Comparison of CFD Simulation of a Hyundai I20 Model with Four Different Turbulence Models

M.Vivekanandan. et. al. Int. Journal of Engineering Research and Application www.ijera.com
ISSN : 2248-9622, Vol. 6, Issue 7, ( Part -1) July 2016, pp.47-53
www.ijera.com 47 | P a g e
Comparison of CFD Simulation of a Hyundai I20 Model with
Four Different Turbulence Models
M. Vivekanandan*, R. Sivakumar**, Aashis. S. Roy***
*(Uttam Industrial Engg. Pvt. Ltd., Tiruchirapalli, Tamil Nadu, India,)
**(Mechanical Engg. Dept., Addis Ababa Science and Technology University, Addis Ababa, Ethiopia,)
***(Industrial Chemistry Dept., Addis Ababa Science and Technology University, Addis Ababa, Ethiopia,)
ABSTRACT
This article describes the CFD analysis of a Hyundai i20 car Model. The focus of this study is to investigate the
aerodynamics characteristics of Hyundai i20 car model and the flow obtained by solving the steady-state
governing continuity equations as well as the momentum conservation equations combined with one of four
turbulence models (1.Spalart-Allmaras 2.k-ε Standard 3.Transition k-kl-ω 4.Transition Shear Stress Transport
(SST)) and the solutions obtained using these different models were compared. Except transition k-kl-ω model,
other three models show nearly similar velocity variations plot. Pressure variation plot are almost similar with
K-ε and transition-SST models. Eddy viscosity plot are almost similar with K-ε and transition k-kl-ω models.
I. INTRODUCTION
Designing the shape of a car is very
important in achieving a good performance
aerodynamically. Computational Fluid Dynamics
has played an important role in a car to determine
the flow features and performance such as lift, drag,
vehicle stability and cross wind parameters. The 3D
numerical simulation method is used for various
turbulence models such as Spalarat Allmaras, k- ε
Standard, Transition K-KL-ω, and Transition Shear
Stress Transport (SST) to simulate the aerodynamic
performance of an i20 car model.
Spalart et al [1] reported a model to solve a
transport equation. It is a one equation model
formulated for the kinematic eddy turbulent viscosity.
The model was designed specifically for aerospace
applications where wall-bounded flows is involved
and is shown to give good results for boundary
layers which are subjected to pressure gradients
adversely. This model is an effective low-Reynolds
number model in its original form. It requires the
viscosity-affected region of the boundary layer to be
resolved properly. It is also gaining popularity in
turbo-machinery applications.
Spalart et al developed a Spalart Allmaras
model to solve aerodynamic flows. It is not
calibrated for general industrial flows, and errors
produced for some free shear flows is relatively
large, especially for plane and round jet flows.
Further, prediction of the decay of homogeneous,
isotropic turbulence cannot be relied upon.
K-ε turbulence model proposed by Lauder
and Spalding [2] is the model most commonly used
in Computational Fluid Dynamics (CFD) for
simulating mean flow characteristics for turbulence
flow conditions. It is a two-equation model and
gives a general description of turbulence by means
of two transport equations. The original objectives of
the k-ε model were to improve upon the mixing
length model and to find an alternative to
algebraically prescribing turbulence length scale in
flows with moderate to high complexity.
The first transports variable determines the
energy in the turbulence and is called turbulence
kinetic energy.
The second transports variable is the
turbulence dissipation which determines the rate
with which the turbulence kinetic energy is
dissipated.
The exact k-ε equations contain many
unknown and immeasurable terms for a practical
approach. The standard k- ε turbulence model used
is based on our best understanding of the relevant
process minimizing unknown set of equation which
can be applied to large number of turbulence
applications.
Transition k-Kl-Ω Turbulence Model is
three-equation turbulence model was developed and
tested by Walters and Leylek [3] first using the far-
field dissipation rate and then later implemented
using the inverse turbulence time scale (ω). The
original concept for the addition of the laminar
kinetic energy (kl) equation was to capture the
effects of bypass transition only but now extended to
the natural and mixed transition cases also.
The transition shear stress transport (SST)
model was developed by Menter [4] to effectively
blend the robust and accurate formulation of the
model in the region near the wall region with the
free stream independence of the model in the far
field. To achieve this, the model is formulated into
equations. The SST model is similar to the standard
model, but includes some refinements such as
RESEARCH ARTICLE OPEN ACCESS

