In-cylinder Flow

Characterisation of Heavy
Duty Diesel Engines Using
Combustion Image
Velocimetry
Henrik W. R. Dembinski
KTH Royal Institute of Technology
School of Industrial Engineering and Management
Department of Machine Design

Doctoral Thesis

Department of Machine Design TRITA – MMK 2013:17

Royal Institute of Technology ISSN 1400-1179
SE-100 44 Stockholm ISRN/KTH/MMK/R-13/17-SE

0
TRITA – MMK 2013:17
ISSN 1400-1179
ISRN/KTH/MMK/R-13/17-SE
ISBN: 978-91-7501-963-5
In-cylinder Flow Characterisation of Heavy Duty Diesel Engines Using Combustion Image
Velocimetry
Henrik W. R. Dembinski
Doctoral Thesis
This academic thesis was approved by Kungliga Tekniska Högskolan and presented for
public review in fulfilment of the requirements for a Doctorial of Engineering in Machine
Design. The public review was held at Kungliga Tekniska Högskolan, Osquldasväg 4,
room Q1, on January 15 at 10:00.

1
Abstract
In-cylinder flow in diesel engines has a large impact on combustion and emission
formation. The flow was examined with cross-correlation of combustion images
for calculating vector plots, called combustion image velocimetry (CIV) in this
work. This technique is used to explain how airflow introduced during induction
affects soot emissions and interacts with injection pressures up to 2500 bar. The
CIV measurements enable flow analysis during the combustion and post-
oxidation phases. The flow velocities inside the cylinder of a heavy duty optical
engine, was measured with a crank angle (CA) resolution of 0.17° at injection
pressures of 200–2500 bar and up to nearly full load (20 bar indicated mean
effective pressure (IMEP)), were investigated with this method. The flow field
results were combined with optical flame temperature and soot measurements,
calculated according to Planck’s black body radiation theory.

At the high injection pressures typical of today’s production standard

engines and with rotational in-cylinder flow about the cylinder axis, large
deviations from solid-body rotational flow were observed during combustion and
post-oxidation. The rotational flow, called swirl, was varied between swirl number
(SN) 0.4 and 6.7. The deviation from solid-body rotational flow, which normally
is an assumption made in swirling combustion systems, formed much higher
angular rotational velocities of the air in the central region of the piston bowl
than in the outer part of the bowl. This deviation has been shown to be a source
for turbulent kinetic energy production, which has the possibility to influence soot
burn-out during the post-oxidation period.

The measured CIV data was compared to Reynolds-averaged Navier–

Stokes (RANS) CFD simulations, and the two methods produced similar results
for the flow behaviour. This thesis describes the CIV method, which is closely
related to particle image velocimetry (PIV). It was found in this work that the
spatial plane in the cylinder evaluated with CIV corresponds to a mean depth of 3
mm from the piston bowl surface into the combustion chamber during
combustion. During the post-oxidation phase of combustion, the measured
spatial plane corresponds to a mean value of the total depth of the cylinder. The
large bulk flow that contributes to the soot oxidation is thereby captured with the
method and can successfully be analysed. The link between changes in in-
cylinder flow and emissions is examined in this work.

i
Acknowledgements
This work was performed under the guidance of Professor Hans-Erik Ångström,
my excellent supervisor at the Royal Institute of Technology (KTH) in Stockholm.
He deserves special thanks for his help and expertise during this industrial PhD
project. I would also like to thank all of the people from Scania and KTH who
have helped me during this project, especially my co-supervisors and steering
committee members: Ernst Winklhofer, Daniel Norling, Anders Björnsjö,
Raymond Reinmann, and Per Stålammar. They have provided valuable ideas,
assistance with my publications, and other support. I appreciate those who have
been involved in the project in the past, including Jonas Holmborn, Andreas
Cronhjort, Magnus Mackaldener, and Per Risberg. Hannan Razzaq and Eric
Baudoin at Scania CV AB kindly helped me with the CFD simulations in this work.
Doctor Paul C. Miles at Sandia National laboratories, that was my opponent at my
Licentiate defence, gave me a very valuable discussion and questions which have
helped me to form this thesis. Thanks Paul, hope to work with you in the future.

The person who deserves the greatest thanks is my wife Helena. I

am grateful for her support and patience, especially when I have been nearly
unreachable inside my “research cloud”, hanging over my laptop writing cool
MatLab programming to quantify and extract flow data from high-speed videos at
midnight... So I think that’s the definition of a pure geek?! Hopefully, this
behaviour does not continue when you become a doctor. I want as well to send a
special thought to my little son Walter, who is 8 months old when I’m writing
this. I also thank my parents and my brother, who have supported me both
before and during my PhD studies.

Last but not least, Scania CV AB and the Swedish energy agency
were the financiers that made this project possible.

ii
List of appended publications
I. An Experimental Study of the Influence of Variable In-Cylinder Flow,
Caused by Active Valve Train, on Combustion and Emissions in a Diesel
Engine at Low λ Operation. Henrik W. R. Dembinski & Hans-Erik
Ångström. SAE paper: 2011-01-1830

II. Optical study of swirl during combustion in a CI engine with different

injection pressures and swirl ratios compared with calculations. Henrik W.
R. Dembinski & Hans-Erik Ångström. SAE paper: 2012-01-0682

III. The effects of injection pressure on swirl and flow pattern in diesel
combustion. Henrik W. R. Dembinski. International Journal of Engine
Research IJER-12-0006

IV. Swirl and Injection Pressure Impact on After-Oxidation in Diesel

Combustion, Examined with Simultaneous Combustion Image Velocimetry
and Two Colour Optical Method. Henrik W. R. Dembinski & Hans-Erik
Ångström. SAE paper: 2013-01-0913

V. In-cylinder flow pattern evaluated with combustion image velocimetry,

CIV, and CFD calculations during combustion and post-oxidation in a HD
diesel engine. Henrik W. R. Dembinski, Hannan Razzaq & Hans-Erik
Ångström. SAE paper: 2013-24-0064

VI. Swirl and Injection Pressure Effect on Post-Oxidation Flow Pattern

Evaluated with Combustion Image Velocimetry, CIV, and CFD Simulation.
Henrik W. R. Dembinski & Hans-Erik Ångström. SAE paper: 2013-01-2577

Other publications
I. The Influence of In-Cylinder Flows on Emissions from Diesel Dual Fuel
Combustion. Fredrik Königsson, Henrik W. R. Dembinski & Hans-Erik
Ångström. SAE paper 2013-01-2509

II. Flow measurements using combustion image velocimetry in diesel

engines. Henrik W. R. Dembinski. Licentiate thesis, Royal Institute of
Technology. TRITA – MMK 2012:03 (2012)

iii
Table of Contents
Abstract ..................................................................................................... i
Acknowledgements ..................................................................................... ii
List of appended publications .......................................................................iii
Other publications ...................................................................................iii
1 Introduction ......................................................................................... 1
1.1 The basic principle of diesel engine combustion ................................... 2
1.1.1 Premixed combustion period ....................................................... 2
1.1.2 Mixing-controlled combustion period ............................................ 3
1.1.3 Post-oxidation period ................................................................. 5
1.1.4 Other diesel combustion modes used in production engines ............ 7
1.2 Emissions formation in a diesel engine ............................................... 8
1.2.1 NOx formation ........................................................................... 8
1.2.2 Hydrocarbons ............................................................................ 9
1.2.3 Soot formation .......................................................................... 9
1.2.4 Soot oxidation ......................................................................... 12
1.2.5 Carbon monoxide .................................................................... 12
1.3 Turbulent flows ............................................................................. 12
1.4 Swirl and tumble flow .................................................................... 16
1.5 Squish flow .................................................................................. 18
1.6 Engine transients .......................................................................... 19
2 Project motivation............................................................................... 21
3 Methodology ...................................................................................... 21
3.1 Experimental equipment ................................................................ 22
3.1.1 Single-cylinder engine with AVT system...................................... 23
3.1.2 Optical engine ......................................................................... 25
3.2 Simulation tools and calculation methods ......................................... 26
3.2.1 CFD numerical modelling .......................................................... 26
3.2.2 1-D simulation modelling .......................................................... 26
3.3 Overview of flow measurement methods .......................................... 27
3.3.1 Hot wire measurements............................................................ 27
3.3.2 Laser Doppler anemometry (LDA) .............................................. 28
3.3.3 Particle image velocimetry (PIV) ................................................ 28
3.3.4 Particle tracking velocimetry (PTV) ............................................ 29

iv
 3.3.5 Combustion image velocimetry (CIV) ......................................... 29
3.4 The CIV method ............................................................................ 30
3.4.1 Cross-correlation applied to CIV ................................................ 30
3.4.2 Comparison of PIV and CIV particle tracing ................................. 34
3.4.3 Reproduction of flame structure................................................. 38
3.4.4 Correlation value ..................................................................... 42
3.5 CIV application results ................................................................... 43
3.5.1 Measured angular velocity during combustion ............................. 44
3.5.2 Flow pattern comparison between CIV and CFD ........................... 46
3.6 Emission spectroscopy ................................................................... 50
3.6.1 CA-resolved soot formation calculation ....................................... 52
4 Results .............................................................................................. 54
4.1 Single cylinder engine tests: Swirl and tumble effects on soot emissions
55
4.1.1 Airflow effects on heat release ................................................... 57
4.2 Optical engine results: Flow effects on soot formation and oxidation .... 60
4.2.1 Injection pressure effect on soot production and oxidation ............ 60
4.2.2 Swirl effect on soot production and oxidation .............................. 62
4.2.3 In-cylinder flow structure at injection and post-oxidation .............. 65
4.2.4 Solid-body deviation caused by injection pressure ....................... 68
4.2.5 Solid-body deviation caused by swirl .......................................... 69
4.3 CFD results: Angular velocity during compression, combustion, and post-
oxidation .............................................................................................. 70
4.3.1 Swirl number, angular momentum, and rotational kinetic energy ... 72
4.3.2 Density distribution .................................................................. 74
4.4 Optical engine results: Fuel injection impact on flow pattern and angular
velocity ................................................................................................ 76
4.4.1 Turbulent kinetic energy production ........................................... 77
5 Discussion ......................................................................................... 84
6 Conclusions ........................................................................................ 85
7 Future work ....................................................................................... 87
8 References ......................................................................................... 88
9 Summary of publications ..................................................................... 95

v
vi
1 Introduction
The compression-ignited engine, named the diesel engine after its inventor
Rudolf Diesel (18 March 1858 – 29 September 1913) [1] [2], has been the main
power source used in heavy-duty vehicles for a long period of time. The first
heavy duty (HD) vehicles with a diesel engine was produced 1923 by MAN and
Benz. Already 1903 diesel engines was fitted into the first ship and 1904 into the
French submarine, the Z. Compared with its competitor the Otto engine,
invented by Nicolaus Otto (14 June 1832 – 26 January 1891), the diesel engine
is more efficient and durable. Additionally, diesel fuel has been historically
cheaper, and have 11% higher energy content than petrol. This is the main
reasons why it is the main power source in HD applications.

Over the decades, since the diesel engine was patented in 1894,
many improvements to the engine have been made until today’s modern Euro VI
HD diesel engine, which is shown in Figure 1. Challenges to further improvement
of the diesel engine include increasingly strict emission legislation and demands
for higher engine efficiency. However, the task of both reducing emissions and
increasing efficiency is not easy because higher engine efficiency does not
automatically mean decreased emissions. A better understanding of the
processes in the engine is one key to solving the emission-efficiency problem.
This work is one small piece of a giant puzzle that many scientists around the
world are trying solve.

Figure 1. A cross-section of a modern Euro 6 HD diesel engine [3].

1
1.1 The basic principle of diesel engine combustion
Diesel combustion is a non-premixed combustion in which the fuel is injected, in
liquid phase, directly into the preheated combustion air. The combustion process
is divided into four phases: ignition delay, premixed combustion, mixing-
controlled combustion (injection), and post-oxidation.

1.1.1 Premixed combustion period

Fuel is injected at high velocity into the hot air in the cylinder, and a short time
delay occurs between start of injection (SOI) and start of combustion (SOC)
called ignition delay. The injected fuel is atomised and mixed with hot air. The
velocity difference between the injected fuel and the hot air increases the
evaporation rate, which amplify the heat and air exchange around the droplets of
fuel [4]. When enough fuel has evaporated and mixed with oxygen at the right
temperature for a sufficient amount of time, ignition of the fuel occurs. Figure 2
shows the injected fuel sprays with the first sign of combustion inside the
cylinder of an optical engine.

The ignition delay,  id , is the chemical reaction time needed for the
fuel to evaporate, decompose, and achieve activation energy in a diesel engine
and is commonly calculated by the Arrhenius correlation [5],

 E 
 id  A  p n  exp  A  , (1)
 R T 
where R is the gas constant and E A is the activation energy. The environmental
conditions that affect the reaction time are temperature (T) and pressure (p). A
and n are constants depending on fuel and, to some extent, the injection and
airflow characteristics [5]. The ignition of fuel is a complex chemical reaction with
many different reaction steps. Detailed reaction models for diesel-like fuels can
be found in [6], and a model that takes the turbulent mixing and history during
self-ignition into account can be found in [7].

2
 Blue colour shows
premixed
combustion

Reflection

Figure 2. Premixed combustion at injection pressure of 2500 bar and load of 20

bar IMEP. The bright area in the lower left is a reflection caused by an outside
light source and has nothing to do with the combustion.

1.1.2 Mixing-controlled combustion period

Immediately after the premixed fuel is consumed, the mixing-controlled
combustion, or diffusion flame, period begins. During this time, the rate of heat
release (RoHR) is governed by the amount of fuel injected per unit of time. A
schematic of the combustion flame, published by John Dec [8], is shown in
Figure 3. The injected fuel core expands in volume as it propagates from the
injector. Hot air is pulled into the injected spray and mixes with the fuel, and the
fuel starts to evaporate and decompose. The lift-off length is the distance from
the injector to the first sign of combustion. The initial soot formation, shown as
the grey area in Figure 3, begins due to a lack of oxygen. Because the λ value,
the ratio of air to fuel, is low inside the flame plume, the rate of combustion is
restricted by the amount of air mixed into the flame. Further downstream in the
flame plume, indicated by the dark blue region, the combustion continues at low
λ conditions resulting in soot formation.

3
 The yellow area in Figure 3 represents the part of the combustion
flame that is easiest to observe. Black-body radiation from the soot creates a
bright yellow light and radiant heat. This radiation contributes to the heat
transfer to the combustion walls and is essential to diesel combustion. If the soot
production is decreased in the flame, the radiation is also decreased. Most of the
thermal nitrogen oxide (NOx) is created on the surface of the flame, shown in
green, where the temperature is high and oxygen is present. Figure 4 shows an
example of the mixing-controlled combustion period in an optical engine with an
injection pressure of 2500 bar and a load of 20 bar IMEP.

Figure 3. Dec’s conceptual diesel combustion model [8].

Figure 4. Mixing-controlled combustion phase in an optical engine at 2500 bar

injection pressure and load 20 bar IMEP.

4
 Comparing the Dec model to images of an optical engine, shown in
Figure 5, clearly demonstrates that the geometry and airflow in the engine
strongly affect the shape of the combustion plume. When only one spray is
injected into the combustion chamber, the flame splits into two halves once it
reaches the edge of the bowl. The swirl direction is marked by the blue arrow in
Figure 5. The flame on the leeward side of the spray moves in the direction of
the swirl. The other flame travels a short distance in the opposite direction
before shifting back in the direction of the swirl. When multiple sprays are
injected into the combustion chamber, as shown in the right image of Figure 5,
the flames are reflected back to the centre of the bowl and interaction between
the flames occur. The interaction between the flames creates local resirculation
zones that enhance mixing. If the geometry is made correct, the reflected flames
form a recirculation zone in the bowl improve the mixing of residual gases and
heat with unburned oxygen. This mixing is of paramount importance for diesel
combustion.

Figure 5. One combustion flame (left) compared with eight combustion flames
(right) at 1500 bar injection pressure. The arrows show the swirl direction,
blue, and the reflected combustion flames, white.

1.1.3 Post-oxidation period

Post-oxidation is also a sort of mixing-controlled combustion, with the difference
that no fuel injection occurs. This phase of combustion is important for reducing
soot emissions. During this period, 25%–45% of the total heat is released, and
most of the soot created during combustion is normally reduced before the
exhaust valves are opened. Historically, rotating swirl motion, which survives the
diffusion-combustion and can thereby influence post-oxidation, has been used to
reduce soot emissions. In Figure 6, the early stages of the post-oxidation period
at high load are shown. The bright soot clouds are slowly rotating due to the
swirl while the soot is oxidized. Increased airflow velocity is beneficial during this
period while this can increase the turbulence production. More turbulence
decrease the soot oxidation time [9].

The post-oxidation period lasts as long as the in-cylinder temperature

is high enough. Therefore, maintaining a high temperature with a high level of
mixing for a sufficient residence time will effectively reduce soot . An earlier SOI
increases the residence time at high temperatures but also increases NOx

5
production. According to the Zeldovich mechanism, the longer residence time
and the higher in-cylinder temperature increase thermal NOx formation. This is
the classic NOx versus soot trade-off problem for diesel engines: reducing one
emission component causes another emission component to increase. During the
expansion stroke, the cylinder volume increases causing a decrease in the
pressure and temperature in the cylinder. The oxidation process rate decreases
with lowered in-cylinder temperature until it nearly stops and the exhaust valves
opens.

Figure 6. Post-oxidation period at 20 bar IMEP.

6
 Slow closure of the injection needle can also increase soot emissions.
At the end of injection (EOI), the injection needle starts to close and the fuel flow
through the injection nozzle decreases due to reduced sac pressure. The fuel
penetration is thereby limited, and the fuel spray creates less turbulence. The
mixing of fuel with air is restricted, and this results in higher soot production as
shown in Figure 7. When the spray velocity decreases at 9.7° ATDC, the bright
soot illumination is visible in the centre of the bowl. When the injection has
ended at 10° ATDC, the bright illumination increases in the middle of the bowl.
This bright section in the middle of the bowl has a high soot content that needs
to be oxidised. This means that a fast needle closure is beneficial to reduce the
late soot production.

9.3° ATDC 9.7° ATDC

10° ATDC 10.3° ATDC

Figure 7. The end of injection period at 500 bar injection pressure.

1.1.4 Other diesel combustion modes used in production

engines
Alternate combustion modes used in diesel engines utilize “low-temperature
combustion”. This type of combustion is characterized by high amounts of
exhaust gas recirculation (EGR) and long ignition delay. Usually all the fuel is
injected into the cylinder before combustion, therefore the fuel has a longer

7
mixing time [10]. This prevents the formation of fuel-rich zones and decreases
soot formation. NOx is also reduced when large amount of EGR lowers the
combustion temperature. Examples of low-temperature combustion modes are
described in [11] [12] [13] [14] [15] [16] [17] [18]. One drawback of this type
of combustion is that it is best suited for low-load operation and is difficult to
control in a fast engine transient. Low-temperature combustion is not considered
in this thesis because fast engine transients and high loads have been studied in
this work.

1.2 Emissions formation in a diesel engine

The Euro 1 emission standards were set in 1992, and emission legislation for HD
diesel engines has become increasingly restrictive culminating in adoption of
current Euro VI standards [19]. The most challenging emissions to reduce from a
diesel engine are particulate matter (PM) and NOx. The permitted levels for these
emissions have decreased greatly to today’s “hard-to-measure” levels. Emissions
are being lowered because of their effect on humans and the environment. Both
soot and NOx irritate the respiratory organs [20]. The harmfulness of NOx was
discovered by tyre manufacturers when ground-level ozone caused rubber tyres
to fracture. It was found that NOx combined with hydrocarbons (HC) from
combustion causes both smog and ground ozone. PM consists of polycyclic
aromatic hydrocarbons (PAH), solid coal, and condensed hydrocarbons. If the
particles are small enough, they pass through the cilia of the airways and enter
the lungs. Exposure to ultra-fine particles increases the risk for cardiovascular
diseases [21], and PAH are considered to be highly carcinogenic.

1.2.1 NOx formation

NOx emissions can be formed by four different mechanisms during combustion
with air: the Zeldovich mechanism, the Fenimore mechanism, the nitrous oxide-
(N2O) intermediate mechanism, and the NNH (nitrogen & hydrogen) mechanism
[4]. The Zeldovich (or thermal) mechanism is the dominating mechanism at
high-temperature combustion over a wide range of λ and is the mechanism that
is most used to explain NOx formation from diesel engine combustion. The
Fenimore mechanism (also called prompt NO) is particularly important in rich
combustion. Fenimore discovered that NO was rapidly produced in a laminar
premixed flame zone a long time before thermal NOx had time to form. The
hydrocarbon radicals react with nitrogen and form amines or cyano compounds.
These compounds are then converted to intermediate compounds that, after a
while, form NO. The N2O-intermediate mechanism is important at very lean, low-
temperature combustion processes, and normally occurs in gas turbines
operating under lean conditions. These three mechanisms contribute to NOx
formation in premixed and diffusion flames. The NNH mechanism is a newly
discovered reaction pathway, and it seems to be particularly important in
hydrogen and methane combustion. NOx is also created if the fuel itself contains
nitrogen, but typical diesel fuel has very low concentrations of nitrogen.

