
net/publication/315542610
CITATIONS READS
0 1,003
3 authors:
Adriano Schommer
Federal University of Santa Catarina
5 PUBLICATIONS 2 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Mario Eduardo Santos Martins on 26 April 2017.
Paulo Soliman
Adriano Schommer
Mario Martins
Federal University of Santa Maria
Abstract
km/h), gives an improvement in performance equating to a gain of
This paper presents the design process of an aerodynamic kit for a approximately 40 points during the dynamic events of a Formula SAE
Formula SAE competition vehicle using CFD with special attention to competition, which means a jump of three places in a typical
the distribution of aerodynamic loads. The methodology for the competition The gains in total times come mainly from the increased
development of concepts is to create a boundary that respects the cornering speed and from reduced braking distance. The increase in
geometric constraints of the vehicle and also complies with FSAE drag forces was not a limitation for the wing project, since the straight
2015/16 Rules. Inside these boundaries different geometries of are short and the maximum speeds are low. The question, if the wings
aerodynamic accessories can be analyzed and several full vehicle work or not, was answered with the observation of positive effects on
models can be created. The initial model is conceived based on the the overall vehicle performance. In 2012 at Formula Student Germany,
literature and then analyzed in CFD to generate another model. The 17 vehicles had an aerodynamic device, either wings or shaped
process repeats until it reaches a model that cannot be considered underfloors. Ten of these vehicles finished in the top 20 and the top
optimum but is close enough, achieving the targets previously defined. five vehicles all employed such devices [4,5,6].
On the numerical simulation, are first presented the equations that
govern it and the reason for its use. The simplifications in the CAD When the dilemma seemed to be over, the FSAE Rules was changed
model are explained and the domain is then discretized in and aerodynamic constraints considerably increased. This work may
predominantly hexahedral elements with orthogonal prismatic cells provide an answer to the reader who wonders if the wings and other
next to wall surfaces. Another difference is that the airfoil profiles used aerodynamic accessories remain being beneficial for a FSAE vehicle
are specifically designed for use on racing vehicles. The most relevant after the rules change for the 2015/16 season. It also describes a design
parameters in multi element profiles (slots, AoA) have been optimized process that uses relatively inexpensive tools with great potential for
and wind-tunnel validated in studies of the literature used. After the repeatability, and can be of great value to other FSAE teams or similar
CFD analysis is completed and the final concept defined, the gains due categories. The design process was used in Formula UFSM FSAE
to the aerodynamic kit are quantified based on lap time simulations. team from the Federal University of Santa Maria, Brazil.
Finally the performance on the track with and without the kit is
evaluated and gg diagrams are made with the obtained data. The paper begins with an overview of CFD, creating a basis for
Additionally flow visualization techniques are used. explaining what the commercial code solves. Then the initial choices
for airfoil profiles and undertray geometries are explained. All
Introduction simulated models are presented and the changes that were made until
achieve the final model are shown. Finally earnings due to the kit are
The first known attempt to run an aerofoil on a racing vehicle have predicted by lap time simulation and the prototype is analyzed on track.
been made in 1956. Michael May mounted an aerofoil above the
cockpit of his Porsche, but he was stopped in the scrutineerings of the Numerical analysis
races that he intended to participate and it is believed that he never ran
with the vehicle. In 1966 Jim Hall ran in New York with his Chaparral Computational Fluid Dynamics or CFD is the analysis of systems
2E with a wing attached, and is credited the first to actually race a involving fluid flow, heat transfer and associated phenomena such as
vehicle with aerofoils fitted [1]. Since then the aerodynamics expanded chemical reactions by means of computer-based simulation. Its use
to practically all major categories of motorsport and areconsidered, in enables a substantial reduction of lead times and costs of new designs
some categories, some of the most determining factors determining [7].
which vehicle will be the fastest. Aerodynamic downforce increases
the tire normal force and, consequently, the maximum friction without
increasing the vehicle’s weight [2].