adverse pressure gradient flows and transonic shock
waves than the standard model.
Sneh Hetawal et al [5] carried out CFD
simulation of a formula SAE car and suggested
modifications to minimize lift and drag.
Z. Y. Chao et al [6] carried out the
transient numerical simulation studies on
aerodynamics of a simple shaped car when they
cross side by side in opposite direction and
concluded that the aerodynamic forces have a
significant effect on the stability in handling the car.
Mattias Olander [7] carried out CFD
Simulation of a Volvo Car in Slotted Walls Wind
Tunnel and reported that the drag and lift coefficient
were lower in the CFD simulations of the wind
tunnel compared to the ordinary CFD simulations.
The results of standard CFD simulations were
opposite to that of real experiments studied in the
tunnel. This was opposite to the real experiments
when compared with the standard CFD simulation.
The reason was mainly because of simplified
geometry of the simulated tunnel and the uniform
boundary condition at inlet, which resulted in a more
uniform flow in the CFD simulations when
compared with experiments.
Fischer et al [8] reported that the variation
between experimental results and that of simulation
is partly due to wind tunnel effects and ground
simulation effects. A CFD study on Sedan scale
model in an open wind tunnel was reported with and
without wind tunnel effect which resulted in
different pressure distribution and hence difference
forces and surface pressure.
Kieffer et al [9] reported the CFD
simulation on Formula Mazda race car using star CD
CFD code. The simulation of turbulent air flow was
performed on the front and rear wings using k-ε
model. Coefficient of lift and drag had a significant
effect on the ground effect. Stalling condition was
observed for angles less than or equal to 12o
, to the
horizontal.
Singh R [10] carried out CFD simulation
studies on NASCAR racing car and reported that lift
decreased because of rotating wheel and moving
ground effect. In contrast, drag increased because of
smaller interaction of the underbody flow with wake
flow.
In this paper, authors report the CFD
simulation solutions of an i20 car model with four
different turbulent models that would enable to
compare and choose a model that suits the
simulation studies better and converge with a real
time solution.
II. COMPUTATIONAL METHOD
For analysis purpose, CFD solver ANSYS
FLUENT is used for calculation. The free steam
temperature is 300 K, which is same as the
environmental temperature. Figure 1 shows the three
dimensional view of the i20 car model.
Figure 1 3D view of the i20 car model
The car profile, boundary condition and
meshes were all created in the pre-processor. The
pre-process is a program that can be used to produce
two and three dimensional models using structure or
unstructured meshes which can consists of variety of
elements such as quadrilateral, and triangular in
regions where greater computational accuracy is
needed, such as the region closed to the profile.
The first step in performing a CFD
simulation should be to investigate the effect of the
mesh size on the solution result. Generally, a
numerical solution becomes more accurate as more
nodes are used but computation time and memory
requirement increases with increase in nodes. The
appropriate number of models can be determined by
increasing the number of nodes until the mesh is
sufficiently fine so that further refinement does not
change the result.
III. RESULTS AND DISCUSSION
Figures 2, 3, 4 and 5 show the pressure cut
plot with Spalart Allmaras, K-Ε, K-KL-Ω and SST
models respectively for turbulent flow. All four
models show a similar pressure variation at the front
end and at the bottom of wheels of the car.
Figure 2 Pressure cut plot with Spalart Allmaras
Model

