You are on page 1of 92

Investigating the rear surface

contamination on an automotive
bluff body when changing the rear
base plane geometry

MSc Mechanical Engineering

Faculty of Computing, Engineering and the Built Environment

School of Engineering and Built Environment

Student Name:

Student Number:

Submission Date:

Supervisor: Jens Lahr

Dissertation submitted to Birmingham City University in partial fulfilment of the


requirements for the degree of Master of Science
Abstract
Abstract

The automotive industry has advanced technologically over the years with the
adaptation of LiDAR systems and reversing vehicle cameras. Whilst these
technologies have developed in their complexity and performance, the problem of
surface contamination has grown alongside this, impeding the device’s performance
in many cases. Surface contamination, on SUVs in particular, is becoming an ever
growing area of aerodynamic research alongside the popular desire to reduce the drag
on vehicles. This project aims to reduce the effects of surface contamination and
simultaneously reduce the aerodynamic drag through the implementation of boat-tail
geometry on the rear of the vehicle, in the form of small chamfers. In industry, boat-tail
geometry has been proven to reduce drag by up to 14% and thereby reduce fuel
consumption by 7%.
This investigation will be conducted using a series of steady-state CFD analyses,
using a Windsor model with a freestream velocity of 30.5𝑚𝑠 −1 . The road spray was
replicated using a particle injector region where water particles with a diameter of
10.6𝜇𝑚 were injected at a velocity of 15.2𝑚𝑠 −1 at the base of the rear of the body and
a particle tracking software was able to plot the particle paths through the domain.
As the chamfer angle was increased from 0° to 30° at 5° intervals, the water
averaged volume fraction per unit area on the rear base plane was calculated as well
as the aerodynamic forces. This project found that the optimal aerodynamic
performance and also minimal surface contamination was at a chamfer angle of 20°.

ii
Acknowledgements
Acknowledgements

I would like to express my gratitude to Dr Jens Lahr for his support, guidance, and
expertise throughout the duration of this project.

I would like to extend my thanks to Professor Adrian Gaylard and Dr Dominic Flynn
from Jaguar Land Rover for their help and knowledge from within the automotive
industry regarding computational fluid dynamics and automotive surface
contamination.

iii
Table of Contents
Table of Contents

1 Introduction ........................................................................................................ 1

1.1 Background and problem definition ............................................................... 1

1.2 Scope of the project ...................................................................................... 2

1.3 Research Question ....................................................................................... 3

1.4 Aim and objectives ........................................................................................ 3

1.5 Rationale ....................................................................................................... 3

2 Literature Review ............................................................................................... 5

2.1 Automotive bluff bodies ................................................................................. 5

2.2 Injector locations ........................................................................................... 6

2.3 Boat-tail geometry in automotive aerodynamics ........................................... 8

2.4 Automotive surface contamination .............................................................. 13

2.5 Particle Parameters..................................................................................... 14

3 Project Management ........................................................................................ 17

3.1 Research Methodology ............................................................................... 17


3.1.1 Philosophy ........................................................................................... 17
3.1.2 Approach ............................................................................................. 18
3.1.3 Methodology ........................................................................................ 18
3.1.4 Strategy ............................................................................................... 18
3.1.5 Timescale ............................................................................................ 18
3.1.6 Technique and procedure .................................................................... 18

3.2 Project Timeframe ....................................................................................... 19

4 Methodology ..................................................................................................... 20

4.1 Bluff Body Geometry ................................................................................... 20

4.2 CFD Domain Configuration ......................................................................... 20

4.3 Meshing Parameters ................................................................................... 22

4.4 CFD Boundary Conditions .......................................................................... 25

iv
Table of Contents
4.5 Lagrangian Particle Tracking ...................................................................... 27

4.6 CFD Post Processing .................................................................................. 27

5 Design Development ........................................................................................ 30

6 Results and Analysis ....................................................................................... 32

6.1 Mesh sensitivity results ............................................................................... 32


6.1.1 Coefficient of pressure on the control body .......................................... 32
6.1.2 Streamlines in the wake of the control body ........................................ 36
6.1.3 Aerodynamic forces acting on the control body ................................... 40

6.2 Surface contamination on the control model using a point cone ................. 41

6.3 Control model results using the full cone nozzle ......................................... 44

6.4 Boat tail geometry implementation .............................................................. 47


6.4.1 Aerodynamic forces when implementing boat-tail geometry ................ 47
6.4.2 Streamlines when implementing boat-tail geometry............................. 50
6.4.3 Surface contamination when implementing boat-tail geometry ............ 52

6.5 Surface Contamination change with different particle densities .................. 59

7 Discussion ........................................................................................................ 62

8 Conclusion........................................................................................................ 65

9 Recommendations and Further work ............................................................. 67

10 References .................................................................................................... 69

11 Appendix ....................................................................................................... 72

11.1 Project Proposal .......................................................................................... 72


11.1.1 Introduction .......................................................................................... 72
11.1.2 Initial Literature Review ........................................................................ 74
11.1.3 Project Management and Methodology ............................................... 77
11.1.4 Conclusion ........................................................................................... 79
11.1.5 References........................................................................................... 80

v
List of Figures
List of Figures
Figure 1: The rear soiling on an production SUV (Davis, 2017) .................................. 1
Figure 2: 2019 Hyundai Santa Fe backup camera covered in raindrops (Linkov, 2019)
............................................................................................................................. 2
Figure 3: Ultra-low carbon trailer for Waitrose and Partners (Gray, 2013) .................. 4
Figure 4: The Windsor Body Geometry (Kabanovs, et al., 2016) ................................ 5
Figure 5: The Windsor body geometry with wheels (Kabanovs, et al., 2019) .............. 5
Figure 6: Injector set up in: (A) Kabanovs, et al., (2016) and (B) Kabanovs, et al., (2019)
............................................................................................................................. 6
Figure 7: 90° cone injector at the rear of a Windsor Body (Kabanovs, et al., 2017b) .. 7
Figure 8: The surface contamination shown with three different injector locations
(Gaylard, 2019) .................................................................................................... 8
Figure 9: Velocity field contours when changing the boat-tail length and boat-tail slant
angle (𝛽) (Hassan, et al., 2018)............................................................................ 9
Figure 10: Windsor geometry with wheels and coordinate system. All measurements
are in mm (Urquhart, et al., 2021) ........................................................................ 9
Figure 11: Dimensions of the: (A) generic SUV model (B) and boat-tail plates
(Krishnani, 2009) ................................................................................................ 10
Figure 12: The pressure contour on the centre plane of the: (A) control body (B) body
with the top and bottom boat-tail angles of 10° to the horizontal (Krishnani, 2009)
........................................................................................................................... 10
Figure 13: The velocity streamlines on the centre plane of the: (A) control body (B)
body with the top and bottom boat-tail angles of 10° to the horizontal (Krishnani,
2009) .................................................................................................................. 10
Figure 14: Windsor Body with: (A) Side Tapering (B) Roof Tapering (Varney, et al.,
2017) .................................................................................................................. 11
Figure 15: The effect of side tapers on the: (A) coefficient of drag (Perry, et al., 2015),
(B) the change in coefficient of drag (Varney, et al., 2017) ................................ 11
Figure 16: The wake structure when changing the small chamfer angles to the
horizontal: (A) 0°; (B) 4°; (C) 12° and (D) 20° (Littlewood & Passmore, 2010) ... 12
Figure 17: Particle diameter distributions at different output pressures from the BETE
MW105 nozzle (Kabanovs, et al., 2016)............................................................. 15
Figure 18: Population of droplets captured at each distance from the nozzle
(Kabanovs, et al., 2017b) ................................................................................... 15

vi
List of Figures
Figure 19: (A) The particle droplet diameter distribution for each nozzle set up and (B)
the particle injector velocity for each nozzle set up ............................................ 16
Figure 20: Saunders' Research Onion layer breakdown (Saunders, et al., 2019) ..... 17
Figure 21: The initial Gantt chart for the project ........................................................ 19
Figure 22: An isometric view of the bluff body ........................................................... 20
Figure 23: A labelled diagram of the domain and bodies of influence on the ZX plane
........................................................................................................................... 21
Figure 24: A diagram of the domain size in terms of 𝐿 along the X axis .................... 21
Figure 25: The different mesh refinement regions and assigned mesh sizing ........... 23
Figure 26: The meshed domain before the Inner BOI and rear base plane meshing
refinement is implemented ................................................................................. 24
Figure 27: The final mesh refinement region ............................................................. 24
Figure 28: (A) The schematic diagram for the nozzle set up and (B) the point cone
nozzle ................................................................................................................. 25
Figure 29: The full cone nozzle set up ....................................................................... 25
Figure 30: The chamfer edges on the rear base plane .............................................. 30
Figure 31: The chamfer length and angle of the boat-tailing shown on the rear of the
Windsor Body ..................................................................................................... 30
Figure 32: The boat-tailing chamfers on the Windsor body at (A) 5°; (B) 10°; (C) 15°;
(D) 20°; (E) 25° and (F) 30° ................................................................................ 31
Figure 33: 𝐶𝑃 distribution along the ZX plane on the control body in the X axis ........ 33
Figure 34: 𝐶𝑃 distribution along the centre of the rear base plane in the Z axis ........ 34
Figure 35: The coefficient of pressure distribution on the rear base plane of (A) the
control model and (B) the steady-state results in the literature (Kabanovs, et al.,
2016) .................................................................................................................. 35
Figure 36: URANS coefficient of pressure distribution (Kabanovs, et al., 2016) ....... 35
Figure 37: The wake structure at different element sizes .......................................... 37
Figure 38: Collapsed lower vortex from the 1 𝑚𝑚 mesh parameter and ................... 38
Figure 39: The particle tracking through the lower vortex cores with different bluff body
geometries (Kabanovs, et al., 2017a) ................................................................ 38
Figure 40: The vortex cores for each meshing parameter and without the mesh
refinement .......................................................................................................... 39
Figure 41: The wake structure for (A) the control model and (B) the literature (Gaylard,
et al., 2017a) ...................................................................................................... 39

vii
List of Figures
Figure 42: The change in 𝐶𝐷 against the rear element size ...................................... 40
Figure 43: The surface contamination spread using a point cone nozzle at the different
meshing parameters .......................................................................................... 42
Figure 44: Deposition pattern from (A) the point cone on the control model and (B) the
literature (Gaylard, et al., 2017a) ........................................................................ 43
Figure 45: The surface contamination spread using a point cone nozzle at 5° chamfer
angles................................................................................................................. 44
Figure 46: The surface contamination spread on the rear base plane of the control
model using a full cone nozzle ........................................................................... 45
Figure 47: Fraction of water as a percentage on the rear base plane from (A) RANS
CFD techniques, (B) URANS CFD techniques and (C) experimental testing ..... 45
Figure 48: Deposition pattern from (A) the full cone on the control model and (B) the
literature (Gaylard, et al., 2017a) ........................................................................ 46
Figure 49: The change in 𝐶𝐷 against the chamfer angle ........................................... 47
Figure 50: The change in 𝐶𝐿 against the chamfer angle ........................................... 48
Figure 51: Lift-to-drag ratio plotted against the chamfer angle .................................. 49
Figure 52: A comparison between 𝐶𝐷 values when changing 𝛼, between the
chamfered model and Perry, et al., (2015) ......................................................... 50
Figure 53: Streamlines on the central plane for each chamfer angle from 𝛼 = 0° to 𝛼 =
30° ...................................................................................................................... 51
Figure 54: Streamlines on the central plane ranging from 𝛼 = 15° to 𝛼 = 20° .......... 52
Figure 55: Water Averaged Volume Fraction per 𝑚2 at different values of 𝛼 ............ 53
Figure 56: Surface contamination on the rear base plane from 𝛼 = 5° to 𝛼 = 30° .... 54
Figure 57: A visualisation of the change in particle path as the chamfer angle increases
........................................................................................................................... 55
Figure 58: Surface contamination on the rear base plane from 𝛼 = 16° to 𝛼 = 19° .. 56
Figure 59: Water Averaged Volume Fraction per 𝑚2 at when 15° ≤ 𝛼 ≤ 20°............ 57
Figure 60: Particle paths at 𝛼 = 15° and 𝛼 = 20° ...................................................... 58
Figure 61: (A) Streamlines on the centre plane with 𝛼 = 30°, (B) particle path with 𝛼 =
30° ...................................................................................................................... 58
Figure 62: A comparison in surface contamination between (A) the bluff body with 𝛼 =
5° and (B) a production vehicle .......................................................................... 59
Figure 63: A comparison between the surface contamination spread between (A)
muddy water and (B) pure water ........................................................................ 60

viii
List of Figures
Figure 64: Particle paths for the (A) rear view of the mud particles, (B) rear view of the
water particles, (C) isometric view of the mud particles and (D) isometric view of
the water particles .............................................................................................. 61

ix
List of Tables
List of Tables
Table 1: The domain and bodies of influence dimensions ......................................... 21
Table 2: The mesh sensitivity test parameters .......................................................... 23
Table 3: The number of elements and node for the final control mesh ...................... 24
Table 4: The different design variation parameters ................................................... 31
Table 5: Comparison between the aerodynamic lift and drag coefficients between the
project results and the existing literature ............................................................ 41
Table 6: Aerodynamic lift and drag coefficients for domain lengths of 8𝐿 and 10𝐿 .... 41
Table 7: Averaged volume fraction for different particles .......................................... 60

x
Nomenclature
Nomenclature

Roman Symbols
Symbol Description Unit
𝐴 Cross-sectional area 𝑚2
𝐵 Blockage ratio -
𝐶𝐷 Coefficient of drag -
∆𝐶𝐷 Change in the coefficient of drag -
𝐶𝐿 Coefficient of lift -
∆𝐶𝐿 Change in the coefficient of lift -
𝐶𝑃 Coefficient of pressure -
𝐶𝑣 Coefficient of velocity -
𝑑𝑝 Particle diameter 𝜇𝑚
𝐹𝑥 Drag force 𝑁
𝐹𝑍 Lift force 𝑁
𝐿 Length of the bluff body 𝑚
𝑚̇ Mass flow rate 𝑘𝑔𝑠 −1
𝑃 Pressure 𝑃𝑎
𝑃∞ Static freestream pressure 𝑃𝑎
𝑅𝑒𝐿 Length based Reynold’s number -
𝑅𝑒𝑝 Particle Reynold’s number -
𝑆𝑡𝑘 Stoke’s number -
𝑣 Velocity 𝑚𝑠 −1
𝑣𝑝 Particle velocity 𝑚𝑠 −1
𝑣∞ Freestream velocity 𝑚𝑠 −1

Greek Symbols
Symbol Description Unit
𝛼 Chamfer angle °
𝜇𝑓 Dynamic viscosity of the freestream fluid 𝑘𝑔𝑚−1 𝑠 −1
𝜌𝑓 Density of the freestream fluid 𝑘𝑔𝑚−3
𝜌𝑝 Particle density 𝑘𝑔𝑚−3
𝜔 Particle injector spray angle °

xi
Nomenclature
Abbreviations
Letters Description
AVF Averaged volume fraction
BOI Body of influence
CFD Computational fluid dynamics
LES Large eddy simulations
LiDAR Light Detection and Ranging
PIV Particle image velocimetry
RANS Reynold’s-averaged Navier-Stokes
URANS Unsteady Reynold’s-averaged Navier-Stokes

xii
Introduction

1 Introduction

1.1 Background and problem definition


In automotive aerodynamics, the rear surface contamination is an important
consideration for many reasons. A build-up of soiling on the rear of a vehicle can come
in many forms, such as snow, water or other particulates and can cause a degradation
in the driver’s rear visibility or even the lack of visibility of the rear number plate or even
brake lights (Gaylard, et al., 2017b). This is mainly seen in the case of square back
SUVs (Jilesen, et al., 2017), it is also visualised in figure 1.

