HEAVY VEHICLE DRAG

REDUCTION USING CFD

TECHNIQUES

Nandeesh Hiremath (1PI09ME063)

Pandit Chinmay Shriram (1PI09ME068)

DEPARTMENT OF MECHANICAL ENGINEERING

Under the guidance of

Dr. T.S. Prahlad

Chair Professor- Fluid Mechanics, Mechanical Engineering

Department

PESIT

In the academic session Feb 2012 June 2013

PEOPLES EDUCATION SOCIETY

INSTITUTE OF TECHNOLOGY
100 Feet Ring Road, BSK III Stage, Bangalore - 560 085. Tel : 26721983
 PEOPLES EDUCATION SOCIETY
INSTITUTE OF TECHNOLOGY
100 Feet Ring Road, BSK III Stage, Bangalore - 560 085. Tel : 26721983

DEPARTMENT OF MECHANICAL ENGINEERING

CERTIFICATE

Certified that the project entitled Heavy Vehicle Drag Reduction

Using CFD Techniques is a bona-fide work carried out by
Nandeesh Hiremath (1PI09ME063) and Pandit Chinmay Shriram
(1PI09ME068) in partial fulfillment for the award of degree of
Bachelor of Engineering in Mechanical department of the
Visveswaraiah Technological University, Belgaum during
the year 2012 - 2013. It is certified that all
corrections/suggestions indicated for Internal Assessment have
been incorporated in the report deposited in the departmental
library. The project report has been approved as it satisfies the
academic requirements in respect of project work prescribed for
the Bachelor of Engineering Degree.

Signature of the guide Signature of H O D

Name : Dr. T.S.Prahlad

Designation: Chair Professor
in Fluid Mechanics

Signature of the examiners

(Examiner 1) (Examiner 2)
 Table of Contents
Sl No. Chapter Page
No.
1. Abstract 1
2. Acknowledgement 2
3. Chapter 1: Introduction 3
1.1 What is aerodynamics? 3
1.2 Scope of HVTT aerodynamics 4
1.3 External flow over HVTT 5
1.4 Factors contributing to the aerodynamic drag 6
1.5 Aerodynamic drag force 8
4. Chapter 2: Literature review 9
2.1 Conservation laws applied to the fluid flow 9
2.2 Navier-Stokes equation 10
2.3 System of equations 11
2.4 Turbulence modeling 11
2.5 Implementation of CFD in vehicle aerodynamics 14
5. Chapter 3: Project details 17
3.1 Concept 17
3.2 Computational environment 17
3.3 Phases involved 18
3.4 CAD model generation 19
3.5 Mesh generation 23
3.6 Problem setup 34
3.7 Grid independence check 35
3.8 Benchmark of CFD results 37
3.9 Introducing modifications 41
6 Chapter 4: Results and Discussions 51
4.1 Comparison of drag coefficient for various attachments 51
4.2 Comparison of drag coefficient values with experiment 52
4.3 Discussion 54
7 Chapter 5: Scope for Future Work 55
8 Chapter 6: Appendix 56
9 Chapter 7: References 58
 Abstract

The aerodynamics of bluff bodies shows a complex feature of the

formation of wake region and vortex shedding. The square base, while
having operational utility has been an obstacle in improving the vehicle
fuel economy. The passage of air over the vehicle surface accounts in its
change of behavior resulting in pressure difference between the front and
the back. This phenomenon results in drag and increases fuel
consumption. Increasing the fuel efficiency of the vehicles, even small,
saves a large quantity of oil so that future generation can benefit.

In India the roads are becoming better and better, and it is easier
for the vehicles to attain speeds of 100kmph to 150kmph. Under these
circumstances, compared to the friction between the tyres and the
ground the aerodynamic forces and moments start becoming more
dominant. The objective of this project is to benchmark the CFD
simulations on the Heavy Vehicle Tractor Trailer with wind tunnel results
so that CFD can be used with confidence for actual vehicles.

The propensity towards computational fluid dynamics has showed

us a new path to solve complex problems with relative ease. The
computational tools like CATIA, Fluent and Gambit are used here to
simulate the wind tunnel experiments. Simple external devices like
fairing, side plates, baffle plates and side skirts are introduced to reduce
the drag. It can be concluded that the CFD tool can be used as an
alternative for the wind tunnel experiments for heavy vehicle drag
reduction.

Page | 1
 Acknowledgement

We take this opportunity to express our sincere gratitude and

respect to all those who guided us in the completion of this project.

We wish to place on record our grateful thanks to Dr.T.S. Prahlad

Chair Professor, Mechanical Engineering, P.E.S.I.T, Bangalore for
providing us with the encouragement and guidance throughout the work.

We are thankful to the thermal panel members involving

Dr.K.N.Seetharamu, Dr.T.R.Seetharam, Dr.Ravichandran, Dr.V.Krishna
and Mr.Sarvanan V. for providing their suggestions and ensuring the
progress of the work.

We are thankful for the PESIT for providing the high configuration
workstations at PACE Lab and the sub-sonic wind tunnel facility. We are
also grateful to another batch of our classmates who took up the project
of wind tunnel testing on tractor-trailer configurations and provided us
the data for comparison.

Page | 2
 Chapter 1

Introduction

1.1 What is aerodynamics?

Aerodynamics is a branch of fluid dynamics which deals with the

motion of fluid and its interaction with solid bodies. The transport
variables can be studied under two frames of references namely,
Lagrangian and Eulerian descriptions. The Lagrangian approach deals with
the flow properties of every fluid particle. It becomes tedious to follow
each and every particle in the flow. On the other hand, Eulerian approach
deals with flow fields within a control volume. By the Eulerian approach, it
becomes feasible to obtain the flow properties in the laboratory co-
ordinates.

The fluid flow is mainly influenced by pressure, temperature,

density and velocity. The mass, momentum and energy conservations are
applied on a control volume to get a set of mathematical equations. This
mathematical model is equivalent to real time wind tunnel
experimentation, provided some assumptions are made for turbulence
flow.

Aerodynamics is mainly classified into two categories namely,

external and internal flow. The internal flow deals with the fluid flow
within a confinement. On the other hand, the external flow deals with
flow over the bodies. The project considers the external flow on Heavy
Vehicle Tractor Trailer (HVTT).

Page | 3
1.2 Scope of HVTT aerodynamics

The interest in truck aerodynamics is mainly driven by the increase

in the price and environmental burden of fossil fuels. In figure 1.1 [2],
the inflation corrected crude oil price per barrel is shown since 1975,
indicating the rapid increase in the last decade. According to [1],
approximately 40% of the fuel consumption of trucks is due to
aerodynamic drag, making aerodynamics an important factor in the
financial and environmental impact of the transport sector. Figure 1.2 [3],
shows an exponential increase in power consumption to overcome
aerodynamic drag as compared to rolling friction at high speeds.
Accordingly, at 70 mph the aerodynamic drag contributes about 65% of
the total energy expenditure. Figure 1.3 [3] illustrates the overall
economic benefit associated with reduction in aerodynamic drag as a
function of vehicle speed.