Figure 3 Pressure cut plot with K-Ε Model
Figure 4 Pressure cut plot with Transition K-KL-Ω
Model
Figure 5 Pressure cut plot with Transition SST
Figures 6, 7, 8 and 9 show the surface
pressure plot with Spalart Allmaras, K- ε, and K-KL-
Ω and SST models respectively for turbulent flow.
Green colour / blue colour shade represents negative
pressure of decreasing order and hence boundary
layer separation at the surface while yellow / orange
/ red colour shade represents positive pressure of
increasing order. More variation in surface pressure
is observed with transition K-KL-Ω turbulence flow
model.
Figure 6 Surface Pressure plot with Spalart
Allmaras Model
Figure 7 Surface Pressure plot with K- ε Model
Figure 8 Surface Pressure plot with K-KL-Ω Model
Figure 9 Surface Pressure plot with SST Model
Figures 10, 11, 12 and 13 shows the
velocity cut plot with Spalart Allmaras, K- ε, K-KL-
All four models show low velocities at the rear end
of the car. K-KL-Ω model showed much lower
velocity compared to the other three models.

Figure 10 Velocity cut plot with Spalart Allmaras
Model
Figures 14, 15, 16 and 17 show the surface velocity
plot with Spalart Allmaras, K- ε, K-KL-Ω and SST
models respectively for turbulent flow. All models
show maximum velocity at the top and minimum
velocity at the rear end of the car. The surface
velocity variations are almost similar with all four
models
Figure 11 Velocity cut plot with K- ε Model
Figure 12 Velocity cut plot with Transition K-KL-Ω
Model
Figure 13 Velocity cut plot with Transition SST
Model
Figure 14 Surface velocity plot with Spalart
Allmaras Model
Figure 15 Surface velocity plot with K- ε Model
Figure 16 Surface velocity plot with Transition K-
KL-Ω Model
Figure 17 Surface velocity plot with Transition SST
Model
Figures 18, 19, 20 and 21 show the eddy
viscosity cut plot with Spalart Allmaras, K- ε, K-KL-
All four models show variations in eddy viscosity

behind the car. K-KL-Ω model showed more
turbulent eddies and hence more variations in eddy
viscosity than the other three models.
Figures 22, 23, 24 and 25 show the eddy
viscosity plot at the surface with Spalart Allmaras,
K-ε, K-KL-Ω and SST models respectively for
turbulent flow. Spalart Allmaras and SST models
show no variations at the surface. K-ε and K-KL-Ω
models show small variations in eddy viscosity on
the sides of the car.
Figure 18 Eddy viscosity cut plot with Spalart
Allmaras Model
Figure 19 Eddy viscosity cut plot with K- ε Model
Figure 20 Eddy viscosity cut plot with Transition K-
KL-Ω Model
Figure 21 Eddy viscosity cut plot with Transition
SST Model
Figure 22 Eddy viscosity plot at the surface with
Spalart Allmaras Model
Figure 23 Eddy viscosity plot at the surface with K-
ε Model
Transition K-KL-Ω Model

Transition SST Model
Figures 26, 27, 28 and 29 show the velocity
trajectory plot with Spalart Allmaras, K- ε, K-KL-Ω
and SST models respectively for turbulent flow.
The trajectory obtained with Spalart Allmaras and
K-KL-Ω models are similar. Flow trajectory
observed with K-ε and transition SST models are
having similarities with mixed stream lines behind
the car.
Figure 26 Flow trajectory with Spalart Allmaras
Model
Figure 27 Flow trajectory with K- ε Model
Figure 28 Flow trajectory with Transition K-KL-Ω
Model
Figure 29 Flow trajectory with Transition SST
Model
Table 1 Average values of Pressure, Velocity and
Eddy viscosity calculated using four different
turbulence Models:
Turbulence
Model
Pressure
Pa
Velocity
ms-1
Eddy
Viscosity
Pa.s
Spalart
Allmaras -255.529 14.964 0.000234
K-Ε -266.533 14.3556 0.000159
Transition
K-KL-Ω -222.889 8.5976 0.000162
Transition
SST -265.724 13.704 0.0000868
The average velocity calculated from the
turbulence models is shown in Table 1. K-KL-Ω
model showed more variation in average pressure
and average velocity compared to other three
models, while Spalart Allmaras and transition SST
models showed maximum and minimum eddy
viscosity values respectively.
Figures 30, 31, 32 and 33 show the
variations in residuals with increase in number of
iterations and convergence of solution for the four
different turbulence models studied.