Figure 1: The rear soiling on an production SUV (Davis, 2017)

In more recent times, technology has advanced, including reversing sensors


and cameras, which in turn will bring more challenges concerning surface
contamination (Kabanovs, et al., 2019) (Jilesen, et al., 2017). According to Gov.uk, you
can receive a one thousand pound fine for having unreadable licence plates and can
incur a one hundred pound on the spot fine for this offence. This is just one reason
why the decreasing the surface contamination spread can ensure you are safe when
driving on the road.
Figure 2 shows how the vision of the driver can be impaired from surface
contamination and how the driver will be affected by this.

1
Introduction

Figure 2: 2019 Hyundai Santa Fe backup camera covered in raindrops (Linkov, 2019)

The vision in the reversing camera in figure 2 is significantly affected and therefore
presents a health and safety issue for the driver practically. Although the technology is
designed to help the driver park, issues can stem from external factors that are
potentially not considered during the research and development stage.

Linkov (2019) discussed that the Cadillac CT6 Rear-Camera Washer can be
implemented in order to clean the cameras and improve visibility, but this project takes
a more preventative and less resource heavy approach when considering this issue.
Although this technology can clean the camera, if the geometry can be optimised so
that the camera does not become dirty in the first place, then this is a more efficient
process to employ.

1.2 Scope of the project


The project will use steady-state computational fluid dynamics (CFD) techniques in
order to investigate how the wake structure will change when varying the boat-tail
chamfers on the back of the vehicle, this in turn will affect how the soiling particles will
be carried in the airflow. A control body will initially be used to make sure the model is
representative of the literature and assumptions will be made for simplicity and to
ensure computational efficiency. Small chamfers will be placed on the rear edges of
the body determined by their length in the X axis and the adjacent angle (𝛼). 𝛼 will vary
from 0° to 30° in increments of 5°. The results taken from the numerical analysis will
be compared to that of the literature. Relevant conclusions will be drawn and an optimal
chamfer angle can be taken when considering the surface contamination spread and
the aerodynamic lift and drag forces.

2
Introduction
1.3 Research Question
What is the effect of implementing boat-tail geometry on the aerodynamic drag and
surface contamination on the rear of an automotive body?

1.4 Aim and objectives


The aim of the project will be to investigate how the rear base plane geometry of an
automotive body will affect the aerodynamic drag and the rear surface contamination
using CFD.

The objectives that will be achieved throughout the course of the project are as
follows:
1. Identify the aerodynamic forces acting on a control automotive bluff body
using CFD techniques.
2. Understand how the aerodynamic flow in the wake will influence the rear
surface contamination through a particle tracking analysis.
3. Conduct steady-state CFD analyses, monitoring the aerodynamic drag
when implementing boat-tail chamfers on the rear of the body.
4. Analyse the rear surface contamination spread on the modified bluff
body using CFD techniques.
5. Evaluate the change in surface contamination and aerodynamic drag at
different boat-tail chamfer angles.
6. Determine the optimal rear geometry parameters required in order to
reduce the aerodynamic drag and minimise rear surface contamination.

1.5 Rationale
Through observing the structure of the wake flow, the path in which the particles can
be carried onto the rear base plane can be determined. Therefore, altering this
behaviour will change the spread in surface contamination. The bodies geometry will
play a vital role in manipulating the wake structure and therefore changing the
aerodynamic behaviour. Using a qualitative analysis of the particle deposition, key
areas can be monitored when considering driver’s visibility.

By altering the wake structure of the aerodynamic flow, this will have an innate effect
on the aerodynamic drag acting on the vehicle and can therefore have a direct effect
on the efficiency of the vehicle, both environmentally and economically. Gray and
Adams were able to design ultra-low carbon lorry trailers for Waitrose and Partners

3
Introduction
back in 2013, using boat-tailing geometry on the top edge of the trailer. This is shown
in figure 3.

Figure 3: Ultra-low carbon trailer for Waitrose and Partners (Gray, 2013)

This meant that they were able to achieve a drag reduction of 14% and this led to
a decrease in fuel consumption by 7%. More orders of this type of trailer were
commissioned later for 2015 to 2017 (Gray, 2013). Although this provides a
performance-based enhancement of the vehicle, it means that by decreasing the drag,
can also have an economical affect when implemented in industry.

4
Literature Review

2 Literature Review

2.1 Automotive bluff bodies


Different bluff bodies have been in the surrounding literature, however, the most
common model used is the Windsor body, shown in figure 4. The Windsor body
represents the key elements of an automotive body and have features proportional to
a small hatchback (Gaylard, et al., 2017a) but is also scalable to represent the square
back features of modern-day production SUVs (Gaylard, 2019).

Figure 4: The Windsor Body Geometry (Kabanovs, et al., 2016)

This type of geometry is really shown to demonstrate the fluid flow strictly over the
body of the vehicle without wheels. There are adaptations of the Windsor body with
wheels that have investigated the surface contamination with the implementation of
tyre geometry (Kabanovs, et al., 2019) as demonstrated in figure 5.

Figure 5: The Windsor body geometry with wheels (Kabanovs, et al., 2019)

There is a strong collaboration within the literature reviewed in recent times, mainly
between Gaylard and Kabanovs, however, Le Good & Gary, (2004), investigates
multiple automotive bluff bodies in detail and does not review the Windsor body, but
instead reviews the Ahmed body and the fundamental fluid flow surrounding the body.
There are clear similarities between the Ahmed body and the Windsor body but from
direct observation, the Windsor body reflects an SUV to a closer extent than the Ahmed
body.

5
Literature Review
The use of simplified automotive bodies reduces computational cost and time, whilst
generating highly representative results (Gaylard, 2019). This is why production
vehicle models are rarely used for generalised analyses but there are case studies
found in the literature where they are required for detailed aerodynamic developments.
An example can be found in Gaylard, et al., (2017b).
The surface contamination shown in this study is demonstrated using a positivist
approach and the context in which the spread in soiling is discussed to little depth. The
film thickness is often analysed but the lack of comparison means that the optimal
performance is not defined numerically but instead can be deemed to be contextual.
This may be due to the generalisation of the study and the simplicity of the geometry.
The interpretation of the surface contamination is discussed with a greater surrounding
context of the location of the rear soiling in Jilesen, et al., (2017).

2.2 Injector locations


The particle inlet parameters are extremely important aspects of the analysis to
consider and there are various injector locations and parameters found within the
literature. Kabanovs, at al., (2016) looks at how the injector is modelled physically using
a BETE MW105 nozzle for their experimental set-up and then aims to replicate this in
the CFD set-up. The main design considerations for the nozzle is the angle of which
the injector will be positioned and the radial plane of which the particles will spray (𝜔)
in figure 6(A). Kabanovs, et al., (2016) uses 70°, shown in figure 6(A) but figure 6(B)
shows how in Kabanovs, et al., (2019) 𝜔 = 90°.

(A) (B)
Figure 6: Injector set up in: (A) Kabanovs, et al., (2016) and (B) Kabanovs, et al., (2019)

It is also important to note that the injectors are not positioned at the same location
but in the literature this is due to the difference in bluff bodies used. As there are wheels
implemented into the geometry in figure 6(B), the location of the nozzle must be moved
back and placed centrally behind the wheels, in order to replicate how the spray comes
off the tyres tangentially to the surface of the tyre, this is why 𝜔 = 90° (Kabanovs, et

6
Literature Review
al., 2019). This injector definition is also used in Kabanovs, et al., (2017b) – shown in
figure 7, and Jilesen, et al., (2017).

Figure 7: 90° cone injector at the rear of a Windsor Body (Kabanovs, et al., 2017b)

This shows how the work of Kabanovs, and the corresponding authors take an
inductive approach as they use proven techniques to build upon. This is proven in the
papers published by the key authors in this field in 2016, 2017 and 2019, where the
results have been considered valid as the same set-up has been used over all articles
but could in fact be deemed as biased.
Looking at the surface contamination on the Ahmed body using snow as the soiling
particle, the injector set-up is defined differently because of the dynamic behaviour of
the snow falling from the top of the domain has to be defined in a different way that
soiling from road spray (Bangalore Narahari & Bharadhwaj, 2021). This type of particle
injector is called a grid-type injector, where the snow will travel through the freestream
velocity before hitting the vehicle and therefore will produce far different results than
those from the rest of the literature.
Gaylard (2019) makes the link between how different injector properties must be
used for different automotive bluff bodies displayed in figure 8, although they still
maintain the injector set-up used in Kabanovs, et al., (2016).

7
Literature Review
Figure 8: The surface contamination shown with three different injector locations (Gaylard,
2019)

Figure 8 shows how when the model uses stilts to support the bluff body, the injector
must sit in the centre plane, otherwise will not be able to generate an accurate
deposition pattern but must sit behind the wheels if they are implemented into the
design. Gaylard, (2019) compares these results of the film thickness to those of
Kabanovs, et al., (2016) in saying that they agree but this could be deemed a biased
stance for the author to take as Gaylard was a co-author for Kabanovs, et al., (2016).
This does not suggest that it impairs the validity of the results, but it could be argued
that external validation would strengthen the results but as this is a new branch of
aerodynamic research, the literature does not yet broadly exist for comparison.
Importantly, these results are backed up by a primary collection of experimental data
using the same assumptions across the literature, which plays an important role in
proving that the data is representative of reality.
The literature shows many examples of the injector location being at the start of the
wake, more development is required when injecting the soiling particles in the onset of
the fluid flow (Schembri Puglisevich, et al., 2016).

2.3 Boat-tail geometry in automotive aerodynamics


Boat-tail geometries have been used in many aerodynamic applications over the
years and have played a large role in drag reduction especially when considering larger
vehicles (Hassan, et al., 2018). Hassan, et al., (2018) shows that the implementation
of the boat-tail structures can result in a reduction in drag by approximately 50% and
looks directly at how the crosswind can affect the vorticity in the wake of the airflow.
However, when using a production vehicle model, the coefficient of drag does not
necessarily reduce, when implementing a rear boat-tail, unless a smooth undercover
is used (Choy, et al., 2021). Hassan, et al., (2018) also investigates how the wake
structure and flow separation is affected when changing the boat-tail length, shown in
figure 9, where the length of the plate is measured in meters.

8
Literature Review

Figure 9: Velocity field contours when changing the boat-tail length and boat-tail slant angle
(𝜷) (Hassan, et al., 2018)

Krishnani, (2009) investigates the effect of implementing top and bottom boat-tail
planes at different angles, creating a cavity. This was also explored by Urquhart, et al.,
(2021), who looked at this same cavity-based mechanism but with other renditions as
shown in figure 10.

Figure 10: Windsor geometry with wheels and coordinate system. All measurements are in
mm (Urquhart, et al., 2021)

Urquhart, et al., (2021), looked at a horizontal measurement of 50 𝑚𝑚 as shown in


figure 10, however, Krishnani, et al., (2009) used a shorter horizontal measurement of
18.75𝑚𝑚 shown in figure 11(A). This was also because the model was far smaller than
the Windsor model at a total length of 432 𝑚𝑚, a 1:12 scale model.

9
Literature Review
(A) (B)
Figure 11: Dimensions of the: (A) generic SUV model (B) and boat-tail plates (Krishnani,
2009)

Krishnani, et al., (2009) found that the model which produced the least drag was
when both the top and bottom plate were positioned 10° to the horizontal. This is shown
in figure 12(B).

(A)

(B)

Figure 12: The pressure contour on the centre plane of the: (A) control body (B) body with
the top and bottom boat-tail angles of 10° to the horizontal (Krishnani, 2009)

Figure 12 shows the pressure scale in the surrounding fluid flow and demonstrates
how the implementation of the boat-tail plates definitely affects the wake structure. This
is shown in greater detail in figure 13(B) where the velocity streamlines are displayed.

(A) (B)

Figure 13: The velocity streamlines on the centre plane of the: (A) control body (B) body
with the top and bottom boat-tail angles of 10° to the horizontal (Krishnani, 2009)

By increasing the angle in which the boat-tail plates are to the horizontal, the higher
the velocity tends to be in the wake.

In contrast, Varney, et al., (2017) investigates the pressure and force coefficients
using ventilated rear geometries. This is geometry is shown in figure 14. This is the
same model that is used as Kabanovs, et al., (2017b). Both of these publications are
in association with Loughborough University so this could be deemed as a bias.

10
Literature Review

(A) (B)

Figure 14: Windsor Body with: (A) Side Tapering (B) Roof Tapering (Varney, et al., 2017)

Both roof and side tapers were investigated separately, at different angles to reduce
the aerodynamic drag. Looking at the side tapers it is reported that at an angle of 0° to
15°, the coefficient of drag reduced by ∆𝐶𝐷 = −0.025 at 15°. Varney, et al., (2017),
compares the results to Perry, et al., (2015) for validation with their value of ∆𝐶𝐷 =
−0.02. Although this is a 20% difference between the results, it shows a definite trend
when considering a similar investigation.