Fig 1.1: Crude oil prices since 1975, corrected for inflation [1]

Page | 4
Fig 1.2: Horsepower required to Fig 1.3: Fuel consumption as a
overcome aerodynamic drag and function of speed for Class 8 tractor
rolling friction as a function of speed trailer [3]
for Class 8 tractor trailer with wind
averaged Cd=0.6 [3]

1.3 External flow over HVTT

The external flow is strongly influenced by the shape of the body.
The external flow is categorized into two regions namely, boundary layer
and inviscid. The boundary layer accounts for a thin layer near to the
body where in the viscous forces are dominant and the flow is rotational.
Beyond the boundary layer the viscous effects are negligible and the flow
is irrotational. For the bluff bodies, the flow separation results in the
formation of wake region having a low pressure as compared to upstream
pressure. This pressure difference results in aerodynamic drag.

In addition to aerodynamic drag caused by the shape of the HVTT,

the gap between the tractor and trailer also plays a major role. In the
gap region the flow recirculation occurs, resulting in a pressure drop and
thereby accounting for 20% of the aerodynamic drag. Similarly, the under
carriage region also accounts for 30% of the aerodynamic drag [4].
Figure 1.4 shows the potential areas for drag reduction.

Page | 5
Fig 1.4: The potential areas for drag reduction [4]

1.4 Factors contributing to the aerodynamic drag

The major factors affecting the aerodynamic drag are: boundary

layer, flow separation, pressure drag and friction drag.

1.4.1 Boundary layer

The boundary layer is defined as a region where in the viscous

effects are dominant. The boundary layer is the result of the shear layers
normal to the surface formed due to no-slip condition and accounts for
the skin friction. For a Newtonian fluid the shear stress is directly
proportional to velocity gradient. The velocity gradient will be steeper in
turbulent flow compared to laminar flow.

1.4.2 Flow separation

Flow separation is characterized by the reversed flow caused by the

adverse pressure gradient. Since the turbulent flow has more energy
compared to laminar flow, the turbulent flow remains attached to the
Page | 6
body for a longer distance. As a result, the turbulent flow will have a
separation point downstream compared to laminar flow.

1.4.3 Pressure drag

In the case of bluff bodies, a low pressure wake region is formed at

the base of the body due to flow separation. The upstream pressure will
be high compared to the downstream pressure. This difference in
pressure results in pressure drag.

1.4.4 Friction drag

The adherence of the fluid to the surface roughness of the surface

strongly influences the fluid-body interaction. With the increase in the
roughness the turbulent frictional stresses increase, thereby increasing
frictional drag.

In the present case, since the surface of HVTT is considered to be

smooth, frictional drag will be smaller. Additionally, the pressure drag and
the flow separation play a dominant role compared to skin friction drag
for bluff bodies.

Page | 7
1.5 Aerodynamic drag force

When the vehicle travels at a high velocity the aerodynamic drag

force will be comparatively higher compared to the rolling friction
between the tyres and ground. The aerodynamic drag is caused by the
pressure distribution and skin friction. The dominance of one or the other
depends upon the shape of the body. Since the dynamic pressure varies
as the square of velocity the aerodynamic drag will have a dominant role
at higher speeds.

Flow
Velocity Drag
Truck

The formula used to define the drag force is as follows.

D = V2 SCd (1.1)

Where,

D = Drag force

= Fluid Density

S = Reference area

V= Free stream velocity

Cd = Drag Coefficient

Page | 8
 Chapter 2

Literature Review

Any real life problem can be transformed into a mathematical model

in the form of partial differential equations. The solution to such
equations can be obtained by either analytical approach or numerical
approach. The limitation on the analytical approach lies in its inability to
solve complex equations. On the other hand, the numerical approach
gives an approximated solution to any complex system of equations with
a quantifiable error. Numerical solution provides reasonably good results
which is always better than having no results at all.

2.1 Conservation Laws applied to fluid flow

In a fluid flow, the flow properties like pressure (P), temperature

(T), density () and velocity ( ) are functions of space and time. In a
control volume approach these fluid properties are defined as field
variables.

The conservation laws cannot be directly applied to a control

volume because these laws are actually defined for a system.
Consequentially, special treatment has to be provided for a control
volume approach. The Reynolds Transport Theorem (RTT) provides a
linkage between the system approach and the control volume approach.
As a result, with the help of RTT all the conservation laws can be applied
to a control volume.

Any real life fluid problem has to obey three conservation laws
namely, the conservation of mass, momentum and energy. The project
deals with the assumption of negligible heat transfer. Hence only
conservation of mass and momentum was considered in this project.

Page | 9
2.1.1 Conservation of mass

The mass flow across a control volume has to be conserved. For a

control volume with no source and sink, the mass influx should be equal
to mass out flux. The application of conservation of mass results in
continuity equation.

=0 (2.1)

2.1.2 Conservation of momentum

The application of Newtons laws of motion gives rise to momentum

equation. Accordingly the rate of change of momentum across a control
volume must be conserved.

(2.2)

2.2 Navier Stokes Equation

The forces acting on a system i.e. the LHS of eq 2.2 can be

classified into traction vector and body forces. The stress tensor maps a
normal vector into a traction vector. For a Newtonian fluid the viscous
stresses are proportional to the rates of deformation. The x, y, z
momentum equations for a Newtonian fluid can be depicted as follows.

( ) + Sm (body force) (2.3)

( ) + Sm (body force) (2.4)

( ) + Sm (body force) (2.5)

Page | 10
2.3 System of equations

For an incompressible fluid, the density being constant we are left

with four unknowns namely; pressure (P) and x-y-z components of
velocity (u, v, w). On the other hand, continuity equation and three
momentum equations form a consistent system to give a unique solution.

Additionally, the continuity equation has no pressure term in itself

and there is no independent variation defined for it in the case of an
incompressible fluid. As a result, the pressure and velocity are coupled.
One such algorithm is SIMPLE (Semi Implicit Method for Pressure velocity
Linkage). In the project, the SIMPLE algorithm embedded in the Fluent
software was implemented.

2.4 Turbulence modeling

2.4.1 What is turbulence?

The fluid flow characterized by the random and chaotic fluctuations
of the flow properties is termed as turbulence. Turbulence exhibits
unsteady flow and always results in high energy dissipation and
momentum diffusion. This random nature of the turbulent flow is dealt in
a statistical manner using Reynolds decomposition. The Reynolds
decomposition is stated as (t) = + (t), wherein the (t) shows the
fluctuating component over the steady mean value . Even in the flows
where in the mean values vary in one or two dimension the turbulent
fluctuations always have a three dimensional spatial character.

The visualizations of turbulent flows reveal rotational flow

structures, so called turbulent eddies with a wide range of length scales.
The smaller eddies detach from the mean flow and transform into large
eddies resulting in vortex stretching. The energy transfer occurs from the
large eddies to small eddy resulting in the energy cascade.