Figure 30 Residuals vs. Iteration graph of Spalart
Allmaras Model
Figure 31 Residuals vs. Iteration graph of K-ε
Model
Figure 32 Residuals vs. Iteration graph of Transition
K-KL-Ω Model
Figure 33 Residuals vs. Iteration graph of Transition
SST Model
IV. CONCLUSIONS
An i20 car model was simulated with four
different turbulent flow models - Spalart Allmaras,
K-ε, K-KL-Ω and SST models respectively. More
variation in surface pressure is observed with
transition K-KL-Ω turbulence flow model. K-KL-Ω
model showed much lower velocity compared to the
other three models. The surface velocity variations
are almost similar with all four models K-KL-Ω
model showed more turbulent eddies behind the car.
K-ε and K-KL-Ω models show small variations in
eddy viscosity on the sides of the car. K-ε and
transition SST models are having similarities with
mixed stream lines behind the car. K-KL-Ω model
showed least average pressure and average velocity
while SST model showed least eddy viscosity.
REFERENCES
[1]. Spalart, PR and Allmaras SR, A One-
Equation Turbulence Model for
Aerodynamic Flows. AIAA Paper, 1992,
92-439.
[2]. Launder BE and Spalding DB, The
numerical computation of turbulent flows.
Comp. Meth. Appl. Mechanics Engr., 1974,
Vol. 3, 269-89.
[3]. Walters D.H. and Leylek J.H. (2003b)
Simulation of Transitional Boundary-Layer
Development on a Highly-Loaded Turbine
Cascade with Advanced RANS Modeling,
Proceedings of ASME Turbo Expo, 2003-
38664.
[4]. Menter FR, Two-Equation Eddy-Viscosity
Turbulence Models for Engineering
Applications. AIAA J., 1994, Vol. 32,
1598-1605.
[5]. Sneh Hetawal, Mandar gophane, Ajay B.K.
and Yagnavalkya Mukkamala,
Aerodynamic study of Formula SAE car,
Procedia Engineering, 97, 2014, 1198 –
1207.
[6]. Zhang Ying-chao and Fu Li-min, Transient
aerodynamic numerical simulation of
simplified shape cars under condition of
opposite meet side by side in tunnel,
Journal of Jilin University, 2006, 03.
[7]. Mattias Olander, CFD Simulation of the
Volvo Cars Slotted Walls Wind Tunnel
Master’s Thesis in Solid and Fluid
Mechanics, Chalmers University of
Technology, Goteborg, Swedon, 2011, 33.
[8]. Oliver Fischer, Timo Kuthada, Jochen
Wiedemann, Patrick Dethioux, Richa Mann
and Brad Duncan, CFD validation study for
a Sedan scale Model in an Open Jet Wind
Tunnel, SAE International, 2008, 1-0385.
[9]. W. Kieffers, S. Moujaes and N. Ambya,
CFD study of section characteristics of
Formula Mazda race car wings,
Mathematical and Computer Modelling,
Vol.43 (11-12), 1275-1287.
[10]. Singh R. CFD Simulation of NASCAR
Racing Car Aerodynamics, SAE Technical
paper, 2008, 01-0659.

Comparison of CFD Simulation of a Hyundai I20 Model with Four Different Turbulence Models

More Related Content

What's hot (19)

Similar to Comparison of CFD Simulation of a Hyundai I20 Model with Four Different Turbulence Models (20)

Recently uploaded (20)

Comparison of CFD Simulation of a Hyundai I20 Model with Four Different Turbulence Models