The math behind the code
With a downforce of 600 N at 60 km/h a lap time could be shorten by
2 seconds of 60 seconds for a usual FSAE endurance course [3. The The codes that are now on the market are very robust, but their
score is much more beneficially sensitive to downforce than is operation still requires a high level of skill and understanding from the
negatively to drag, mass and CoG and a simple front and rear wing operator to obtain meaningful results in complex situations [7].
configuration, providing relative low downforce (1000 N at 110
Governing equations 𝐷𝐸𝑥 𝜕[𝑢(−𝑝 + 𝜏𝑥𝑥 )] 𝜕(𝑢𝜏𝑦𝑥 )
𝜌 =[ +
𝐷𝑡 𝜕𝑥 𝜕𝑦
The governing equations of a fluid flow are based on the conservation 𝜕(𝑢𝜏𝑧𝑥 ) 𝜕𝑞𝑥
laws of physics. + ] 𝛿𝑥𝛿𝑦𝛿𝑧 − 𝛿𝑥𝛿𝑦𝛿𝑧
𝜕𝑧 𝑥
+ 𝑆𝐸 (5)
Rate of increase of momentum of fluid particle = Sum of forces on Turbulence is a chaotic and random state of motion in which the
fluid particle: velocity and pressure change continuously with time within substantial
regions of flow. This is developed at high Reynolds numbers. The
𝐷𝑢 𝜕(−𝑝 + 𝜏𝑥𝑥 ) 𝜕𝜏𝑦𝑥 𝜕𝜏𝑧𝑥 random nature of a turbulent flow precludes computations based on a
𝜌 = + + + 𝑆𝑀𝑥 complete description of the motion of all fluid particles. For most
𝐷𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 (2)
engineering purposes it is unnecessary to resolve the details of the
p – pressure; τij – viscous stress component acting in the j-direction on turbulent fluctuations and only the effects are usually sought. [7]. A
a surface normal to i-direction; SMx – source of x-momentum per unit turbulence model is a computational procedure that allows the
volume per unit time. quantification of these effects.
This is the equation that defines the x-component of momentum. The Standard k-ε
equations for the y and z-components are analogous to this.
A k-ε model is a two-equation model in which transport equations are
Energy (First Law of Thermodynamics) solved for the turbulent kinetic energy k and its dissipation rate ε. In
other words, a k-ε model focuses on the mechanisms that affect the
Rate of increase of energy of fluid particle = Net rate of heat added to turbulent kinetic energy [7]. This model is widely used by its
fluid particle + Net rate of work done on fluid particle. recognized robustness, ease of convergence and reduced
computational consumption. The equations of the turbulent kinetic
The net rate of work done on a fluid particle in the element by a surface energy and its dissipation for the model are:
force is equal to the product of the force and velocity component in the
direction of the force. Taking as an example the x-component: 𝜕 𝜕
(𝜌𝑘) + (𝜌𝑘𝑢𝑖 )
𝜕𝑡 𝜕𝑥𝑖
𝜕[𝑢(−𝑝 + 𝜏𝑥𝑥 )] 𝜕(𝑢𝜏𝑦𝑥 ) 𝜕(𝑢𝜏𝑧𝑥 ) 𝜕 𝜇𝑡 𝜕𝑘
[ + + ] 𝛿𝑥𝛿𝑦𝛿𝑧 = [(𝜇 + ) ] + 𝐺𝑘 + 𝐺𝑏
𝜕𝑥 𝜕𝑦 𝜕𝑧 (3) 𝜕𝑥𝑗 𝜎𝑘 𝜕𝑥𝑗
− 𝜌𝜀 − 𝑌𝑚 + 𝑆𝑘 (6)
δx, δy and δz – the sides of a small element of fluid.
Again, the equations for the y and z-components are analogous to this.
𝜕 𝜕
The net rate of heat transfer to the fluid particle due to heat flow in the (𝜌𝜀) + (𝜌𝜀𝑢𝑖 )
𝜕𝑡 𝜕𝑥𝑖
x-direction:
𝜕 𝜇𝑡 𝜕𝜀
= [(𝜇 + ) ]
𝜕𝑞𝑥 1 𝜕𝑞𝑥 1 𝜕𝑥𝑗 𝜎𝜀 𝜕𝑥𝑗
[(𝑞𝑥 − 𝛿𝑥) − (𝑞𝑥 − 𝛿𝑥)] 𝛿𝑦𝛿𝑧
𝜕𝑥 2 𝜕𝑥 2 𝜀2
𝜕𝑞𝑥 + 𝐶1𝜀 (𝐺𝑘 + 𝐶3𝜀 + 𝐺𝑏 ) − 𝐶2𝜀 𝜌
=− 𝛿𝑥𝛿𝑦𝛿𝑧 𝑘
𝑥 (4) + 𝑆𝜀 (7)
qx – heat flux vector in x-direction. Gk – generation of turbulence kinetic energy due to the mean velocity
gradients; Gb – the generation of turbulence kinetic energy due to
Then the rate of increase of energy in x-direction of a fluid particle is buoyancy; Ym – contribution of the fluctuating dilatation in
equal to the sum of equations (3) and (4): compressible turbulence to the overall dissipation rate; C1ε, C2ε and C3ε
– constants; σk, σε – turbulent Prandtl numbers for k and ε; Sk and Sε –
pre-defined source terms.