Page | 11
2.4.2 Turbulence model
The Reynolds decomposition leads to additional variables. In order
to provide a closure for the additional variables the turbulence model
comes into picture. The most commonly and widely used model is k-
model (Launder and Spalding, 1974). The k- model considers the
turbulent kinetic energy and its dissipation. Another variant of this model
is realizable k- model (rke). The rke model is more suitable for the flows
encountering flow separation and wake region [5]. Additionally, the rke
model has shown reasonably good results for the analysis on simplified
GTS model compared with experimental results [6]. In the project the
same rke model has been used.

2.4.3 Wall functions

Traditionally, there are two approaches to modeling the near wall
region. In one approach, the viscosity-affected inner region (viscous sub
layer and buffer layer) is not resolved. Instead, semi-empirical formulae
called wall functions are used to bridge the viscosity-affected region
between the wall and the fully turbulent region. The use of wall functions
obviates the need to modify the turbulence models for the presence of
the wall.

In another approach, the turbulence models are modified to enable

the viscosity affected region to be resolved with a mesh all the way to the
wall, including the viscous sub-layer. This is termed as near wall
modeling. As a result, the near wall modeling requires a very fine mesh
to resolve the steeper gradients near the wall. This method is not
economical and would require large computational power which was
beyond our reach. Consequently, the project was restricted to the use
wall functions which are economically viable and within our
computational limits.

The use of wall functions imposes a limit on the near wall cell size.
The near wall cell size is defined by a non-dimensional term y+. Figure
2.1 shows the various regions and the corresponding y+ values of a
turbulent region. The term u+ (u/u) is plotted against y+ (ln(uy/v)).

Page | 12
Fig 2.1: Subdivisions of near wall region

The wall functions are not valid for a y+<5, as a result the wall
functions can be applied provided the near wall cell size exceeds the size
of viscous sub-layer. The ideal range for the near wall cell size is 30 <
y+< 60.Consequently, it would be sufficient enough to concentrate on the
buffer region and the log-law region for the effective capture of turbulent
boundary layer close to the wall.

In order to find the near wall cell size that lies within the acceptable
range, the standard correlations available in [5] and [7] for a flat plate
were used. The sides of the truck surface were assumed to be a flat plate
and the following correlations were used to find the approximate near
wall cell size.

Cfl = for ReL = 1.86 million (2.1)

W=Cfl* v2 (2.2)

U = (2.3)

Y+ = (2.4)

Page | 13
2.5 Implementation of CFD in vehicle aerodynamics
2.5.1 What is CFD?

Computational fluid dynamics is the branch of fluid dynamics which

deals with predicting fluid flows and heat transfer using computational
methods. The virtual formulation of a physical phenomenon using
numerical methods and algorithms mainly involving flow problems is the
key objective of CFD.

2.5.2 How does CFD work?

CFD codes are developed from the numerical algorithms that are
used to solve the fluid problems. These CFD codes have three basic
elements namely, Pre-Processor, Numerical Solver and Post-Processor.

a) Pre-processor
It mainly involves the process of providing inputs of flow problems
to the CFD using software interface and subsequently transforming it into
a form suitable for numerical solver. This stage involves following
processes,

i. Definition of Computational domain:

The topology of geometry is mainly prepared using software tools
such as CATIA. The geometry for HVTT is prepared by using CATIA V5R20
software package. The modeling of the HVTT involved replicating the
exact tractor-trailer assembly present at the wind tunnel lab by covering
all the surface details. This model is an input to the meshing software
such as Gambit. The ease and accuracy of this geometry creation tool
goes a long way in establishing accurate results.

ii. Grid generation:

This is the important step in CFD. The computational domain is
divided into several smaller, non-overlapping sub-domains. These smaller

Page | 14
units are called elements. In the CFD code used here a finite volume
method is used. The entire domain is divided into smaller volumes by
maintaining the connectivity. The accuracy of the CFD results depends on
the number of such small volumes/cells, and generally with increase in
the number of cells, the accuracy increases. For the optimal results, the
meshing should be finer where there are large variations in flow
properties and can be coarser where there are small variations.

The software package used for grid generation is GAMBIT. The

primary step is to form a surface mesh on to the geometry. The domain is
then meshed with 3D elements either tetra or hex elements or
combination of both of the elements is used.

iii. Selection and definition of fluid properties:

The definition of fluid properties is important as all the results
depend on the extent of accuracy to which these properties are defined.
These properties for the given computation domain are defined at this
stage.

iv. Defining boundary conditions:

The boundary conditions are important in order to initialize the CFD
computations. These are defined at the cells which coincide with or touch
the domain boundary.

b) Numerical Solver
The CFD uses solver such as FLUENT. There are three distinct
numerical solution techniques, namely finite volume method, finite
element method and finite difference method. FLUENT uses finite volume
method. It is the process of numerical algorithm calculation involving the
use of finite volume cells.

Page | 15
 The numerical algorithm consists of the following steps:

i. Integration of the governing equation of the fluid flow over all the
finite volume cells.
ii. Discretization: conversion of resulting integral equations into
system of algebraic equations.
iii. Solution of an algebraic equation by an iterative process.

c) Post-processor
The data visualization and result analysis is the key step in CFD.
The results of simulation can be obtained by the following

i. Line and contour plots

ii. Vector plots
iii. Particle tracking
iv. 2D and 3D surface plots
v. XY plots and graphs of results

Page | 16
 Chapter 3

Project Details

3.1 Concept

The heavy duty vehicles experiences significant drag when they

travel at high speeds. The formation of wake in the base region and the
recirculation in the gap and under-carriage region contribute a major part
to the aerodynamic drag. The project delineates the use of CFD tool in
determining the aerodynamic drag coefficient followed by drag reduction.
The exact wind tunnel environment is simulated including the walls of the
wind tunnel. The various add-ons like fairing, side plates, baffle plates
and side skirts were implemented to reduce the drag.

3.2 Computer environment

The CFD simulations of a three dimensional configuration require

enormous computational power. A large number of iterative calculations
are viable only through the use of powerful computers. The configuration
of the computers determines the complexity of the problem that can be
solved. A computer with more number of parallel processors will handle
more challenging problems and the computational time will also be
reduced.

In the present project, the CFD simulations are carried on the

workstations in the PACE lab at PESIT. The configuration of the
workstations is as follows,

i. Processor: 64 bit Xeon Quad Core Processor

ii. Processor speed: 3.5GHz
iii. RAM: 12GB
iv. Graphics card: 1GB Nvidia
Page | 17
 The configuration of these workstations imposed a limit on the
number of elements they could handle. On an average, the maximum
number of elements they could handle was 25lakhs. The workstations
took approximately 38 hours to simulate the results for each test.

3.3 Phases involved

The project is carried at various phases, starting from the CAD

model-mesh generation followed by validation of results and the
implementation of external devices. The following chart shows the phases
involved sequentially. Each phase is dealt in detail in the following
sections.
CAD Model Generation

Mesh Generation

Grid Independence Check

Benchmark of CFD Results

Introducing Modifications

Fairing Side Plates Baffle Plates Side Skirts

Page | 18
3.4 CAD model generation

The CAD model was generated in CATIA V5R20 software. The

geometry involves a simplified model of the present day tractor trailer
systems. The model is in exact correlation with the scaled wind tunnel
model with certain approximations for the ease of analysis.