8
 As earlier mentioned, the Zeldovich mechanism is the dominating
NOx mechanism in diesel combustion. The extended Zeldovich [4] mechanism, as
it is called today, has three reactions:

O + N2 ↔ NO + N

N + O2 ↔ NO + O

N + OH ↔ NO + H
The first reaction has a very high activation energy due to the strong triple bond
in the N2 molecule. The Zeldovich mechanism needs heat, time, oxygen, and
nitrogen to form NOx. Reducing or removing any of these components decreases
the NOx formation rate. The Fenimore mechanism is linked to the combustion
chemistry of hydrocarbons.

1.2.2 Hydrocarbons
HC emissions are usually low from conventional diesel combustion, but some of
the fuel that is injected is converted into different HC:s that are not present in
the fuel. For example, small amounts of methane, formaldehyde, and aromatics
have been found in the exhaust gases. PAHs are formed under fuel-rich
conditions, and they are important precursors in soot formation. Another source
of HC comes from the injector needle sack volume [5]. When the injection ends,
some fuel is left in the needle sack volume and is ventilated to the combustion
chamber during the expansion stroke. When the cylinder pressure decreases
during the expansion stroke the evaporating fuel that is left in the sack is
evacuated into the cylinder. Because the pressure and temperature decrease
during the expansion, the HC:s are not fully combusted before the exhaust valve
opens. By reducing the sack volume, the HC emissions can be reduced.

1.2.3 Soot formation

PM, or soot, is formed in diffusion combustion and is kinetically controlled. PM
starts to form after PAHs have formed by pyrolysis of fuel under rich conditions
at high temperatures. Acetylene is the most important precursor for PAH. Each
particle then starts to form in the nucleation phase from gas-phase reactants.
Nucleation does not contribute significantly to the total soot mass but is
important when it provides sites for surface growth. In Figure 8, the different
steps of PM formation are shown. The surface growth is the process of adding
mass to the surface of a nucleated soot particle, and most of the mass is added
during this phase. The residence time has a large impact on the total soot mass
that is created. The particles then collide and coagulate to form bigger particles,
which reduces the number of particles and increases the particle size for the
same total particle mass.

9
 Figure 9 shows a diesel flame that is cut in half where the fuel is
entering from the nozzle. The first PAHs form in the soot precursor formation
zone, and particles grow as their distance from the nozzle increases [22]. In the
outer part of the flame, oxidation occurs both in the precursor zone and in the
soot oxidation zone where oxygen is present. The particle growth is time-
dependent, which means that longer time in the oxygen-poor combustion-plume
produces larger particles. Soot formation has also been found to be strongly
dependent on the oxygen entrainment in the lift-off length. The soot production
increases with increased temperature due to decreased air entrainment into the
spray [23]. Conversely, increased temperature during post-oxidation effectively
reduces soot.

Figure 8. Schematic plot of soot formation in combustion [9].

Figure 9. Schematic plot of soot formation in a diesel flame [24].

Zheng et al. measured particle nanostructure, fractal dimension, and
size at various crank angles (CA) in an HD diesel engine using a total cylinder
sampling system followed by high-resolution transmission electron microscopy

10
(TEM) and Raman scattering spectrometry [25]. They found that the particle size
varied with CA. Particles start small, expand to their maximum size in the early
diffusion combustion phase, and decrease in size as combustion proceeds. If the
injection pressure increases, the PM size during combustion decreases [26].
Figure 10 is a TEM image of particles captured inside a diesel flame during
combustion in an engine. The size of the soot particles is directly affected by the
injection pressure. The time for surface growth and later coagulation inside the
flame decreases with the flame velocity, thereby decreasing the size of the
particle. With smaller particles, the post-oxidation is affected by decreased
oxidation time.

Figure 10. Soot particle size at different injection pressures, [26].

PM varies in size distribution and consists of different compositions (PAH,
carbon, HC, oil, etc.), which makes the emission hard to quantify. In this work,
PM is called smoke or soot and is measured by filter smoke number (FSN) or
optically with the 2-colour method unless another method is mentioned. These
two methods are simple approximations to quantify diesel particles and are
widely accepted in the field.

11
1.2.4 Soot oxidation
Most of the PM that forms during combustion is converted from carbon or
hydrocarbon to combustion products during the post-oxidation process. Soot is
also oxidised during the precursor, nuclei, and particle stages in the soot
formation stages. However, the production of soot is normally much higher than
its oxidation in this phase.

Soot oxidation is kinetically controlled, and the time for oxidation is

highly dependent on temperature mixture and mixing rate. According to [27],
the temperature needs to be higher than 1300 K for particulate soot oxidation.
The oxygen attaches to the PM surface (absorption) and then detaches with a
fuel component from its surface. It has been shown that the dominant
contributors to soot oxidation in diesel engines are radical species, especially
hydroxyl (OH) [27]. OH is most dominant at rich and stoichiometric conditions,
and oxygen and OH contribute to oxidation at lean conditions. OH production
increases with increased temperature, and this can reduce soot in the post-
oxidation phase [28]. As stated earlier, increased temperature during diffusion
combustion increases the soot production due to less air mixing into the lift-off
region. Conversely, a higher temperature in the post-oxidation phase reduces
soot significantly.

1.2.5 Carbon monoxide

CO emissions are usually low and develop from incomplete or cold combustion in
diesel engines. Increased EGR, late combustion phasing, lack of oxygen, and too-
high swirl number (SN) can increase the CO emissions. HC and CO can easily be
treated in an oxidation catalyst mounted after the engine at an exhaust
temperature above 250°C–300°C.

1.3 Turbulent flows

Most of the flow in nature and in engineering applications is turbulent, and
turbulent flow is the dominant flow type in internal combustion engines.
Turbulence can be observed all around us. For instance, stirring our drink is
turbulent, the cumulus clouds are turbulent, and the boundary layer around a
sailboat is turbulent. Without turbulence, many things would not work. For
example, the combustion in a spark ignited (SI) premixed engine would be too
slow, and the maximum possible engine speed would be limited to a couple of
revolutions per minute. Turbulence is characterised by its irregular random
nature with circular motions (eddies) in all three dimensions, as shown in Figure
11. It behaves randomly in time and space and always occurs at high Reynolds
numbers (Re). Re is defined as

 u d ud
Re   , (2)
 

12
where µ is the dynamic viscosity, ρ is the density, d is the pipe-flow diameter or
a characteristic length for the problem, u is the free-stream velocity, and  is
the kinematic viscosity.

Figure 11. A side view of a turbulent boundary layer near a surface. The arrow
marks the flow direction [29].
Turbulence cannot maintain itself, and it depends on its environment
to obtain energy. Turbulent flows are generally shear flows. If the energy supply
is shut off, the turbulence quickly dissipates and is transformed to heat. As
turbulence dissipates, the large eddies lose their kinetic energy first to smaller
length scales and then to heat. The rate of conversion of turbulence into heat by
molecular viscosity is called the dissipation rate, ε. In a diesel engine, both large-
and small-eddy turbulence is created during the inlet stroke. Only the large flow
structures can survive for a longer time in the cylinder and have a chance to
affect combustion when the inlet valves (the energy supply) are closed. The
mean small-eddy turbulence lifetime is much shorter than the time for induction
and compression [30].

The transition from laminar to turbulent flow is poorly understood

and does not occur at a specific Re at which point the flow is either one or the
other. The scientist Osborne Reynolds (1842–1912), who first experimented with
high Reynolds numbers in pipe flow, reported that the flow becomes turbulent at
a critical Re ≈ 13,000. His experiments were repeated in the 1970s in
Manchester where Reynolds’ original experimental apparatus still exists. The
experimenters determined a critical Re « 13,000, much less than Reynolds
results. The reason why it differs can be found in that the flow is very sensitive
to external disturbances during the transition from laminar to turbulent flow.
Small external vibrations from for example traffic easy disturbs the transition
region and the flow become turbulent at lower Re numbers. Back in Reynolds
days those outside disturbances can be expected to be much smaller and thereby
the reason why he reached higher Re numbers before turbulence occurred.

The circular motions in the turbulent layer increase the mixing of the
reactants and products significantly, which increases combustion velocity. The

13
fuel, oxygen, free radicals, and heat from the combustion are mixed and exposed
to each other more rapidly compared with laminar combustion. A wide range of
length scales exists in turbulent flow, from the biggest dimensions of the flow
field to the diffusive-length scales of molecular viscosity. The smallest scales,
called Kolmogorov’s microscales, have relatively small time scales, and this
makes them statistically independent of the big and relatively slow vortices. The
large eddies lose most of their kinetic energy during one turnover and the energy
goes into smaller length scales (called dissipation). This means that turbulence is
a strongly damped non-linear stochastic system [31].

A schematic sketch of a turbulent boundary layer near a wall is

shown in Figure 12. From the stochastic boundary layer profile, a time-averaged
thickness profile can be plotted that describes the mean thickness of the
boundary layer,  (x) . The instantaneous velocity, u , can be divided into two

velocity components, the mean, U or u , and the fluctuating part u  . The

Reynolds decomposition is

ui  Ui  ui . (3)

In Figure 12, u  and v are the fluctuating velocities in the x- and y-directions.
The z-direction velocity, w , is not shown in this figure.

The velocity components are zero at the wall, and outside the
boundary layer the velocity is the same as the free-stream velocity, U  . A time-
averaged velocity profile for the turbulent layer can be plotted that describes the
behaviour of the mean velocity. Compared with the laminar case, the turbulent-
velocity profile has a higher flow velocity near the wall. The difference is
explained by the fact that the transverse transport (transport in the y-direction)
of momentum and vorticity in laminar flow is driven by the viscous shear stress
in the fluid. In the turbulent case, the transverse transport is driven by
convection and the fluctuating turbulent eddies. This creates higher velocity,
momentum, and friction near the wall in the turbulent case. The drag coefficient,
CD, is also higher compared to the laminar case.

14
 U

u’

y v’
x

Figure 12. Turbulent boundary layer with time-averaged thickness, turbulent

velocity, and laminar velocity profiles.
The flow in engines is not stationary. To simplify calculations in
engine flow, the velocity components can be considered stationary for a small
period of time (or for a small crank angle window). The mean velocity for
stationary flow is [31]

1 t0 T
U i  lim
T  T  t0
ui dt . (4)
The mean value of the fluctuating part (velocity) is zero by definition:

t0 T
 u  U i dt  0 .
1
ui  lim i (5)
T  T t0
In turbulent flow, convection dominates over molecular diffusion.
Turbulence has fast shifts in pressure and velocity. Two basic equations describe
the motion of the gas: the mass conservation equation and the momentum
equation. Considering a control volume, the mass in the control volume ( M cv ) is
the mass transported into the volume minus the mass transported out of the
volume:

M cv
  m   m . (6)
t in out
The mass can also be described in terms of velocity (U) and density (  ), which
yields the mass conservation, or continuity, equation [32]:


   U   0 , (7)
t
where

3

 . (8)
i 1 xi
The momentum equation, or Navier-Stokes equation, is derived from Newton’s
second law and relates the fluid particle acceleration to the surface forces and
body forces. The Navier-Stokes equation is [32]:

15
 DU 1
  p    2U . (9)
Dt 
The Navier-Stokes equation, together with the continuity equation, describe the
conservation of mass, momentum and energy in a flow field. With the restriction
of an incompressible flow field, the energy equation can be neglected to describe
the flow. In compressible flow, such as supersonic flow or when heat transfer is
involved, the energy equation cannot be ignored.

When velocity differences initiated by fluid motion exist between the

fluid particles, shear stress is generated. When a particle with a mass is
transported in a flow with a different velocity than the particle, shear stress
occurs and this accelerates or decelerates the particle. The total mean shear
stress for two-dimensional flow is

U
   uv , (10)
y
where   uv is the turbulent stress, or Reynolds stress, and the other
component is the viscous stress. Reynolds stress is an internal stress that acts
on mean turbulent flow. The viscous stress is acting on the particle by the fluid
viscosity.

Vorticity is when a fraction of the fluid rotate around itself, in smaller

or bigger volutes. If the vorticity is zero, the fluid can still move in a curve, but it
does not rotate around its own axis. Vorticity has the dimension of frequency
(1/s) and is a good source of turbulence production. Vortisity can be described as

 w v   u w   v u 
    u    e x    e y    e z . (11)
 y z   z x   x y 

1.4 Swirl and tumble flow

Airflow inside the engine cylinders is commonly characterised by swirl, tumble,
and turbulence intensity. Swirl and tumble are large-scale vortices that can exist
inside the cylinder, either on their own or in combination. The vortices are
created during the inlet stroke when the piston moves down and the inlet air
passes over the inlet valves. Depending on the inlet port design, swirl and/or
tumble vortices are created and conserved (or dissipate slowly) in the cylinder
when the inlet ports are closed. Unlike small-scale vortices, these large-scale
vortices do not dissipate quickly and survive for an extended time. Swirl survives
longer than tumble and can thereby affect the post-oxidation process. Friction
against the cylinder walls and within the airflow causes the swirl flow to slowly
dissipate. Swirl, Swirl , is the angular velocity around the cylinder centre axis, and
tumble, Tumble , is the angular velocity perpendicular to the cylinder axis. Both
variables can be normalised to the engine crankshaft angular velocity, Engine . The
resulting dimensionless numbers, swirl number (SN) and tumble number (TN),
are

16
 Swirl
SN  and (12)
Engine
Tumble
TN  . (13)
Engine
When both swirl and tumble exist, they combine to create one large
vortex. Normally, swirl is used in direct injection (DI) diesel engines and tumble
in SI engines. With variable valve actuation (VVA) or by blocking one inlet port, it
is possible to control SN and TN. VVA is used in light duty (LD) engines [33] [34]
but is not yet common in HD diesels, although research and development is in
progress [35]. Port designs on diesel engines have historically been very
important [36] [37]. SN has been an important factor for good combustion and
low smoke at moderate injection pressures. Recently, research on the injection
system using higher injection pressure and EGR has shown significantly
decreased emissions. Although SN has a demonstrable effect on emissions and
combustion [38], SN did not change appreciably until the introduction of today’s
high-injection pressures. Today, some manufacturers produce quiescent
combustion systems with nearly no swirl and this is believed to lower the heat
transfer.

During the engine cycle – from the time when swirl is created and the
inlet valves are closed – the swirl rotational velocity changes during compression
and combustion. During compression, the swirl rotational velocity increases at
the end of the compression stroke when the airflow is forced into the piston
bowl. The radius is reduced while the momentum is conserved leading to
increased angular velocity. When the piston moves down again, the opposite
happens. The flow also slows down due to the friction against the combustion
chamber walls.

If the swirl flow is assumed to be conserved for a small time window,

the swirl vortex conservation can be observed if the continuity equation (7) and
momentum equation (9) are rewritten in polar-cylindrical coordinates for the
mean flow. The details can be seen in [39]. The tangential equation from this
momentum equation describes the conservation of mean flow angular
momentum. If a fluid element is moved outside its normal track around the
cylinder axis (increased radius) and maintains its angular velocity in the swirl, it
will have an angular momentum loss compared with the surrounding fluid
particles. The centripetal force acting on the fluid element is lower than the
opposite net pressure force caused by the mean radial pressure gradient. The
result is that the fluid is forced into its normal equilibrium orbit. This
demonstrates why swirling flow in a cylinder can be stable and does not dissipate
into smaller-scale turbulence as fast as unstructured turbulent flows when the
energy source is shut off. Thus, it is easy to understand why swirling flows in an
engine are often modelled as a solid-body rotation. Even if PIV measurements
show that a perfect solid-body rotation does not exist [40] [41], this can still be
a reasonable assumption in modelling flow before combustion.

17
 Tumble is of paramount importance for an SI engine to increase combustion
velocity. The tumble vortex is transformed into small-scale turbulence around top
dead centre (TDC), due to the geometric change of the combustion chamber
during compression. In the SI engine, the fuel and air are premixed before
combustion, and a spark ignites the mixture. The flame front propagates through
the premixed air and fuel. Initially, the flame propagation is laminar with a
velocity around 0.3 m/s, but later it becomes turbulent with a velocity on the
order of 10–80 m/s depending on the turbulence intensity in the cylinder [5].

Turbulence is also of great importance for diesel combustion. The spray

creates a lot of turbulence, which speeds up the mixing process. Higher injection
pressure means increased turbulent kinetic energy.

1.5 Squish flow

The squish region is the flat area located on the outer radius of the piston. This
area is normally designed to have a small clearance to the cylinder head, and in
a diesel engine there are several reasons for this. First of all, as much of the air
trapped in the cylinder as possible should be located in the piston bowl cavity at
TDC to contribute to the combustion when the fuel is injected [42]. For a given
compression ratio (a higher compression ratio gives higher efficiency to a certain
level), the heat transfer at TDC is relatively high and by decreasing the volume
outside the bowl the heat transfer is also lowered. During the late stage of the
compression, just prior to TDC, the volume that is above the squish area (the
squish volume) rapidly decreases. The air trapped in this volume is forced
towards the centre of the cylinder and creates a flow. This squish flow can then
affect the combustion depending on when the fuel injection starts.

The squish flow affects the swirl in the bowl, and depending on how
strong the swirl flow is, the squish flow contributes in different ways. According
to [39] and [42], in an LD engine at moderate SN, the squish flow flows into the
bowl horizontally and creates a rotating vortex in the bowl as illustrated in Figure
13. At high SN, the squish is deflected from the horizontal track to follow the
bowl geometry down into the piston cavity. This results in a change in the
direction of the created vortex in the bowl. This flow behaviour is valid for deep
piston bowl designs found in LD engines. In HD engines, the bowls are normally
shallower with a smaller squish band. This changes the flow behaviour, and it can
be assumed that the squish flow has a smaller impact on the flow field in an HD
engine due to the smaller squish band.

18
Figure 13. Squish flow interaction with swirl flow in the piston bowl [39]. This is
a numerical simulation of flow velocity at different SN, here called RS. Flow
velocity (Sp) is expressed as a fraction of mean piston speed.

1.6 Engine transients

For fast engine load build up, called transient, it is a challenge to keep the
exhaust emissions at legal levels. In Figure 14, a transient load build up for a
turbo diesel engine is plotted as IMEP and inlet plenum pressure at 1000 rpm.
This transient is the basis for the later selected load points in this work. The
requested load, the blue line, is the desired value, and the red line shows the
actual load on the engine. The reason for the slow load increase is that the air
supply from the turbocharger, the green line, needs time to build up the boost
pressure. The amount of fuel must be restricted during the boost pressure build-
up before it is at a normal operating level. If the fuel mass is not restricted, the
in-cylinder lambda (λ) will fall below a critical level and the engine will start
producing high levels of soot emissions.

The engine electronic control unit (ECU) controls low λ operation.

Different models are implemented in the ECU to predict the amount of oxygen
trapped in the cylinder allowing the right amount of fuel to be injected for each
cycle during the transient. As shown in examples of control system models [43]
[44] [45] [46], the λ cannot pass below a critical value (typically λ = 1.25–1.30
for an engine without a diesel particle filter) before smoke emissions increase
rapidly. Besides resulting in a disappointed driver, who wants engine power as
fast as possible, the engine operates at unfavourable conditions for a longer time
period. As long as the air is restricted, an EGR engine cannot use EGR because

19
the turbo pressure is too low. The in-cylinder mean temperature increases due to
less gas mass trapped in the cylinder resulting in higher NOx emissions, and the
engine efficiency is negatively affected if the SOI needs to be later to maintain
NOx emissions at acceptable levels. If the combustion system can maintain low
emissions of PM and NOx at a lower λ, the available exhaust energy and engine
torque will be increased. This results in a faster build-up of boost pressure and
engine torque.

25 2.5
Load 3 Load 4

20 2.2

Load 2

Inlet pressure [bar]

15 1.9
Load 1
IMEP [bar]

10 1.6

5 Req. Load IMEP [bar] 1.3

IMEP engine out [bar]
Inlet pressure [bar]
0 1
-1 0 1 2 3 4 5
Time [s]

Figure 14. Turbo diesel engine transient from 3 bar IMEP to full load. The
requested load, actual load, and inlet pressure are plotted versus time.
The investigations performed in the publications attached to this
thesis are based on the engine transient in Figure 14. Stationary load points
were selected along the actual transient load curve (the red line) to enable more
in-depth studies. The selected load points indicated in Figure 14 were tested in a
single-cylinder engine, an optical engine, and in computational fluid dynamics
(CFD) simulations with the same boundary conditions. The advantage of this
approach is that modifications to in-cylinder flow and injection parameters can be
easily studied in a controlled way.

20
2 Project motivation
In-cylinder bulk flow rotation in the form of swirl has been shown to have an
impact on soot emissions. Today, much of the research is focused on the
injection system, and soot emissions have been successfully depressed with
increased injection pressures. As a result, some HD engines are manufactured
with quiescent combustion chambers with no swirling motion. However, little is
known about how today’s high injection pressures in combination with swirl
influence the combustion and emissions of an HD engine. LD engines normally
have some variability in swirl levels. For example, the swirl can be increased
during an engine transient, and the increased swirl combined with high injection
pressure reduces soot emissions. This is also the case in HD engines, but the
mechanisms behind the reduced soot emissions are not well understood.

The purpose of this study is to determine what the in-cylinder flow

looks like and how it affects the combustion and emissions at critical engine
operation points in an HD diesel engine. Furthermore, the study seeks to create
an understanding of how the flow fields in the cylinder arise, what affects the
flow, and how the flow can be changed. The expected outcome of the project is
the elucidation of parameters that describe why changes in the in-cylinder flow
have an effect on measured engine-out emissions. The specific aims of this study
are to

1. Determine how injection, swirl, and tumble affect

measured engine-out emissions and heat release.

2. Examine how changes in injection pressure and swirl

influence the in-cylinder flow behaviour at different stages
in the cycle.

3. Explain why injection pressure, swirl, and tumble affect

soot emissions.

4. Compare the trends observed experimentally in the optical

engine to computer simulations.

3 Methodology
Fast load increase means low λ operation with less remaining oxygen for post-
oxidation of the fuel. Therefore, the demands on the mixing process in the
cylinder are higher. The procedure for the work presented here was to first
measure what is happening during an engine transient and examine when the
critical load points occur in a six-cylinder turbo diesel engine. Then, the critical
points were repeated in a single-cylinder engine with an active valve train (AVT)
that made it possible to allow variable in-cylinder airflow. The simulation tool GT-
POWER was used with constant flow-rig measurements to quantify the airflow
before combustion. To investigate how the flow field influences the combustion,

21
optical engine measurements were taken using a high-speed camera. The flow
inside the cylinder during combustion was quantified from the captured optical
engine images using a cross-correlation program. In-cylinder soot formation and
temperature were calculated with the 2-colour method. Simulated, measured,
and processed data on flow quantities, emissions, and combustion were
combined to examine the airflow effect on diesel combustion during transient and
then compared with Reynolds-averaged Navier-Stokes (RANS) CFD simulations.

3.1 Experimental equipment

Three different engines were used in this experiment: a six-cylinder
turbocharged production engine, a single-cylinder engine, and an optical engine.
The three test engines have nearly the same combustion system layout. The
common engine specifications (injection system, bowl geometry, cylinder head
design, etc.) are listed in Table 1. Some deviation in system layout was
inevitable due to the different engine constructions. The differences between the
test engines are listed in Table 2. The obvious difference is that bore and stroke
differ slightly between the engines. The reason is that 127/154 mm is the old
engine configuration, and 130/160 mm is the new configuration. The optical
engine was updated with the new bore but not the new stroke. The engine layout
is based on a Scania DL Euro 5 combustion system equipped with a common rail
XPI injection system capable of injection pressures up to 2,500 bar. This is a
swirl-supported combustion system with a re-entry combustion bowl design that
does not need any extra after-treatment system for Euro 5 emission standards
when EGR is used. It has a 4-valve cylinder head with a centrally placed 8-hole
injector.

Table 1. Engine specifications.

Common test engine data
Compression ratio 17.3:1
No. of valves 4
Injection system Scania common rail XPI
Injector holes 8
Spray angle [deg]
(° between cyl.head and spray) 16
Injector hole diameter
(inner/outer) [mm] 0.187 / 0.163
Max Injection pressure [bar] 2500

Table 2. Test engine individual specifications.

Engine type Optical engine Scania single cylinder Scania DL Euro 5 engine
Bore/stroke [mm] 130/154 127/154 130/160
Connecting rod [mm] 255 255 255
Valve system Camshaft Active valve train Camshaft

22
 The six-cylinder engine was used to measure the transient behaviour,
and the data provided was used to set up a single-cylinder and optical-engine
test series. The single-cylinder engine was equipped with a Lotus AVT system.
With this system, the valves were hydraulically controlled enabling the valve
profiles to be shifted during operation. With different valve profiles, the airflow in
the cylinder can be modified over a wide range. The airflow was quantified with
SN, TN, and normalised turbulence intensity (NTI). The 1-D simulation program,
GT-POWER, was used to calculate these quantities. Two different tested cylinder
heads were used and measured in a constant flow rig. Swirl and tumble at valve
lifts from 1 mm to 15 mm were measured at 1 mm increments for each valve
individually and with the two valves together. Valve profiles were created in
MATLAB. These were then used in the Lotus AVT system and in GT-POWER. In
this way, the SN, TN, and NTI could be varied. The tested load points in the
single-cylinder engine were then repeated in the optical engine. The combustion
images were captured with a high-speed Phantom v7.3 camera and evaluated
using LaVision DaVis 7.2 PIV and AVL thermovision software.

3.1.1 Single-cylinder engine with AVT system

The Lotus AVT system is a fully hydraulic system with one hydraulic cylinder
coupled to each valve in the cylinder head as shown in

Figure 15. Hydraulic oil was supplied with a pressure of approximately 200 bar to
each side of the piston inside the hydraulic cylinder making the piston move. The
oil flow was controlled by a servo valve that directs the oil to one side of the
piston at a time. In this way, the engine valves are controlled and a modelled
valve profile can be used. During engine operation, it is possible to change the
valve profiles. Some examples of valve profiles used in this work are plotted in
Figure 16. This system enables SN variations between 0.4 and 6.7 and TN from
0.5 to 4.0 with the cylinder head configuration used in this work.

23
 a

Figure 15. The Lotus AVT valve actuator with inlet valve (a) and the actuators
on the cylinder head (b)

14 15 mm std
10 mm std Std. engine profile
5 mm std
12
15 mm step
10 mm step
10 5 mm step
Valve lift [mm]

0
300 350 400 450 500 550
CAD
Figure 16. Lift profile versus crank angle degree (CAD) for some of the tested
valve profiles.

24
3.1.2 Optical engine
The optical engine layout is shown in Figure 17 with the two different tested
piston bowls. A piston extension connects the original piston to the optical piston
that is fitted into a liner. The Phantom v7.3 camera is installed next to the
engine, and the combustion light is transferred to the camera by a mirror
mounted inside the piston extension. The engine is capable of running with
cylinder pressures up to 160 bar. A titanium clamping ring was mounted above
the piston glass to fix it. This restricted the field of view to a diameter of 80 mm,
compared with the total cylinder bore of 130 mm. Two different shapes of the
piston bowl glass were tested, a bowl-shaped piston bowl and flat piston bowl. A
schematic of the spray path is plotted in the two piston bowls in Figure 17. The
optical engine had a normal camshaft valve mechanism. To change the in-
cylinder airflow, two different cylinder heads were used. To further extend the
possible airflow in the cylinder of the optical engine, one of the inlet ports could
be blocked.

Figure 17. Principal layout of the optical engine is shown on the left. On the
right, the two tested piston-bowl shapes are shown with schematic of the spray.
Due to the long piston extension, the effective compression ratio is
lower than the geometrical compression ratio and decreases with cylinder
pressure. At 160 bar, the distance between cylinder head and squish area on the
piston increased by 1.5 mm compared with atmospheric pressure. To
compensate for the lower compression ratio, the boost pressure and inlet
temperature were increased so the motoring cylinder pressure at TDC in the
optical engine was equal to the single-cylinder engine. The λ was slightly higher
in the optical engine compared with the single-cylinder engine. The increase in
inlet temperature compensated for the increased ignition delay in the optical
engine because only one combustion event was performed during the
measurement.

25
3.2 Simulation tools and calculation methods

3.2.1 CFD numerical modelling

AVL FIRE 2010.1 software was used to perform RANS CFD simulations in this
work. A 45° sector of the engine cylinder was modelled corresponding to the
sector for one injection hole. The spray-oriented mesh, shown in Figure 18,
consisted of about 58,000 cells at TDC and about 214,000 cells at bottom dead
centre (BDC). The FIRE module ESE Diesel mesh tool was used. The turbulent
flow was modelled using the k-zeta-f turbulence model [47] [48], and the spray
droplet break-up was modelled with the WAVE model [47] [49]. The 3-Zones
Extended Coherent Flame Model (ECFM-3Z) [47], [50] was used for combustion
modelling. Boundary conditions were set to replicate the performed engine test
as closely as possible, shown later in Table 3. The simulation started at inlet
valve closure (IVC), -136° after TDC (ATDC), and ended at exhaust valve
opening (EVO), 124° ATDC. In the CFD, SN was set at BDC according to the
measured swirl data recalculated with GT-POWER from steady-state flow bench
tests. The assumption of a centred swirl was used to enable sector simulation of
1/8 of the total combustion chamber volume.

Figure 18. Sector mesh of in-cylinder fluid domain at TDC.

3.2.2 1-D simulation modelling

To simulate the outer gas exchange system, GT-POWER was used with a 1-D
approach. In a pipe that contains a pulsating flow, the velocity in the axial
direction in the pipe is significantly higher than the velocity and flow in the cross-
sectional plane. By modelling the flow in the axial (x-) direction only, the
continuity, momentum, and energy equations can be greatly simplified. This is
the fundamental approach of 1-D simulation. GT-POWER handles the flow in the
different objects with a space-discredited form of the continuity, momentum, and

26
energy equations. More information about 1-D simulation in GT-POWER can be
found in [55].

GT-POWER was used to calculate the swirl (SNGT) and tumble (TNGT)
numbers at BDC. Those numbers were then used as boundary conditions in the
CFD calculations. In Publication I, calculation of swirl number by GT-POWER is
compared with the widely accepted Thien method [51]. GT-POWER assumes
solid-body rotation in the cylinder and uses flow-rig data to calculate SN. The
flow-rig data for respective valve lift is used by the program to estimate the SNGT
and TNGT. The program calculates the air mass flow passing into the cylinder
depending on the pulsating pressure and flow in the inlet system and the valve
lift. With the calculated airflow, the momentum contribution from the airflow
moving into the cylinder is added to the rotating air mass in the cylinder.

Measured engine data was compared with simulated engine data to fine-
tune the model. More on how this can be done is shown in [52]. The built-in
“flow” model was used to estimate in-cylinder airflow and heat transfer, and the
combustion model “CombDIJET” was used. The flow model uses a simplified
cylinder geometry and a k-ε model to estimate, for example, normalised
turbulence intensity. More information about how GT-POWER handles in-cylinder
flow can be found in [53] [54] [55]. The tumble algorithm in GT-POWER is quite
similar to the calculation method for swirl, and this is shown in detail in [53].

3.3 Overview of flow measurement methods

There are many different techniques available to visualize and measure flows in
gaseous fluids. Some common methods are hot wire measurements, laser
Doppler anemometry (LDA), phase Doppler anemometry (PDA), particle image
velocimetry (PIV), and particle tracking velocimetry (PTV).

3.3.1 Hot wire measurements

Hot wire measurement is a common technique that allows measurement of both
the mean velocity and the turbulent velocity wherever the probe is placed. By
heating a thin wire, normally made of platinum-coated tungsten, and placing it in
the gas stream (turbulent or laminar), the energy that is needed to maintain a
constant temperature is proportional to the gas mass flow that passes over the
wire. By choosing a thin wire, the probe’s response time can be kept low so both
the fluctuating part (u’) and the mean part (U) of the turbulent velocity can be
measured [41]. It is possible with this method to measure turbulence up to the
50 kHz range. However, hot wire measurements around TDC in an engine
produce large errors with respect to turbulence intensity levels. The measured
values have been reported to be a factor two or greater than those obtained with
LDA, according to publication [41].

27
3.3.2 Laser Doppler anemometry (LDA)
With LDA, two intercepting laser beams are directed to a small measurement
volume of interest. The examined flow is seeded with small particles that follow
the flow. When a particle crosses this small volume, the beams light up the
particle and cause interference fringes to appear in the sample volume. The
periodic light reflection is used to calculate the particle velocity that passes the
small measurement volume. This gives just one velocity vector for the current
volume. For the entire flow field in an engine, the measurement volume needs to
be moved. This is done by focusing the intercepting beams on another location.
As a result, vector data cannot be captured at the same time for all interesting
locations. Because the flow is normally turbulent in an engine, with its randomly
nature, measurements unevenly distributed over time are not desirable.

3.3.3 Particle image velocimetry (PIV)

In PIV, the flow is seeded with particles as with LDA. The technique is an indirect
velocity measurement because it measures the particle movement and not the
flow itself. The particles are illuminated by a thin laser sheet, and a camera takes
two images (either a double exposure in a single image or two separate images)
with a small timeframe in between. The particle movement in the plane can then
be estimated between the images. By first dividing the images into smaller
interrogation windows, a mean value of the movement of particles inside these
windows can be estimated with statistical methods such as auto- or cross-
correlation. It is assumed that all particles inside an interrogation window move
homogeneously between the frames. A small interrogation window is, therefore,
important to obtain good resolution of the flow and to visualize turbulent
fluctuations with a good correlation value. With known spatial mean movement
of the particles inside the window and the time delay between exposures, the
mean velocity can be calculated for the interrogation window. When the laser
sheet is thin and one camera is used, only the 2-D spatial movement can be
resolved. There is also 3-D PIV, which uses two cameras and can resolve all
three velocity components. Normally, the cameras used are charge coupled
device (CCD) types that take two double exposures (on the same frame) before
the image can be read out from the sensor element. The time for reading the
information from the sensor limits the frame rate to around 100 readings per
minute on normal CCD cameras [56], which is too slow for the time scale of the
turbulent part of the flow. To catch the fluctuating velocity part of the flow, a
number of exposures need to be captured, and then a probability density
function (PDF) is applied to calculate u’. If high-speed cameras with
complementary metal-oxide semiconductor (CMOS) sensors are used instead,
the number of exposures can be increased up to the kHz range. This is called 4-D
PIV, and this technique has been made possible due to recent developments in
camera and computer technology that have led to rapid increases in
performance.

28
3.3.4 Particle tracking velocimetry (PTV)
PTV is a similar technique to PIV. Both use seeding particles and laser sheets to
light up the particles. The difference is that PTV identifies every single particle in
the first image and tries to find the same individual particles in the next
exposure, while PIV uses cross-correlation of integration volumes. Therefore, the
seeding does not need to be as dense as with PIV. In PTV, probability functions
are used to guess where the different particles have moved.

3.3.5 Combustion image velocimetry (CIV)

None of the above-mentioned techniques can easily measure flow during full-load
combustion due to the intense black body radiation from soot. All other possible
light sources are easily drowned due to the broad black body illumination
spectrum, as seen in figure 39 (soot radiation spectrum is between candle and
photo lamp). With CIV, the soot radiation is used as tracking source, which was
also done in [77] & [85]. As in the case with PIV, two pictures are captured with
a high speed CMOS camera within a small timeframe. The images are divided
into interrogation windows and cross-correlated to calculate the mean velocity
vector between two pictures (from t to t + Δt) for every window:

V  V x2  V y2 . (14)
The main difference between CIV and PIV is that in CIV no seeding particles are
introduced into the cylinder to be illuminated by a laser. The natural light from
the combustion and the light gradients that occur in the cylinder are used as
tracers when the cross-correlation is made. In Figure 19, two images and the
result of the evaluation are shown. The colour images are converted to grayscale
images before cross-correlation. A colour scale on the resulting vector plot
indicates the mean velocity in every “box.” The arrows indicate the direction of
the flow, and the arrow length indicates the velocity magnitude. Black dots
indicate missing or erroneous data or zero velocity.

vec2/B00080.vc7

20
-10
18
-20 16
[ (ux )2 + (uy )2 ]1/2 (m/s)

14
-30
Pic 1: At 25.90° ATDC
12
y (mm)

-40
10

-50 8

6
-60
4
-70
2

Pic 2: At 26.07° ATDC

10 20 30 40 50 60 70
x (mm)

Figure 19. Evaluated combustion pictures at 25.9 and 26.1° ATDC and the
resulting velocity field.

29
3.4 The CIV method
To answer the question of how in-cylinder flow affects the combustion and
subsequent emissions, the flow field during the combustion period must be
determined. The CIV method can be used to obtain flow data in an engine, as
shown in [77]. In the earlier work CIV was applied to relatively low loads at low
time resolution. In this work CIV is operated up to full load with a high time
resolution, enabling caption of high flame velocity gradients. In this section, the
CIV technique is described.

3.4.1 Cross-correlation applied to CIV

To extract velocity information from PIV/CIV measurements it is necessary to
introduce statistical methods, for tracing the movement of "groups of particles",
or more general, of coherent structures in time. These groups are identified by
structures in the images caused by the radiation pattern from the glowing soot
particles. The mathematics applied to such structures are described for
(mathematical) particles in this section. The images are divided into small
regions called interrogation windows, shown in Figure 20. A correlation plane is
introduced in the interrogation window, and the highest correlation peak
corresponds to the most likely displacement of the particle ensemble in the
interrogation window.

Figure 20. Particle displacement in an interrogation window. The position of one

particle is captured at three time steps, t, t’, and t”.
The idea of cross-correlation is to combine two signals and determine
the location of maximum overlap. Consider two signals that have a displacement
t in one direction but are otherwise equal. The cross-correlation of the two
signals f(t) and g(t), where t can be time or distance, is [57]

(15)

where is the complex conjugate of f(t) and * is the convolution. The

convolution describes the amount of overlap the two signals have when they are
shifted over each other. The convolution is defined by

(16)

30
By letting and and applying equation (16) to equation (15), the
cross-correlation equation becomes

(17)

where is the distance between the two signals. When cross-correlation is

applied in two dimensions and the number of points (particles), N, for each signal
increases, the number of operations increases by N4. Therefore, the number of
operations increases rapidly and Fourier transforms are used in PIV and CIV
evaluations.

In PIV or CIV, two images are divided into interrogation windows and
each window contains a number of particles as shown in Figure 21. The particles
inside one interrogation window (intensity field) are assumed to move in the
same direction and with the same velocity and thus have a constant
displacement d. The particles in each window are combined into one function that
describes their individual positions. Cross-correlation of the functions from the
two interrogation windows is applied, and the total displacement of the particles
is calculated as shown in Figure 22.

Figure 21. Particles in an interrogation window at time t and t’. Particles x1-x3 in
intensity field I shift to new positions at t’, resulting in intensity field I’ [56].

Figure 22. Cross-correlated interrogation window RII’. The particles cross-

correlated between the two interrogation windows and the resulting
displacement d of the particles is shown.

31
 The resulting cross-correlated interrogation window contains
fluctuating background noise, , the convolution of the mean intensities, , and
the self-correlated peak, , as seen in Figure 23. The self-correlated peak (only
one per interrogation window) represents the overall displacement of all particles
in the interrogation window. If a peak is not so sharp, the correlation value has a
lot of background noise or the particles inside the window have large differences
in velocity. Unmatched particles in the window also increase the error, and this
comes from particles leaving or entering the interrogation window. The earlier
assumption was that all particles inside one window have the same velocity, but
some differences normally occur. A typical correlation peak for the CIV technique
is shown in Figure 24. Red indicates high correlation values, and the white dot
indicates the highest value. More details about the cross-correlation mathematics
can be found in [56].

Figure 23. The correlation peaks resulting from the cross-correlation of two
images in an interrogation window. RF is the fluctuating background noise, RC is
the convolution of the mean intensities, and RD is the self-correlated peak.

32
Figure 24. An example of the correlation peak obtained with the CIV technique
in one interrogation window.

To increase the accuracy of the resulting vector fields, different

methods can be applied. Overlapping interrogation windows can be used to
overcome the problem of particles from one image leaving the interrogation
window in the second image. A 50% overlap of the windows has been shown to
effectively reduce correlation errors [58]. Also, a multi-pass (decreasing size)
window can be used in combination with an overlapping window (normally found
in PIV software). The vector fields are calculated with a number of iterations with
a decreasing interrogation window size. In this work, a window size of 64 x 64
pixels is used, which is iterated down to 12 x 12 pixels. The outcome is a more
precise result. A first-reference vector field is calculated with a window size of
64 x 64 pixels, called “1st pass” in Figure 25. In the second pass, half the window
size is used and a new vector is calculated. The reference vector, calculated in
the 1st pass, is used as a best-choice window shift. This means that the window
shift is adaptively improved to calculate more accurate vectors in the next pass.
The interrogation window is shifted to maintain as many particles as possible
inside the window. A smaller interrogation window can then be used with
accurate results. Using this method, the same particles are consistently
correlated with each other and are not leaving or entering the interrogation
window.

33
 Figure 25. Multi-pass interrogation window [59].

3.4.2 Comparison of PIV and CIV particle tracing

The size of soot particles created during diffusion combustion is on the order of
10–30 nm [26] [25] depending on injection pressure. The particle size is also
very much dependent on the time of combustion [25]. The particles are small in
the first phase of the diffusion combustion and largest around EOI. The sizes
decrease after EOI and during the post-oxidation [25]. Due to its small size,
tracing an individual soot particle is not possible. However, the flame structure is
traceable. As can be seen in Figure 19, the structure created by the radiating
soot particles can be cross-correlated between two images.

3.4.2.1 Soot particle oxidation rate

The time scale for oxidation of particles is much larger than the time step
between the exposures. According to the theory of soot oxidation presented in
Heywood [5], the soot oxidation rate of a particle can be estimated with the
equation

(18)

With a small starting soot particle radius = 5 nm, a particle density

3
~ 2 g/cm [5], and an oxidation rate = 10 g/cm2s [5] (at 2200 K in-
-3

cylinder temperature), the time to oxidise one particle is on the order of 1 ms

34
(6° CA at 1000 rpm). The time between the images taken with the camera is
µs (0.168° CA at 1000 rpm), which is much faster than the soot oxidation of
individual particles. This leads to the conclusion that statistically, the same
particles are captured in both images. This allows cross-correlation of the
images.

3.4.2.2 Soot particle flow tracing

In comparison to PIV measurements, a normal seeding particle diameter is about
10 µm [60], which is 103 times larger than the soot particles created during
combustion (~10 nm). The density of the seeding particles is between
= 4.2 g/cm3 for titanium dioxide (common for combustion PIV measurements)
and = 1.14 g/cm3 for nylon [61]. The soot particle density, ~ 2 g/cm3, is
within that range. ([62] reported = 1.84 ± 0.1 g/cm3 for soot). Assuming the
particle density is constant, smaller particles follow the path of flow more
accurately.

To determine how closely a particle follows a flow, the Stokes drag

law can be applied. Normally, PIV seeding particles are assumed to be spherical,
and the aerodynamic equivalent diameter of the particle, d p , is the diameter of
the particle. A soot particle is not normally spherical but is instead a complex
shape that should increase the drag even more. To keep the calculations simple,
the soot particles are assumed to be spherical as well. The Stokes drag law also
assumes laminar flow. The flow in the cylinder is not laminar, but near the
particle the flow can be assumed to be laminar if the velocity between the fluid
and the particle is low enough. A lower velocity difference means that the
particle follows the flow path better. The particle Reynolds number, Re p , should
be small enough for the laminar flow assumption and is calculated by

(19)

where is mean velocity of air around the particle, is the particle diameter,
and is the kinematic viscosity of the fluid (the surrounding air). In normal PIV
measurements the mean velocity round the particle, , can be assumed to be
0.2 m/s and = 10 µm. At atmospheric pressure and 20°C, = 0.13 for a
PIV particle. For a soot particle with = 20 nm in an engine with 100 bar
cylinder pressure and 1000°C cylinder temperature, = 0.0021. is much
lower in the case with soot particles.

At very low Re, the particle response time  p (or relaxation time) can
be calculated according to the equation

p
 p  d 2p . (20)
18

35
Particle response time is a measure of the particle’s inertia, where is the
particle density and  is the dynamic viscosity of the surrounding fluid. The
diameter term is squared, which means that the response time is very sensitive
to the diameter of the particle. When the particle response time is known, the
Stokes number, St , can be estimated with the equation

(21)

which is the ratio between the particle response time, , and the Kolmogorov
time scale, , (the turbulence time scale). The Kolmogorov time scale is the
turnover time for the smallest turbulent eddy in fully developed turbulence. If
St  0, the particle follows the flow perfectly. If St  ∞, the particle does not
follow the flow at all. If St < 0.1, the tracing accuracy error is below 1%
according to [63]. The Kolmogorov time scale can be hard to estimate in a
combustion engine with large pressure and temperature variations. Instead, the
minimum size of turbulent eddies that can be traced by a particle with a certain
weight and size can be estimated. The turbulent eddies can be expressed as
turbulence frequency (or cut frequency). By allowing a slip, s, between the fluid
velocity and particle velocity, the cut frequency can be expressed as [64]

(22)

The scalar component s of a slip velocity vector is the relative difference

between the particle velocity and the surrounding velocity of the fluid and
can be written as [64]

(23)

36
 In Figure 26, the turbulent response of an engine soot particle with
= 10 nm and = 30 nm is compared with a typical PIV particle with = 10
µm in the engine environment (100 bar cylinder pressure and 1000°C) and in the
ambient air environment (1 bar and 20°C). The results show that the soot
particle, due to its size, follows smaller turbulence length scales (higher
turbulence frequencies) compared to a normal PIV seeding particle. The higher
pressure that occurs in the engine also affects the ability to resolve smaller
length scales. As can be seen in the figure, the soot particles can, in theory,
follow very high turbulence frequencies (in the MHz range). This is normally
irrelevant when the highest interesting frequency bandwidths are in the kHz
range. Therefore, it can be assumed that the glowing soot particles in the engine
follow the in-cylinder flow much better than a normal PIV particle.

0.9

0.8

0.7

0.6
Relative Slip

0.5

0.4

0.3

0.2 Soot particle in engine, 10 nm

Soot particle in engine, 30 nm
0.1 PIV particle at atm p, 10 µm
PIV particle in engine, 10 µm
0
-2 0 2 4 6 8 10
10 10 10 10 10 10 10
Turbulent Frequency [kHz]

Figure 26. Soot and PIV particle response to turbulence frequency.

37
3.4.3 Reproduction of flame structure
In CIV, the traced soot particles are much smaller than what the camera can
resolve. The spatial resolution that the imaging system can achieve is between
1.5 line pairs per mm (mm-1) (85 mm lens with 256 pixel x 256 pixel resolution)
and 5.3 mm-1 (300 mm lens with 256 pixel x 512 pixel resolution). Consequently,
individual particles cannot be resolved, but the structure of the flame and the
glowing soot clouds are traceable. In the experiments, a typical aperture setting
(F-number) was between 8 and 11 during combustion. The exposure time was
set between 3 µs and 4 µs. Sufficient depth of field is important for later cross-
correlation of the flame structure. For optimal image quality, the F-number
should be set around 8. Aberrations occur when the F-number is too small, and
the image quality is limited by diffraction when the F-number is too large. The
depth of field is strongly limited at small F-numbers and long focal length.
Aberrations in the images can occur from refraction errors in the surface of the
lens element, which increases the blurredness. Diffraction is unavoidable when
light is travelling through an opening with a limited diameter (the aperture). The
diffraction pattern appears when light waves travel through the limited diameter
and creates illumination variations around a reproduced point light source. These
variations cause blurriness in the image.

When the aperture was set to 11 for the 85 mm lens, the depth of
field was ~0.1 m (circle of confusion = 25 µm) at a distance of 1.5 m from the
combustion chamber. For the 300 mm lens at an aperture of 11, the depth of
field was ~0.01 m. In both cases, the depth of field was in the range required to
reproduce the flame structure correctly.

Illumination intensity of the structures varies with soot

concentrations in the cylinder, and these structures create sufficient information
for cross-correlation. Before the raw images from the colour camera can be
imported to a PIV program, the images need to be converted to grayscale
images. Only the intensity peaks in the images are cross-correlated. In Figure
27, raw images of the diffusion combustion period (left) and the post-oxidation
period (right) are shown in the top row. The signal intensity within the 64 pixel x
64 pixel yellow frame is plotted in the second row. In both of the 3-D signal
graphs, the signal strength is high with sharp peaks. These peaks survive the
time step between the images (27 µs) as discussed earlier. When the signal area
is reduced to 12 pixels x 12 pixels (the red frame), the peaks are still
pronounced, as shown in the last row of Figure 27. Therefore, it is still possible to
cross-correlate the images with such small interrogation windows. It is
necessary, however, to use the PIV program’s built-in intelligent functions with
overlapping interrogation windows to be able to follow the peaks between the
frames at high flow velocities and small interrogation windows.

38
 64x64 64x64

12x12 12x12

Figure 27. Signal intensity plots at 7° ATDC (left) and at 17° ATDC (right) with
64 pixel x 64 pixel (yellow frame) and 12 pixel x 12 pixel (red frame) windows.

39
 In Figure 28, the background noise pixel values are plotted for a
section of the first image taken before the combustion starts. The amplitude of
the noise (with the camera built-in noise compensation switched on) is 5–15
pixel values for the 14-bit image. Compared to the highest possible pixel value of
16,383, the background noise is very low. The background noise comes from
stray light and noise from the electronics (random noise, fixed pattern noise, and
banding noise [65]). To compensate for some of this noise, a black reference
picture (with the lens cap mounted) was taken before the experiment began.
Some noise in the picture is inevitable because of statistical fluctuations in the
dark signal. The number of photons that fall toward every pixel fluctuates, which
is called photon noise. The dynamic range is a measure of the range between the
highest and lowest exposure the sensor can register correctly (contrast range):

(24)

S is the stray light factor (random light from the surroundings), which limits the
lowest possible dark signal. For the Phantom v7.3 camera, the dynamic range is
60 dB.

Figure 28. Noise pixel values of an image taken before combustion.

Variations in luminosity and signal strength in the images enable

mathematical calculation of peak values and cross-correlation between images,
but problems with error readings can occur if the signal is weak. Between the
flames, the light intensity is reduced compared with the signal from the flame
area. In Figure 29, peaks of around 500 pixel values to 1500 pixel values can be

40
observed in the area between the flames. The background noise was on the
order of 15 pixel values, which means that any movement of the structure
between the flames can be traced by the software. It will be shown later in
section 3.4.4 that the correlation value will decrease in this area. A filter is
applied to remove weak correlations from the resulting vector image. The
interrogation window with weak correlation is marked with a dot.

Figure 29. Surface plot of the area between flames at 7° CA ATDC.

41
3.4.4 Correlation value
The correlation value is a measure of the similarity between interrogation
windows from the first image compared with the second image. A value close to
1.0 means that the particle patterns in the windows are nearly the same. A
correlation between 0.7 and 0.8 is a good value for PIV measurements [59]. In
the left plot of Figure 30, which is a cross-correlation at the early injection
period, most of the interrogation windows have a correlation value of 0.7 or
higher. However, some areas, especially in the middle of the cylinder and in
some interrogation windows in the outer part of the cylinder, have lower
correlation values. In the middle, there are less light gradients because most of
the combustion occurs in the outer part of the combustion chamber. Less
information is available to cross-correlate thereby resulting in a lower correlation
value. In the outer part of the combustion chamber where we find low correlation
values, the flame velocity is high. Here, the problem is that the traceable light
gradient travels quickly and risks going outside the interrogation window area.
To overcome this problem, the time step between the image exposures can be
shortened. However, if the time between exposures is too short then the lowest
velocities cannot be captured correctly. This happens when the displacement of
the light gradients is so small that the camera cannot register the movement
(the same active pixels as in the previous image). In the post oxidation phase,
the velocities are significantly lower. This results in higher correlation values, as
shown in the right plot of Figure 30, and the lowest values are not located in any
special part of the cylinder.

A trade-off between measuring the highest velocity of interest and

the lowest velocity is inevitable. At low injection pressures, the velocity
difference at different CAD is not as big, and higher correlation values are
achievable over a broader CA for the same settings. At high injection pressures,
a decision needs to be made as to which velocity will be prioritised. In Figure 31,
the resulting vector fields for the correlation values in Figure 30 are shown. The
corresponding grayscale raw images can be observed in the background. Large
differences in velocity occur between injection and post-oxidation at 1500 bar
injection pressure and SN = 1.2.

42
 1
Injection Post-oxidation

0.7

Correlation value
0

Figure 30. Correlation values at early injection and post-oxidation with 1500 bar
injection pressure.

Figure 31. Resulting vector field during injection and post-oxidation.

3.5 CIV application results

Vector graphs of the flow are produced in 2-D with CA resolution from the CIV
technique. The results can then be used to calculate angular velocity and kinetic
energy etc. that can be compared with CFD simulations. This section investigates
the deviation from cycle to cycle of the CIV measurements, which is important
for analysis of the flow characteristics. Also a comparison between CIV and CFD
is made in this section.

43
3.5.1 Measured angular velocity during combustion
To quantify the rotational flow in the cylinder that survives into post-oxidation
and can affect the soot oxidation, angular velocity versus piston bowl radius was
plotted as shown in Figure 32. This can be a good measure of the in-cylinder flow
pattern and how it behaves at different CA, SN, and injection pressures.
Deviation from solid-body rotation can be a large contributor to turbulence
production that can affect the combustion [66] and, therefore, the post-
oxidation. From the resulting plot with vector fields, shown in the left plot of
Figure 32, the angular velocity was extracted for each respective radius with
respect to the geometrical centre of the piston bowl. In the right figure, the
information of every interrogation window is shown in the form of angular
velocity and its radial position. A mean angular velocity profile was then
calculated (the blue line in the graph). An imaginary solid-body rotational
angular velocity line is also shown in the figure at a level of SN = 4. As can be
seen in the figure, the measured flow values deviate from solid-body rotation
during the post-oxidation part of the combustion. The graph shows the angular
velocity at 26° ATDC. The fuel injection ends at 10° ATDC.
vec2/B00080.vc7

ui e 20
-10 2000
18 respective vector angl.vel.
1800 mean value angl.vel.
-20 16
1600
[ (ux )2 + (uy )2 ]1/2 (m/s)

14 1400
-30 ri
12 1200
omega [rad/s]
y (mm)

-40 Solid body rotation

10 1000
ez
-50 8 800

6 600
-60
4 400

-70 200
2
0
0 5 10 15 20 25 30 35 40
10 20 30 40 50 60 70
Radius [mm]
x (mm)

Figure 32. CIV vector field plot (left) and angular velocity versus piston bowl
radius for each interrogation window with the mean angular velocity at 26°
ATDC (right). The red line shows the theoretical solid-body rotational angular
velocity at SN = 4.

3.5.1.1 Standard deviation of angular velocity at combustion and

post-oxidation
Cycle-to-cycle variations between combustion events are inevitable.
Characteristically, the coefficient of variation(CoV) IMEP for diesel combustion is
on the order of 1%. For optical engine measurements with single combustion
events, it is slightly higher. Figure 33 shows three angular velocity profiles at
different CA intervals with standard deviation bars. The standard deviations were
calculated from 10 individual recorded combustion events at 10 bar IMEP, 1500
bar injection pressure, and SN = 1.2. It was observed that the standard
deviations for small radii were higher. This is because there are fewer

44
interrogation windows for smaller radii, and swirl centre that is not centred in the
cylinder and its stochastic behaviour between cycles. Nevertheless, the standard
deviation was small enough to capture the trends of angular velocity behaviour
at different CA intervals.

The angular velocity also changed with the CA. For a SOI at -1° ATDC
and EOI at 10° ATDC, the angular velocity during injection was lower than the
velocity directly after EOI for small radii, as shown in Figure 33. At later CA, the
green curve in Figure 33, the velocity decreased for small radii and increased at
radii between 15 mm and 27 mm. It was also observed that the standard
deviation was larger at lower SN than at high SN cases. At low SN, the large-
scale swirling structure is not as uniform as in the case with high SN, and this is
assumed to be the explanation for larger standard deviation in low SN cases. The
conclusion was that the CIV technique repeatability was sufficiently good for in-
cylinder flow observations and evaluation. Also the standard deviation was small
enough that the flow trends, shown as angular velocity, were possible to
evaluate even if only one combustion event was recorded.

2500
6.5-8° ATDC
10-11.5° ATDC
19.5-21° ATDC
2000
Mean angular velocity [rad/s]

1500

1000

500

0
0 5 10 15 20 25 30 35 40 45
Radius [mm]

Figure 33. CIV-measured mean angular velocity and standard deviation over ten
cycles at 1500 bar injection pressure. Data are shown for three CAD intervals.

45
3.5.2 Flow pattern comparison between CIV and CFD
The CIV method is assumed to show the in-cylinder flow closest to the piston
bowl glass during the diffusion combustion period. A comparison of CIV and CFD
results at EOI is presented in Figure 34. The CFD velocity vector results are
plotted as mean values between the bowl surface and different depths along the
cylinder axis (1 mm, 3 mm, 7 mm, and the total depth from piston to cylinder
head). The mean velocity is calculated from the bowl surface to the evaluated
depth. A section that is marked in the CIV plot corresponds to the compared 45°
CFD slice. A red arrow marks the flame stagnation point in both the CIV and the
CFD 3 mm plot. Because the maximum viewable diameter is 80 mm in the
optical engine, the CFD plots are also restricted to this value in Figure 34 and
Figure 35. It is clear in Figure 34 that the 3 mm CFD results correspond well to
the CIV results, both in flow structure and in absolute velocity values. At CFD 1
mm, the velocities are higher than the CIV results and at CFD 7 mm the flow
pattern structure differs from the CIV results. The total mean CFD velocities are
largely affected by the spray core, which is not seen in the CIV result. The
outcome is that the CIV results correspond to a mean velocity pattern calculated
at a depth between 0 and 3 mm from the piston bowl surface at the end of the
injection period.

During the post-oxidation phase, the remaining glowing soot particles

are moving around in the combustion chamber with low luminous flux. The
luminous flux decreases due to less soot remaining in the cylinder at post-
oxidation compared to the diffusion combustion phase and to the lower
temperatures that occur with later CAD. The CIV cross-correlation is performed
on the light structure. This means that the depth at which glowing soot particles
are visible increases, and flow at larger depths can be measured with CIV. In
Figure 35, a comparison between CIV and CFD during the post-oxidation phase is
made at a different depth than for the injection phase. At 1 mm depth, the CFD
velocities seem to be too high and do not match the flow profile in the CIV case.
A flow directed into the cylinder centre is seen in the CFD results, but not in the
CIV. At 7 mm and 15 mm, the velocity and structure match the measured case
better, but the “total mean” CFD case also seems to correspond well with the
measured CIV figure.

It can also be seen in the CIV measurements that the swirl centre
does not match the geometrical centre. In the CFD, this is assumed to be the
case at the start of calculation (IVC) when a symmetrical solid-body rotational
flow is applied. It has been shown that this asymmetrical swirl exists at IVC and
survives up to combustion TDC [67] [40]. The swirl offset has also been shown in
[68] to survive into the post-oxidation phase, which is also shown here. A CFD
model that also simulates the inlet stroke with the correct inlet geometry would
increase the accuracy between the CIV and CFD calculations (but the results are
still useful). The CIV-measured flow pattern captures the mean total bulk flow in

46
the cylinder. It can also be stated that the swirling motion is the main flow
pattern at different depths in the cylinder.

CIV

CIV
Layer velocity
-15 40

40 m/s 0 m/s 35
Layer velocity Layer velocity Layer velocity
-10
3 mm
-15 -15 40 -15 40 40
30
-5
-10
1 mm -10
7 mm
35
-10
Tot. mean
35 35

30 30 30 25
-5 -5 -5
0
25 25 25
0 0

Y
0 20
Y

20
Y

20
Y

20
5 5 5 Lay er velocity
5
-15 40
15 15
-10
15
40 m/s
35
15
10 10 10
10 10 30
10
-5
10
CFD CFD CFD
25
15 15 5 15
0
5
10
5
Y

20
5
15
20 0
0 10 20 30 4020 0 2010 0 15 0
10 20 30 40 0
CFD 5
10
10 20 30 40
X X 15 X
20
5

0
0 m/s
0 10 20 30 40
X
20 0
0 10 20 30 40
X

Figure 34. Injection phase, 10° ATDC, at 1500 bar injection pressure and
SN = 3.4. The measured flow velocity is shown in the top row, and the CFD
results are shown in the bottom row. CFD results are plotted as mean velocity
over different distances from bowl surface.

Flame CIV

Layer velocity
-15 25

25 m/s 0 m/s
-15
Layer velocity
-15 25
Layer velocity
-15 25
Layer velocity -10 7 mm
25
20
-10
1 mm -10
20 3 mm
-10
Tot mean
20
-5
20
-5 -5 -5
15
15 15
0 15
0 0 0
Y
Y

Y
Y

5 5 5
10 10 5 10
Lay er velocity 10
-15 40
10 10 10
-10 35 25 m/s
5 5
10
CFD CFD CFD
5 30
15 15 15-5
25
0
5
Y

20
20 0 20 5 0 0
0 10 20 30 40 20 0 10 20 30 40 0 10 20 30 15 40 15

X X 10

15
X 10
CFD
5
0 m/s
20
0 10 20 30 20 40
0
0
X
0 10 20 30 40
X

Figure 35. Post-oxidation phase, 20° ATDC, at 1500 bar injection pressure and
SN = 3.4. The measured flow velocity and CFD results are shown. CFD results
are plotted as mean velocity over different distances from bowl surface.

47
3.5.2.1 Angular flow velocity comparison
A comparison of the angular velocity between the measured optical engine data
and the calculated CFD data can be seen in Figure 36. The measured results are
compared with the CFD results in the form of a mean velocity value over the
entire combustion chamber depth and as a mean velocity value at 3 mm depth.
The outer piston bowl edge is located at the 40 mm radius, which is also the limit
for the CIV data in Figure 36 and Figure 37. As shown in Figure 34, the 3 mm
results seem to correlate well with the CIV results during the injection event.
During post-oxidation, the results from the 7 mm depth to the total combustion
chamber depth seem to correlate well to the CIV results. At low SN, both the 3
mm and mean curves correspond to the measured angular velocity profile during
the injection event, (Figure 36a). Later in the cycle during the post-oxidation
phase, the 3 mm case follows the measured profile well, as shown in Figure 36b
and Figure 37b. In Figure 36b, both CFD and measured data drop slightly in
angular velocity at a radius around 15 mm for CFD and 10 mm for CIV. The
reason why this occurs is not fully understood, but the total mean angular
velocity (the green dots in Figure 36b) does not show this dip. It can be assumed
that the velocity in the upper part of the bowl is slightly higher. This gives an
increased mean value and has no effect on the mean CFD curve in Figure 36b.
For the high SN case shown in Figure 37, the angular velocity and its deviation
from solid-body rotation are higher both during injection and post-oxidation
phases compared to the low SN case. The measured values are similar to the
CFD values, as was seen in the low SN case. One difference is that the dip in
velocity during the post-oxidation event at 15 mm for the low SN case is not
seen in the high SN case.

8° to 11.4° ATDC 19.5° to 22.9° ATDC

3000 3000
Mean vel.profile
Mean vel.profile
2500 2500 3 mm
3 mm Meas. SN 1.2, Injp. 1500bar
Meas. SN 1.2, Injp. 1500bar
omega [1/rad]

2000 2000
omega [1/rad]

1500 1500

1000 1000

500 500
A B
0 0
0 20 40 60 80 0 20 40 60 80
Radius [mm] Radius [mm]

Figure 36. Angular velocity (omega) vs. radius at injection pressure 1500 bar
and SN = 1.2. CFD results labelled “Mean vel.profile” and “3 mm” are compared
with measured CIV results.

48
 8° to 11.4° ATDC 19.5° to 22.9° ATDC
3000 3000
Mean vel.profile Mean vel.profile
2500 3 mm 2500 3 mm
Meas., SN 3.4, Injp. 1500bar Meas., SN 3.4, Injp. 1500bar
2000
omega [1/rad]

2000

omega [1/rad]
1500 1500

1000 1000

500 500
A B
0 0
0 20 40 60 80 0 20 40 60 80
Radius [mm] Radius [mm]

Figure 37. Angular velocity (omega) vs. radius at injection pressure 1500 bar
and SN = 3.4. CFD results labelled “Mean vel.profile” and “3 mm” are compared
with measured CIV results.

3.5.2.2 Optical distortion

The piston bowl curvature is a source for optical distortion, as shown in Figure
38. The comparison with CFD strengthens the earlier statement that the flame is
near the bowl curvature. This gives low optical distortion. The results show that
during injection and combustion, the flame is attached to the piston bowl glass.
Any compensation for the glass curvature is, therefore, not necessary.

Figure 38. Photo through the piston bowl glass of a checked pattern mounted
flush to the cylinder head, left picture. In the right picture, the checked pattern
is following the bowl surface curvature.

49
3.6 Emission spectroscopy
Emission spectroscopy can be used to extract temperature and soot data from a
radiating flame by applying Planck’s theory of black body radiation. Radiation
from a known body can be approximated as a black body radiator. An ideal black
body radiator has a known spectral distribution for a certain temperature, as
shown in Figure 39. Some of the radiation is in the visible light. When the
temperature changes on the black body radiator, the spectral distribution is also
changed. The temperature can, therefore, be estimated by comparing two known
wavelengths that the radiator sends out. Soot inside a combustion chamber is
not an ideal black body radiator and is often referred to as a grey body radiator
when its emissivity throughout the radiation spectrum is not constant [69]. Its
emissivity depends on the wavelength. To determine the temperature of a non-
black body emitter from its spectral intensity, the spectral emissivity needs to be
calculated. This can be estimated with Wein’s law [69]. With known temperature,
the KL factor, a measure of soot, can then be estimated.

In this work, the same Phantom high-speed colour camera images

used in CIV measurements were used to determine the two wavelengths that
resulted in spatial temperature and soot data. The AVL Thermovision software
[70] [71] was used for these calculations with a known camera response, shown
in Figure 40, and a tungsten lamp as the calibration source. Normally, this
process requires two cameras or half-pass mirrors and optical band-pass filters
as described in [72]. However, this technique can be used with just one
calibrated colour camera, as shown in [73], [74], and [75], when the camera has
a colour filter attached to the image chip. The calibration source used in this
thesis was a tungsten lamp with a temperature of 2624 K. It was placed on top
of the piston bowl glass to calibrate the whole measurement chain with mirrors,
optics, and camera. In [76], it was suggested that the recorded temperature and
KL (optical soot density) factor is a mean of the whole-flame volumetric
emission. Pastor et al. compared simultaneous 2-colour recordings and laser-
induced incandescence (LII) [76]. With LII, the spatial position is known and can
be compared with the 2-colour method in which the spatial measurement
position is not known. The results indicated that instantaneous LII soot
distributions showed increased variation between different injections than the 2-
colour method. This is because LII only measures a thin plane where the laser
sheet is introduced. For further details about the 2-colour method, see [75].

50
Figure 39. Example of the spectral distributions of radiation for different bodies.

Figure 40. Colour and Spectral Response Curve for the Phantom v7.3 camera
used in the measurements. Quantum efficiency (QE) is the percentage of
photons hitting the pixels that produce charge carriers for different
wavelengths.

51
3.6.1 CA-resolved soot formation calculation
A total soot signal was calculated using the KL factor obtained from the
Thermovision software. In equation (25), x i is the KL factor in pixel i, and the
total numbers of pixels with a valid signal (> 0 and not overexposed) is n. The
total area in the cylinder that gives a signal ( Apix ) is multiplied with the mean KL
factor. A measure of the soot that is in the line of sight in the piston bowl is the
result of the calculation, as shown in the top row of Figure 41 (“total soot”).

 x 
i 1
i
(25)
Tot.soot   A pix
n
In Figure 41 the soot area graph corresponds to the total area that contains valid
soot information and can be seen as a measure of how scattered the in-cylinder
soot is. The soot concentration is the mean KL factor calculated on all active
pixels (the soot area Apix ).

Variations between repeated measurements are unavoidable. In

Figure 41, individual combustion cycles together with mean total soot (calculated
on 10 individual measurements) are shown. Some of the variations can be
related to the randomness between the combustion events due to its fluctuating
nature. Variations can also be attributed to measurement deviation and soot on
the piston glass, as can be seen in the yellow curve in Figure 41. The
repeatability is, however, sufficient to see the response when changes in
injection pressure and SN are applied.

52
 4
x 10
8
Mean total soot
7 Individual cycles

6
tot soot [Kl*area]

0
0 5 10 15 20 25 30
CAD

3000 40
Mean Kl factor
2500 Individual cycles
Soot concentration

30
soot area [mm ]
2

2000

1500 20

1000
10
500

0 0
0 10 20 30 0 10 20 30
CAD CAD

Figure 41. Total soot, mean KL factor (“soot concentration”), and soot area CAD
resolved for individual cycles and the mean value. 1500 bar injection pressure,
SN = 1.7, load 10 bar IMEP. The injection event is marked with a black arrow.

53
4 Results
This project seeks to explain how the in-cylinder flow influences the combustion
and emissions behaviour in a HD diesel engine. The purpose is to study the in-
cylinder flow field and determine how it affects the combustion and emissions at
critical engine operation points. To answer these questions, CIV and 2-colour
methods were applied to optical engine measurements. When both CIV and 2-
colour methods are applied to the same high-speed images, simultaneous results
in 2D vector field, temperature, and soot spatial data can be acquired with CA
resolution (Figure 42). Both the diffusion combustion period and the post-
oxidation period can be studied. Important information from the on-going
combustion process can thereby be extracted to explain why emissions are
influenced when changes in swirl and injection pressure are applied. The results
obtained with the previously described techniques are discussed in this section.
This section is a summary of the results presented in earlier publications that are
included at the end of this thesis. vec2/B00080.vc7

Flame -10 ui e Flow 20

18
-20 16

[ (ux )2 + (uy )2 ] 1/2 (m/s)

14
-30 ri
12
y (mm)

-40
10
ez
-50 8

6
-60
4
-70
2

10 20 30 40 50 60 70
Soot x (mm) Temp

Figure 42. Measured in-cylinder flow, soot, and temperature during post-
oxidation, 10° CA after EOI.

54
4.1 Single cylinder engine tests: Swirl and tumble effects
on soot emissions
Swirl is known historically to reduce soot emissions. Today’s high injection
pressures reduce soot emissions effectively. Still, in transients at low λ problems
with soot emissions occur. Therefore, an investigation of different in-cylinder SN
and TN is of interest at low λ operation. The load points, listed in Table 3, were
selected from the six-cylinder turbo engine transient, shown in Figure 14,
repeated in the single cylinder engine. SN and TN are imposed by the AVT
system. Every load point was tested with different valve settings, called “No. of
tested airflow settings” in Table 3. Load 1 was tested at four different injection
pressures (named a, b, c and d) with the same fuel mass per cycle and APMAX
(12° ATDC). Load 2 had 41 different airflow settings to increase the resolution of
the resulting plot.

Table 3. Load points tested in the single cylinder engine.

Load 1 a,b,c,d Load 2 Load 3 Load 4
Rail press. [bar] 500, 1000, 1500 & 2000 2200 2500 2500
SNgt (at BDC) 0.4 - 6.7 0.4 - 6.7 0.4 - 6.7 0.4 - 6.7
TNgt (at BDC) 0.5 - 4.0 0.5 - 4.0 0.5 - 4.0 0.5 - 4.0
No. of tested air-flow settings 12 x 4 41 12 12
Lambda 1.20 - 1.25 1.12 1.10 1.20
Inlet press. Rel. [bar] 0.09-0.17 0.23 - 0.43 1.06-1.27 1.94-2.36
SOI [°ATDC] -11, -6, -4, -3 -5.8/-2.8 (pilot/main) -2 -2
IMEP [bar] 10 13 20 20
EGR [%] 0 0 0 30

In Figure 43, the engine-out soot plots are shown for the four load
points tested in the single cylinder engine. At high injection pressures, SN (called
Swirl in the graphs) has a large impact on soot emissions independent of load.
TN (called Tumble in the graphs) seems to have a negative effect on soot
emissions. The emissions are scaled by colour, where blue is low and red is high.
Increasing SNgt decreases smoke emissions at high-injection pressures. In
Publication I, it was shown that this was not valid for low injection pressures.
When SN is too high at 500 bar injection pressure, ”over-swirling” occurs with
increased soot as result (see Publications I and III). Increasing TNGT increases
soot emissions by producing an asymmetrical swirl vortex. In the upper left
graph in Figure 43, comparison of low and high tumble at the same swirl level
(SN = 2.5, marked with a black arrow) shows a 50% increase in soot emissions.

The main conclusion from Figure 43 is that swirl affects measured

tailpipe soot emissions strongly, even at injection pressures up to 2500 bar. With
today’s high injection pressures, the SN can be increased to much higher levels
before soot emissions stop decreasing.

55
 Injection pressure 1500 bar, load 1c Load 2 Inj.p. 2200 bar pilot
4
0.5
4 Smoke [FSN]
0.25
one valve 0.45 3.5

two valves
3.5 0.4
3
0.2
3 0.35
2.5
0.3

Tumble
2.5
Tumble

2 0.15
0.25

2
0.2 1.5
Smoke [FSN]
0.1
2v std
1.5 0.15
1 1v std dnr1
1v std dnr3
0.1 2v std dlift
1
1,500 bar 0.5 2v trap 0.05
0.05 2,200 bar 1v trap dnr1
1v trap dnr3
0.5 0
0 0 1 2 3 4 5 6 7
0 1 2 3 4 5 6 7 Swirl
Swirl
25 2.5
Load 3 Load 4

20 2.2

Load 2

Inlet pressure [bar]

15 1.9
Load 1
IMEP [bar]

10 1.6

5 Req. Load IMEP [bar] 1.3

IMEP engine out [bar]
Inlet pressure [bar]
0 1
-1 0 1 2 3 4 5
Time [s]
High load lambda 1.2 EGR 30%, load 4
High load lambda 1.1, load 3 0.5
0.25
4 Smoke [FSN]
4 Smoke [FSN]
one valve 0.45
one valve
two valves
two valves 3.5 0.4
3.5 0.2

3 0.35
3

0.15 0.3
2.5
Tumble

2.5
Tumble

0.25

2 2
0.1 0.2

1.5 1.5 0.15

0.05 0.1
1 1
2,500 bar 2,500 bar 0.05
0.5 0.5
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Swirl
Swirl

Figure 43. The centre plot shows the transient from 3-bar IMEP to full load. The
four load points tested in a single-cylinder engine are marked on this graph
with arrows leading to the corresponding emission plots at different in-cylinder
airflows. The colour scale indicates the filter smoke number at different flows.

56
4.1.1 Airflow effects on heat release
Changes in SN can be observed in the RoHR trace. In Figure 44, the RoHR for
load case 3 at 20 bar IMEP is plotted for three different SN. Even at the highest
injection pressure of 2500 bar, there is a large impact on the RoHR during the
diffusion flame period. At high SN, the RoHR is higher than at low SN in the early
diffusion flame period. According to [77], the air mixing into the flame increases
more when swirl is used than when no swirl is used. The RoHR is higher because
more air is entered into the flame and increases the combustion rate. In [78], it
was suggested that fuel sprays, like those found in modern diesel engines, have
mixing-limited vaporization. This means that the fuel jet and mixing processes,
not the transport processes at the droplet surface, control the vaporization. If the
mixing process is enhanced, a larger amount of hot air enters the
spray/combustion plume, and a larger amount of the injected fuel can be
combusted per unit time.

During the post-oxidation period, the RoHR decreases with increased

SN. The reason for this is suggested to be related to faster soot oxidation. The
described RoHR response to SN is valid for all tested load cases and injection
pressures. In Figure 44, images of the combustion for the SN = 3.4 case are
shown with green arrows that indicate the corresponding CAD. At 9 CAD during
diffusion combustion period, the CIV flow field is shown. High flow velocities of
over 70 m/s in the outer part of the piston bowl enhance the mixing during
combustion. Directly after EOI, the high velocities from the injection dissipate
quickly and do not continue into the post-oxidation phase. However, as shown
later in section 4.2.4, the high velocities during injection do affect the in-cylinder
flow drastically during post-oxidation.

57
 Inj_mean/CN318_Inj.vc7
60
SNw=3.4037
-10
50
-20

[ (ux )2 + (uy )2 ]1/2 (m/s)

40
-30

y (mm)
-40 30

-50
20
-60

10
-70

-80
0
10 20 30 40 50 60 70 80
x (mm)

2500 bar inj. pres

~2 bar ABS inlet p
Lambda 1.1
Shaped piston
High SN-Head

Injection
period
Figure 44. RoHR and combustion images at load 20 bar IMEP and 2500 bar
injection pressure. The RoHR is plotted for three different SNgt. The images
were taken at SNgt = 3.4. The green arrows indicate the CAD corresponding to
the images.

4.1.1.1 Airflow effects on ignition delay

The ignition delay was affected when different airflows were applied. The delay
differed by up to 0.7 CAD for the same SOI. The difference cannot be fully
explained with the Arrhenius correlation. The change in ignition delay is
suggested to be related to the turbulence level during the premixed combustion
event. Ignition delay at 10 bar IMEP and 1000 bar injection pressure can be seen
in Figure 45 for two cases. The left plot shows the ignition delay for a range of
SN (swirl) and TN (tumble) for a continuously firing engine. The right plot shows
the ignition delay for a motored (without combustion) engine with one single
combustion event performed every 3 minutes to minimise heat transfer effects.
The ignition delay is defined as the CAD between the start of the fuel injection
and positive RoHR. The red areas show a long ignition delay and the blue areas
show a short ignition delay. Both short and long ignition delays can be observed

58
at SNgt = 2.5 as indicated by the black arrow in the left plot in Figure 45. Tumble
seems to affect ignition delay strongly. Tumble is a good source for TDC
turbulence production when the tumble vortex is dissipated into smaller vortices
(turbulence) caused by the geometrical change of the combustion chamber. In
[79], [80], [81], and [82], it was reported that the turbulence intensity from the
inlet port caused shorter combustion periods, shorter ignition delays, and lower
emissions. In the data presented in this work, the trend of decreasing ignition
delay is more associated with tumble and soot emissions are more associated
with swirl.

Cont. firing [CAD] Single cycle [CAD]

4 Igndelay [cad] 4
4.8
one valve 2.5
3.5 two valves 3.5
4.6
2.4
3 3
4.4

2.3
2.5 2.5
Tumble
Tumble

4.2

2 2.2 2 4

1.5 2.1 1.5 3.8

1 2 1 Igndelay 1C [cad] 3.6

one valve
two valves
0.5 0.5 3.4
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Swirl Swirl

Figure 45. Ignition delay for load 10 bar IMEP at 1,000 bar injection pressure.
Ignition delay is plotted for a range of swirl and tumble for a continuously firing
engine (left) and a motored engine without combustion except for one single
engine cycle, (right).
According to the literature [5], higher SN should give a longer
ignition delay due to the higher heat transfer during compression and resulting
lower compression temperature. In this work it was indicated that the ignition
delay is short where the turbulence intensity (TI) is high at start of combustion
(SOC). This is created during tumble dissipation into TI near TDC. But much TI is
created during the inlet stroke. This TI does not survive to TDC, as shown in
Publication I, because it is small scale vortexes that is transferred from small-
scale turbulence to heat. With higher TI during the inlet stroke, the heat transfer
from the cylinder head, liner, and hot valves to the inlet air increases. The inlet
valve temperature is approximately 300–400°C. A higher initial temperature in
the cylinder at IVC produces a higher compression temperature and a shorter
ignition delay. Calculations of ignition delay with the Arrhenius correlation using
measured cylinder pressure cannot alone explain the difference in ignition delay.
The measured ignition delay values range from 1.9° to 2.6° CA (0.7° CA
difference) compared with 2.04° to 2.12° CA (0.08° CA difference) in the
calculated case.

To minimize the effect of heat transfer from the engine to the inlet
air during the inlet stroke, thereby affecting the starting temperature before

59
compression, single combustion tests were performed. The inlet pressure was set
to the same level as in the standard case, but the inlet, cooling water, and oil
temperatures were all set to 40°C. The engine was motored without injection or
combustion except for one single cycle so that the heating from the engine was
kept to a minimum. Three measurements were taken at every test point with
three minutes of motoring (without combustion) between the measurements. A
mean value was calculated for every test case, plotted in Figure 45, and
compared with the continuously firing engine (load point 1b). The ignition delay
trends was the same in the motored and the standard case. This leads to the
conclusion that ignition delay is affected not only by pressure and temperature,
but also by the airflow in the cylinder. Higher tumble affects the SOC. When the
piston is near TDC, the global tumble flow is transformed to small-scale turbulent
vortices. If the turbulence is increased during the ignition delay, the hot air and
active species mixing through the pre-ignition and reaction zones are also
increased [66]. This means that the mixing process of air and fuel is more
intensive and the reaction time decreases. Therefore, tumble has an effect on
ignition delay.

4.2 Optical engine results: Flow effects on soot formation

and oxidation
The tail pipe soot emissions are strongly affected by the in-cylinder flow and
injection pressure, as shown in section 4.1. To investigate why the engine-out
soot is affected, soot information was extracted with CA resolution in the optical
engine with the 2-colour method. It was found that the amount of soot observed
during combustion does not reflect measured engine-out soot emissions. The
reduction rate during post-oxidation and “end of signal” could instead be directly
coupled to measured engine-out soot. Therefore, the oxidation period is
important for reducing soot emissions.

4.2.1 Injection pressure effect on soot production and

oxidation
Injection pressure has a large effect on measured engine-out soot emissions, as
shown in Table 4. Increased injection pressure gives lower tail pipe soot but not
necessarily lower soot production during combustion. When injection pressure
increases from 500 bar to 1000 bar, the “total soot” increases by 50% during
combustion, as shown in Figure 46. This is because more fuel is injected and
combusted per unit time, which increases soot production. However, the engine-
out soot emission is lowered from 1.22 FSN to 0.49 FSN. When the injection
pressure increases to 1500 and 2000 bar, the maximum registered soot signals
from the 2-colour measurements are the same or lower than the 1000 bar case,
as shown in Figure 46. Table 4 shows that tailpipe soot decreases continuously
with increased injection pressure. The increased soot production during
combustion does not explain why soot emissions are lowered with increased
injection pressure.

60
 As explained earlier in the work, injection pressure affects the size of
the soot particles [26]. Higher injection pressure gives smaller particles, as
shown in Figure 10. However, the total particle mass produced during the
diffusion combustion period is not necessarily lowered with increased injection
pressure. In Figure 46, it can be seen that increased injection pressure results in
an earlier CAD when the “total soot” signal is zero. Tailpipe soot emissions, listed
in Table 4, can be correlated with the “end of signal” and the slope of the soot
oxidation curve. The reason for the decreased engine-out soot can be connected
to the size of the soot particles created during diffusion combustion. The soot
particles are formed inside the rich diesel flame. Higher injection pressures
reduce the time the particles are in the flame and thereby also the time it can
grow. In the post-oxidation phase, smaller particles are combusted faster, as
discussed in section 3.4.2, resulting in lowered tailpipe PM.

Table 4 Smoke emissions for different injection pressures at 10 bar IMEP.

Injection
pressure 500 bar 1000 bar 1500 bar 2000 bar
Smoke [FSN] 1.22 0.49 0.19 0.16
4
x 10
14
500 bar
~50% 1000 bar
12 1500 bar
increase
2000 bar
10
total soot (KL*area)

6 EOI

0
0 10 20 30 40 50 60 70 80
CAD

Figure 46 The total soot at different injection pressures. For each case, the
amount of fuel, 110 mg/cycle, was the same as well as CA50. This resulted in
different start- but the same end of injection (marked in the graph).

61
 The increase in oxidation rate (negative signal) caused by the
injection pressure can be clearly seen in Figure 47, where the derivative of the
“total soot” signal is plotted. The blue line (500 bar) has the lowest rate during
both soot formation and oxidation. With higher injection pressure, the rate
increases and the soot oxidation is higher per CAD.

8000
500 bar
1000 bar
6000 1500 bar
2000 bar
4000
EOI
Tot soot rate

2000

1.22 FSN
-2000

0.49 FSN
-4000

0.16 FSN 0.19 FSN

-6000
-10 0 10 20 30 40 50 60
CAD

Figure 47. Total soot derivative at different injection pressures.

4.2.2 Swirl effect on soot production and oxidation

With smaller soot particles the oxidation time is thereby lowered. Swirl affects
measured engine-out soot, as seen earlier. The flow field during post-oxidation is
suggested to be the main contributing source of lowered soot emissions. In
Figure 48, swirl was changed from SN = 1.7 to SN = 6.4, and injection pressure
was set to 1500 bar in both cases. The tailpipe soot was lowered from 0.19 FSN
to 0.04 FSN as a result. The “total soot” signal is slightly higher during injection
with higher SN. This can indicate better mixing during diffusion combustion and
more fuel burnt per unit time, which increases soot production as discussed in
4.1.1. The RoHR curves in Figure 49 indicate that SN increases the RoHR rate in
the early diffusion combustion phase. This is valid for all tested injection
pressures. Higher soot production cannot explain why the measured engine-out
soot decrease is so drastic. The large difference in soot between high and low SN
in Figure 48 occurs during the post-oxidation phase. With higher SN, the “total
soot signal” decreases to zero with a steeper slope of the soot oxidation curve. If
it is assumed that the particle size created during combustion is the same in both
cases (same injection pressure), the faster oxidation rate can be attributed to

62
the in-cylinder flow and resulting turbulence production. Turbulence is known to
affect oxidation rate. This indicates that the post-oxidation phase is of paramount
importance for the measured tailpipe soot.

4
x 10
15
LowSNhead 2v
HighSNhead 1v

Engine-out soot:
Low SN = 0.19 FSN
10
High SN = 0.05 FSN
Tot. soot (Kl*area)

0
-5 0 5 10 15 20 25 30
CAD

Figure 48. Total soot at 10 bar IMEP and 1500 bar injection pressure for low SN
(1.7) and high SN (6.4).
450
500 bar SN 6.7
Load 1 500 bar SN 0.4
400 1000 bar SN 6.7
1000 bar SN 0.4
350 2000 bar SN 6.7
2000 bar SN 0.4
Heat Release [J/CAD]

300

250

200

150

100

50

-50
-4 -2 0 2 4 6 8 10 12 14 16
Crank-angle°
Figure 49. RoHR for load point 1 at different injection pressures and SN. The
arrows show the RoHR change when SN is increased.

63
 Figure 50 shows that the temperature of the glowing soot in the
cylinder was higher up to 15° ATDC for the high SN case. With higher
temperature during diffusion combustion, more soot is created, and this explains
why higher soot concentrations are observed in the first part of the combustion
phase. Higher temperature during post-oxidation increases the oxidation rate
[23], but in the soot oxidation phase the temperature was lower for the high SN
case. Soot oxidation can be established down to 1300 K according to [42], which
indicates that the temperature difference between the two cases cannot explain
the lowered soot emissions. The only other thing that can influence the soot
oxidation is the in-cylinder flow pattern in form of turbulence.

2100K
2200 2000
LowSNhead 2v
2100 HighSNhead 1v 1500
Temp [K]

2
2000 mm 1000

1900 500

1800 0
0 10 20 30 0 10 20 30
CAD CAD
2200K 2300K
1000 250
800 200

600 150
2

2
mm

mm

400 100

200 50

0 0
0 10 20 30 0 10 20 30
CAD CAD
Figure 50. Mean temperature (Temp) of active pixels and area plots for
temperatures exceeding 2100, 2200 & 2300 K, CAD resolved for low SN (1.7)
and high SN (6.4). Load 10 bar IMEP at injection pressure 1500 bar.

64
4.2.3 In-cylinder flow structure at injection and post-
oxidation
The flow pattern inside the cylinder is clearly coupled to the measured levels of
engine-out soot. To understand this connection, the flow field was extracted
during the combustion and post-oxidation phases with the CIV technique. In
Figure 51, natural flame images and flow patterns are shown for injection
pressures of 200 bar and 2000 bar just before EOI. In the natural flame images,
top row, the variation in injection pressure gives a large difference in penetration
length. The 200 bar case has strongly restricted penetration as well as lift-off
length. In the 2000 bar case, the spray core and its lift-off length are much
longer. The spray forces the rich diffusion combustion to the outer part of the
piston bowl where a bright soot cloud can be observed. In the 200 bar case, the
cloud is seen in the central part of the bowl. When the flow field is extracted for
the two tested injection pressures, differences can be clearly seen. With high
injection pressure, the flow is redirected back to the central region of the piston
bowl and creates mixing. This does not occur in the 200 bar case. The flow
follows the swirl vortex with weak penetration, and no flow towards the piston
bowl centre is redirected back. Not surprisingly, this poor mixing results in high
soot emissions.

During the post-oxidation phase, the earlier observed difference in

velocity and flow pattern survives, as shown in Figure 52, even if the same SN
and fuel mass are applied in both cases. At 2000 bar, much higher velocities are
observed in the centre region of the bowl compared to the 200 bar case. The 200
bar case seems to have near-zero velocity in the centre regions. The natural
flame images also show a different pattern and seem to have a denser, but not
so bright, soot cloud rotating in the piston bowl. The observable soot cloud
expands more towards the edge of the piston bowl, as can be seen in Figure 52
(left), compared to the diffusion combustion soot cloud shown in Figure 51 (left).
The conclusion is that the kinetic energy introduced by the injection influences
the in-cylinder flow velocities even after EOI during the post-oxidation phase.
This is valid for swirling combustion systems.

65
 200 bar 2000 bar

55 m/s 0 m/s
Figure 51. Combustion images during diffusion combustion near EOI. CIV
results in the form of flow vectors at 200 bar and 2000 bar injection pressure.
Load 1, SNGT = 3.4.

66
 200 bar 2000 bar

55 m/s 0 m/s
Figure 52. Combustion images during post-oxidation, 10° after EOI. CIV results
in the form of flow vectors at 200 bar and 2000 bar injection pressure.
SNGT = 3.4.

67
4.2.4 Solid-body deviation caused by injection pressure
Swirl introduced during the inlet stroke can be assumed to be of the
solid-body type. In a real engine, the flow deviates some from this behaviour
[40] but it is still a good assumption that the flow behaves in this way before fuel
injection. As shown in this work, the post-oxidation flow is strongly affected by
injection pressure. To investigate this behaviour, a comparison of the angular
velocity profile for injection pressures from 200 bar up to 2000 bar was made.
Figure 53 shows the angular velocity at EOI. All cases had the same SN and
injected fuel mass. The effect of injection pressure on the angular velocity profile
is clear. With increased injection pressure, the angular velocity increases in the
central regions of the piston bowl. At low injection pressure, 200 bar, the angular
velocity profile has a near solid-body rotation behaviour. The velocity in the outer
part of the piston bowl, near the 40 mm radius, does not seem to be affected
due to the changes in injection pressure. Only the angular velocity in the central
region is strongly affected. Injection pressures of 500 bar, 1000 bar, and 1500
bar fall between the 200 bar and the 2000 bar case, and this strengthens the
assumption that injection pressure is the main force that creates the deviation in
angular velocity from solid-body rotation.

3000
2000 bar
1500 bar
2500 1000 bar
500 bar
200 bar
2000
Omega [rad/s]

1500

1000

500

0
0 5 10 15 20 25 30 35 40 45
Radius [mm]

Figure 53. Measured angular velocity (omega) profiles at EOI at various

injection pressures at Load 1. The CIV results are calculated according to Figure
32.

68
4.2.5 Solid-body deviation caused by swirl
The angular velocity was also strongly influenced when SN was varied and the
injection pressure was kept constant at 1500 bar, as shown in Figure 54. The
evaluation was made during the early post-oxidation phase. With higher SN, the
overall velocities are higher compared to the low SN case. In the central regions,
the angular velocity is high for both SN cases, but for radii larger than 5 mm, the
high SN case has higher angular velocities. This higher angular velocity is
assumed to be the main contributor to the large reduction in measured tailpipe
soot at increased SN that is shown in Table 4. Large deviation in angular velocity
creates shear in the rotating fluid, and this can be a large contributor to
turbulence production during post-oxidation.

3000
LowSNhead
SN = 1.7 2v 1500bar
HighSNhead
SN = 6.7 1v 1500bar
2500

2000
Omega*vol [rad/s*m3]

1500

1000

500

0
0 5 10 15 20 25 30 35 40 45
Radius [mm]
TimeMeanQF_Vector_122-142/B00001_Avg V.vc7 TimeMeanQF_Vector_122-142/B00001_Avg V.vc7
30 30
80 80

70 70 25 25

60 60
(m/s)

[ (Avg)2 + (Avg)2 ]1/2 (m/s)

20 20
2 1/2

50 50
[ (Avg) + (Avg) ]
y (mm)

y (mm)

40 15 15
40
2

30 30 10 10

20 20
5 5
10 10
Velocity SN = 1.7 Velocity SN = 6.7
0 0
10 20 30 40 50 60 70 80 10 20 30 40 50 60 70 80
x (mm) x (mm)

Figure 54. Angular velocity profile & vector field at post-oxidation, 16°–19°
ATDC. Load 1, 1500 bar injection pressure, and two SN settings.

69
4.3 CFD results: Angular velocity during compression,
combustion, and post-oxidation
The kinetic energy introduced by the fuel spray and combustion can explain how
the deviations in angular velocity profile occur. The flow field was computed for
the cylinder domain using CFD. Angular momentum, kinetic energy, and SN was
calculated from the CFD results. Three cases were studied: “Spray OFF,” a
motored engine without injection; “Comb. OFF”, fuel injection with the
combustion model off; and “Comb. On”, fuel injection with combustion. Injection
pressure was set to 1500 bar, injected fuel was 110 mg/cycle, SOI at −1° ATDC,
EOI at 10° ATDC, and SN = 3.4 (solid-body type). In Figure 55, the angular
velocity profile at −70° ATDC is shown. The profile indicates nearly solid-body
rotational flow, which was the CFD model boundary condition at IVC. At 0° ATDC
directly after injection starts, the flow has been influenced by the squish. The
squish area starts at a radius of ~40 mm and continues to 65 mm. In this region,
the flow has a lower angular velocity compared to the bulk flow in the piston
bowl. The decreasing angular velocity in the squish region comes from the wall
friction that is dominant at TDC. The distance between the head and piston is
1.35 mm. The highest velocity is observed in the outer part of the bowl, near 40
mm, where the bowl is the deepest.

At 10° ATDC, a large impact on the angular velocity can be seen in

the centre of the bowl. In the figure, the CIV-measured angular velocity is
plotted with black stars. Here the CIV results are directly compared to the
angular velocity profile that corresponds to a mean value over the total bowl
depth calculated from the CFD results. The observed angular velocity trend is the
same for the CIV results as for the CFD results. A significant increase in velocity
is seen in the centre region of the piston bowl for the cases with injection. The
case with combustion has the largest velocity increase, and the velocity increase
for the case with only injection (no combustion) is nearly as large. The case
without injection is not influenced, and the angular velocity profile is more similar
to solid-body rotation. This was also observed in measurements with low
injection pressures in Figure 53. One important observation in Figure 55 at 10°
ATDC is that the “Spray OFF” case has higher angular velocity at the bowl rim
(radius greater than 32 mm) compared to the other two cases with injection. It
seems like there is a redistribution of angular velocity from the outer part of the
bowl to the inner part in the injection cases.

In the 20° ATDC case, which corresponds to the post-oxidation, the

large velocity deviation between the injection and non-injection cases is still
significant. There is also still a deviation between the combustion ON and the
combustion OFF cases. Therefore, both the force from the injection and the
expansion from the combustion and density redistribution have an impact on the
angular velocity profile. At 40° ATDC in Figure 55, the angular velocity profiles
with injection (with and without combustion) do not deviate as much as at 20°
ATDC. Both cases have a reduced peak velocity in the centre region, but the

70
combustion ON case has decreased the most. This decrease is not only caused by
the skin wall friction but also by the shear that occurs from the deviation in
angular rotation at different radii. This shear makes the large-scale swirl vortex
dissipate to vortices with smaller length scales that produce turbulence, as seen
later in the work. At 80° ATDC, the dissipation of the angular velocity in the
centre region of the piston bowl continues. A large deviation from solid-body
rotation can still be observed in the injection cases. The case without injection
does not change noticeably and retains the solid-body like angular velocity
profile, which does not contribute to any turbulence production.
-70 CAD 0 CAD
600 600
SN 3.4, comb ON SN 3.4, comb ON
SN 3.4, spray OFF SN 3.4, spray OFF
500 SN 3.4 Comb OFF 500 SN 3.4 Comb OFF

400 400
omega [1/rad]

omega [1/rad]
300 300

200 200

100 100

0 0
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
Radius [mm] Radius [mm]

10 CAD 20 CAD
3000 3000
Measured Measured
SN 3.4, comb ON SN 3.4, comb ON
2500 SN 3.4, spray OFF 2500 SN 3.4, spray OFF
SN 3.4 Comb OFF SN 3.4 Comb OFF

2000 2000
omega [1/rad]

omega [1/rad]

1500 1500

1000 1000

500 500

0 0
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
Radius [mm] Radius [mm]

40 CAD 80 CAD
3000 3000
SN 3.4, comb ON SN 3.4, comb ON
SN 3.4, spray OFF SN 3.4, spray OFF
2500 SN 3.4 Comb OFF 2500 SN 3.4 Comb OFF

2000 2000
omega [1/rad]

omega [1/rad]

1500 1500

1000 1000

500 500

0 0
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
Radius [mm] Radius [mm]

Figure 55. Angular velocity profiles at different CAD. The CFD results are shown
for the cases with no injection, injection without combustion, and injection with
combustion and compared with measured CIV results in 10° and 20° ATDC
cases.

71
4.3.1 Swirl number, angular momentum, and rotational
kinetic energy
A large increase in the angular velocity in the central part of the piston bowl
caused by the injection has been shown in both the CFD and CIV-measured
results in the previous figures. The SN does not increase with injection and
combustion, as can be seen in Figure 56. The SN is calculated according to
equation (29), where the angular velocity ( ) is calculated from the
relationship between the total in-cylinder angular momentum ( ) and moment of
inertia ( ) as shown in equations (26), (27), and (28). is the mass of the
respective volume element in the mesh grid, and is the radius at which the
volume element is located. All calculations are made with the geometrical centre
of the piston as a starting point. corresponds to the angular velocity around
the geometrical centre point.

(26)

(27)

(28)

(29)

The SN increases during compression due to the reduced radius when

the air is forced into the piston bowl. At TDC, all cases have equal SN. At SOI,
the two cases with injection start to deviate from the case without injection,
labelled “spray off” in Figure 56. The reason why SN decreases faster in the
cases with injection can be seen in the angular momentum plot in Figure 56.
Some of the SN drop can be related to the increased mass (a 5% increase in this
case) in the cylinder when fuel is added, which increases the moment of inertia.
The fuel is injected radially without any contribution to angular momentum. With
no contribution of angular momentum and increased moment of inertia, the SN
decreases. The SN drop is steepest for the case with combustion. The angular
momentum decreases faster with combustion than without injection because of
friction losses. Even if the velocity is much higher in the central part of the piston
bowl, the velocity decreases slightly in the outer part of the bowl, as shown in
Figure 55 that compares the cases without and with injection. The decrease is
much smaller compared to the increase. A small difference in angular velocity at
large radii has a larger impact on the angular momentum than angular velocity
differences at small radii (see equation (27)). Additionally, more of the total in-
cylinder mass is located at larger radii. This means that a minor change in
velocity in the outer part of the piston bowl has a larger impact on angular
momentum compared to when the velocity is increased in the central part of the
bowl.

72
 When this deviation from solid-body rotation caused by the injection
occurs, the rotational kinetic energy is affected as can be seen in Figure 57. The
rotational kinetic energy ( ) is

(30)

The rotational energy increases with injection and combustion, and this
contributes to the deviation from solid-body rotation. The redistribution of
velocity is mainly driven by the kinetic energy in the spray. The difference in
density is not the main driver because the same phenomenon occurs
independently of the combustion. However, the density has an impact on how
large the deviation in angular velocity will be.
Swirl number angmom -3
4.5 x 10 Angular momentum
1.5
spray off
spray off
comb off
comb off
4 comb on 1.4 comb on
angular momentum [kg m/s]
2

1.3
3.5

1.2
SNmass

1.1

2.5
1

2
0.9

1.5 0.8
-150 -100 -50 0 50 100 150 -150 -100 -50 0 50 100 150
CA [°] CA [°]

Figure 56. Swirl number (SN) (left) and angular momentum (right) calculated
from the CFD results at 1500 barKinetic
injection
energy
pressure and initial SN = 3.4.
0.45
spray off
comb off
0.4 comb on

0.35
Kinetic energy [J]

0.3

0.25

0.2

0.15

0.1
-150 -100 -50 0 50 100 150
CA [°]

Figure 57. Rotational kinetic energy calculation from CFD results at 1500 bar
injection pressure and initial SN = 3.4.

73
4.3.2 Density distribution
The density ( ) distribution was compared for the cases with and without
combustion to explain the difference between the two cases. The results from the
case without spray were used to determine how the spray and the combustion
affect the density distribution in the cylinder. The density distributions for the no
combustion case ( ) and the combustion case ( ) are calculated as

(31)

(32)

The radial distribution in density difference is plotted in Figure 58 CA

resolved, and large differences are observed between the two cases. Without
combustion ( ) the density is high in the outer part of the piston bowl,
around 35–50 mm from the injector, due to the cooling effect of the evaporating
sprays. With combustion ( ) the opposite occurs. The density is lower where
the combustion takes place and the temperature is highest. In the central part of
the bowl, the density is high because no combustion (no temperature increase)
occurs in this region during injection. After EOI (10° ATDC), the increased
density in the centre is still observed at 20° ATDC but the density difference
slowly levels out in the combustion chamber. At large radius (60 mm), near the
cylinder liner, the density is also higher. No combustion occurs near the liner, so
there is no temperature increase. The small distance between the cylinder head
and the squish plane (1.35 mm distance at TDC) provides large cooling surfaces
that lower temperature in the region.

The main findings are that the density distribution is strongly affected
by combustion and shows large deviation from the case with just injection. Still,
the deviation from solid-body rotation is similar between the cases with and
without combustion. The large deviation from solid-body rotation is mostly
caused by the redistribution of angular velocity driven by the kinetic energy from
injection, not by the combustion-driven flow that causes density differences.

74
 3
[kg/m ]
10
60

8
50

6
40
radius [mm]

30
2

20
0

10 -2

0 -4
-10 0 10 20 30 40 50 60
CA [°]
3
[kg/m ]
10
60

8
50

6
40
radius [mm]

30
2

20
0

10 -2

0 -4
-10 0 10 20 30 40 50 60
CA [°]

Figure 58. Density difference between the case without combustion, , and
with combustion, . CFD results at 1500 bar, SN = 3.4.

75
4.4 Optical engine results: Fuel injection impact on flow
pattern and angular velocity
Now we can understand why the flow is accelerating in the central regions of the
bowl when injection occurs. Figure 59 shows a measured flow pattern during
injection at 20 bar IMEP and 2500 bar injection pressure at SN = 6.4. The fuel is
injected with a velocity of around 700 m/s in the direction of the green straight
arrow. Its trajectory is affected by the swirling flow in the clockwise direction,
which makes the fuel spray and flame curve move slightly to the leeward side of
the swirling flow. When the flame hits the outer part of the bowl, it is redirected.
A stagnation point, marked with a red arrow, occurs where the central point of
the flame hits the bowl surface. The flame flow is directed back to the central
part of the piston bowl, as illustrated with the green curved arrows. The flame
has a measured velocity of around 70 m/s at the outer diameter of the bowl. The
closest flame on the leeward side is also redirected, and interference between
the two flames occurs as marked with a white arrow. The flow back to the central
region now has an offset from where the fuel was injected.

If assuming that one particle that follows the flow can be traced from
the injection back to the central part of the combustion chamber, how many CAD
does this particle need to travel this distance? From injection to the piston bowl
edge, only 0.5 CAD are needed, and 3.5 to 5 CAD are required to travel from the
bowl rim to the centre (depending on injection pressure). The injection period in
this case was 18 CAD. This means that the fuel parcels from the injection have
plenty of time to make this orbit and influence the swirling flow in the central
part of the bowl. A small, but important, decrease in angular velocity is observed
in the outer part of the bowl and in the squish band as shown in Figure 55. The
velocity in the central part of the combustion chamber is affected not only by the
reverse flow induced by the injection but also by a redistribution of mass from
the outer part of the piston bowl to the central part. The spray and the flame
force the bulk flow to move towards the centre of the piston bowl. The spray
evaporates on its way to the piston edge, and a large amount of air is pulled into
the spray. This flow forces the gas in the outer part of the bowl to move towards
the centre. For a certain angular momentum, the angular velocity increases with
decreasing radius. These phenomena are assumed to be the main driving forces
for the redistribution of the angular velocity.

76
 Offset

Figure 59. In-cylinder flow during combustion at 10° ATDC with 20 bar IMEP,
2500 bar injection pressure, and SN = 6.4. White arrows indicate the
interference flow between two flames. The highest velocity, ~70 m/s, occurs
near the rim of the bowl. The velocity vectors are measured CIV results.

4.4.1 Turbulent kinetic energy production

It was assumed earlier that the deviation in solid-body rotation of the in-cylinder
flow creates turbulence that can influence the post oxidation. Turbulence is
produced if kinetic energy is released to turbulent kinetic energy. There are
several questions to answer about turbulent kinetic energy. How does the kinetic
energy change with different SN and injection pressures? Does the kinetic energy
introduced by the spray survive into the post-oxidation phase? Does the energy
captured in the swirling vortex dissipate into turbulent energy or is the energy
conserved?

77
 The mean kinetic energy per unit mass, KE, was estimated with
equation (33) for every vector plot captured in the optical engine. A mean value
of the KE was calculated for the entire observable area in the piston bowl (80
mm diameter) and plotted in Figure 60. The velocity component in the z-
direction was excluded in the calculation because it was not measured. The
kinetic energy equation is

(33)

Spatial KE was also calculated for low and high SN during post-oxidation, which
was averaged between 19.4°–22.8° ATDC (20 vector images) as shown in Figure
60.

In Figure 60, it is clear that both injection pressure and SN affect KE.
Injection starts at −7° ATDC at 500 bar injection pressure and at −2° ATDC at
1000 bar injection pressure. Both end at 10° ATDC, marked with a blue dashed
line. During injection, the KE is much higher in the 1000 bar cases because of
higher injection velocity. SN also affects KE at injection in both the 500 bar and
1000 bar cases. Directly after EOI, the KE drops significantly. For the 500 bar
and 1000 bar cases at SN = 1.7, KE drops to nearly the same levels at 15°
ATDC. This means that the KE introduced by the fuel injection dissipates in just 5
CAD. KE for SN = 6.4 remains at a higher level than the KE for SN = 1.7 during
post-oxidation. KE is highest for the case with 1000 bar and SN = 6.4. It is also
higher than the KE at 500 bar and SN = 6.4 during the remainder of the post-
oxidation period. This was not the case for the low SN case. The KE introduced
by the injection for SN = 6.4 and 1000 bar causes large differences in angular
velocity in the piston bowl, as shown in 4.2.4. Increased velocity in the piston
bowl caused by high SN in combination with high injection pressure increases the
KE that remains into post-oxidation phase, as can be seen in the lower images in
Figure 60. At 1000 bar injection pressure, the KE is much lower for SN = 1.7
than for SN = 6.4.

78
 500
500 bar SN=1.7
450 1000 bar SN=1.7
500 bar SN=6.4
400 1000 bar SN=6.4
Mean kinectic energy [m /s ]
2 2

350

300

250

200

150

100

50

0
0 5 10 15 20 25 30
CA [°]
1000 bar SN = 1.7 1000 bar SN = 6.4

300 m2/s2

0 m2/s2

Figure 60. Mean kinetic energy measured in an optical engine with the CIV
technique at load 1. The dashed line indicates EOI. The mean kinetic energy
images are averaged between the black solid lines.

KE from the injection is conserved in the central part of the

combustion bowl at high SN and high injection pressures. This is not true for low
SN because the KE from injection dissipates in just 5 CAD. The single question
remaining, then, is whether this KE producing any post-oxidation turbulent
kinetic energy.

79
 Turbulent kinetic energy (TKE) per unit mass is estimated from the
vector field with equation (34). (equation (35)) is the RMS deviation from
an averaged vector field in the x- and y-directions for 4.4 CAD (20 vector field
images) where is the mean velocity over images (in this case 20) and is
the velocity in the interrogation window for image .

(34)

(35)

In Figure 61, the mean velocity vector field (averaged between 19.4°
and 22.8° ATDC) and TKE during the post-oxidation phase are shown for
SN = 1.7 at 500 bar and 1000 bar injection pressures. As seen in the KE plot
during the post-oxidation phase, the mean KE does not deviate between the 500
bar and 1000 bar cases. In Figure 61, TKE is slightly higher in the 1000 bar case
than in the 500 bar case. The mean velocity is on the same order for both cases.
The conclusion is that at low SN the TKE introduced by the injection dissipates
quickly and the influence on the mean velocity and late TKE is minor.

80
 TKE[m2/s2] Mean velocity [m/s]
500 bar
SN = 1.7

25 m/s

0 m/s
1000 bar
SN = 1.7

25 m/s

0 m/s

Figure 61. Turbulent kinetic energy and mean velocity during post-oxidation,
19.4°–22.8° ATDC, for two different injection pressures at the same load. The
velocity vectors are CIV results.

81
When the SN is increased for the same injection pressure (1000 bar), both TKE
and mean velocity are strongly influenced as shown in Figure 62. In the central
part of the piston bowl, significant TKE exists in the high-SN case even after the
injection-introduced TKE has dissipated. The white areas in the figure indicate
TKE higher than 25 m2/s2.

TKE[m2/s2] Mean velocity [m/s]

1000 bar
SN = 1.7

25 m/s

0 m/s
1000 bar
SN = 6.4

25 m/s

0 m/s

Figure 62. Turbulent kinetic energy and mean velocity during post-oxidation,
19.4°–22.8° ATDC, for two different SN at the same load. CIV results are shown.

Turbulence production is possible at late CAD because of the slowly dissipating

velocity gradients. This indicates that high injection pressure together with high
swirl can produce more late-cycle TKE than independent increases in SN or
injection pressure. This is why increased SN at high injection pressures causes
lower engine-out soot emissions. TKE increases the soot oxidation rate [9].

TKE produced by the angular velocity gradients caused by injection is

concentrated in the central part of the combustion bowl where most of the

82
oxidation takes place. The difference in the TKE produced with combustion and
the TKE produced without spray is calculated from the CFD data as

(36)
and plotted in Figure 63. By subtracting the turbulence for the case without
spray from the cases with combustion, squish and other displacement effects can
be eliminated. The contribution from the spray and combustion can be seen in
Figure 63. Slightly after SOI, TKE increases in the central regions of the bowl at
the same time as the angular velocity profile starts to deviate from solid-body
rotation. The bowl radius is 40 mm. At EOI (10° ATDC), TKE is still increasing.
The TKE slowly starts to decrease after EOI, but it is still high at later CA. When
injection pressure and SN are varied, as shown in Publication VI, both
parameters have an impact on TKE and this can be seen in the CIV
measurements. With increased injection pressure and swirl, TKE increases. This
turbulence production can be related to the mechanical flow distribution caused
by the spray in combination with swirl. The location of TKE in the central part of
the bowl is beneficial for the post-oxidation because this is where most of the air
and residuals trapped in the cylinder are found near TDC.

2 2
[m /s ]

60 12

10
50
8

40
6
radius [mm]

4
30

2
20
0

10 -2

-4
0
0 20 40 60 80 100
CA [°]

Figure 63. The CFD-computed turbulent kinetic energy difference per unit
mass, , is plotted for 1500 bar injection pressure and SN = 3.4.

83
5 Discussion
Soot oxidation is affected by temperature, pressure, oxygen access, fuel
composition, turbulence levels, OH radical concentration, and other factors. All of
these parameters change continuously in a diesel engine during the post-
oxidation period. Therefore, it is very difficult to isolate the effects of individual
parameters on the oxidation process. Much research on laboratory-flames has
been done where it is easier to isolate the different effects. Still, it is hard, or
nearly impossible, to measure how the turbulence affects the oxidation rate.
Although it is well known that turbulence does affect oxidation, measurements of
how much the oxidation rate is affected when the turbulence level is increased
do not exist, especially inside an expanding cylinder with a rotating flow
structure at different concentrations and local turbulence levels.

In this work, it was found that engine-out soot emissions decrease

with increased TKE levels inside the cylinder during the post-oxidation phase. A
large amount of TKE was produced by the differences in angular velocity of the
in-cylinder flow caused by the injection. This turbulence increased the mixing of
soot particles with in-cylinder gas. Quantification of the effects of turbulence is
difficult because there is no direct measurement available. Mathematical
modelling of the oxidation process is useful for isolating parameters to determine
their contribution. One approach is to use a transported probability density
function (tPDF) [83] [84] for extracting an ensemble-averaged value of the soot
oxidation rate from local reaction zones with detailed kinetic mechanisms. The
modelling is complex and is a PhD thesis by itself.

This work highlighted the importance of late TKE production for

successful soot oxidation and identified injection and swirl as the source for TKE
production. The flow-field behaviour was determined with the CIV technique,
with its specific possibilities and limitations. Flow data was combined with the 2-
colour method for simultaneous soot and temperature information. The region
that is examined with CIV is located near the surface of the access window
during diffusion combustion. This is due to optical density of diesel flames. With
the CIV method it is the flame surface that is examined, at diffusion combustion.
This is where the interaction between the bowl curvature and flow occurs. During
the post-oxidation phase, the remaining soot cloud in the cylinder is not as dense
as the flame in the diffusion flame period and the CIV correlation was done on a
larger depth into the combustion chamber. The CFD simulations showed also that
the flow is more homogenous at different depths into the piston bowl in this
phase. CIV measurements during post-oxidation is thereby related to a mean
flow over the total bowl depth.

As stated in this thesis, the SN decreases with injection and

combustion. The SN was calculated from the CFD results where the entire flow
domain was known. The total flow domain was not measured with CIV due to
limited optical access to the combustion chamber (the field of view was only 80
mm in diameter out of the total bore diameter of 130 mm). Therefore, the SN

84
calculated from the CIV measurements corresponded only to the flow in the
piston bowl. The CFD simulations revealed that a small decrease in angular
velocity in the outer part of the bowl greatly decreased the total angular
momentum, which lowers the total SN. The injection caused a redistribution of
angular momentum in the cylinder when swirl was present. The SN is thereby
not a good measure of the in-cylinder flow when the flow deviates from solid-
body rotation. The angular velocity profile gives a better understanding of the
phenomena that occur at high injection pressures and can explain why TKE
increases during the post-oxidation phase. The link between TKE and soot
oxidation is well known but is not easily quantified, as discussed earlier.

6 Conclusions
The purpose with this work was to investigate flow and flame interaction and
how this affects soot emissions. Therefore CIV was selected as a practical means
to study flame motion in the piston bowl at realistic high load operation points in
a optical diesel engine. The CIV method as applied for this purpose has been
described. This technique was used to explain how swirl motion introduced
during induction affects soot emissions and interacts with fuel sprays injected at
pressures up to 2500 bar. The comparison between the CIV technique and CFD
simulations showed the same trends in flow behaviour for equal boundary
conditions.

Using the soot created during combustion as a tracer, the CIV

technique can obtain flow-field data during full-load engine operation when
normal PIV measurement is not possible because the bright combustion light
drowns out the introduced laser light. Glowing soot structures in a diesel engine
are excellent tracers for cross-correlation. The particles created during
combustion survive much longer (approximately 35 times longer in this work)
than the time between two exposures with the high-speed camera. The same
particles are thereby captured in the two images that are cross-correlated to
calculate the vector field. The soot particle follows a flow much better than a
normal PIV particle and follows thereby smaller length scales of the turbulence. A
10–30 nm soot particle can theoretically follow turbulence frequencies up to MHz
range. For most applications, frequencies in the kHz range provide enough
resolution for turbulence approximations.

This work examines diesel combustion at low λ. The most important

findings are summarized:

1. A correlation of swirl, tumble, and injection pressure with

engine-out soot emissions was found. Increasing SN at
relatively high injection pressures produced lower soot
emissions.

2. Increasing TN increased the soot emissions and shifted the

swirl vortex off-centre to produce asymmetrical effects on

85
 the vortex geometry. It is suggested that this offset is not
beneficial and negatively influences the engine-out soot
emissions.

3. Both measured CIV results and CFD simulations showed

deviation from solid-body rotational flow during fuel
injection, diffusion combustion, and post-oxidation for
swirl-supported diffusion combustion systems. This
deviation is created by the kinetic energy from the
injection and scales with injection pressure.

4. Solid-body rotational airflow was shown to be a good

assumption in CFD calculations before the start of
injection. Increased injection pressure produces higher
rotational velocity in the central part of the piston bowl. In
the outer part of the bowl, the rotational velocity
decreases slightly compared to the case without injection.

5. The SN decreases during injection due to increased mass

in the cylinder as well as to decreased total angular
momentum caused by turbulent dissipation. The rotational
kinetic energy caused by the injection increases and is the
contributor to the angular velocity difference at different
radii. The injected flame trajectory is slightly bent to the
leeward side by the swirling bulk flow in the cylinder.
When the flame reaches the piston bowl edge, the flame is
redirected back to the central bowl regions with an offset
and forces the bulk flow toward the bowl centre. When the
mass is moved to a smaller radius, the angular velocity
increases for a given angular momentum.

6. The large density difference caused by combustion

increases the angular velocity slightly in the central
regions of the bowl. However, the kinetic energy
introduced by the spray has a larger effect on the
deviation from solid-body rotation than the density
difference.

7. The angular velocity deviation from solid-body rotation is a

source for turbulent kinetic energy (TKE) production. The
TKE is produced from the shear stress that occurs
between the rotating layers in the cylinder with different
angular velocities. When TKE is produced, the angular
velocity decreases in the central regions of the piston bowl
at later CA thereby decreasing the angular momentum
due to dissipation and frictional losses. The angular
momentum decreases faster with injection than without

86
 injection because of increased mass and increased TKE
production. In the case without injection, no large angular
velocity deviation occurs. Therefore, no TKE production
caused by shear can take place. TKE is of paramount
importance for post-oxidation of the remaining soot in the
cylinder and has a large impact on engine-out soot
emissions.

7 Future work
The results of this work indicate that the flow velocity of the in-cylinder gas can
be reduced in the outer part of the piston bowl with increased SN. During
compression, the in-cylinder air velocity is at its maximum around 10 m/s, but
during combustion the velocity of the flame that hits the piston bowl is on the
order of 70 m/s. The bulk flow temperature is around 1000 K during compression
and around 2300 K during combustion. This means that a small change in the
velocity of the flame that hits the piston bowl can produce a larger difference in
heat transfer than the same change in airflow velocity during compression. In
this work, both CIV experiments and CFD simulations show that increased swirl
decreases the flame velocity, publication V. The engine simulations show that
this velocity decrease reduces heat transfer during the combustion event, which
is the most critical part of the entire engine cycle with the highest temperatures
and pressures. This heat transfer decrease is larger than the heat transfer
increase created by higher SN during compression. More engine tests should be
performed to verify this observation.

The CIV technique is a very useful method to extract flow patterns

from natural flame images taken in an optical engine. This technique can be
further developed and integrated into an image evaluation software. The
technique can also be developed for use with other optical accesses to the engine
or with an endoscope. For example, the squish band could be observed through
the cylinder head. Better high-speed cameras with higher spatial and time
resolution would allow resolution of smaller length scales to gain more insight
into complex small-scale turbulence.

The 2-colour method was used to extract the soot density with CA
resolution in this work. When the signal disappears, post-oxidation is still active.
The light from the combustion was too weak at late CA, and evaluation with the
2-colour method was not possible. A light amplifier could be applied to gain
information at late CA. With the light amplifier, the remaining soot could be
studied, and CIV could be used to extract velocity information. This could provide
a better understanding of the late post-oxidation phase in the cycle and the final
engine-out particulate emissions.

87
8 References
1. Diesel, Rudolf. Method of and apparatus for converting heat to work. US.Pat:
542.846 1895.

2. Diesel, Rudolf. Internal combustion engine. US.pat: 608.845 1898.

3. [Online] Scania CV AB. [Cited: 12 09 2013.] http://www.scania.com/.

4. Turns, Stephen R. An Introduction to Combustion, Concepts and

Applications. ISBN 978-007-108687-5.

5. Heywood, John B. Internal Combustion Engine Fundamentals international

edition. 1988. ISBN: 0-07-100499-8.

6. H. Wang, E. Dames, B. Sirjean, D. A. Sheen, R. Tangko, A. Violi, J. Y. W.

Lai, F. N. Egolfopoulos, D. F. Davidson, R. K. Hanson, C. T. Bowman, C. K.
Law, W. Tsang, N. P. Cernansky, D. L. Miller, R. P. Lindstedt. A high-
temperature chemical kinetic model of n-alkane (up to n-dodecane),
cyclohexane, and methyl-, ethyl-, n-propyl and n-butyl-cyclohexane oxidation at
high temperatures, JetSurF version 2.0, September 19, 2010
(http://melchior.usc.edu/JetSurF/).

7. Pires da Cruz, A.; Baritaud, T.; Poinsot, T. Turbulent Self-Ignition and

Combustion Modelling in Diesel Engines. SAE 1999-01-1176.

8. Dec, John E. A Conceptual Model of DI Diesel Combustion Based on Laser-

Sheet Imaging. SAE970873.

9. Warnatz, J, Maas, U and Dibble, R.W. Combustion, Physical and Chemical

Fundamentals, Modeling and Simulation, Experiments, Pollutant Formation 4th
Edition. ISBN-10 3-540-25992-9.

10. Bobba, Mohan; Genzale, Caroline; Musculus, Mark. Effect of Ignition

Delay on In-Cylinder Soot Characterististics of a Heavy Duty Diesel Engine
Operating at Low Temperature Conditions. SAE 2009-01-0946.

11. Kook, Sanghoon; Bae, Choongsik; Miles, Paul C.; Choi, Dae; Bergin,
Michael; Reitz, Rolf D. The Effect of Swirl Ratio and Fuel Injection Parameters
on CO Emission and Fuel Conversion Efficiency for High-Dilution, Low-
Temperature Combustion in an Atomotive Diesel Engine. SAE paper 2006-01-
0197.

12. Kook, Sanghoon; Bae, Choongsik; Miles, Paul C.; Choi, Dae; Picknett,
Lyle M. The Influence of Charge Dilution and Injection Timing on Low-
Temperature Diesel Combustion and Emissions. SAE paper 2005-01-3837.

13. Minato, Akihiko; Tanaka, Tsuneo; Nishimura, Terukazu. Investigation

of Premixed Lean Diesel Combustion with Ultra High Pressure Injection. SAE
paper 2005-01-0914.

88
14. Kanada, Tomohiro; Hakozaki, Takazo. PCCI Operation with Early
Injection of Conventional Diesel Fuel. SAE paper 2005-01-0378.

15. Rate-Limiting Processesin Late-Injection, Low-Temperature Diesel

Combustion Regimes. Miles, P. C.; Choi, D.; Pickett, L. M. s.l. : Thiesel
Conference, 2004.

16. Kimura, S.; Ogawa, H.; Matsui, Y.; Enomoto, Y. An experimental analysis
of low-temperature and premixed combustion for simultaneous reduction of NOx
and particulate emissions in direct injection diesel engines. s.l. : IMechE Int J
Engine Research, 2002. Vol. 3, 4.

17. Kimura, Shuji and et.al. New Combustion Concept for Ultra-Clean and
High-Efficiency Small DI Diesel Engines. SAE 1999-01-3681.

18. A Parametric Study of Low-Temperature Late-Injection Combustion in a HSDI

Diesel Engine. Choi, Dae; Miles, Paul C.; Yun, Hanho; Reitz, Rolf D. 4, s.l. :
JSME International Journal, 2005, Vol. 48.

19. COMMISSION REGULATION (EU) No 582/2011 of 25 May 2011 implementing

and amending Regulation (EC) No 595/2009 of the European Parliament and of
the Council with respect to emissions from heavy duty vehicles (Euro VI)....
[Online] [Cited: 20 12 2011.] http://eur-
lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2011:167:0001:0168:en:PDF.

20. Toxikologi, Institutionen för. Toxikologi. Uppsala Universitet, 1981. ISBN

91-506-0281-0.

21. Jungnelius, Sten; Svartengren, Magnus. Hälsoeffekter av trafikavgaser.

Rapport från Yrkesmedicinska enheten 2000:3. ISSN: 1401-0550.

22. Kook, Sanghoon; Pickett, Lyle. Soot Volume Fraction and Morphology of
Conventional, Fischer-Tropsch, Coal-Derived and Surrogate Fuel at Diesel
Conditions. SAE Int. J. Fuels Lubr. 5(2):2012, doi:10.4271/2012-01-0678.

23. Tree, D.R.; Svensson, K.I. Soot processes in compression ignition engines.
Progress in Energy and Combustion Science 33 (2007) 272–309.

24. Siebers, Dennis L. Chapter 5, Recent Developments on Diesel Fuel Jets

Under Quiescent Conditions. [book auth.] C Arcoumanis and T. Kamimoto. Flow
and Combustion in Reciprocating Engines. Berlin Heidelberg : Springer-Verlag,
2009.

25. Zheng Li; Chonglin Song; Jinou Song; Gang Lv; Surong Dong; Zhuang
Zhao. Evolution of the nanostructure, fractal dimension and size of in-cylinder
soot during diesel combustion process. Combustion and Flame, Volume 158,
2011, Pages 1624-1630.

89
26. Kook, Sanghoon; Zhang, Renlin; Szeto, Kevin; Pickett, Lyle M.;
Aizawa, Tetsuya. In-Flame Soot Sampling and Particle Analysis in a Diesel
engine. SAE Int. J. Fuels Lubr. 6(1):2013, doi:10.4271/2013-01-0912..

27. Glassman, Irvin; Yetter, Richard A. Combustion, 4th edition. San Diego :
Academic press, 2008. ISBN: 9780120885732.

28. Bergman, Miriam; Golovitchev, Valeri I. Application of Transient

Temperature vs. Equivalence Ratio Emission Maps to Engine Simulations. SAE
paper: 2007-01-1086.

29. Gad-el-Hak, M.
http://media.efluids.com/galleries/turbulence?medium=198. [Online] [Cited: 20
December 2011.]

30. Winterbone J.H & Horlock D.E. The Thermodynamics and Gas Dynamics of
Internal Combustion Engines Volume II. ISBN 0-19-856212-8.

31. Tennekes, H.; Lumley, J.L. A first course in turbulence. s.l. : Mit Pr, 1972.
ISBN 0 262 20019 8.

32. Pope, Stephen B. Turbulent flows. Cambridge University Press, 2000. ISBN:
978-0-59125-6.

33. Schwoerer, John; Kumar, Krishna; Ruggiero, Brian; swanbon, Bruce.

Lost-Motion VVA Systems for Enabling Next-Generation Diesel Engine Efficiency
and After-Treatment Optimization. SAE paper 2010-01-1189.

34. Tomoda, T; Ogawa, T; Ohki, H; Kogo, T; Nakatani, K; Hashimoto, E.

Improvement of diesel engine performance by variable valve train system. s.l. :
Int. J Engine Res. Vol. 11. DOI: 10.1243/ 14680874JER586.

35. Deng, Jiamei; Stobart, Richard. BSFC Investigation Using Variable Valve
Timing in a Heavy Duty Diesel Engine. SAE 2009-01-15-25.

36. Suzuki, Takashi. The Romance of Engines. 1997. ISBN 1-56091-911-6.

37. Antila, Eero; Imperato, Matteo; Kaario, Ossi; Larmi, Martti. Effect on
Intake Channel Design to Cylinder Charge and Initial Swirl. SAE paper 2010-01-
0624.

38. Genzale, Caroline L.; Reitz, Rolf D.; Wickman, David D. A Computational
Investigation into the Effects of Spray Targeting, Bowl Gemetry and Swirl Ratio
for Low-Temperature Combustion in a Heavy-Duty Diesel Engine. SAE paper
2007-01-0119.

39. Miles, Paul C. Chapter 4, Turbulent Flow Structure in Direct-Injection, Swirl-

Supported Diesel Engines, p.173 - p.256. [book auth.] C. Arcoumanis and T.
Kamimoto. Flow and Combustion in Reciprocating Engines. Springer-Verlag DOI:
10.1007/978-3-540-68901-0_4, 2008.

90
40. Nordgren, Henrik; Hildingsson, Leif; Johansson, Bengt; Dahlén, Lars;
Konstanzer, Dennis. Comparison Between In-Cylinder PIV Measurements, CFD
Simulations and Steady-Flow Impulse Torque Swirl Meter Measurements. SAE
2003-01-3147.

41. Reuss, David L. Cyclic Variability of Large-Scale Turbulent Structures in

Directed and Undirected IC Engnie Flows. SAE paper 2000-01-0246.

42. Andersson, Öivind. Diesel Combustion Handbook of Combustion. 415-440.

s.l. : Lund University, Department of Energy Sciences, Lund, Sweden, 2010.

43. Alberer, Daniel; del Re, Luigi. On-line Abatement of Transient NOx and PM
Diesel Engine Emissions by Oxygen Based Optimal Control. SAE paper 2010-01-
2201.

44. Tufail, K.; Winstanley, T.; Karagiorgis, S.; Hardalupas, Y.; Taylor, M.
K. P. Characterisation of Diesel Engine Transient Pumping-loss and Control
Methodology for Transient Specific Fuel Consumption (SFC). SAE 2009-01-2748.

45. Kirchen, Patrick; Obrecht, Peter; Boulouchos, Konstantinos. Soot

Emission Measurements and Validation of a Mean Value Soot Model for Common-
Rail Diesel Engines during Transient Operation. SAE paper 2009-01-1904.

46. Darlington, Alex; Glover, Keith; Collings, Nick. A Simple Diesel Engine
Air-Path Model to Predict the Cylinder Charge During Transients: Strategies for
Reducing Transient Emissions Spikes. SAE paper 2006-01-3373.

47. CFD Solver, AVL FIRE 2010.1 user manual, Edition 11/2010.

48. Hanjalić, K.; Popovac, M.; M., Hadžiabdić. A robust near-wall elliptic-
relaxation eddy-viscosity turbulence model for CFD. International Journal of Heat
and Fluid Flow, Volume 25, Issue 6, pp. 1047-1051, December 2004.

49. Reitz, R. D. Modeling atomization processes in high-pressure vaporizing

sprays. Atomisation and Spray Technology, Vol. 3, pp. 309-337, 1988.

50. Colin, O.; Benkenida, A. The 3-Zones Extended Coherent Flame Model
(ECFM3Z) for Computing Premixed/Diffusion Combustion. Oil & Gas Science and
Technology - Revue de l'IFP, Vol. 59, No. 6, pp. 593-609, 2004.

51. Thien, G. Entwicklungsarbeiten an Ventilkanälen von Viertakt-

Dieselmotoren. ÖIZ Gratz 1965, Vol.9.

52. Dembinski, Henrik; Lewis, Clive. Miller-cycle on heavy duty diesel

engines. Stockholm : KTH, 2009. Master Thetis, MMK2 2009:1 MFM124.

53. Morel, T.; Rackmil, I.; Keribar, R.; Jennings, M. J. Model for Heat
Transfer and Combustion in Spark Ignited Engines and Its Comparison with
Experiments. SAE paper 880198.

91
54. Uzkan, Teoman; Borgnakke, Claus; Morel, Thomas. Characterization of
Flow Produced by High-Swirl Inlet Port. SAE paper 830266.

55. Morel, Thomas; Keribar, Rifat. A Model for Predicting Spatially and Time
Resolved Convective Heat Transfer in Bowl-in-Pistin Combustion Chambers. SAE
paper 850204.

56. Markus Raffel; Christian E. Willert; Steve T. Wereley; Jürgen

Kompenhans. Particle Image Velocimetry, A Practical Guide. Springer-Verlag
Berlin Heidelberg, 2007, 2007. ISBN: 9783540723080.

57. http://mathworld.wolfram.com/Cross-Correlation.html. [Online] [Cited: 28

06 2013.]

58. Hart, Douglas P. The Elimination of Correlation Errors in PIV Processing. 9th
International Symposium on Applications of Laser Techniques to Fluid Mechanics,
Lisbon, Portugal, July, 1998.

59. Product-Manual for DaVis 8.1. LaVision GmbH, Göttingen, 2012. Document
name: 1003005_FlowMaster_D81.pdf.

60. [Online] 03 06 2013. http://www.dantecdynamics.com/.

61. TSI. [Online] [Cited: 14 06 2013.]

http://www.tsi.com/uploadedFiles/Product_Information/Literature/Spec_Sheets/
SeedParticles_2980461.pdf.

62. Choi, MY; Hamins, A; Mulholland, GW; Kashiwagi, T. Simultaneous

optical measurement of soot volume freaction and temperature in premixed
flames. Combustion & Flame. 1994;99: s174-186.

63. Tropea, Cameron; Yarin, Alexander L.; Foss, John F. (Eds.). Springer
Handbook of Experimental Fluid Mechanics. Springer verlag. ISBN 978-3-540-
33582-5.

64. H.E. Albrecht; M. Borys; N. Damaschke; C. Tropea. Laser Doppler and

Phase Doppler Measurement Techniques. Berlin, Heidelberg, New York :
Springer-Verlag, 2003.

65. [Online] 17 06 2013. http://www.cambridgeincolour.com/tutorials/image-

noise.htm.

66. Beér, J.M. and Chigier, N.A. Combustion Aerodynamics. London : Applied
Science, 1972.

67. Petersen, Benjamin; Miles, Paul. PIV Measurements in the Swirl-Plane of

a Motored Light-Duty Diesel Engine. SAE 2011-01-1285.

92
68. Ge, Hai-Wen; Reitz, Rolf D.; Willems, Werner. Modeling the Effects of In-
Cylinder Flows on HSDI Diesel Engine Performance and Emissions. SAE 2008-01-
0649.

69. Bakenhus, Marco; Reitz, Rolf D. Two-Color Combustion Visualization of

Single and Split Injections in a Single-Cylinder Heavy-Duty D.I. Diesel Engine
Using an Endoscope-Based Imaging System. SAE paper: 1999-01-1112.

70. AVL. [Online] 25 10 2012. www.avl.com.

71. AVL List GmbH, “ThermoVision Advanced” , AVL Product Guide, 2004.

72. Hampson, Gregory; Reitz, Rolf. Two-Color Imaging of In-Cylinder Soot

Concentration and Temperature in a Heavy-Duty DI Diesel Engine with
Comparison to Moltidimentional Modeling for Single and Split Injections. SAE
paper 980524.

73. Larsson, Anders. Optical Studies in a DI Diesel engine. SAE 1999-01-3650.

74. Svensson, Kenth; Mackrory, Andrew; Richards, Michael; Tree, Dale.

Calibration of an RGB, CCD Camera and Interpretation of its Two-Color Images
for KL and Temperature. SAE: 2005-01-0648.

75. Schmidradler, Dieter. Temperaturmessung im Verbrennungsraum eines

Dieselmotors mittels RGB-Kamera. Wien 1999. Doktorial thesis, Technischen
Universität Wien, Matrikelnummer 8826156.

76. Pastor, J V; Garcia, J M; Pastor, J M; Buitrago, J E. Analysis Methodology

of Diesel Combustion by Using Flame Luminosity, Two-Colour Method and Laser-
Induced Incandescence. SAE 2005-24-012.

77. Reitz, Rolf D; Bergin, Michael J; Oh, Seungmook; Miles, Paul C;

Hildingsson, Leif; Hultqvist, Anders. Fuel Injection and Mean Swirl Effects on
Combustion and Soot Formation in Heavy Duty Diesel Engines. SAE paper 2007-
01-0912.

78. Siebers, Dennis L. Recent Developments on Diesel Fuel Jets Under

Quiescent Conditions, Page 257-308. [book auth.] C Arcoumanis and T.
Kamimoto. Flow and Combustion in Reciprocating Engines. Berlin Heidelberg :
Springer-Verlag, 2009.

79. Shimoda, Masatoshi; Shigemori, Masashi; Tsuruoka, Shingo. Effect on

Combustion Chamber Configuration on In-cylinder Air Motion and combustion
Characteristics of D.I. Diesel Engine. SAE paper 850070.

80. Shigemori, Masashi; Tsuruoka, Shingo; Shimoda, Masatoshi.

Development of a Combustion System for a Lighit Duty D.I. Diesel Engine. SAE
paper 831296.

93
81. Shiozaki, T; Suzuki, T; Shimods, M. Observation of Combustion Process in
D.I Diesel Engine via high Speed Direct and Schlieren Photography. SAE paper
800025.

82. Suzuki, Takashi; Shiozaki, Tadakazu. A New Combustion System for the
Diesel Engine and Its Analysis via High Speed Photography. SAE paper 770674.

83. Jaishree, Jaishree. Lagrangian and eulerian probability density function

methods for turbulent reacting flows. 2011. Dissertation in Mechanical
Engineering, The Pennsylvania State University.

84. Kung, E. H.; Haworth, D. C. Transported Porbability Density Funktion

(tPDF) Modelling for Direct-Injection Internal Combustion Engines. s.l. : SAE Int.
J. Engines. Vol. 1, 1. 2008-01-0969.

85. Ikuo Yamaguchi, Toshio Nakahira and Masanori Komori. An Image

Analysis of High Speed Combustion Photographs for D.I. Diesel Engine with High
Pressure Fuel Injection. SAE paper 901577, 1990

94
9 Summary of publications
A summary of each of the appended publications is provided in this section. All
measurements, GT-POWER simulations, and writing of the publications were
completed by the author of this thesis and reviewed by Professor Hans-Erik
Ångström. The co-author of publication 5, Hannan Razzaq, completed the CFD
simulations and the author of this thesis, Henrik Dembinski, completed the post-
processing of the CFD data.

9.1.1.1 Publication 1
An Experimental Study of the Influence of Variable In-Cylinder Flow, Caused by
Active Valve Train, on Combustion and Emissions in a Diesel Engine at Low λ
Operation. Henrik W. R. Dembinski & Hans-Erik Ångström. SAE paper: 2011-01-
1830

The paper describes an investigation of how in-cylinder flow variation affects

combustion and measured engine-out emissions in a single-cylinder HD engine.
The variation of the in-cylinder flow was imposed by a Lotus active valve train
(AVT), which made it possible to vary SN from 0.4 to 6.7 and TN from 0.5 to 4
by changing the valve profile. The investigated load points were chosen to reflect
critical operation conditions that occur in an engine transient on a turbo diesel
engine. Injection pressure was varied between 500 bar and 2500 bar, load varied
from 10 bar to 20 bar IMEP, and EGR was set to 0% or 30%. The main
conclusion was that measured tailpipe soot emissions can be strongly reduced,
even at high injection pressures up to 2500 bar, when swirl was varied at all
tested loads both with and without EGR. It was also found that RoHR was
affected with higher heat release (HR) rate at diffusion combustion period for all
tested injection pressures. This suggests that the air mixing during combustion
increases. During the post-oxidation phase, increased SN reduced the HR rate
faster than low SN case, and this affected the oxidation rate of remaining soot.

9.1.1.2 Publication 2
Optical study of swirl during combustion in a CI engine with different injection
pressures and swirl ratios compared with calculations. Henrik W. R. Dembinski &
Hans-Erik Ångström. SAE paper: 2012-01-0682

This is the first paper that introduces the use of the CIV method with an optical
engine. The load points from Publication 1 were repeated in an optical engine,
and the combustion was filmed with a high-speed camera. The images were
cross-correlated with Lavision PIV software, and flow-field data were extracted
for the post-oxidation period. It was observed that injection pressure strongly
affected the flow field even during the post-oxidation period. The results showed
a large deviation from solid-body rotation due to the kinetic energy introduced by
the injection when swirl was used. CA-resolved SN was calculated for the
observed volume in the cylinder. The CIV results were compared with GT-POWER

95
1-D simulations and CFD calculations. The CFD and CIV produced similar results,
but the 1-D simulations over-predicted the SN at high swirl.

9.1.1.3 Publication 3
The effects of injection pressure on swirl and flow pattern in diesel combustion.
Henrik W. R. Dembinski. International Journal of Engine Research IJER-12-0006

In this journal article, the diffusion combustion and post-oxidation events were
captured with the CIV technique. Injection pressure and swirl were varied to
determine how changes in the flow field affected engine-out soot emissions. Two
different bowl geometries were tested in the optical engine. It was found that
measured SN in the piston bowl increased with injection pressure due to
redistribution of angular momentum in the piston bowl. Variations in swirl,
tumble, and injection pressure altered the engine-out soot. Tumble offsets the
swirl vortex so that the injected fuel sprays are exposed asymmetrical to the
rotating vortex in the cylinder. This was assumed to be the explanation for
increased soot emissions at high TN. As first indicated in publication 2, the post-
oxidation flow vortex deviates strongly from solid-body rotation and this was
assumed in this journal to be a central mechanism for post-oxidation.

9.1.1.4 Publication 4
Swirl and Injection Pressure Impact on After-Oxidation in Diesel Combustion,
Examined with Simultaneous Combustion Image Velocimetry and Two Colour
Optical Method. Henrik W. R. Dembinski & Hans-Erik Ångström. SAE paper:
2013-01-0913

This publication combined CIV with the 2-colour method. By applying Planck’s
theory of black body radiation, it was possible to extract temperature and soot
measurements from the radiating flame. Flow field, soot, and temperature data
were collected during combustion and post-oxidation to examine how the flow
field affects soot formation and oxidation. By increasing injection pressure from
500 bar to 1000 bar, the maximum KL was amplified during combustion by 50%,
but the measured tailpipe soot decreased from 1.22 FSN to 0.49 FSN. The solid-
body deviation during the post-oxidation phase of the combustion increased for
the 1000 bar case. The results indicated that he flow field during the late part of
the cycle has thereby a strong impact on tail pipe soot emissions. The amount of
soot created during the diffusion combustion has less impact on the total tail pipe
soot compared to the flow field effects during post-oxidation.

9.1.1.5 Publication 5
In-cylinder flow pattern evaluated with combustion image velocimetry, CIV, and
CFD calculations during combustion and post-oxidation in a HD diesel engine.
Henrik W. R. Dembinski, Hannan Razzaq & Hans-Erik Ångström. SAE paper:
2013-24-0064

96
The flow structures simulated with CFD and measured with CIV are compared in
this publication. Large-scale swirl vortices were evaluated with the CIV
technique, and the results were compared to the CFD results at different
distances from the piston bowl surface. The flow field according to CIV is shown
to resemble the flow quite near the optical piston bowl surface during the
diffusion combustion period in the CFD results. During the post-oxidation period,
the CFD-calculated mean velocity in the total depth from cylinder head to piston
surface agreed with the measured CIV data.

9.1.1.6 Publication 6
Swirl and Injection Pressure Effect on Post-Oxidation Flow Pattern Evaluated with
Combustion Image Velocimetry, CIV, and CFD Simulation. Henrik W. R.
Dembinski & Hans-Erik Ångström. SAE paper: 2013-01-2577

This paper explains the mechanism behind why the in-cylinder flow deviates from
solid-body rotational flow when high injection pressures are used with swirl. The
main findings show that with increased injection pressure and swirl, the angular
velocity increases in the centre of the piston bowl while decreasing slightly in the
outer region. When injection starts, the total angular momentum decreases
slightly, but the total rotational kinetic energy increases significantly. The
redistribution of the angular velocity is caused by the driving force from the
injection. The SN also decreased with high injection pressure in the CFD
calculations where the entire flow domain was known.

97