Standard k-ω
𝜕 𝜕 𝜕 𝜕𝑘
(𝜌𝑘) + (𝜌𝑘𝑢𝑖 ) = (𝛤𝑘 ) + 𝐺𝑘 − 𝑌𝑘 + 𝑆𝑘
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑗 𝜕𝑥𝑗 (8)
Figure 1: CAD model example
𝜕 𝜕 𝜕 𝜕𝜔
(𝜌𝜔) + (𝜌𝜔𝑢𝑖 ) = (𝛤𝜔 ) + 𝐺𝜔 − 𝑌𝜔 + 𝑆𝜔
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑗 𝜕𝑥𝑗 (9) The main elements that were kept: the tires, the bodywork, the radiator,
the firewall and the head restraint. Where there is an accumulation of
triangulated tubes a tangent outer surface was modeled for
simplification. The main hoop and the main hoop bracings were kept
Shear-Stress Transport k-ω because they were considered tubes that are very exposed to the flow
resulting in considerable interaction especially with the rear wing. The
The k-ω SST model was developed to effectively blend the robust and driver was dimensioned according to the 95th percentile male template
accurate formulation of the k-ω model in the near-wall region with the [11].
freestream independence of the k-ε model in the far field. The standard
k-ω model and a transformed k-ε model are both multiplied by a Computational domain
blending function and both models are added together. The blending
function is designed to be one in the near-wall region, which activates In experiments, for simulation of an open road condition, the cross
the standard k-ω model, and zero away from the surface, which sectional area of the wind tunnel should generate a blockage ratio of
activates the transformed k-ε model [9]. less than or equal to 0,2%. The blockage ratio is calculated using the
ratio of the projected frontal area to the cross sectional area of the wind
As there is no definitive turbulence model considered superior to all tunnel [8, 10]. It is important not to oversize the domain to save
analyzes, caution is required in the selection. A good approach is to computational processing. Due to computational limitations were
use k-ε model for initial iterations and then solve with k-ω SST. It is made only half vehicle simulations, which saves processing but makes
also common to use k-ε for the optimization stage and the k-ω SST for it impossible to analysis the vehicle behavior during yaw situation.
the analysis of the final model.
The first step of the finite volume method is to divide the domain into
discrete control volumes. The governing equations already presented
are then integrated along each of those control volumes created to yield
a discretized equation at its nodal points. Discretized equations must
be set up at each of the nodal points in order to solve a problem. For
control volumes that are adjacent to the domain boundaries the general
discretized equation is modified to incorporate boundary conditions.
The resulting system of linear algebraic equations is then solved to
Figure 2: Domain size based on vehicle dimensions (for better visualization
obtain the distribution of a general property at nodal points [7].
figure is not in scale)
Were searched sections that best suited the position they would have
in the vehicle. For example, with a profile of less camber the ground
effect is better used throughout ventral extension, which is desirable
for the main profile of the front wing. Because of the low speed FSAE
circuits a large acceleration immediately after the leading edge is
important. A thick and high cambered profile is desirable in this
situation, this was considered to choose the main profile of the rear
Figure 4: Prism layer generated at the rear wing's flaps
wing. The effect of flaps is similar to having a wing with only one
element but with an extreme camber. But in this wing with only one
Boundary Conditions and Solver element would be severe flow detachment which would spoil their
efficiency. The increase in kinetic energy due to the ducts (called slots)
The boundary conditions and the solver setup main characteristics are formed between a profile and another maintains the flow attached. Due
presented in Tables 1 and 2. to this function of flaps special sections were also chosen for them. In
Benzing’s [13] book are presented geometric parameters (slots,
Wings relative angles of attack and sizes) investigations of multi elements
profiles and correlated with the efficiency of the wing.
Wings are devices that generate aerodynamic downforce due to a
pressure gradient which is derived from a speed difference between
surfaces. The price paid by the gain in downforce is the increase in
drag. Studies show that in a FSAE vehicle with a full aerodynamic
End plates influence of other vehicle parts in the wings, the whole optimization
was done with half vehicle simulations.
When we have high and low pressure fields the air tends to migrate
from the high to the low pressure field. This causes vortexes which Table 3: Vehicle concepts developed and the changes (highlighted in red)
between them
result in increased drag. It also reduces the pressure gradient which
decreases generated downforce. End plates attenuate this effect since Model Main Features
reduces this migration.