The figure 3.1 and 3.2 shows the CAD model of HVTT followed by
the figure 3.3 showing the dimensions of the model.

Fig 3.1: CAD model of HVTT

It can be observed from the figure 3.1 and figure 3.2 that the
model involves all the aspects of the present day tractor-trailer with
simplifications. The HVTT is a ten-wheel model with wheel diameter of
61mm and having a gap of 40mm between the tractor and the trailer.
The model dimensions are 893mm (L) x 140mm (W) x 216.5mm (H).

Page | 19
Fig 3.2: Side view of HVTT- All dimensions in mm

Fig 3.3: Bottom view of HVTT- All dimensions in mm

For the ease of analysis certain approximations to the model were

made. Firstly, the supports provided to the wind tunnel model for fixing it
to the base have not been included in the CAD model. Secondly, the small
gap in the undercarriage region between the tractor and the trailer is
filled completely as it would be difficult to provide meshing in the small
gap of 5.5mm. Lastly, the filleting provided to the wind tunnel model was
ruled out initially because of the complexities involved in providing a
curvature while meshing. But, the initial analysis showed drastic deviation
in the result; hence we had to implement an approximate filleting of 3mm
in order to exactly correlate with the wind tunnel model.

Page | 20
 Further, the external devices were modeled and were attached to
the base model. Figure 3.4 shows the various attachments applied to the
model in the CFD analysis. Firstly, the fairing of triangular shape with a
dimension of 105 (L) x 48.5 (H) was attached. It was followed by the side
plates covering half the gap between the tractor-trailer. Thirdly, the baffle
plates were introduced on the trailer face at an interval of 35mm. Lastly,
the side skirts were introduced along the length of the trailer.

Fairing

Fig 3.4a: Fairing attachment

Side Plates

Fig 3.4b: Fairing with side plates; side plates cover half the gap between
tractor-trailer

Page | 21
 Baffle Plates

Fig 3.4c: Fairing with baffle plates; baffle plates are provided at an
interval of 35mm

Side Skirts

Fig 3.4d: Fairing with the side skirts; side skirts along the length of trailer

Page | 22
3.5 Mesh Generation

The generation of surface mesh and volume mesh is an important

step in CFD. The software used for the generation of mesh is GAMBIT
2.2.30. It has the ability to form a surface mesh on a HVTT model for
various element sizes. The volume mesh is then generated around the
model in the specified computational domain. The 2D elements used for
surface meshing are either quads or trias or combination of both. The
volume mesh is either tetra or hex or combination of both in which
prismatic layers are formed at the tetra and hex junction for the smooth
transformation.

The mesh generation for HVTT was done in two steps.

a) Domain extent
b) Grid generation
i. Unstructured grid
ii. Structured grid

3.5.1 Deciding the domain extent

The primary objective is to decide the domain extent. As the main

objective of simulations was to exactly replicate the wind tunnel
conditions, the test domain section of the wind tunnel is taken as the
domain extent for CFD. The test section is of 2000*600*600 cubic
millimeters. Accordingly, the CFD simulations were performed on the
HVTT. The pressure contours obtained for this domain extent is shown in
the figure 3.5.

Page | 23
 Interference of Inlet
Condition

Fig 3.5: Pressure contours for domain extent of 2000mm x 600mm x

600mm

When this pressure contours are analyzed it is observed that the

contours were interfering with the inlet boundary conditions. As a result it
was not possible for CFD solver to take into account a complete variation
in flow parameters. The results were greatly affected because of this
interference as there was no sufficient space for the flow to be developed.
The coefficient of drag obtained was 1.44 which was too high compared
to the experimental value of 0.75.

There was a need to remove the interference of the inlet boundary

condition. In order to accomplish this, the length of the domain was
increased from 2000 mm to 8000 mm. The cross-sectional area was
maintained the same i.e. 600*600 square millimeters. This is important
consideration as the mass flow rate across the domain should not be
affected. The new computational domain now obtained has
8000*600*600 cubic millimeter as its extent.

Page | 24
 For this new domain extent, the CFD simulations were performed.
The pressure contours then developed was observed. Figure 3.6 shows
the pressure contour plot for the new domain extent. It can be seen from
the given pressure contour plot that the interference which was observed
earlier was removed. As a result the CFD solver could take into account a
complete variation in the flow parameters without interfering with the
inlet boundary conditions. The coefficient of drag obtained was 0.85 as
compared to 1.44 which was observed for the previous domain extent.
So, with the increase in the domain length, solver was able to show
improvement in the drag coefficient, because interference got nullified
and the flow was developed completely.

Fig 3.6: Pressure contours for domain extent of 8000mm x 600mm x

600mm

The domain extent length was further increased, to check if there

would be any change in drag coefficient. So, the domain length was
increased from 8000 mm to 12000 mm. There was no significant change
Page | 25
in the drag coefficient value. Hence, it was concluded that the domain
extent with 8000 mm as its domain length would be sufficient enough to
take into account a complete variation in flow parameters. The increase in
length could only increase the number of elements and increase the
solver time. So, it was finally decided that the optimum domain extent to
be 8000*2000*2000 cubic millimeters.

The HVTT length (L) is 893 mm. The HVTT was placed at a distance
of 2L (approximately 2000 mm) from the upstream inlet boundary and a
distance of 5L (approximately 5000 mm) was provided in the
downstream. Also, the ground clearance of 25mm was provided in
contrast to the experimental setup, where in the model was placed on the
rotating table with no clearance. This approximation is essential during
meshing and is not expected to affect the comparison. The domain extent
with HVTT is shown in the figure 3.7.

2L L 5L

Fig 3.7: Placement of HVTT in the control volume

Page | 26
3.5.2 Grid generation

Once the domain extent was fixed the next step was to form a
surface mesh on to the HVTT and form a volume mesh in the
computational domain. Initially, the mesh was generated with
unstructured type because of the ease of meshing. Later we shifted to
structured type as it offered more feasibility in localized refinement.

a) Unstructured mesh
The unstructured mesh was formed using tetra elements in the
chosen domain. As for the computation accuracy and for restricting the
number of elements to the optimum level, the mesh was generated in
such a way that it was coarser at the areas where there was lesser
variations in the fluid properties and finer in the areas where there was
higher variations in the fluid properties.

As the areas near the truck surface and near the walls are
subjected to the boundary layer effects, they are under the influence of
large flow property variations. Hence, it was required to increase the
number of cells near these areas. Also, the wake region of HVTT at the
downstream side would also be subjected to large property variations,
because of the recirculation and eddy formation. Accordingly the mesh
was generated with higher number of elements in the wake region.

The near wall cell size was chosen to be 5mm. The meshing is done
in such a way that element size increased from HVTT surface to the bulk
of the fluid followed by a gradual decrease to the wall. This was done by
providing a specific growth ratio. The maximum element size was
restricted to 80 mm and the growth ratio provided was 1.2. The wake
region had elements with a size of 10 mm. The grid thus obtained is
shown in the figure 3.8. This grid when subjected to the computation
gave a Cd of 0.89.