Figure 15(A) shows the relationship between the side taper angles and the effect on
the coefficient of drag at 0° yaw. This is the case for both Perry, et al., (2015) and
Varney, et al., (2017). This can be considered an inaccuracy in the methodology as
this is not representative of what would happen in reality. However, it is one of many
assumptions of the model and can be deemed a limitation.

(A) (B)

Figure 15: The effect of side tapers on the: (A) coefficient of drag (Perry, et al., 2015), (B) the
change in coefficient of drag (Varney, et al., 2017)

11
Literature Review
Both papers are published in the same Journal of Passenger Vehicles – Mechanical
Systems, in association with Loughborough University and both include Passmore, so
the correlation between the work done in both sets of literature will be biased. As
Varney, et al., (2017) cites Perry, et al., (2015) this highlights this biased. The validation
for both sets of results would be much stronger if both publications were from separate
institutions and different author collaborations.
Passmore is also a co-author of Littlewood & Passmore (2010), which is a similar
technical paper, exploring the optimisation of roof trailing edge geometry for drag
reduction. This is an adaptation of boat tailing but specifically looking at the top
chamfer. This is also significant as the geometry being used is a simple square back
Windsor Body rather than a simplified SUV body, meaning that the complexity of the
model is reduced in order to focus on the effect of a simple change rather than
focussing on how close the value of 𝐶𝐷 is to reality, but rather ∆𝐶𝐷 . It is found that the
optimal chamfer angle is 12° to the horizontal, shown in figure 16(C).

(A) (B)

(C) (D)

Figure 16: The wake structure when changing the small chamfer angles to the horizontal:
(A) 0°; (B) 4°; (C) 12° and (D) 20° (Littlewood & Passmore, 2010)

There is an obvious visible change in the wake structure when the chamfer angle
increases, in terms of the vortex cores. At a chamfer angle of 12°,the drag reduction is
4.4% and is achieved by changing the pressure configuration on the rear base plane

12
Literature Review
(Littlewood & Passmore, 2010). The credibility of this study is limited by the age of the
publication but, Passmore is in fact a co-author in Perry, et al., (2015) which also found
12° to also be an optimal chamfer angle with approximately 7.4% drag reduction, taken
from figure 15(A).

2.4 Automotive surface contamination


The majority of the research surrounding surface contamination is collaborative
work between a series of authors, mainly Gaylard, Kabanovs and Passmore. These
authors take a positivist approach to the results and use a deductive methodological
approach in their work. This could potentially cause a bias in their methodology and
therefore results. However, as this is a topic of automotive aerodynamics is not
researched into as much depth as others – such as, force manipulation mechanisms
on vehicles, it is important that the there is a strong foundation to build upon, in terms
of the numerical approaches and their corresponding assumptions.
Kabanovs, et al., (2017) discusses how the wake structure has a large effect of the
rear surface contamination and how the wake carries the different soiling particles onto
the rear base plane. This can be investigated experimentally and numerically using
CFD techniques (Gaylard, 2019) (Gaylard, et al., 2017a) (Gaylard, et al., 2017b)
(Kabanovs, et al., 2019) (Kabanovs, et al., 2017b) (Kabanovs, et al., 2016) (Schembri
Puglisevich, et al., 2016). As the majority of the literature looking at surface
contamination agrees on this, it suggests that CFD techniques are highly
representative of reality. The results in the literature, that use both experimental and
computational techniques, visualise the results clearly from the numerical methods.
This is accomplished using clear legends that enables the reader to view the results in
greater detail and it can quantify the film thickness with ease. It is much less efficient
to visualise this experimentally. For example, Gaylard, et al., (2017b) shows the
surface contamination physically by using laser light sheet illumination and ultra violet
dye, which is far harder to implement than using a post processor.
Gaylard, et al., (2017b), a review paper, discusses many aspects of automotive
surface contamination and the challenges that the industry is currently facing. It
highlights the fact that the surface contaminant used in the majority of research is water
but explains how this is not highly representative of the true nature of the problem and
how this is an arear that needs to be investigated further. Gaylard, et al., (2017b) is
one of few publications that discuss surface contamination on wing mirrors and
highlights Bannister (2000) who investigates how the wing mirror casing itself can
13
Literature Review
break up a spray and the airflow can cause surface contamination on the mirror face
itself and cause a blockage in vision. Bannister (2000) is a more intricate exploration
of surface contamination as it identifies a specific part of a vehicle. However, due to
the age of this publication, this limits the validity of the investigation as greater
improvement in methodological approaches have been refined in recent years.

2.5 Particle Parameters


When investigating particle tracking in CFD, it is important to recognise that the
particles must have defined parameters in order to reflect reality, one being the particle
diameter which can be determined using the Stoke’s Number which is defined in
Kabanovs. et al., (2016) and in equation 1.
𝜌𝑝 𝑑𝑝 2 𝑣∞
𝑆𝑡𝑘 = (1)
18𝜇𝑓 𝐿
where,
𝜌𝑝 is the density of the particle
𝑑𝑝 is the particle diameter
𝑣∞ is the free stream velocity
𝜇𝑓 is the dynamic viscosity of the free stream fluid
𝐿 is the characteristic length of the model

When 𝑆𝑡𝑘 ≫ 1 it is said that the particles will be unaffected by the fluid flow, when
𝑆𝑡𝑘~1 the particles will follow the fluid flow with approximately 1% error and when
𝑆𝑡𝑘 ≪ 1 the particle will follow the streamlines of the fluid flow, this could be
approximately 0.01 (Crowe, et al., 1988).
Kabanovs, et al., (2016) uses a variety of particle diameters and can be considered
the most in depth analysis of how the surface contamination is affected by the different
values of 𝑑𝑝 . Figure 17 shows when looking at the experimental set-up, using a
physical BETE MW105 nozzle, that you can use different pressures to inject the
particles into the airflow. At these different pressures, a different distribution of 𝑑𝑝 is
plotted. All showing the same bell shape, with the highest population of 19.58𝜇𝑚,
25.60𝜇𝑚 and 40.2𝜇𝑚 for 11𝑀𝑃𝑎, 4.5𝑀𝑃𝑎 1𝑀𝑃𝑎 respectively, meaning that the lower
the outlet pressure of the nozzle, the larger the particles will be.

14
Literature Review

Figure 17: Particle diameter distributions at different output pressures from the BETE
MW105 nozzle (Kabanovs, et al., 2016)

However, when calculated, this will mean that the Stoke’s number will be less than
1 for all of these cases but for the higher pressures, the particles will be smaller and
therefore follow the streamlines closer as the Stoke’s number is proportional to 𝑑𝑝 2 .
This is a large factor to consider when generating an accurate model.
Kabanovs, et al., (2017a) also looks closely at how the particle diameter distribution
affects how the surface contamination forms on the rear of the automotive body, but in
a different sense than Kabanovs, et al., (2017b), instead it discusses how the diameter
distribution changes the further from the nozzle the particles get, displayed in figure
18.

Figure 18: Population of droplets captured at each distance from the nozzle (Kabanovs, et
al., 2017b)

15
Literature Review
It is different from the rest of Kabanovs’ investigations as this article uses multiple
nozzles in the CFD analyses all set to different particle injector velocities and this has
the best way of replicating computationally what is happening as the particles leave a
real nozzle in the experimental approach. These velocities are shown in figure 19(B).

A B

Figure 19: (A) The particle droplet diameter distribution for each nozzle set up and (B) the
particle injector velocity for each nozzle set up

16
Project Management

3 Project Management

3.1 Research Methodology


When considering how the project will be conducted Saunders’ research onion,
shown in figure 20, shows how this can be broken down into a series of layers. Each
layer is considered and provides a structured pathway for the project and is often used
in surrounding literature as the benchmark for research methodology.

Figure 20: Saunders' Research Onion layer breakdown (Saunders, et al., 2019)

Each layer contains the different subcategories that will be considered within the
project and has then using each layer the research and methodology will become more
refined.

3.1.1 Philosophy
This project primarily uses a positivist philosophy. This means that the conclusions
drawn from the project will be taken from the data collected rather than the context of
the results. However, the context behind the data is an important consideration which
is classed as realism. This is because, when we look at the surface contamination, it
could be measured by the film thickness of the soiling, however, if the soiling is mostly
on the bodywork, this is not as detrimental as if all the soiling were to cover the rear
windscreen. This is why positivism should not be the only philosophy to take during
this investigation. Although the literature all uses positivism, it could be improved if the
quality of the data was a greater consideration, rather than just conducting a purely
quantitative study.

17
Project Management
3.1.2 Approach
A deductive approach will be used as there is a hypothesis that this project will be
working towards where the research implies that using boat-tail geometry will affect
the surface contamination. This approach is widely used in the literature and is mostly
applicable in most scientific investigations.

3.1.3 Methodology
A quantitative study will be conducted using one method. This is due to time and
resource constraints. If the project were to be conducted using a wider range of
resources such as Particle Image Velocimetry (PIV) technology, greater accuracy
could be achieved using a wind tunnel, however one method will be used and then
compared to the literature that have used such technologies for experimental validation
previously.

3.1.4 Strategy
The strategy taken from Saunders’ research onion will be an experimental strategy
as the data will be collected using an experimental technique, more specifically a
numerical study.

3.1.5 Timescale
The timescale refers to when the data will be collected and the chosen category for
this layer is the cross-sectional timescale, where the data will be collected using current
techniques and assumptions considered at this moment in time. This is opposed to a
longitudinal timescale where the data is collected over a long period of time which is
not applicable for this project.

3.1.6 Technique and procedure


The technique used for this project will be a primary CFD numerical study where the
results will be validated using those from the literature. Traditionally, CFD studies are
conducted and are verified using a primary wind tunnel experiment. However, as this
project aims to replicate a control model found within the literature that has been
previously validated using the author’s primary wind tunnel data, as long as the control
data is collected, the parametric study can also be considered to be true as long as the
model stays consistent. This is however a limitation that must be considered when
looking at the findings.

18
Project Management
3.2 Project Timeframe
The initial project timescale discussed in the proposal, detailed in the appendix,
estimated that the CFD analyses would take a total of three weeks to complete, but
failed to consider the time required to complete the mesh sensitivity testing. This has
had a detrimental effect on the structure of the project and meant that a secondary
parameter could not be investigated as a result of this. The secondary parameter to be
investigated was how the yaw angle of the bluff body would affect the way in which the
surface contamination would deposit on the rear base plane. Figure 21 highlights the
section of the project where the CFD analyses were planned to happen.

Figure 21: The initial Gantt chart for the project

In reality the section highlighted in the red box in figure 21 was much broader than
expected and needed to include a section for the mesh sensitivity test and extraction
of the figures from the post-processor.
In contrast, the time allocated for the literature review allowed for an extensive time
period to suitably select and filter the papers required to achieve a well-rounded study
and enabled a sufficient array of literature to be reviewed and critically analysed in
order to capture the most suitable methodologies and parameters that should be used
for this study.

19
Methodology

4 Methodology

4.1 Bluff Body Geometry


The automotive bluff body used in this study will be the Windsor Body dimensions
from Kabanovs, et al., (2016). This will be generated using DesignModeller inside the
ANSYS Inc Workbench application. This software allows for parametric changes to be
made within the workbench and will mean the rear geometry can be changed without
having the generate multiple CAD models externally. Figure 22 shows the Windsor
Body model generated in DesignModeller.
𝟎. 𝟏𝟗𝟒𝟓𝒎

𝟎. 𝟐𝟖𝟗𝒎

𝟏. 𝟎𝟒𝟒𝒎

Figure 22: An isometric view of the bluff body

The model is cut in half in the ZX plane in order to reduce computational cost and
improve time efficiency due to the model being symmetrical in this plane. This is
compensated for within the set-up, where a symmetrical boundary condition can be
applied to ensure no data is lost because of this.

4.2 CFD Domain Configuration


A domain is constructed around the body for the air to flow through, as well as a
two-stage wake-refinement region. This is so that the two wake refinement regions can
be meshed to a finer degree as bodies of influences and can solve to a higher degree
of accuracy in these areas. Figure 23 shows these regions in respect to the Windsor
Body, also referred to as the ‘Car Body’.

20
Methodology

Domain

Inner BOI Car


Body
BOI

Figure 23: A labelled diagram of the domain and bodies of influence on the ZX plane

6𝐿

1.5𝐿

0.5𝐿

8𝐿
𝑥=0

Figure 24: A diagram of the domain size in terms of 𝑳 along the X axis

The dimensions for the domain have been taken from Krishnani (2009). Although
this is a smaller domain than used in more industrial based articles discussed in the
literature review, it also shows that accurate results can also be obtained using this
smaller domain length, whereas the other domain dimensions are based on real-life
wind tunnels. As the domain is defined in terms of 𝐿 it makes this model scalable
therefore keeping the effect of the blockage ratio, minimal. Table 1 shows these
dimensions of the domain, BOI and inner BOI when 𝐿 = 1.044𝑚.

Table 1: The domain and bodies of influence dimensions

Length (x) Height (z) Width (y)


Body
[𝒎𝒎] [𝒎𝒎] [𝒎𝒎]
Domain 8352 2088 2088
BOI 1500 360 220
Inner BOI 500 360 220

A study was conducted in order to see if increasing the length of the domain to 10𝐿
had a prominent effect on the fluid flow. The domain was increased in front and behind
the body both by a factor of 𝐿, and it found that this had a minimal effect on the
aerodynamic forces beyond the dimensions in table 1. Therefore, an overall domain
length of 8𝐿 was suitable for this investigation.

21
Methodology
The blockage ratio (𝐵) is the measure of how much the bluff body is blocking the
domain and as such interfering with the airflow. 𝐵 is calculated using equation 2.
Blockage cross sectional area (2)
𝐵 = 100 ×
Domain cross section area
The blockage cross sectional area is calculated to be,
Blockage cross sectional area = (0.1945 × 0.289) + (0.05 × 0.0075) (3)
= 0.0566𝑚2
And the domain frontal area is,
Domain cross sectional area = 2.088 × 2.088 (4)
4.3597𝑚2

Therefore 𝐵 can be calculated to be,


0.0566 (5)
𝐵 = 100 × = 1.298%
4.3597

This is far lower than the blockage ratios used in the reviewed literature, ensuring
accuracy of the results as this will minimise the risk of the freestream velocity being
accelerated around the bluff body and affecting the surface pressures on the body.