Same FW/RW
Foot plates are mounted perpendicular to the end plates and prevent configuration: main
the air from spilling under the end plate, they are used mainly in the element plus 1 flap;
front wing. As the wingspan is restricted by the width of the vehicle The flap of FW is
the footplates can be exchanged with a full wingspan. With foot plates located only in the
of a size equivalent to 1,8% of the wing chord was observed a gain of inner area of the tire
3,6% in downforce compared to using the full wingspan [1]. (rules restriction); No
undertray; Additional
single element on FW;
Six end plates.
Undertray
The idea of the undertray is based on the Venturi effect. The flow is
The profiles going
accelerated to the throat where it reaches full speed. This increase in
through the whole
speed causes a drop in pressure resulting in downforce. After the throat
front consists now of 1
flow must be slowed down to free stream velocity.
main element plus 1
The undertray also reduces drag by decreasing the wake after the flap; FW main
vehicle, the smaller is the wake less energy is expended to move the element has less
vehicle against the air [14]. As the downforce is directly proportional camber and thickness
to the area, the search is for maximum throat area without for best use of ground
compromising the nozzle and diffuser. The main parameters that must effect; Additional
be dimensioned are: the angle of the nozzle, throat ground clearance elements has now 1
and the angle of the diffuser. The downforce is inversely proportional flap and the profiles
to ground clearance if we analyze the Bernoulli’s principle. But as are different from the
Bernoulli does not consider the viscosity of the fluid that is true up to previous one.
a critical point, when it begins to occur the flow detachment. In the
No additional
literature these optimal points were sought through CFD analysis [1,
elements; FW has a 2nd
15]. As the undertray had to adapt to a chassis that was already
flap in the inner area
designed, these parameters had to be reassessed, but the values found
of the tire and just 4
in previous studies served as initial guess.
end plates; RW has 3
flaps; st undertray is a
CFD results profile ventral surface
working on ground
The previous prototype complied with the 2014 rules and because of
effect.
that had an aerodynamic area many times greater than the maximum
allowed for 2015. For this reason the values achieved previously not
served as a parameter for the new kit. Based on the literature and some
benchmarking the goal set was 500 N of total downforce at 52 km/h
with approximately 50% of the load on each axle, if there is an
imbalance should be in favor of front. A loss of 30-40% of downforce
was observed comparing the simulation of a wing in free stream
against the same wing mounted in the vehicle [6]. Because of this huge
The wings are the Figure 6 shows the visual difference between speeds below the front
same from model 4; wing of the models 2, 3 and 4. The ground clearance is the same and
Undertray is now it´s based on the chord of the wing [1]. This demonstrates that a profile
composed of a nozzle, with less thickness and camber takes best advantage of the ground
a flat throat and a effect. The large increase in the wing efficiency (less drag and more
diffuser. downforce) is also due to the flap extending from side to side of the
vehicle.
Table 4: Results Figure 7 shows the difference of the flow between the undertray
models. The model 5 has the first undertray that generates significant
Model 1 2 3 4 5 6 7 8 downforce. The model 6 has a complex undertray geometry, the nozzle
FW -Lift (N) 172 184 220 254 232 194 246 218 extends up near the front wing. It proved its ineffectiveness for 2
FW Drag (N) 76 81 72 66 70 62 64 64 reasons: it generated very little downforce and had great negative
Undertray -Lift (N) 0 0 0 8 62 30 64 84 influence on the front wing. The nozzle near the wing causes a sort of
RW -Lift (N) 200 216 202 230 198 186 198 226 suction effect causing the flow separation and thus spoiling the wing
RW Drag (N) 94 92 86 82 78 66 72 84 efficiency. Models 7 and 8 are the same but with different throat
Total -Lift (N) 372 400 422 492 492 410 508 528 ground clearance. The ground clearances are: 20 mm for model 7 and
Car CL 1,52 1,63 1,66 2,08 2,36 1,68 2,3 2,39 30 mm to 8. With 20 mm loss of downforce due to the detachment of
Car CD 1,21 1,22 1,15 1,27 1,09 1,07 1,14 1,2 the flow caused by the viscosity is observed. But 20 mm is an
% Front 51,5% 51,6% 52,6% 53,2% 53,0% 52,4% 53,4% 52,3% impractical ground clearance due to chassis rollover. In conclusion, to
get the maximum downforce at the analyzed speed the lowest feasible
ground clearance can be used. Another important observation is that
the more the undertray accelerates the flow, generating more
downforce, more flow tends to separate in the front wing, spoiling their
efficiency. This becomes clear if we look at the differences in models
4, 7 and 8 in figures 6, 7 and 8. All models have the same front wing
configuration. The numbers in table 4 also demonstrate that when
undertray is becoming more efficient the front wing ends up becoming
less.