Page | 27
Fig 3.8: Unstructured gird

The accuracy of results is directly dependent on the number of

elements, and increases with the increase in the elements. The numbers
of elements in the computational domain were increased. But, due to the
computational limitation, to increase the number of elements without
affecting the results, the flow analysis was restricted to half truck rather
than the full truck, thereby increasing the number of elements and
simulating a symmetrical flow.

The boundary layer near the truck surface needs to be captured

sufficiently. From the flat plate analysis, the thickness of the boundary
layer was calculated and it was found to be 15 mm. The primary aim was
to provide sufficient number of elements in this boundary layer to capture
the flow well. The growth ratio was maintained in such a way that for this
unstructured mesh, there were 3 elements within the boundary layer of
15 mm.

Page | 28
b) Structured mesh
The unstructured mesh with the tetra elements was not able to
capture enough elements within the boundary layer. Hence, in order to
increase the number of elements within the boundary layer and to
maintain the optimum level of elements at the areas of large gradients,
structured meshing was considered.

The structured meshing involves dividing the complete

computational domain into smaller volumes mostly cubical volumes. Each
volume is then meshed individually. The meshing starts from one end and
continues further. As hex elements give more accuracy in the results as
compared to the tetra elements, it was important to use hex elements
wherever possible in the computational domain. The smaller volumes
were created near the HVTT surface, to account for boundary layer
thickness. These volumes were then meshed with the hex elements. By
maintaining sufficient aspect ratio for the hex elements, the number of
elements within the boundary layer was now increased.

Figure 3.9 shows the structured grid around the HVTT. The
optimum near wall cell size was calculated by knowing the non-
dimensional parameter called the y+ value. In order to capture the
turbulent boundary layer the y+ should fall between 30 and 60.
Accordingly the near wall cell size was chosen to be 0.5mm.

Fig 3.9a: Structured grid

Page | 29
 Wall

Fig 3.9b: Structured grid Near wall cells

The structured mesh shown in the figure 3.9 has near wall cell size
of 0.5 mm. The growth ratio of 1.2 could capture ten layers within the
boundary layer. This number was sufficiently large to capture the flow
variations within the boundary layer as compared to unstructured
meshing which could capture only three layers within the boundary layer.
As the volume above this boundary layer up to the bulk of the flow has
lesser flow property variations, it made sense to provide lesser number of
elements in this volume. Hence, growth ratio was given in such a way
that there were six layers with varying height in bulk of the flow.
Additionally, in order to capture the boundary layer effects on the wind
tunnel wall, the growth ratio was given in such a way that there were
seven elements from the bulk of the flow to the wind tunnel wall with
smaller cells near the wall. Consequently, the turbulent velocity profile
was effectively captured and is shown in the figure 3.10.

Page | 30
 Wall

Fig 3.10: Turbulent velocity profile

The drag coefficient obtained with the structured meshing has a

value of 0.78 as compared to 0.89 obtained from unstructured mesh.
Consequently, the deviation in the value of drag coefficient was reduced
from 28% to 13% as compared to experimental values. Hence, it was
concluded that by sufficiently capturing the turbulent boundary layer and
with the use of hex elements with optimum aspect ratio as obtained from
structured grid, improvement in the value of drag coefficient was
possible.

Figure 3.11 represents the meshing just above the top surface of
the tractor. It is found that the CAD model of the HVTT has sharp corners
and the edges. Due to these sharp corners, the flow got separated from
the leading edge. This flow separation phenomenon is supported by the
velocity contours obtained from the structured grid as shown in the figure
3.12. It can be observed that the flow is not attached to the surface of
the tractor part of the HVTT.

Page | 31
Fig 3.11: Meshing near the sharp corners

Fig 3.12: Velocity contours of HVTT with sharp corners

Page | 32
 When the CAD model was cross-verified with the wind tunnel model
it was found that the model had filleting all around and there were no
sharp corners. Also, the painting applied to the model provided extra
smoothness at the curvatures. Hence, it was thought of introducing
filleting at the sharp corners, mainly at the leading edge. The filleting
provided has a radius of curvature of 3 mm. With the introduction of
filleting as a modification, to avoid the flow separation the structured
mesh was regenerated. The CFD computation was done for this modified
mesh. The figure 3.13 gives the velocity contours obtained after the
filleting. A drastic change in the velocity contours can be observed from
the figure. The flow reattaches at an early stage after being separated at
the leading edge. As a result, the deviation in the drag coefficient was
reduced from 13% to less than 5%, after the implementation of fillets.

Fig 3.13: Velocity contours of HVTT with filleting

Page | 33
3.6 Problem setup
CFD simulations require solvers like Fluent. The problem was set up
in the Fluent and the appropriate boundary conditions were provided. In
order to simulate the wind tunnel results the exact operating conditions
were provided. Approximate values for parameters involved in the k-
turbulence model were assumed based on the literature [5].

Figure 3.14 shows the boundary conditions imposed on the model

and the Table 3.1 shows the various parameters set up in the Fluent.

Top Wall
Pressure Outlet

Velocity Inlet
Bottom Wall

Fig 3.14: Boundary conditions

Table 3.1: Parameters set up in Fluent

Operating pressure 100798.3 Pa
Inlet Velocity 30.6 m/s
Outlet Pressure 100798.3 Pa
Turbulent Intensity 0.05 %
Turbulent Length Scale 0.06 m
Solver Implicit
Type of analysis Unsteady
Time step 1 second

The entire CFD simulation was carried at zero degree yaw with a
symmetry plane. Since the simulation for the entire domain requires
more computation which was beyond our reach, we had restricted our
analysis on half-truck thereby simulating a symmetry flow.

Page | 34
3.7 Grid independence check

The grid independence check is a method to ensure correctness of

the CFD results. Generally, in the initial stages a coarse mesh will be
generated in order to have the convergence at a faster rate. Later on, the
mesh will be refined in the regions of high gradients step by step.
Subsequently a stage will be reached when there will be no significant
change in the results between the current mesh and the previous mesh.
At this stage it can be said that the mesh has passed grid independence
check.

3.7.1 Effect of domain extents

The domain extents for the CFD simulations were taken to be the
same as that of the test section of the wind tunnel. The cross section area
of the control volume was fixed in order to maintain the same mass flow
rate of the fluid as compared to wind tunnel testing. On the other hand
there was feasibility in varying the length of the control volume in order
to check the effect of domain extents. Consequently, the changes
observed in the drag coefficient is depicted in table 3.1

Table 3.2: Effect of domain extents

Test Domain Extents No. of Elements Cd
Test No. 9 2000 x 600 x 600 9 lakhs 1.45
Test No.15 8000 x 600 x 600 12 lakhs 0.85
Test No. 20 12000 x 600 x 600 16 lakhs 0.89
*Test section dimensions are 2000 x 600 x 600
*Test conducted on a bare model

From the table 3.1 it can be observed that the Test No. 15 showed
a larger change in the Cd value compared to Test No. 9 and there was no
much significant change in the Cd with the further increase in domain
extents. As a result the domain extent of 8000x600x600 was chosen for
Page | 35
further analysis. Additionally, this domain extent is in exact correlation
with [8] which says that the length of the control volume should be
approximately five times the length of the model.

3.7.2 Unstructured grid

During the initial stages the unstructured grid was implemented

because of the greater ease in meshing. The element size was gradually
increased from the surface to the bulk of the fluid. The near wall cell sizes
ranged from 10mm to 5mm for coarser to finer mesh. But no significant
change in the Cd was observed because the boundary layer was not being
effectively captured. It was possible to capture hardly three elements
within the boundary layer. The further refinement in the near wall cell
size was limited by the computational power as the number of elements
had already reached the limits. Hence we had to opt for structured grid
wherein local refinement was possible.

3.7.3 Structured grid

The near wall cell size cannot be kept on reducing as there is a limit
for it in order to satisfy the Standard Wall Functions. The limit to the near
wall cell size is expressed in terms of y+ values. Additionally, the k-
model also imposes a condition that the y+ has to range from 30 to 60.
Accordingly, the approximate near wall cell sizes were calculated for
varying y+ values and the corresponding Cd values are shown in table
3.2.

Table 3.3: Near wall cell size

Test Y+ Near-wall Cell No. of Cd % Deviation
Size Elements Cd
Test No. 31 90 2 mm 16 lakhs 0.82 9.3
Test No. 33 22 0.5 mm 16 lakhs 0.78 4
*Test conducted on a bare model
Page | 36
 From the table 3.2 it can be observed that for a near wall cell size
of 0.5 mm the percentage deviation in Cd from the experimental value is
4% less than for the near wall cell size of 2mm (9.3%). Also, the Test No.
33 had ten layers within the boundary layer as compared to eight layers
in the Test No. 31. The combined effect of smaller cell size and more
number of layers within the boundary layer has shown reasonably good
correspondence with the experimental values. As a result, the parameters
involved in the Test No. 33 were chosen for further analysis.

3.8 Benchmark of CFD results

Good engineering practice suggests that prior to using an analysis
technique on a new configuration, one should benchmark (validate) the
technique against a known (respected) test case similar to the new
configuration. If no suitable test case exists, then cross referencing with
another analysis technique, such as a wind tunnel, is essential. The
benchmark test process is the process of numerical analysis performed on
a case which is replica of the real time testing or previous results of
numerical simulations. While performing the benchmark testing, the
results of the test will be further compared with other available results.
Benchmarking is done in order to achieve desired level of confidence and
accuracy in the results.

3.8.1 Wind tunnel facility

The wind tunnel facility available at PESIT was utilized to obtain the
experimental values by another batch of students. The wind tunnel is an
open type suction wind tunnel with a test section dimension of 2000mm x
600mm x 600mm.

The lead edge of the wind tunnel is placed approximately at the half
length of the test section. The model is fixed on to a rotating table and
the wheels touch the base. The testing is done at a fan rpm of 800 that
Page | 37
corresponds to 30.6 m/s. The strain gages embedded inside the balance
mounted on rotating table measure the drag force, side force and yaw
moments. Additionally, the pressure ports are provided which will read
the local static pressure. The entire output is displayed on to a digital
display. All the wind tunnel readings are obtained from [9].

893 mm
Test Section

Truck

2000 mm

Fig 3.15: Wind tunnel facility at PESIT

Page | 38
3.8.2 Benchmarking
The results obtained from CFD are validated with the experimental
values in terms of drag coefficient and the coefficient of pressure. The
benchmarking of the CFD results was done with the base model having
fairing as an attachment.

Cp Plot - CFD Vs Exp

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2
-0.2

-0.4

-0.6

Cp-CFD
Cp

-0.8
CP-Exp
-1

-1.2

-1.4

-1.6

-1.8
X/L

Fig 3.16: Cp Plot Validation

Cp = (3.1)

Where,

P = local static pressure

P = Free stream pressure

= Fluid density

V = Free stream velocity

Page | 39
 Figure 3.16 shows the variation of Cp on the top surface of HVTT.
The Smooth curve in blue color shows the Cp plot obtained from CFD and
the red spots denote the localized Cp values obtained from wind tunnel
experiment. It can be observed that there is a good correspondence
between the CFD and the experimental values.

Table 3.4: Drag Coefficient Validation

Test No. of Cd - CFD Cd - Exp %
Elements Deviation
Test No. 43 16 lakhs 0.67 0.65 3.07

Table 3.4 shows the validation of drag coefficient obtained from

CFD with the wind tunnel experiment. Test No. 43 was conducted on a
model with fairing attachment. The Cd-CFD is shown as 0.67 as compared
to Cd-Exp of 0.65 with a 3.07 % deviation. The CFD slightly over predicts
the Cd value that may be due to the approximations involved in the
turbulence model.

3.8.3 Conclusion from benchmarking

From the simulation results and the setting shown above, the
conclusion can be drawn from the benchmarking process for the HVTT
model used. The results of the simulation agree well with experimental
results. If the results are to be made completely reliable, including model
yaw, the CFD simulation is needed without using any symmetry plane.
For complete domain simulation, very competitive computer configuration
and power will be required, which is beyond the scope of the present
project work.

Page | 40
3.9 Introducing modifications

This section delineates the post processed results for the bare
tractor-trailer followed by the changes observed after the implementation
of the external devices. The external devices are fairing, side plates,
baffle plates and side skirts.

3.9.1 Bare tractor-trailer

a) Pressure contour

Fig 3.17: Pressure contour for bare HVTT

The pressure contours from figure 3.17 shows the uniform pressure
distribution at the far-field of the model. The red region at the front of the
tractor part shows the stagnation pressure and the vast spread yellow-
green region shows the ambient pressure.

Page | 41
b) Velocity contour

Fig 3.18a: Velocity contour for bare HVTT

Wake Length = 320mm

Fig 3.18b: Velocity contour in the wake region of bare HVTT

Page | 42
 Symmetry

Tractor Trailer

Fig 3.18c: Velocity contour in the gap region of bare HVTT Top view

Symmetry Line

Fig 3.18d: Velocity contour in the under carriage region of bare HVTT
Bottom view

Page | 43
 The velocity contours from figure 3.18 shows the flow velocity
around the model. The blue region shows the zero velocity flow that
occurs in the regions of negligible dynamic pressure. The recirculation in
the wake region and the gap region can be prominently seen. The wake
length is observed to be 320mm. Also, the nascent bubble formation that
may lead to Vortex Street can be observed in the wake region. Lastly, the
undercarriage region shows large recirculation. The recirculation in the
wheel base region can be prominently observed.

c) Cp plot

Cp Plot
0
0 0.2 0.4 0.6 0.8 1 1.2
-0.2

-0.4

-0.6

-0.8
Cp

CFD-No Fairing
-1

-1.2

-1.4

-1.6

-1.8
X/L

Fig 3.19: Cp plot on top surface for bare HVTT

The Cp variation is plotted on the top surface of the HVTT. It shows

a maximum value of -0.2 at the stagnation point. The Cp value remain
almost constant on the trailer surface and slightly dips at the trailing end

Page | 44
of the trailer that may be because of the adverse pressure gradient
caused by flow separation. The steep fall in the Cp values occur on the
top surface of the tractor region and in the gap region.

3.9.2 Fairing attachment

a) Velocity contours

Fairing Attachment

Fig 3.20a: Velocity contour for fairing attachment

Page | 45
 Symmetry

Tractor Trailer

Fig 3.20b: Velocity contour in the gap region for fairing attachment
Top view

The velocity contours from figure 3.20 shows the flow velocity
around the model. The blue region shows the zero velocity flow that
occurs in the regions of negligible dynamic pressure. In the leading edge
the flow reattaches quickly due to the presence of fairing. The
recirculation in the gap region can be prominently seen.

Page | 46
b) Cp plot

Cp Plot
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2
-0.2
-0.4
-0.6
Cp

-0.8
Cp-CFD _on the Fairing
-1
-1.2
-1.4
-1.6
-1.8
X/L

Fig 3.21: Cp plot on top surface for fairing attachment

The Cp variation is plotted on the top surface of the HVTT. It shows

a maximum value of -0.2 at the stagnation point. The Cp variation is
almost similar to the bare model plot. The difference lies in the region
above the tractor and the gap region. The steeper fall in the Cp is nullified
by the effect of the fairing.

Page | 47
3.9.3 Side plates

The side plates were implemented mainly to prevent the cross flow
in the tractor-trailer gap. Since the project deals with symmetry flow
simulation, the cross flow will be comparatively negligible compared to
flows at various yaw angles. The flow recirculation in the gap region can
be observed from the figure 3.22.

Symmetry

Tractor Trailer

Side plate

Fig 3.22: Velocity contour plot in the gap region for side plates Top view

Page | 48
3.9.4 Baffle plates

The baffle plates were also implemented mainly to prevent the

cross flow. Since the project deals with symmetry flow simulation, the
cross flow will be comparatively negligible compared to flows at various
yaw angles. The flow recirculation in the gap region can be observed from
the figure 3.23

Symmetry

Trailer
Tractor

Baffle Plates

Fig 3.23: Velocity contour in the gap region for baffle plates Top view

Page | 49
3.9.5 Side skirts

The side skirts were implemented mainly to prevent the cross flows
in the undercarriage region. The recirculation regions in the undercarriage
region were analyzed with this attachment. The flow recirculation in the
under carriage region can be observed from the figure 3.24

Symmetry

Fig 3.24: Velocity vector in the under carriage region for side skirts
Bottom view

Page | 50
 Chapter 4

Results and Discussion

The main purpose of the project was to benchmark the CFD results
and observe the variation in the drag coefficient with the implementation
of the external devices. From the chapter 3.8.2 it can be concluded that
there is a strong correspondence between the CFD results and the wind
tunnel results as far as Cp distribution is considered.

4.1 Comparison of drag coefficient for various

attachments

In the chapter 3.9 various modifications were introduced followed

by their respective contour plots and the vector plots. The purpose of
implementing these devices was to check their effects on the drag
coefficient. Accordingly, these external devices have shown drag
reduction ranging from 2.98 % to 7.46%. The table 4.1 shows the drag
coefficient values for various attachments.

Table 4.1: Drag Coefficient values for various attachments

Test Model Cd-CFD % Cd Reduced
Test No.33 Bare model 0.78 -
Test No.43 Fairing 0.67 -
Test No.45 Side Plate 0.65 2.98
Test No.48 Baffle Plate 0.65 2.98
Test No.46 Side Skirts 0.62 7.46

From the table 4.1 it can be observed that the bare model offers a
Cd of 0.78. With a simple fairing attachment the Cd has reduced to 0.67
showing a reduction in Cd by 14.1%. Since the present day trucks come

Page | 51
with the fairing, the fairing model has been chosen as a base model for all
the further comparison. Accordingly, the model with the side plates offers
a Cd of 0.65 showing a reduction by 2.98 %. We can also observe that
the baffle plates also show the same reduction in Cd. Lastly, the side
skirts shows the maximum reduction in the Cd by 7.46 % and offers a Cd
of 0.62.

4.2 Comparison of drag coefficient values with

experiment

The drag coefficient values obtained from CFD are compared with
the experimental values. Within the purview of the project this
comparison is limited to the symmetry flow at zero degree yaw. The table
4.2 shows the comparison between the CFD and experimental values. The
table 4.3 shows the bar chart of the relative Cd values for CFD and
experiment.

Table 4.2: Comparison of Drag Coefficient values with experiment

Test Model Cd-CFD Cd-Exp %Deviation
Test No.33 Bare Model 0.78 0.75 4
Test No.43 Fairing 0.67 0.65 3.07
Test No.45 Side Plate 0.65 0.62 4.8
Test No.48 Baffle Plate 0.65 0.62 4.8
Test No.46 Side Skirts 0.62 0.61 1.6

Page | 52
Table 4.3: Relative Cd-CFD comparison with Cd-Exp

Comparison of Drag Coefficient - CFD Vs Exp

0.9
0.8
0.7
0.6
0.5
Cd

0.4 Cd-CFD
0.3 Cd-Exp
0.2
0.1
0
Bare Model Fairing Side Plate Baffle Plate Side Skirts
Models

From the table 4.2 it can be observed that the deviation in the CFD
results as compared to experimental results is less than 5% for all the
cases. This deviation of less than 5% is reasonable is within the purview
of acceptable error. The table 4.3 shows the relative comparison of the
Cd-CFD and the Cd-Exp. The blue bars show the Cd-CFD and the red bar
shows the Cd-Exp. From the chart it can be observed that the CFD
slightly over predicts drag.

However, considering the complexity of the configuration, it can be

said that good benchmarking has been done for CFD both in terms of
pressure distribution and drag coefficient without and with drag reducing
attachments. Thus, there should be reasonable confidence in using CFD
for full scale vehicles with proper simulation of road conditions.

Page | 53
4.3 Discussion

The deviation and over prediction in Cd values may be due to the

approximations involved with the turbulence model and also wind tunnel
test uncertainties. The mathematical model does not involve exact
correspondence with the practical scenario as certain inherent
assumptions will be involved for the ease of analysis.

Within an acceptable range the CFD results are in good

correspondence with the experimental values. The full domain simulations
involving full HVTT model might provide still more accurate results,
especially at yaw. Also, the project emphases on the relative drag
reduction and not the absolute values. As a result the observed deviation
of less than 5% is a good comparison.

Finally, it can be concluded that CFD can be used as an alternative

for the wind tunnel approach for the simulation of heavy vehicles with
bluff configuration. CFD can be used to simulate the exact flow conditions
on the road including the relative motion between the vehicle and the
road without any wind tunnel wall effect.

Page | 54
 Chapter 5

Scope for Future Work

The project mainly takes into account the gap region and the
undercarriage region. These two regions collectively contribute for 50% of
total aerodynamic drag. The attachments such as the side plates, baffle
plates and the side skirts were provided in order to improve the drag in
these two regions.

The front region and the base region of the truck correspond to
overall aerodynamic drag of 50%. These two regions have not been taken
for the drag improvements in the current project. As a future work the
attachments can be implemented in the front and the base region.
Additional devices like vortex strake devices and the momentum injection
devices like roller pin can be analyzed using CFD.

A more intrinsic study can be made for a wind averaged drag

coefficient. The actual road conditions can be provided with the inclusion
of the relative motion between the model and the ground followed by the
relative rotary motion between the wheels and the road.

The project mainly involved with the flow analysis at zero degree
yaw. As the analysis was restricted to half-model flow, due to
computation limitations, it was not possible to analyze the model at yaw
angles. With better computers a more complex full domain simulations
can be made.

Page | 55
 Chapter 6

Appendix
Figures and Tables
Description Page
No.
Fig 1.1 Crude oil prices since 1975, corrected for inflation 4

Fig 1.2 Horse power required to overcome aerodynamic drag 5

and rolling friction for Class-8
Fig 1.3 Fuel Consumption as a function of speed for Class-8 5
Fig 1.4 The potential areas for drag reduction 6
Fig 2.1 Subdivisions of near wall region 13
Fig 3.1 CAD model of HVTT 19
Fig 3.2 Side view of HVTT 20
Fig 3.3 Bottom view of HVTT 20
Fig 3.4a Fairing attachment 21
Fig 3.4b Fairing with side plates 21
Fig 3.4c Fairing with baffle plates 22
Fig 3.4d Fairing with side skirts 22
Fig 3.5 Pressure contours- domain extent 2000*600*600 24
Fig 3.6 Pressure contours- domain extent 8000*600*600 25
Fig 3.7 Placement of HVTT in the control volume 26
Fig 3.8 Unstructured grid 28
Fig 3.9a Structured grid 29
Fig 3.9b Structured grid-Near wall cells 30
Fig 3.10 Turbulent velocity profile 31
Fig 3.11 Meshing near the sharp corners 32
Fig 3.12 Velocity contour of HVTT with sharp corners 32
Fig 3.13 Velocity contour of HVTT with filleting 33
Fig 3.14 Boundary conditions 34
Fig 3.15 Wind tunnel facility 38
Page | 56
Fig 3.16 Cp plot validation 39
Fig 3.17 Pressure contour of bare HVTT 41
Fig 3.18a Velocity contour of bare HVTT 42
Fig 3.18b Velocity contour in the wake region of bare HVTT 42
Fig 3.18c Velocity contour in gap region of bare HVTT 43
Fig 3.18d Velocity contour in the undercarriage region of HVTT 43
Fig 3.19 Cp plot on top surface of bare HVTT 44
Fig 3.20a Velocity contour for fairing attachment 45
Fig 3.20b Velocity contour in gap region for fairing attachment 46
Fig 3.21 Cp plot on top surface for fairing attachment 47
Fig 3.22 Velocity contour plot in gap region for side plates 48
Fig 3.23 Velocity contour plot in gap region for baffle plates 49

Fig 3.24 Velocity contour in the undercarriage region for side 50

skirts

Table 3.1 Parameters setup in FLUENT 34

Table 3.2 Effect of domain extents 35
Table 3.3 Near wall cell size 36
Table 3.4 Drag coefficient validation 40
Table 4.1 Drag coefficient values for various attachments 51
Table 4.2 Comparison of drag coefficient values with experiment 52
Table 4.3 Relative Cd-CFD comparison with Cd-Exp 53

Page | 57
 Chapter 7

References

1. W. H. Hucho, Aerodynamik des Automobiles. Vogel Verlag, 1981

2. P.M. van Leeuwen, Computational Analysis of Base Drag Reduction
Using Active Flow Control, Master of Science Thesis, TUDELFT-Delft
University of Technology, November 2009.
3. Rose McCallen, Richard Couch, Juliana Hsu, Fred Browand, Mustapha
Hammache, Anthony Leonard, Mark Brady, Kambiz Salari, Walter
Rutledge, James Ross, Bruce Storms, J.T.Heineck, David Driver,
James Bell and Gregory Zilliac, Progress in Reducing Aerodynamic
Drag for Higher Efficiency of Heavy Duty Trucks (Class 7-8)
4. Richard M. Wood, A Discussion of a Heavy Truck Advanced
Aerodynamic Trailer System, SOLUS-Solutions and Technologies LLC
5. Fluent 6.26 database, http://jullio.pe.kr/fluent6.1/help/html/ug/node450.htm
6. Kambiz Salari, Mary McWherter-Payne, Computational Flow
Modeling of a Simplified Integrated Tractor-Trailer Geometry, Sandia
National Laboratories.
7. C.P.Kothandaraman, S.Subramanyan, Heat and Mass Transfer Data
Handbook, 7th edition, New Age International Publishers, 2010.
8. V.Malviya, R.Mishra, J.Fieldhouse, Comparative Computational
Analysis of Drag-Reducing Devices for Tractor-Trailers, University of
Huddersfield.
9. Praneet Udyavar, Sourabh Rai, Akhil Kumar, Dr.T.S.Prahlad (Guide),
REDUCTION OF AERODYNAMIC DRAG ON A TRACTOR TRAILER
THROUGHWIND TUNNEL TESTS, PESIT, Bangalore, 2013.
10. David Pointer, Tanju Sofu, Jimmy Chang, David Weber, Applicability
of Commercial CFD Tools for Assessment of Heavy Vehicle
Aerodynamic Characteristics, Argonne National Laboratory.

Page | 58
11. Subrahmanya P. Veluri, Christopher J. Roy, Anwer Ahmed, Rifki Rifki,
Preliminary RANS Simulations and Experimental Study of a
Simplified Tractor/Trailer Geometry, 24th Applied Aerodynamics
Conference, California, 2006.
12. D. Computational Investigation of Aerodynamics of Rail Coal Cars
and Drag Reducing Add-On Devices, Heavy Vehicle Systems
Optimization Program, FY Annual Report, 2004.
13. H.Versteeg, W.Malalasekra, An Introduction to Computational Fluid
Dynamics - The Finite Volume Method, Second edition, Pearson,
2013.
14. Jiyuan Tu, Guan Heng Yeoh, Chaoqun Liu, Computational Fluid
Dynamics-A Practical Approach, Second edition, Butterworth-
Heinemann Publisher, 2007.
15. David C. Wilcox, Turbulence Modeling for CFD, Second print, DCW
Industries, 1994.
16. Yunus A. Cengel, John M. Cimbala, adapted by S Bhattacharyya,
Fluid Mechanics-Fundamental and Applications, 7th reprint, Tata
McGraw Hill, 2009.
17. John D. Anderson Jr., Computational Fluid Dynamics-The Basics with
Applications, McGraw Hill.
18. NPTEL online CFD lectures, http://nptel.iitm.ac.in/courses/112105045/1

Page | 59