4.3 Meshing Parameters

When considering the meshing, a preliminary starting point must be found before
considering a mesh sensitivity test. The predominant meshing method used within the
surface contamination literature is a hex-dominant method which creates a series of
elements aligned so that the streamlines can follow a more level path than when a
tetrahedral method is used. The mesh was constructed using the Mechanical software
from ANSYS Inc and the meshing software used was the default in the Mechanical.
A mesh sensitivity test must be conducted to ensure that the results generated in
the simulation are independent of the meshing parameters. Figure 25 shows the
different mesh sizing parameters used before considering the mesh sensitivity test.

22
Methodology

Domain body Rear base plane


sizing of 75𝑚𝑚 meshing parameter

Inner BOI
BOI sizing of Car body face
Meshing
15𝑚𝑚 sizing of 10𝑚𝑚
parameter

Stilts face sizing


of 1 𝑚𝑚

Figure 25: The different mesh refinement regions and assigned mesh sizing

The sections of the domain depicted in red show what sections of the model will be
changed in the mesh sensitivity test. This consists of the inner BOI and the rear base
plane. Table 2 shows the different element sizes that will be used in the mesh
independence test.

Table 2: The mesh sensitivity test parameters

Mesh sensitivity Rear base plane face sizing Inner BOI body sizing
test iteration [𝒎𝒎] [𝒎𝒎]

1 10 10
2 5 5
3 4 4
4 3 3
5 2 2
6 1 1

By keeping these element sizes the same for each iteration, this will eliminate the
chance of instabilities in the mesh being caused by the growth rate and will help keep
the hex dominant meshing stay in a lattice like structure.
Before this test takes place, the inner BOI will be meshed to the same degree as
the BOI itself with a body sizing of 15 𝑚𝑚, and the rear base plane will be meshed as
part of the car body face sizing of 10 𝑚𝑚. The preliminary mesh generated before the
mesh sensitivity test is shown in figure 26.

23
Methodology

Figure 26: The meshed domain before the Inner BOI and rear base plane meshing
refinement is implemented

The element number and nodes are shown in table 3. This is significant as the higher
the number of nodes and elements, the higher the degree of accuracy of the simulation.
However, this can come at a computational cost and impede the time efficiency of the
project.

Table 3: The number of elements and node for the final control mesh

Elements Nodes
1.153 × 106 1.042 × 106

These areas have been chosen as this is where the rear wake will sit, and it is
important that this area solves accurately in order to plot the streamlines and particle
tracking to a high level of detail for the streamlines and particle tracking. This is
highlighted in figure 27.

Figure 27: The final mesh refinement region

24
Methodology
4.4 CFD Boundary Conditions
When defining the boundary conditions for the CFD set up, the same freestream
velocity (𝑣∞ ) as Kabanovs, et al., (2016) is used since this literature is the closest
representation to the model used in this project. As 𝑣∞ is 30.5𝑚𝑠 −1 in Kabanovs, et al.,
(2016), the rest of the parameters must be taken from the same source as the particle
injector velocity (𝑣𝑝 ) must be of a similar ratio to reflect reality, this is why 𝑣𝑝 is 15.2𝑚𝑠 −1
at the same location and injector set up as shown in figure 6(A).

A B

Figure 28: (A) The schematic diagram for the nozzle set up and (B) the point cone nozzle

In order to simulate the particle injector, figure 28(B) shows a point nozzle being
used in order to represent the BETE MW105 nozzle used in Kabanovs, et al., (2016)
and Gaylard (2019), however this was proven to show an irregular spray pattern and
instead, a full nozzle must be used to represent what was happening during the
experimental set up that could not be reflected when using a point nozzle. Figure 29
shows the full nozzle being used with a radial plane of 0.189 𝑚𝑚 (Gaylard, 2019).

Figure 29: The full cone nozzle set up

25
Methodology
Using this value of 𝑣∞ the length based Reynold’s number for the fluid flow can be
determined and the turbulence of the airflow can be found using equation 6.
𝜌𝑓 𝑣∞ 𝐿 (6)
𝑅𝑒𝐿 =
𝜇𝑓 (NASA, 2014)

1.184 × 30.5 × 1.044


𝑅𝑒𝐿 = = 2.083 × 106
1.81 × 10−5

where the density and dynamic viscosity of air is taken at 25°𝐶 to be 1.184𝑘𝑔𝑚−3 and
1.81 × 10−5 𝑘𝑔𝑚−1 𝑠 −1 respectively, according to The Engineering Toolbox (n.d.).
Therefore, the model can be considered turbulent as 𝑅𝑒𝐿 > 20,000 (NASA, 2014).

The mass flow rate (𝑚̇) of the particles will also determine how the particles will
enter the model and is also defined using Kabanovs, et al., (2016) where it was taken
from the manufacturers specification. This will be 0.00171 𝑘𝑔𝑠 −1 when the particle
diameter is 10.6𝜇𝑚, which was found using the Stoke’s number equation discussed
during the literature review using equation 1.

18 ∙ 𝑆𝑡𝑘 ∙ 𝜇𝑓 ∙ 𝐿 (7)
𝑑𝑝 = √
𝜌𝑝 ∙ 𝑣∞

18 × 0.01 × 1.81 × 10−5 × 1.044 (8)


𝑑𝑝 = √ = 10.6𝜇𝑚
997 × 30.5

where the data is taken from air and water at 25°𝐶 (The Engineering Toolbox, n.d.).
In the literature, the CFD models are set up to replicate the wind tunnel experimental
set up, which means that the air is moving relative to the body instead of the body
moving through the air, which is how a car would move in reality. A limitation of this is
that the ground floor in the wind tunnel would be stationary and this would have an
effect on the aerodynamic flow beneath the bluff body and therefore affect the velocity
of the air beneath the body. In order to combat this, a moving ground wall will be
implemented underneath the body, where the velocity of the wall is defined using
Cartesian coordinates, where it will be moving at the same magnitude and direction as
𝑣∞ .
The outlet pressure is discussed in the literature where the relative pressure will be
0𝑃𝑎, meaning the pressure will not cause the freestream velocity to accelerate through

26
Methodology
the domain. A symmetry plane will be defined in the ZX plane, down the centreline of
the model as discussed previously.

4.5 Lagrangian Particle Tracking


Lagrangian particle tracking will be the numerical technique used in order to plot the
soiling particles accurately in the streamlines of the CFD. If this technique was not
employed, the particle velocities would be assumed to be equal to the freestream
velocity. As Lagrangian particle tracking acknowledges that there are drag and
buoyancy forces acting on the particles in the wake, among other forces that are
assumed to be negligible in this study (Xiao, et al., 2020), these will need to be
considered in the set-up of the particle tracking.
When considering the drag force acting on the particle as it travels through the fluid
flow, the Schiller-Naumann method can be used to determine what 𝐶𝐷 value must be
selected to correctly plot the particle tracking. This 𝐶𝐷 value is dependent on the
Reynolds number of the particle (𝑅𝑒𝑝 ). If 𝑅𝑒𝑝 > 1000 then 𝐶𝐷 = 0.44 (Crowe, et al.,
2011) (Crowe, et al., 1988). This can be determined by equation 9.
𝜌𝑝 𝑑𝑝 |𝑣 𝑖 𝑝 − 𝑣∞ |
𝑅𝑒𝑝 = (9)
𝜇𝑓
where 𝑣 𝑖 𝑝 is the particle velocity at any given iteration.
𝑣 𝑖 𝑝 is calculated in the CFD software by using the velocity value from the previous
iteration and can therefore calculate the next value using this iterative method (Fluid
Mechanics 101, 2020). As this is the only unknown variable for equation 9 at this point,
we can use the first value of 𝑣𝑝 as 15.2𝑚𝑠 −1 to calculate in equation 10, where the
particle diameter will be taken from section 4.4.
997 × 10.6 × 10−6 × |15.2 − 30.5|
𝑅𝑒𝑝 = = 8933
1.81 × 10−5 (10)
∴ 𝐶𝐷 = 0.44
The buoyancy forces will be considered using a density difference model, defined
in the set-up application.

4.6 CFD Post Processing


The Post processor application allows for complex results to be shown
simultaneously if necessary and was used to show and calculate the following:
• The aerodynamic lift and drag forces
• The streamlines of the airflow
27
Methodology
• The particle tracking streamlines
• The surface contamination spread on the rear of the body
• The pressure contours

The coefficient of lift (𝐶𝐿 ) and drag (𝐶𝐷 ) was used by generating two expressions
respectively shown below in equations 11 and 12.

2 ∙ 𝐹𝑍 (11)
𝐶𝐿 =
𝜌𝑓 ∙ 𝐴 ∙ 𝑣∞ 2
2 ∙ 𝐹𝑥 (12)
𝐶𝐷 =
𝜌𝑓 ∙ 𝐴 ∙ 𝑣∞ 2

(Domel, et al., 2018)

where,
𝐹𝑍 is the lift force
𝐹𝑥 is the drag force
𝜌𝑓 is the density of the fluid
𝐴 is the blockage cross sectional area

The coefficient of pressure (𝐶𝑃 ) and velocity (𝐶𝑣 ) will also be used in the post
processor. This is because in aerodynamics, dimensionless measures are generally
used in order to keep the models scalable and to easily make comparisons in the plots.
Both can be used for compressible and incompressible flows. The equations for 𝐶𝑃 and
𝐶𝑣 are given by the following equations,
𝑃 − 𝑃∞ (13)
𝐶𝑃 =
0.5 ∙ 𝜌𝑓 ∙ 𝑣∞ 2
𝑣 (14)
𝐶𝑣 =
𝑣∞

(Gudmundsson, 2014) (Merriam-Webster, n.d.)

where,
𝑃 is the measured pressure
𝑃∞ is the static freestream pressure
𝑣 is the measured velocity.

The surface contamination spread will also be calculated using a volume average
fraction of the water where it will be determined how much water will come into contact
with the rear base plane per 𝑚2 . This will give a general trend of where the particles

28
Methodology
will land on the rear base plane and how many will do so. In order to ensure the
particles are able to come to rest on the rear base plane, the perpendicular and parallel
restitution coefficients will be set to zero in order to take away the particles velocity
upon impact and will, in essence, stick to the rear of the body.

29
Design Development

5 Design Development
The chamfers will be applied to the rear edges highlighted in green in figure 30 to
ensure a full uniform boat-tail is applied to the rear. As there is a symmetrical plane
positioned down the centre plane of the Windsor Body, only the three faces need to be
selected.

Figure 30: The chamfer edges on the rear base plane

The trailing edges of the bluff body will be chamfered with a constant horizontal
length of 0.05 𝑚𝑚 as used in Urquhart, et al., (2021). Krishnani (2009) uses two
different horizontal lengths for the top and bottom chamfers, but as their model is a
1:12 scale model and the Windsor body is a 1:4 scale model, this will scale up to
approximately 50 𝑚𝑚 again, much like Urquhart, et al., (2021). The chamfer angle (𝛼)
shown in figure 31, will vary from 0° to 30° in increments of 5° as stated in section 1.2.
This will be the input parameter for the study.

Figure 31: The chamfer length and angle of the boat-tailing shown on the rear of the
Windsor Body

30
Design Development
Table 4 shows the different angles of 𝜶 used in this study for each design variation.

Table 4: The different design variation parameters

Chamfer Length 𝜶
Design Iteration
[𝒎] [°]

Control Body 𝟎 𝟎
1 𝟎. 𝟎𝟓 𝟓
2 𝟎. 𝟎𝟓 𝟏𝟎
3 𝟎. 𝟎𝟓 𝟏𝟓
4 𝟎. 𝟎𝟓 𝟐𝟎
5 𝟎. 𝟎𝟓 𝟐𝟓
6 𝟎. 𝟎𝟓 𝟑𝟎

The different chamfers are shown in figure 32 all with a chamfer length of 𝟎. 𝟎𝟓𝒎,
which can be represented as approximately 𝟎. 𝟎𝟒𝟖𝑳 for scalability.

A B

C D

E F

Figure 32: The boat-tailing chamfers on the Windsor body at (A) 5°; (B) 10°; (C) 15°; (D) 20°;
(E) 25° and (F) 30°

31
Results and Analysis

6 Results and Analysis

6.1 Mesh sensitivity results


Before looking at how the mesh affects the surface contamination on the control
body it is important to make sure that the fluid flow is refined to a high enough standard
that the particles can be introduced to the flow because this is what will determine the
path of the particles when incorporated into the model.
The results of the mesh sensitivity results were used in order to refine the mesh to
a sufficient degree where the results are independent of the mesh size. A topological
and numerical approach is taken so that the results can be fully visualised and
analysed to ensure a detailed enough mesh is selected and used in the dynamic
modelling further into the project.
After conducting the mesh sensitivity testing it was found that the element sizes of
1 𝑚𝑚 and 2 𝑚𝑚 solved without converging to an RMS value of U-Mom to 1 × 10−5
over a period of suitable iterations. This could be due to a number of issues in the
mesh but could mostly be down to the difference in mesh sizing between the BOI and
the Inner BOI from 15 𝑚𝑚 down to 1 𝑚𝑚 and 2 𝑚𝑚. As the growth rate is defaulted to
a scale of 1.2, this could cause large distortions in the mesh that will decrease the
element quality and thus lead to convergence issues. The results for 1 𝑚𝑚 and 2 𝑚𝑚
will be disbanded for this reason but will be shown in the results section for comparison.

6.1.1 Coefficient of pressure on the control body


Figure 33 shows 𝐶𝑃 across a polyline intersection between the car body and the
plane for different rear element sizes. It is important to look at the coefficient of
pressure because the aerodynamic forces are found using Bernoulli’s equation and if
the characteristic regions of low and high pressure change significantly, then so will
the behaviour of the airflow in these regions.

32
Results and Analysis

1.5

1.0

0.5

0.0
CP

-0.5

-1.0

-1.5

-2.0
0.00 -0.10 -0.20 -0.30 -0.40 -0.50 -0.60 -0.70 -0.80 -0.90 -1.00 -1.10
X Axis [m]

3 mm 4 mm 10 mm

Figure 33: 𝑪𝑷 distribution along the ZX plane on the control body in the X axis

The 𝐶𝑃 spread across the polyline shows vary little change but this is mostly due to
the nature of only changing the rear element size at −1.044 𝑚. Therefore, to get a
better representation of the change in 𝐶𝑃 in the mesh sensitivity region at the rear, the
coefficient of pressure must be found in the area of interest. Figure 34 shows the 𝐶𝑃
across the back face in the vertical direction on ZX plane.
The spike at the front of the body at X = 0 𝑚, is the high pressure when the air hits
the front of the body.

33
Results and Analysis

0.30

0.25
Z Axis [m]

0.20

0.15

0.10

0.05
-0.30 -0.25 -0.20 -0.15 -0.10 -0.05 0.00
CP

3 mm 4 mm 10 mm

Figure 34: 𝑪𝑷 distribution along the centre of the rear base plane in the Z axis

This is a more prominent change in the coefficient of pressure and although the do
not follow each other perfectly 10 𝑚𝑚 element size is still a greater outlier than the
3 𝑚𝑚 and 4 𝑚𝑚 element sizes, particularly at the bottom of the rear base plane as this
is where the particles will be injected into the airflow and will be carried away in the
wake. As expected the coefficient of pressure remains negative along this plane and
thus will induce a drag force as this pressure will pull the car against the movement of
the vehicle.
The trend in 𝐶𝑃 can then be investigated on a more topological point of view as
shown in figure 35, where the 𝐶𝑃 distribution on the back face for the element size of
3 𝑚𝑚 shows a direct correlation to that of the steady-state results in the literature. This
is a significant comparison because it is broadly discussed in literature that there are
many limitations to using Reynold’s-Averaged Navier-Stokes (RANS) analyses in
surface contamination due to the detail in the airflow not being captured (Kabanovs, et
al., 2016), but it does produce representative results for a simpler investigation. Ideally
Large Eddy Simulations (LES) would be ideal.

34
Results and Analysis

A B

Figure 35: The coefficient of pressure distribution on the rear base plane of (A) the control
model and (B) the steady-state results in the literature (Kabanovs, et al., 2016)

The trends are closely aligned to one another except in the magnitude of the 𝐶𝑃 . This
could be because the ground boundary for the control model uses a moving wall to
show the car moving relative to the ground in reality. However, as the distribution is
close to the literature at an inner BOI 3 𝑚𝑚 body sizing, this is an acceptable level of
detail for the investigation at this stage.
Transient or Unsteady Reynolds-Averaged Navier-Stokes (URANS) CFD analyses
are able to capture bilateral instabilities in the 𝐶𝑃 on the rear of the body as shown in
Kabanovs, et al., (2016), captured in figure 36.

Figure 36: URANS coefficient of pressure distribution (Kabanovs, et al., 2016)

Figure 36 accurately shows how using a half body model would not have been
suitable for a transient analysis because of these bilateral instabilities in the rear. As
transient analyses solve for a number of timesteps, but steady-state captures an

35
Results and Analysis
average result pattern. Thereby, transient analyses can capture more accurate results
but at the cost of computational and time efficiency.

6.1.2 Streamlines in the wake of the control body


The wake structure is a vital aspect of the analysis to get right. By changing the
meshing parameters particularly in the region of the wake, more detail can be captured
in the more important parts of the aerodynamic analysis. A topological analysis can be
used to visually compare the wake structure and vortex core locations and then the
coefficients of lift and drag can be compared as well, in order to see the mesh
independence as the element sizes decrease.
The streamlines generated in the post processor have been plotted on the centre
plane which is on the ZX plane. Looking at streamlines on a 2D plane makes it easier
to analyse the results comparatively against one another. Figure 37 shows the
structure of the wake where the colour of the streamlines is defined by 𝐶𝑣 in order to
show the magnitude of the acceleration of the air around the body.

36
Results and Analysis

𝟏𝟎 𝒎𝒎 𝟓 𝒎𝒎

𝟒 𝒎𝒎 𝟑 𝒎𝒎

𝟐 𝒎𝒎 𝟏 𝒎𝒎

Figure 37: The wake structure at different element sizes

The vortex structures in the wake of the car body are similar to one another as the
mesh size decreases, however, there are slight differences that are important to spot
when looking at the lower vortex cores. For the 1 𝑚𝑚 element size, although it was
decided that these results were discarded as not converged to the degree of accuracy
that has been decided for this project, the lower vortex core disappears and there is a
detachment that should not be present. This is highlighted in figure 38.

37
Results and Analysis

Figure 38: Collapsed lower vortex from the 𝟏 𝒎𝒎 mesh parameter and

The lack of vortex core in this region is significant because this aerodynamic
mechanism is responsible for carrying the majority of the soiling particles onto the rear
of the base plane. This is shown in Kabanovs, et al., (2017a) in figure 39.

Figure 39: The particle tracking through the lower vortex cores with different bluff body
geometries (Kabanovs, et al., 2017a)

Therefore, if this lower vortex structure collapses as it is shown on the 1 𝑚𝑚


streamlines, this will affect how the particles will be carries onto the rear base plane.
Figure 40 shows how the vortex cores have changed location for each meshing
parameter, this will show where the mesh independence lies. The 1 𝑚𝑚 and 2 𝑚𝑚
results have not been included.

38
Results and Analysis

3 𝑚𝑚
4 𝑚𝑚
5 𝑚𝑚
10 𝑚𝑚
15 𝑚𝑚

Figure 40: The vortex cores for each meshing parameter and without the mesh refinement

It is clear to see that there is very little variation in the upper vortex core and that
there is not too much variation in the location in the lower vortex cores too. Form a
purely observational point of view, 3 𝑚𝑚 and 4 𝑚𝑚 show to be similar, although there
is still not much variation.
Looking at the formation of the streamlines at the 3 𝑚𝑚 parameter, in figure 41(A),
there is a close similarity in the structure of the wake but there are some differences in
the set-up, such as the ground floor is moving at the Inlet Velocity for the control model
and is stationary in the literature. Also, the literature shows a URANS result rather than
RANS in this project, which as previously discussed means there will be unsymmetrical
behaviour in the wake.

A B

Figure 41: The wake structure for (A) the control model and (B) the literature (Gaylard, et al.,
2017a)

39
Results and Analysis
This is interesting as it seems that the moving ground seems to have a large effect
on the lower vortex, although the core locations are in the same place. The literature
would benefit from having the streamlines coloured as the velocity of 𝐶𝑣 in order to help
critically analyse these results in more detail.

6.1.3 Aerodynamic forces acting on the control body


The value of 𝐶𝐷 stabilises at around 4 𝑚𝑚 and 3 𝑚𝑚, at values of 0.2463 and 0.2448
respectively, this is shown in figure 42. As these values have a percentage difference
of 0.7%, the mesh independence can be shown at this point instead of a percentage
difference of 2% from 5 𝑚𝑚 to 3 𝑚𝑚. Although 3 𝑚𝑚 and 4 𝑚𝑚 show highly similar
values, due to the resolution on the rear of model, the smaller value will be used in
order to detail the results to a higher degree, but it is still true that the results are
independent of the mesh size at this element size.

0.285

0.28

0.275

CD 0.27
0.2664

0.265

0.26

0.255

0.25
0.2498

0.245
0.2463
0.2448
0.24
10 9 8 7 6 5 4 3
Element size [mm]

Figure 42: The change in 𝑪𝑫 against the rear element size

As this final value of 𝐶𝐷 = 0.2448 is in the right order, we can say that the
aerodynamic model is correct but must be compared to existing literature, when using
the same geometry, comparing the 𝐶𝐿 values too. This comparison can be seen in
table 5.
40
Results and Analysis
Table 5: Comparison between the aerodynamic lift and drag coefficients between the project
results and the existing literature

Source 𝑪𝑫 𝑪𝑳

Control Body 0.2488 −0.1255


Kabanovs, et al., (2016) 0.276 −0.1
Gaylard (2019) 0.286 −0.107

There is a difference between these results but although they are all using the same
bluff body, the control body does have the moving ground floor which will affect the
airflow underneath the vehicle, which could be the reason for the discrepancies.
However, there is an overall agreement in the values of both 𝐶𝐷 and 𝐶𝐿 , therefore we
can move forward using the 3 𝑚𝑚 rear element sizing on the inner BOI and rear base
plane.
When looking at the effect of the domain length on the aerodynamic drag and lift,
table 6 shows there is little variation between the results and so the domain length of
8𝐿 will be used.

Table 6: Aerodynamic lift and drag coefficients for domain lengths of 𝟖𝑳 and 𝟏𝟎𝑳

Domain length in terms 𝑪𝑫 𝑪𝑳


of 𝑳

8𝐿 0.2488 −0.1255
10𝐿 0.2438 −0.1221

For the coefficient of drag the percentage difference is only 2.01% and the
coefficient of lift percentage difference is 2.71%. Therefore, it can be said that by
changing the domain length, there would be very little effect on the lift and drag
coefficients.

6.2 Surface contamination on the control model using a point cone


The point cone was used to inject the particles into the simulation at the start of the
project before it was found that it did not produce the correct results and meant that
the particles were not dispersed in a manner that they had been in the CFD results in
the surrounding literature. Figure 43 shows the surface contamination spread using
the point cone for the different meshing parameters.

41
Results and Analysis

𝟏𝟎 𝒎𝒎 𝟓 𝒎𝒎

𝟒 𝒎𝒎 𝟑 𝒎𝒎

𝟐 𝒎𝒎 𝟏 𝒎𝒎

Figure 43: The surface contamination spread using a point cone nozzle at the different
meshing parameters

Compared to the literature, this is not what the results were expected to show. The
refined deposition patter that starts to emerge at 4 𝑚𝑚 is not directly affected by the
mesh independence as the deposition pattern is determined by the refinement of the
streamlines in the wake. As it was decided that the best rear element size to carry on
forwards with was the 3 𝑚𝑚 size, we dismiss the latter 2 𝑚𝑚 and 1 𝑚𝑚 results.
However, it is clear that when you decrease the mesh size on the rear base plane, this
improves the resolution. This can be classed as a limitation within the study as this

42
Results and Analysis
may increase the chances of having an instability in the mesh that has already been
proven to appear when reducing the mesh in this region.
Using the 3 𝑚𝑚 rear element sizing will help keep the resolution as detailed as
possible whilst also generating results which converge in an efficient manner whilst
maintaining accuracy. However, reflecting back on the literature, figure 44 shows that
there are clear differences when using the same model and boundary conditions when
comparing these deposition patterns to those of Gaylard, et al., (2017a).

A B

Figure 44: Deposition pattern from (A) the point cone on the control model and (B) the
literature (Gaylard, et al., 2017a)

Again, the literature shows results from a URANS analysis, however, this pattern
does not show up again in the literature for steady-state analyses. This means that
there is an error in the set-up as discussed earlier in section 4.4 where a full cone must
be used with particle dispersion in order to capture an accurate particle injection region.
Another important consideration is that the film thickness, shown in figure 44(B), can
only be achieved through conducting a transient analysis, because the film needs to
build up over a specified timeframe. Therefore, this will hinder how closely we can
compare figure 44(A) and figure 44(B), however, the overall trends are still
comparable.
The chamfer angle of 5° was implemented at this stage, although the results were
thought to be incorrect, in order to see how the deposition pattern would be affected if
at all. The deposition pattern is shown in figure 45 where the boat-tailing is shown to
have a pronounced effect on the soiling pattern.

43
Results and Analysis

Figure 45: The surface contamination spread using a point cone nozzle at 𝟓° chamfer angles

The surface contamination spread shows a significant change in the deposition


pattern when the boat-tailing is implemented. The key changes come from the
concentration of the surface contamination lies closer to the centre line than on the
control body and also the particles are directed higher up the rear base plane. The
aerodynamic flow has changed how the surface contamination spread lies on the rear
surface and has reduced the amount of soiling particularly on the edges closer to the
top of the base plane. This is a significant finding as this shows that by implementing
boat-tailing geometry, the surface contamination pattern does change. But the quality
of this change will be assessed later on when the model is generating results to a
higher degree of accuracy.

6.3 Control model results using the full cone nozzle


When using the full cone set up with the particle dispersion command switched on,
this allowed for more realistic and comparable results to be generated. As the mesh
sensitivity test was completed in order to ensure the aerodynamic flow was accurate,
this was not needed to be tested again for different mesh sizes as this will only really
have an effect on the detail in the rear base plane at lower element sizes when using
3 𝑚𝑚. Figure 46 shows the surface contamination spread shown on the rear base
plane of the control model using the refined cone parameters.

44
Results and Analysis

Figure 46: The surface contamination spread on the rear base plane of the control model
using a full cone nozzle

This surface contamination spread is far closer to that of the literature with an
averaged water volume fraction per unit area of 2.38 × 10−5 𝑚−2. This will now be set
as the benchmark for the surface contamination numerically and this will be used to
calculate how much the surface contamination increases or decreases per unit area.
As the literature mostly shows this type of result as film thickness, there are now true
values to look at for comparison. However, we can look at Kabanovs, et al., (2016)
where he uses fraction of water as a percentage, which is the closest measure to that
used in figure 46.

A B

Figure 47: Fraction of water as a percentage on the rear base plane from (A) RANS CFD
techniques, (B) URANS CFD techniques and (C) experimental testing

45
Results and Analysis
The results shown in figure 46 and figure 47(A) are directly comparable due to the
bluff bodies being identical and the computational set up also being the same. It is
obvious that by changing the nozzle to a full cone has generated more representative
results and in turn has led to more comparable data. The surface contamination shown
in figure 46 has a deposition pattern localized around the bottom centre of the rear
base plane, much like that for figure 47(C). Comparing the results to the experimental
data is far more valuable as the experimental data, in most cases, is far more realistic
and is broadly used for validating numerical solutions. As these two sets of data in
agreement as well as the results from figure 47(A), this control model and particle
injector set up is suitable to carry on forwards with.
The spray patterns is symmetrical due to the symmetry plane in the ZX plane, but
this may not be the case in reality but is an assumption that has been used in order to
maintain a higher level of computational efficiency. Figure 48 shows the comparison
between the control data and Gaylard, et al., (2017a) and how the results are similar
and therefore this model is proven further to be suitable.

A B

Figure 48: Deposition pattern from (A) the full cone on the control model and (B) the
literature (Gaylard, et al., 2017a)

There are still differences, but the trend is still relatively the same when considering
that the legends are showing different variables. These differences are expected due
to the difference in rear wake structure shown in figure 48 where the streamlines do
not match directly but it was discussed that this could be due to the moving ground
wall boundary that was not implemented in Gaylard, et al., (2017a).

46
Results and Analysis

6.4 Boat tail geometry implementation


When implementing the boat-tail geometry the aerodynamic flow will have changed
because of this variation in the rear geometry and in turn changes how the drag and
lift forces will act as a result of the coefficient of pressure spread changing.

6.4.1 Aerodynamic forces when implementing boat-tail geometry


The aerodynamic forces changed significantly when the boat-tail geometry was
used on the rear of the model. Figure 49 shows how the aerodynamic drag was
affected as the chamfer angle was increased.
0.25
0.245

0.23

CD 0.21

0.19

0.17

0.163

0.15
0 5 10 15 20 25 30
Boat-Tail Angle [°]

Figure 49: The change in 𝑪𝑫 against the chamfer angle

Looking back to equation 12 in section 4.6, the coefficient of drag is proportional to


the drag force acting on the body. As the X axis shown drag in the negative direction
this drag force would have to be converted from a negative value to a positive in order
to represent the direction of the drag force.
The maximum drag was found to be from the control model with no boat-tailing as
suggested by the literature, however, the minimum amount of drag was found at 20°

47
Results and Analysis
chamfering with a 𝐶𝐷 value of 0.163. This is a 33.47% decrease in drag but after 20°,
the drag increases again. This is the trend shown in the literature and particularly in
figure 15(A) by Perry, et al., (2015). Although the minimum drag shown by Perry, et al.,
(2015) is at 12°, the parabolic shape is still shown and can be justified by using different
bluff bodies and different chamfering geometries.
Looking at the coefficient of lift when the lift force is acting in the positive Z axis, this
can be shown in figure 50 where a similar trend is shown.
-0.1
-0.127

-0.15

-0.2

CL -0.25

-0.3

-0.35
-0.352

-0.4
0 5 10 15 20 25 30
Boat-Tail Angle [°]

Figure 50: The change in 𝑪𝑳 against the chamfer angle

The maximum 𝐶𝐿 magnitude at 20°, much like 𝐶𝐷 showing an optimal chamfer angle
when considering the aerodynamic forces. This is because a lower drag value and a
greater negative lift coefficient will be favourable. As discussed in the literature, a lower
coefficient of drag will generate economic and environmental gain in terms of a
reduction in fuel consumption and a greater negative coefficient of lift means that there
is a greater downforce acting on the bluff body. An increase in downforce in the
automotive industry is generally considered to be a good thing as it will increase tyre
grip and therefore improve handling. An optimal value can be determined by looking
at the lift-to-drag ratio in figure 51 where the ratio between the lift and drag force will

48
Results and Analysis
be plotted for each chamfer angle, where the highest value will be deemed as the
optimal value.

-0.5

-1
Lift-to-Drag Ratio

-1.5

-2

-2.161

-2.5
0 5 10 15 20 25 30
Boat-Tail Angle [°]

Figure 51: Lift-to-drag ratio plotted against the chamfer angle

The optimal value found from figure 51 can therefore be taken as 20° when
considering the aerodynamic performance of the car body. When comparing the 𝐶𝐷
values directly to Perry, et al., (2015), both results have parabolic shape, but reach a
minimum value at different angles of 𝛼, as shown in figure 52.

49
Results and Analysis

0.29

0.27

0.25

0.23

CD
0.21

0.19

0.17

0.15
0 5 10 15 20 25 30
Chamfer Angle [°]

Chamfered Model Perry, et al., (2015)

Figure 52: A comparison between 𝑪𝑫 values when changing 𝜶, between the chamfered
model and Perry, et al., (2015)

Differences in the set-up include the moving wall boundary as the ground in the
chamfered model and Perry, et al., (2015) uses only side chamfers in order to reduce
the drag.
Overall, the four edge chamfering is able to reduce 𝐶𝐷 by 33.47% at 20°, which is a
significantly higher drag reduction than Perry, et al., (2015) with 7.4% at 12°.

6.4.2 Streamlines when implementing boat-tail geometry


The streamlines were also captured on the central plane and using the same post
processing state, a comparison of the different streamlines is shown in figure 53.

𝟓° 𝟏𝟎°

50
Results and Analysis

𝟏𝟓° 𝟐𝟎°

𝟐𝟓° 𝟑𝟎°

Figure 53: Streamlines on the central plane for each chamfer angle from 𝜶 = 𝟎° to 𝜶 = 𝟑𝟎°

The boat-tailing chamfers redirect the airflow and centralise the wake, causing the
top vortex to collapse. The streamlines generated by the 20° chamfers, where 𝐶𝐷 is
the lowest, shown that the lower vortex is closest to the top chamfer. This has caused
the airflow to pass over the top face and merge with streamlines at the top of the lower
vortex, which could be the reason that less drag is present at this chamfer angle. This
collapsing of the top vortex was displayed in figure 9, by Hassan, et al., (2018). This
was found at lower values of 𝛼 but the chamfers were peresent in the form of angled
plates causing a rear cavity which will have an affect on the streamlines when
comparing the results to fiigure 53.
The area of stagnation narrows towards 20° and therefore will limit the region of
streamlines which will cause the soiling particles to deposit on the rear of the body.
However, after 𝛼 = 20° towards 30° the wake structure opens up again meaning that
there is a greater area of turbulence emerging and thus will cause more drag to act on
the body and more deposition will occur. The coefficient of velocity is greater at the
chamfered corners at greater values of 𝛼, which is significant because this is another
sign of flow separation and can result in pressure changes in these regions too.

51
Results and Analysis
Figure 54 shows the breakdown of how the streamlines change from 𝛼 = 15° to 𝛼 =
20° in order to breakdown how the streamlines behave up until the point of minimal
drag.

𝟏𝟓° 𝟏𝟔°

𝟏𝟕° 𝟏𝟖°

𝟏𝟗° 𝟐𝟎°

Figure 54: Streamlines on the central plane ranging from 𝜶 = 𝟏𝟓° to 𝜶 = 𝟐𝟎°

At the lower values of 𝛼 in figure 54, the top vortex is clear but starts to reduce as
the 𝛼 increases, until it nearly merges into the streamlines above the region of
stagnation that are flowing at a higher 𝐶𝑣 value.

6.4.3 Surface contamination when implementing boat-tail geometry


The surface contamination changes significantly with the implementation of boat-
tailing as shown in figure 55. There is a distinct pattern between the surface
contamination and aerodynamic forces when you change 𝛼.

52
Results and Analysis

1.E-04

2.38E-05
Water Averaged Volume Fraction per m^2

1.E-05

1.E-06

1.E-07

1.E-08
0 5 10 15 20 25 30

Boat-Tail Angle [°]

Figure 55: Water Averaged Volume Fraction per 𝒎𝟐 at different values of 𝜶

This graph shows that the amount of water landing on the rear base plane per unit
area decreases significantly after 𝛼 = 15° and has a value of 0 at 𝛼 = 25°. Due to this
graph being plotted using a logarithmic scale, a data point ≠ 0 and therefore cannot
be shown. The water averaged volume fraction (AVF) drops two orders of magnitude
from 𝛼 = 15° to 𝛼 = 30°, which is displayed visually in figure 56.

53
Results and Analysis

5° 10°

15° 20°

25° 30°

Figure 56: Surface contamination on the rear base plane from 𝜶 = 𝟓° to 𝜶 = 𝟑𝟎°

The deposition pattern can be seen to shift vertically as 𝛼 increases, and the
concentration of the water AVF tends to localise around the centre plane, up until when
the surface contamination peaks off at 15°. Due to the particle numbers not being as
accurate to real life, there may still be deposition at 25° in reality but for this purpose
where the particle tracking must be simplified in order to keep the solution
computationally efficient, this is an aspect where assumptions may affect the results.

54
Results and Analysis

Control Model

𝟓° 𝟏𝟎°

Figure 57: A visualisation of the change in particle path as the chamfer angle increases

Figure 57 shows how the particle paths end further up the model as the chamfer
angle increases making the localisation of the surface contamination sit further up the
centre line. There still a high level of concentration in the localised areas but the
translation is interesting because when looking at how this would affect a vehicle in
reality, different components sit at different places in the vertical plane so this would
affect them in different ways. For example, there would be less effect on the rear
windscreen when looking at the control model surface contamination, but it would be
covering most of the centre of the rear windscreen when 𝛼 = 10°.
As there is such a large difference in surface contamination from 𝛼 = 15° to 𝛼 = 20°,
figure 58 shows how this reduction in deposition happens and the point where high
concentration of particles tapers off.

55
Results and Analysis

16° 17°

18° 19°

Figure 58: Surface contamination on the rear base plane from 𝜶 = 𝟏𝟔° to 𝜶 = 𝟏𝟗°

As of 𝛼 = 18° the surface contamination concentration tends to disappear in the top


centre of the rear base plane. This decline in the surface contamination can be shown
in figure 59, where the data displays an exponential trend.

56
Results and Analysis

1.6E-05

1.4E-05
Water Averaged Volume Fraction per m^2

1.2E-05

1.0E-05

8.0E-06

6.0E-06

4.0E-06

2.0E-06

0.0E+00
15 16 17 18 19 20
Boat-Tail Angle [°]

Figure 59: Water Averaged Volume Fraction per 𝒎𝟐 at when 𝟏𝟓° ≤ 𝜶 ≤ 𝟐𝟎°

The exponential decay of this graph is interesting because this displays how the
water AVF per unit area tends towards zero at 19° although this is a limitation of the
set-up of the simulation, it shows that at this chamfer angle, the surface contamination
will start to tend towards its lowest value. The particle paths will determine whether the
particle will land on the rear base plane and figure 60 shows the difference between
the paths of the particles at 𝛼 = 15° and 𝛼 = 20°. The particle coefficient of velocity is
given by equation (15).
Velocity
Particle Coefficient of Velocity = (15)
Particle Injection Velocity
Meaning that the streamlines will show how much the particles accelerate in the fluid
flow in respect to the velocity they were injected at, which is 15.2𝑚𝑠 −1. Interestingly,
the particle velocity changes to be approximately 2.2 times the particle injection velocity
just after they leave the injector, and as 𝑣∞ is nearly this, it shows that the particles are
being swept into the freestream effectively.

57
Results and Analysis

𝟏𝟓° 𝟐𝟎°

Figure 60: Particle paths at 𝜶 = 𝟏𝟓° and 𝜶 = 𝟐𝟎°

There is a distinct difference between the particle paths shown in figure 60, at 15°
the particles flow back into the region of stagnation more easily than at 20°. Looking
back at the figure 54, it is clear that in the centre of the region of stagnation, the
coefficient of velocity of the streamlines is approximately 0.3 when 𝛼 = 15° but 0 at 20°,
this suggest that the turbulence in this area is more chaotic when the particles enter
the streamlines at this point. Also, the fact that the top vortex disappears at 20° will
mean that there is very little airflow counteracting the lower vortex so the particles can
be transported into the freestream easier.

When looking at a chamfer angle of 30°, when the surface contamination reappears,
the particle path shows that when the velocity increases in the region of stagnation,
the particles find it harder to escape this airflow and are more inclined to be deposited
on the rear of the vehicle, but the effect of this is lower than when the chamfer angles
are much lower.

A B

Figure 61: (A) Streamlines on the centre plane with 𝜶 = 𝟑𝟎°, (B) particle path with 𝜶 = 𝟑𝟎°

At the bottom of the model the freestream velocity path detaches much earlier than
on the previous design iterations, creating a layer of air for the air to It is important to

58
Results and Analysis
consider that these are not the entirety of the particle paths generated by the particle
tracking software but are just 75 paths chosen to show a generalised behaviour of the
particles.
Interestingly the surface contamination spread at 𝛼 = 5° is largely similar to that
shown in Gaylard, et al., (2017b) in figure 62(B), published by the IMechE.

(A) (B)

Figure 62: A comparison in surface contamination between (A) the bluff body with 𝜶 = 𝟓°
and (B) a production vehicle

The significance of this is that it proves that although this is a RANS solution using
a very simplified automotive body, incredibly similar results can still be obtained,
especially when the chamfers have been implemented, making the model closely
represent the body of the production vehicle in figure 62(B).

6.5 Surface Contamination change with different particle densities


After changing the density of the soiling particle so that it is closer to representing
mud particles. Using a density of 1910𝑘𝑔𝑚−3 from aqua_calc.com, the deposition
pattern has changed marginally, as shown in figure 63.

59
Results and Analysis

A B

Figure 63: A comparison between the surface contamination spread between (A) muddy
water and (B) pure water

Although this is a very crude representation a mud particle, it does show that by
changing the density of water at 25°𝐶, from 997𝑘𝑔𝑚−3 to 1910𝑘𝑔𝑚−3 the momentum
of the particles entering the freestream changes dramatically and the particle paths will
change as a result of this. Table 7 shows the quantitative study showing that the
surface contamination has overall decreased when the density of the particles were
increased, however, this is only a percentage reduction of 8.82%, when nearly doubling
the density.

Table 7: Averaged volume fraction for different particles

Soiling Particle Density Averaged Volume


[𝒌𝒈𝒎−𝟑 ] Fraction

Water 𝟗𝟗𝟕 𝟐. 𝟑𝟖 × 𝟏𝟎−𝟓


Mud 𝟏𝟗𝟏𝟎 𝟐. 𝟏𝟕 × 𝟏𝟎−𝟓

The particle paths show different patterns when looking onto the rear and
isometrically in figure 64, but this difference still follows a similar pattern.

60
Results and Analysis

A B

C D

Figure 64: Particle paths for the (A) rear view of the mud particles, (B) rear view of the water
particles, (C) isometric view of the mud particles and (D) isometric view of the water particles

Figures 64(A) and 64(C) show how broad the particle paths are when using mud
particles, this demonstrates why the surface contamination around the outer radial
area has a higher amount of deposition than the water state. In reality there are mud
particles combined with water particles, displayed in this section and the reality of this
is that these could be considered to be more realistic results but are of course averaged
snapshots in how the surface contamination could be developed using water with
different densities.

61
Discussion

7 Discussion
This project has successfully been able to demonstrate how the surface
contamination and aerodynamic drag can be optimised between a chamfer angle of
18° to 20°. However, this has been achieved using a vast array of assumptions and
simplifications within the model. Kabanovs, et al., (2016) and Gaylard, et al., (2017a)
discuss in detail how the results are far more accurate when using URANS calculations
rather than RANS. Seeing as this project uses explicitly RANS solutions, the reality of
the results may be different to that shown in section 6. This does not mean that the
trends shown are not valid but the detail within the results could be improved upon.
Another reason why the results may not correlate to the literature as well as they could,
could be because the mesh refinement at the front of the model could be improved
upon as much as it has been in the rear. The reason this was done, was because the
results have been collected on the rear of the body and the detail needed to be
captured in the rear, in order to make sure these results were as accurate as that could
be. However, the elemental size at the front of the model could have had an effect on
the airflow across the top of the body. This could have been investigated by
implemented another wake refinement region around the entirety of the body, but this
was considered to be too computationally costly for this project for the refinement of
results that are quite similar to that of Kabanovs, et al., (2016) and Gaylard (2019).

When considering how the aerodynamic forces are found, the best way in which this
can be achieved is experimentally, but unfortunately due to resourcing and time
limitations, experimental results were not able to be obtained for comparison. One of
the most commonly used techniques for collecting physical aerodynamic data is by
using PIV technology, where the detail and accuracy of the results can be far more
accurate than using some smaller wind tunnels with force measuring instruments.
Looking at how the CFD software has been used in order to collect the data for the
aerodynamic forces, it is possible that the complexity of the model has hindered how
accurate these coefficients may be when comparing these numerically to a production
vehicle. As the 𝐶𝐷 and 𝐶𝐿 values are similar but not identical to those of the literature
this could be improved upon in order to ensure that the wake structure is closer to that
of the RANS results found in the literature. However, as previously mentioned this
could be down to the moving wall ground boundary that was implemented so that it
would replicate the relative movement between the car body and the road. When
applying this logic, when we consider that the CFD set-up in Kabanovs, et al., (2016)

62
Discussion
was mirrored to that of the experimental wind tunnel set-up – meaning the stilts must
be used to suspend the Windsor body in the air, maybe the stilts did not need to be
implemented in this investigation in order to improve the accuracy of the model and
wheels could have been incorporated instead.

The rear surface contamination has been measured by finding the averaged volume
fraction of the soiling particles per unit area instead of the conventional method of
displaying the film thickness of the soiling. This is another limitation of RANS
calculations. Not only this but when looking at 𝛼 = 25°, there is no surface
contamination present on the rear base plane. This is the result of not injecting as many
particles as there would necessarily be in reality due to this making the model too
computationally expensive. Although the parameters used to replicate the physical
BETE MW105 nozzle used in Kabanovs, et al., (2016), they are still not enough
particles that would be present when considering the amount of particles being
deposited from two wheels. This is another limitation of using a simplified automotive
body that could be improved by using a more detailed model like that from Kabanovs,
et al., (2019). The surface contamination was investigated at intervals of 5° but there
was more detail to be collected around 𝛼 = 20°, but due to time constraints and disk
memory, the data was unable to be processed and the results for 𝛼 = 21° to 24° was
unable to be captured but this may have displayed some valuable data when looking
at the 𝐶𝐷 and 𝐶𝐿 values. Importantly, the transition from 𝛼 = 15° to 20° was captured
for the surface contamination where it was found that the surface contamination could
be considered to be negligible after 𝛼 = 18°, whereas this could’ve been found to be
20° if this in depth investigation were not to be investigated.

A brief investigation took place in altering the density of the water particles to a
generalised value of 1910𝑘𝑔𝑚−3 from 997𝑘𝑔𝑚−3 and found that because the particles
were denser, they inherently carried more momentum and were able to travel further
through the wake and created higher concentrated regions on the base plane,
although, there was not much change in the overall trend. This crude change in the
particle parameters needed to have more considerations in place in order to sufficiently
model this realistically. Such considerations must be the mass flow rate, the particle
diameter and the composition of soil. Simply altering the density of water to a new
value of mud defined from the internet, is not enough of an investigation into whether
muddy water acts in the way shown in section 6.5 but it does show that there is a trend
to be found that when the particle density increases, the surface contamination

63
Discussion
decreases – in this case by 8.82%. There is not any published data on the effect of
particle density change on the surface contamination, this is why these values cannot
be explicitly compared to result found in the literature and therefore cannot be
considered valid.

64
Conclusion

8 Conclusion
After the data was obtained and analysed, it is clear that the boat-tail geometry had
a significant effect on the aerodynamic forces acting on the body. The optimal value 𝛼
for both the aerodynamic drag and downforce was found to be 20°. The significance of
this is that both of the forces are optimal at this angle, making it easy to determine
which model is the most aerodynamically acceptable design to implement into industry.
This is with a value of ∆𝐶𝐷 of 0.082 and ∆𝐶𝐿 of −0.479, which are considered
significantly high changes in the aerodynamic research field.

When looking at the surface contamination, there is a substantial drop in the water
AVF on the rear base plane after 𝛼 exceeds 18°, therefore using this particle injection
method, there is little surface contamination at this point. Although, the injection
method could be improved upon when generating a more detailed bluff body using
wheel geometry and more detailed frontal components. There was however no surface
contamination using this set-up at 25°, meaning this would be the ideal value, but in
reality with a greater number of particles entering the system, there may still be some
surface contamination, hence why the results at 𝛼 = 20° are just as valuable as that of
𝛼 = 25°.

As the surface contamination tends to move in the positive Z direction as the


chamfer angle increases, this will move the surface contamination away from the area
where the licence plates will be and onto the windscreen. When considering the safety
of the driver, it is more important for the driver to be able to see out of the rear
windscreen than it is for the licence plates to be visible. This is why it is important to
perform a qualitative analysis on the surface contamination spread rather than just
investigating at what design iteration shows a lower value of the water AVF on the rear
base plane.

When investigating a change in density of the soiling particles from 997𝑘𝑔𝑚−3 to


1910𝑘𝑔𝑚−3 the surface contamination spread has changed and a soiling particle AVF
per unit area has decreased by 8.82% but this needs to be investigated in further detail.

This project aimed to investigate the optimal chamfer angle for boat-tailing geometry
in terms of the surface contamination and aerodynamic performance and it can be
concluded that as the optimal chamfer angle will be 20° due to this generating the

65
Conclusion
lowest lift-to-drag ratio, with −2.61 and there being a low surface contamination spread
for this design iteration.

66
Recommendations and Further work

9 Recommendations and Further work


Throughout the course of this project, it was found that the spray pattern was in fact
changed by the boat-tail geometry but with a single particle parameter set, with a fixed
𝑑𝑝 value and only using water as the soiling particle at 25°𝐶. If time restrictions were
eliminated, it would be valuable to investigate how the density of the particles could
affect the surface contamination spread over a range of values, instead of using a
density change to refine a particle that has a different chemical composition to that
used in the set-up. As discussed in section 2.5, the particle diameter is not a constant
value, and has a certain bell curve, distribution pattern with a given standard deviation.
Investigating how this would work on a four edge chamfered body would change how
surface contamination spread. At 𝛼 = 20°, using different particle diameters might
suggest that there would be more surface contamination and a parametric study could
be conducted to find out of this is the case. This could be achieved by defining multiple
injector locations using different values of 𝑑𝑝 and then particles with a range of
densities could be defined in the set-up.

The effect of the yaw angle of the surface contamination was initially a part of this
investigation but due to the fact that the model was cut in half to reduce the
computational load, the yaw angle could not simply become an input parameter as this
would make the model asymmetric. This could be a study conducted in the future as
this would change the fluid flow around the body but would mean that the mesh would
have approximately double the element size and another mesh sensitivity test would
have to be conducted to accurately plot this change. The yaw angle is something highly
applicable to real life scenarios as cars do not simply drive in a straight line. It would
be beneficial to conduct an analysis using a dynamic mesh where the model would
move from a yaw angle of 0° to 4° during a transient analysis in order to see how the
film thickness would build up as this change where to take place.

Typically, the chamfer angles on the top and bottom of the rear base plane are not
kept the same in automotive design as shown in Krishnani (2009), where the top and
bottom angles are separate parameters. This would be investigated in the future and
could prove that the value of 𝛼 with the least amount of drag could be achieved at
different angles, when implementing a moving wall as the ground boundary.

The Windsor model makes for a more realistic shape than the Ahmed body used in
Krishnani (2009) but is not as realistic as the simplified SUV models used in the

67
Recommendations and Further work
majority of the surface contamination literature, where wheels are implemented.
Usually, the particles will leave the wheel tangentially in reality, but through using stilts
to support the model, this is not represented as accurately as it could be. It would be
valuable to implement wheel geometry into the design and see how the chamfering
could affect the surface contamination then. Although, the injector location would have
to be changed in order to accommodate this more complex model as shown in figure
8 by Gaylard (2019).

68
References

10 References
Aqua-Calc, Density of mud. [Online]
Available at: https://www.aqua-calc.com/page/density-table/substance/mud-coma-
and-blank-packed
[Accessed September 2022].
Bangalore Narahari, T. S. & Bharadhwaj, Y., 2021. Snow modeling to predict surface
contamination on cars, s.l.: Chalmers University of Technology.
Choy, H. A. et al., 2021. Numerical Study on the effect of boat tail shape on
aerodynamic drag of SUV. International Journal of Automotive Technology, 22(1), pp.
165-172.
Crowe, C. T., Chung, J. N. & Troutt, T. R., 1988. Partiles mixing in free shear flows.
Progress in Energy and Combustion Science, 14(3), pp. 171-194.
Crowe, C. T., Schwarzkopf, J. D., Sommerfield, M. & Tsuji, Y., 2011. Multiphase Flow
with Droplets and Particles. Second ed. New York: CRC Press.
Davis, B., 2017. 2017 Range Rover SVAutobiography Dynamic review (video). [Online]
Available at: https://performancedrive.com.au/2017-range-rover-svautobiography-
dynamic-review-video-1013/
[Accessed August 2022].
Domel, G. A. et al., 2018. Shark skin-inspired designs that improve aerodynamic
performance. Journal of the Royal Society Interface, 7 February, 15(139), p.
20170828.
Fluid Mechanics 101, 2020. [CFD] Lagrangian Particle Tracking. [Online]
Available at: https://www.youtube.com/watch?v=jdq6puyvQ7E
[Accessed August 2022].
Gaylard, A., 2019. Vehicle Surface Contamination, Unsteady Flow and Aerodynamic
Drag, s.l.: University of Warwick.
Gaylard, A. et al., 2017a. Simulation of rear surface contamination for a simple bluff
body. Journal of Wind Engineering and Industrial Aerodynamics, Volume 165, pp. 13-
22.
Gaylard, A., Kirwan, K. & Lockerby, D., 2017b. Surface contamination of cars: A
review. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of
Automobile Engineering, 231(9), pp. 1160-1176.
Gray, S., 2013. Gray & Adams. [Online]
Available at: https://www.gray-adams.com/case-study-waitrose-ultra-low-carbon-
trailer/
[Accessed August 2022].

69
References
Greifzu, F. et al., 2016. Assessment of particle-tracking models for dispersed particle-
laden flows implemented in OpenFOAM and ANSYS FLUENT. Engineering
Applications of Computational Fluid, 10(1), pp. 30-43.
Gudmundsson, S., 2014. The Anatomy of the Airfoil. In: General Aviation Aircraft
Design. s.l.:s.n., pp. 235-297.
Hassan, M., Badlani, D. & Nazarinia, M., 2018. On the effect of boat-tails on a simplified
heavy vehicle geometry under crosswinds. Journal of Wind Engineering and Industrial
Aerodynamics, Volume 183, pp. 172-186.
Jilesen, J., Gaylard, A. & Escobar, J., 2017. Numerical Investigation of Features
Affecting Rear and Side Body Soiling. Passenger Cars - Mechanical Systems, 10(1),
pp. 299-308.
Kabanovs, A., Garmory, A., Passmore, M. & Gaylard, A., 2017b. Computational
simulations of unsteady flow field and spray impingement on a simplified automotive
geometry. Journal of Wind Engineering and Industrial Aerodynamics, Volume 171, pp.
178-195.
Kabanovs, A., Garmory, A., Passmore, M. & Gaylard, A., 2019. An investigation into
the dynamics of wheel spray released from a rotating tyre of a simplified vehicle model.
Journal of Wind Engineering and Industrial Aerodynamics, Volume 184, pp. 228-246.
Kabanovs, A., Garmory, A., Passmore, M. & Gaylard, A., 2019. Investigation into the
dynamics of wheel spray released from a rotating tyre of a simplified vehicle model.
Journal of Wind Engineering & Industrial Aerodynamics, Volume 184, pp. 228-246.
Kabanovs, A. et al., 2017a. A Parametric Study of Automotive Rear End Geometries
on Rear Soiling. Passenger Cars - Mechanical Systems, 10(2), pp. 553-562.
Kabanovs, A. et al., 2016. Experimental and Computational Study of Vehicle Surface
Contamination on a Generic Bluff Body, s.l.: SAE Technical Paper.
Kabanovs, A. et al., 2016. Experimental and Computational Study of Vehicle Surface
Contamination on a Generic Bluff Body, s.l.: SAE International.
Krishnani, P. N., 2009. CFD study of drag reduction of a generic sport utility vehicle,
Sacramento: American Society of Mechanical Engineers.
Le Good, G. & Garry, K., 2004. On the Use of Reference Models, Detroit, Michigan:
SAE World Congress.
Linkov, J., 2019. The Race to Protect Car Sensors From Their Biggest Foes: Dirt and
Weather. [Online]
Available at: https://www.consumerreports.org/car-maintenance/protect-car-sensors-
from-dirt-and-weather/
[Accessed August 2022].
Littlewood, R. & Passmore, M., 2010. The Optimization of Roof Trailing Edge
Geometry of a Simple Square-Back, s.l.: SAE Technical Paper.

70
References
Merriam-Webster, n.d. Merriam-Webster.com Dictionary. [Online]
Available at: https://www.merriam-
webster.com/dictionary/coefficient%20of%20velocity
[Accessed May 2021].
NASA, 2014. Glenns Research Center. [Online]
Available at: https://www.grc.nasa.gov/www/BGH/reynolds.html
[Accessed 2021].
Perry, A.-K., Passmore, M. & Finney, A., 2015. Influence of Short Rear End tapers on
the Base Pressure of a Simplified Vehicle. SAE International Journal of Passenger
Vehicles - Mechanical Systems, 8(1), pp. 317-327.
Saunders, N. K., Lewis, P. & Thornhill, A., 2019. Research Methods for Business
Students. 8th ed. London: Pearson Education.
Schembri Puglisevich, L. et al., 2016. Application of CFD to Predict Brake Disc
Contamination in Wet Conditions. SAE International Journal of Passenger Cars -
Mechanical Systems, 9(2), pp. 800-807.
The Engineering Toolbox, n.d. Air - Density, Specific Weight and Thermal Expansion
Coeffciient vs. Temperature and Pressure. [Online]
Available at: https://www.engineeringtoolbox.com/air-density-specific-weight-
d_600.html
[Accessed July 2022].
The Engineering Toolbox, n.d. Air - Dynamic and Kinematic Viscosity. [Online]
Available at: https://www.engineeringtoolbox.com/air-absolute-kinematic-viscosity-
d_601.html?vA=25&units=C#
[Accessed July 2022].
Urquhart, M., Varney, M., Sebben, S. & Passmore, M., 2021. Drag reduction
mechanisms on a generic square-back vehicle using an optimised yaw-insensitive
base cavity. Experiments in Fluids, 62(241).
Varney, M., Passmore, M. & Gaylard, A., 2017. The Effect of Passive Base Ventilation
on the Aerodynamic Drag of a Generic SUV Vehicle. SAE International Journal of
Passenger Cars - Mechancial Systems, 10(1), pp. 345-357.
Volvo, 2017. Volvo Trucks’ Latest Concept Vehicle Tests a Hybrid Powertrain for Long
Haul Transport. [Online]
Available at: https://www.multivu.com/players/uk/8047351-volvo-trucks-concept-
hybrid-powertrain/
[Accessed August 2022].
Xiao, W., Jin, T., Dai, Q. & Fan, J., 2020. Eulerian–Lagrangian direct numerical
simulation of preferential accumulation of inertial particles in a compressible turbulent
boundary layer. Journal of Fluid Mechanics, Volume 903.

71
Appendix

11 Appendix

11.1 Project Proposal

11.1.1 Introduction
Background and problem definition
In automotive aerodynamics, the rear surface contamination is an important
consideration for many reasons. A build-up of soiling on the rear of a vehicle can come
in many forms, snow, water or other particulates and can cause a degradation in the
driver’s rear visibility or even the lack of visibility of the rear number plate or even brake
lights (Gaylard, et al., 2017b). This is mainly seen in the case of square back SUVs
(Jilesen, et al., 2017). In more recent times, technology has advanced, including
reversing sensors and cameras, which in turn will bring more challenges concerning
surface contamination (Kabanovs, et al., 2019) (Jilesen, et al., 2017).

Research question
What is the effect of implementing boat-tail geometry on the surface contamination
on the rear of an automotive body?

Aim and objectives


The aim of the project will be to investigate how the rear base plane geometry of an
automotive body will affect the rear surface contamination using Computational Fluid
Dynamics (CFD) at different yaw angles.

The objectives that will be achieved throughout the course of the project are as follows:
1. Identify the aerodynamic forces acting on a control automotive bluff body
using CFD techniques.
2. Understand how the aerodynamic flow in the wake will influence the rear
surface contamination through a particle tracking analysis.
3. Conduct transient CFD analyses, monitoring the aerodynamic drag when
implementing boat-tail chamfers.
4. Analyse the rear surface contamination spread on the modified bluff body
using CFD techniques.
5. Evaluate the change in surface contamination and aerodynamic drag at
different yaw angles.

72
Appendix
6. Determine the optimal rear geometry parameters required in order to reduce
the aerodynamic drag and minimise rear surface contamination.
Rationale
Through observing the structure of the wake flow the path in which the particles can
be carried onto the rear base plane can be determined and therefore by altering this
behaviour will change the spread in surface contamination. The bodies geometry will
play a vital role in manipulating the wake structure and therefore changing the
aerodynamic behaviour. Using a qualitative analysis of the particle deposition, key
areas can be monitored when considering driver’s visibility.
By altering the wake structure of the aerodynamic flow, this will have an innate effect
on the aerodynamic drag acting on the vehicle and can therefore have a direct effect
on the efficiency of the vehicle, both environmentally and economically. This will aim
to be done by implementing small boat-tail chamfers on the rear base plane and it has
been shown that this is possible through existing literature, but by investigating the
impact this will have on the rear soiling will determine how applicable this application
will be when both the drag and surface contamination are investigated simultaneously.

73
Appendix
11.1.2 Initial Literature Review
Automotive bluff bodies for surface contamination
Different bluff bodies have been in the surrounding literature, however, the most
common model used is the Windsor body, shown in figure 1. The Windsor body
represents the key elements of an automotive body and have features proportional to
a small hatchback (Gaylard, et al., 2017a) but is also scalable to represent the square
back features of modern day production SUVs (Gaylard, 2019).

Figure 1: The Windsor body geometry (Kabanovs, et al., 2016)

This type of geometry is really shown to demonstrate the fluid flow strictly over the
body of the vehicle without wheels. There are adaptations of the Windsor body with
wheels that have investigated the surface contamination with the implementation of
tyre geometry (Kabanovs, et al., 2019) as demonstrated in figure 2.

Figure 2: The Windsor body geometry with wheels (Kabanovs, et al., 2019)

There is a strong collaboration within the literature reviewed in recent times, mainly
between Gaylard and Kabanovs, however, Le Good & Gary, (2004), investigates
multiple automotive bluff bodies in detail and does not review the Windsor body, but
instead reviews the Ahmed body and the fundamental fluid flow surrounding the body.
There are clear similarities between the Ahmed body and the Windsor body but from
direct observation, the Windsor body reflects an SUV to a closer extent than the Ahmed
body.

74
Appendix
The use of simplified automotive bodies reduces computational cost and time, whilst
generating highly representative results (Gaylard, 2019). This is why production
vehicle models are rarely used for generalised analyses but there are case studies
found in the literature where they are required for detailed aerodynamic developments.
An example can be found in Gaylard, et al., (2017b) published by the IMechE.
The surface contamination shown in this study is demonstrated using a positivist
approach and the context in which the spread in soiling is discussed to little depth. The
film thickness is often analysed but the lack of comparison means that the optimal
performance is not defined numerically but instead can be deemed to be contextual.
This may be due to the generalisation of the study and the simplicity of the geometry.
The interpretation of the surface contamination is discussed with a greater surrounding
context of the location of the rear soiling in Jilesen, et al., (2017).

Boat-tail geometry in automotive aerodynamics


Boat-tail geometries have been used in many aerodynamic applications over the
years and have played a large role in drag reduction especially when considering larger
vehicles (Krishnani, 2009) (Hassan, et al., 2018). Hassan, et al., (2018) shows that the
implementation of the boat-tail structures can result in a reduction in drag by
approximately 50% and looks directly at how the crosswind can affect the vorticity in
the wake of the airflow. However, when using a production vehicle model, the
coefficient of drag does not necessarily reduce, when implementing a rear boat-tail,
unless a smooth undercover is used (Choy, et al., 2021). Hassan, et al., (2018) also
investigates how the wake structure and flow separation is affected when changing the
boat-tail length, shown in figure 3, where the length of the plate is measured in meters.

75
Appendix
Figure 3: Velocity field contours when changing the boat-tail length (B*) and boat-tail slant
angle (𝜷) (Hassan, et al., 2018)

Krishnani, (2009) investigates the effect of implementing top and bottom boat-tail
planes at different angles and found that the combination which produced the least
drag was when both the top and bottom plate were positioned 10° to the horizontal.
This is shown in figure 4.

(A)

(B)

Figure 4: The pressure contour on the centre plane of the: (A) control body (B) body with
the top and bottom boat-tail angles of 10° to the horizontal (Krishnani, 2009)

Figure 4 shows the pressure scale in the surrounding fluid flow and demonstrates
how the implementation of the boat-tail plates definitely affects the wake structure. This
is shown in greater detail in figure 5 where the velocity streamlines are displayed.

(A) (B)

Figure 5: The velocity streamlines on the centre plane of the: (A) control body (B) body with
the top and bottom boat-tail angles of 10° to the horizontal (Krishnani, 2009)

By increasing the angle in which the angle of the boat-tail plates are to the horizontal,
the higher the velocity tends to be in the wake.

76
Appendix
11.1.3 Project Management and Methodology
Research Philosophies
The project will take a deductive research approach initially to justify the results. A
positivist approach will also be taken throughout the project as the primary data
collected through analytical techniques will be observing the quantitative elements of
the project. However, it has to be considered that the spread in surface contamination
is a qualitative measure as different areas of the rear base plane can be affected
differently by a build-up of soiling.

Ethical review
There are no ethical issues to consider throughout this project and upholds the
university’s guidelines.

Methodology
This section outlines the processes that will be undertaken throughout the project
and describes each process, stating which software will be used, if applicable. The
project will be predominantly conducted using ANSYS Workbench which hosts a
multitude of software able to complete each process of the CFD analyses.

Project Timeline
The project timeline has been detailed in an attached Gantt chart outlining the
details of each task and the predecessors for each task when applicable. The duration
of each task has been allocated and the timescale of the project has been highlighted
also. External commitments have been considered and this is reflected within the
project plan.

CAD Model
The Windsor body will be drawn up in DesignModeller in ANSYS Workbench, where
the rear boat-tail chamfers will be defined as input parameters. This will mean that
each design point can be saved and solved for different chamfer angles. These angles
will range from 0° to 25° in increments of 5°.

77
Appendix
Domain Meshing
The In order to simulate the CFD analysis accurately the domain must be set to an
appropriate scale, which can be defined using the values determined in the literature.
However, the meshing software used in ANSYS Mechanical is different to that
described in the literature, therefore, a mesh independence test will have to be
conducted on a simple case. This will ensure accuracy in the model.

CFD Boundary Conditions


The boundary conditions of any CFD analysis is key to obtaining the correct results.
The type of CFD analysis and turbulence model will be identified in this step. This will
be using ANSYS CFX where the conditions for the fluid flow will be defined as well as
the particle injection parameters can also be defined. This will include the particle
dimensions and density as well as injection location.

Post processing
Using ANSYS CFX Postprocessor, the drag forces, coefficient of velocity and
coefficient of pressure can be defined and plotted graphically across the Windsor body,
once the simulation has solved. The particle tracking will also be viewing in this process
and using an isosurface, the film thickness and surface contamination spread. These
results can be withdrawn as output parameters and can be saved for each design point.

78
Appendix
11.1.4 Conclusion
This proposal outlines the issues surrounding surface contamination on the rear
base planes of automotive vehicles and describes how this can be modelled using
CFD techniques using the Windsor body. The literature review discusses how the
addition of a boat-tail geometry can affect the aerodynamic flow specifically in the wake
of the vehicle. This change in wake structure will affect how the soiling particles travel
through the wake and may affect the surface contamination on the rear of the body.
The boat-tail geometry will be implemented in the form of small equal chamfers on
the rear on the edges of the base plane and will range from 0° to 35° in intervals of 5°
as reflected in the literature review. The results of this investigation will determine the
optimal angles for reducing the aerodynamic drag and surface contamination spread
when considering the implementation of new technologies in the automotive industry.
The proposal outlines the research methodology and processes required to
complete the project, including an in-depth project plan attached in the form of a Gantt
chart. The proposal also addresses the fact that the project is not ethically
compromising.

79
Appendix
11.1.5 References
Choy, H. A., Park, S. O., Moon, H. G., Kim, M. S., & Cho, J. S., 2021. Numerical Study
on the effect of boat tail shape on aerodynamic drag of SUV. International Journal of
Automotive Technology, 22(1), pp. 165-172.

Gaylard, A., 2019. Vehicle Surface Contamination, Unsteady Flow and Aerodynamic
Drag, s.l.: University of Warwick.

Gaylard, A., Kabanovs, A., Jilesen, J., Kirwan, K. & Lockerby, D. A., 2017a. Simulation
of rear surface contamination for a simple bluff body. Journal of Wind Engineering and
Industrial Aerodynamics, Volume 165, pp. 13-22.

Gaylard, A., Kirwan, K. & Lockerby, D., 2017b. Surface contamination of cars: A
review. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of
Automobile Engineering, 231(9), pp. 1160-1176.

Hassan, M., Badlani, D. & Nazarinia, M., 2018. On the effect of boat-tails on a simplified
heavy vehicle geometry under crosswinds. Journal of Wind Engineering and Industrial
Aerodynamics, Volume 183, pp. 172-186.

Jilesen, J., Gaylard, A. & Escobar, J., 2017. Numerical Investigation of Features
Affecting Rear and Side Body Soiling. Passenger Cars - Mechanical Systems, 10(1),
pp. 299-308.

Kabanovs, A., Garmory, A., Passmore, M. & Gaylard, A., 2019. Investigation into the
dynamics of wheel spray released from a rotating tyre of a simplified vehicle model.
Journal of Wind Engineering & Industrial Aerodynamics, Volume 184, pp. 228-246.

Kabanovs, A., Varney, M., Garmory, A., Passmore, M. & Gaylard, A., 2016.
Experimental and Computational Study of Vehicle Surface Contamination on a Generic
Bluff Body, s.l.: SAE International.

Krishnani, P. N., 2009. CFD study of drag reduction of a generic sport utility vehicle,
Sacramento: American Society of Mechanical Engineers.

Le Good, G. & Garry, K., 2004. On the Use of Reference Models, Detroit, Michigan:
SAE World Congress.

80

You might also like