Figure 8 shows the symmetry planes velocities. The idea behind the
additional elements is to use a larger area to generate greater
Figure 7: Models 5, 6, 7 and 8 ground plane downforce. However, in model 2 it may be observed that the additional
element causes flow acceleration in the low speed field of the main higher than that calculated for the engine cooling system have the heat
element and hence decreases the pressure gradient of the main element, exchange necessary for the maintenance on its temperature.
reducing the downforce. On balance it is noted that the larger area used
does not compensate the loss of downforce on the main element. The For comparison purposes it was made a simulation with all the rear
same was observed for the additional elements in the front wing. The wing profiles replaced by Selig 1223. This aeronautical profile is
additional element of model 3 has a flap and is smaller to try not to widely used in FSAE for its good coefficient of lift at low Reynolds
interfere too much in the main element. As expected it really less number.
accelerated the low speed field, but guided the rear flap to have great
flow detachment. On models 8, 7 and 4 it can be seen that the
effectiveness of the wing is quite sensitive to the position of the main’s
element leading edge. As the wing is already at the maximum allowed
height and length, it is preferable to decrease the size of the wing and
place it over backward and upward. Thus its leading edge is exposed
to a cleaner flow from the space between the main hoop and the driver's
helmet. The result was that the model 8 generates almost the same
downforce of model 4 with a smaller area and one less flap. In Figure
8 is visible that the rear wing in model 8 accelerates more the flow over
a larger area.
After analyzing the results the team has chosen design 8 for the 2015
prototype. With the design defined the final simulation was made. The
model includes the radiator to check if the mass flow is sufficient for
engine cooling. To accommodate some new chassis members the
bodywork is slightly different from that used in the optimization. For
this model was used the k-ω SST turbulence model, final results are
shown in table 5.
Element
FW Lift FW Drag Undertray RW Lift RW Drag
base size Mesh type Car CL Car CD
(N) (N) Lift (N) (N) (N)
(mm)
50 Trimmer 205 58 54 210 77 2,17 1,05
30 Trimmer 201 60 52 211 80 2,15 1,07
25 Trimmer 198 63 51 210 83 2,15 1,08
25 Polyhedral 196 62 44 222 80 2,16 1,06
The software calculates just one lap time and extrapolate this for 19
laps would not be appropriate. For a more realistic scenario the actual
times of the 19 laps were arranged in ascending order, then was
calculated the average difference between the times. This average was
added consecutively to the simulated time until it reach 19 laps. The
sum of 19 laps is the total endurance time. Assumptions:
Figure 12: g-g diagrams
1. The static score does not change;
The analysis of the diagram shows that the vehicle had higher lateral
2. If a simulated time with aero is larger than non-aero but smaller
accelerations as well as longitudinal decelerations while showing
than the real one, the actual score will be applied. This process
smaller longitudinal accelerations. In this test the aerodynamic kit
aims to eliminate a false gain due to a real performance under the
resulted in a 12,8% increase in maximum lateral acceleration, a 7,4%
vehicle's ability.
increase in the maximum longitudinal deceleration and decreased
8,3% at maximum longitudinal acceleration. A 4% decrease in the
average lap time was observed with the use of the kit.
Yaw rate
Using a gyro sensor also logging in the vehicle ECU differences in yaw
rate with and without the kit were analyzed. Yaw rate is important to
quantify how agile the car is. Surely the aerodynamic kit increases the
yaw moment of inertia which reduces the yaw. What was evaluated is
if the extra grip compensates this increase in moment of inertia and
enhances vehicle’s agility.
Flow visualization
The wool tuft technique was used to analyze two phenomena observed
in CFD of great interest:
1. If the layer remains attached until the last flap of the rear wing;
2. If the layer is detached in the last flap of the front wing;
Figure 14:FW CFD velocity vectors vs. FW on track with wool tufts
It was observed that the layer remains attached to the last flap of the
rear wing. In CFD occurs recirculation from the first flap of the front
wing, indicating the detachment. On the track was observed same
phenomenon, the wool tufts had chaotic movements. The MoTeC i2
Pro software was used and the captured videos were synchronized with
the logged data to ensure that the analysis was performed at the same
speed of the simulation.
Summary/Conclusions
The key learnings developed with this work: