Final Project :

Characterization of Diesel Sprays

in the Eindhoven High Pressure Cell





















Thomas GIMENES
EI3 Energtique & Environnement
Final Project 10/04/06 25/08/06
Dept. Mechanical Engineering
Thermo Fluids Engineering
Combustion Technology
Introduction

Diesel engine designers are challenged by the need to deal with ever more stringent
emissions standards while at the same time improving engine efficiency. In order to achieve these
goals, a better understanding of the diesel combustion and emissions process is critical. Within the
several issues raised by this need of investigation, the spray characteristics (dynamics, combustion
and soot formation) are studied on the Eindhoven High Pressure Cell (EHPC) facility of the
Technical University of Eindhoven (TU/e). For my final project, I have been involved in this
characterization on various tasks which are reported below.

The first part is a presentation of the Combustion Technology Section and the experimental
set up EHPC which is a high pressure combustion vessel with optical access where diesel sprays are
injected and studied. The main advantages is that it is possible to simulate engine-like conditions.
The optical techniques (Schlieren and Mie scattering) are also presented.

In Part II, investigations about the cooling phase after pre-combustion in the EHPC are
presented. It includes the modelling of cooling with conduction and convection in comparison with
experimental results and investigations on the pressure measurements.

In Part III, the results of the computation of the Siebers and Hiroyasu models for maximum
liquid penetration (liquid length) in diesel sprays are presented. This part includes the computation,
theoretical explanations for both models and a comparison between numerical results and
experiments performed on the EHPC in order to validate these models for liquid length prediction.

In Part IV, a literature study about soot diagnostic is presented. It includes an overview of
the different optical techniques for soot characterization (size, distribution, temperature) in a high
pressure and high temperature environment for further research.


Acknowledgements

First of all I would like to thank Professor Philip de Goey for having accepted me for my
final project in the Section Combustion Technology. Furthermore, I am grateful towards Carlo
Luijten and Peter Frijters who have supervised my project and helped me, as well as Roel Peters for
the experiments. I also thank all the people from Combustion Technology who have helped me and
have contributed to my good stay in TU/e.
































Part I

EHPC Experiments
TU/e - Combustion Technology
High Pressure Cell Set Up

















Final Project : Characterization of Diesel Sprays I-2
Technical University of Eindhoven TU/e
Mission
The TU/e intends to be a research-driven, design-oriented university of technology, with the
primary objective of providing young people with an academic education within the engineering
science & technology domain. The main pillars on which the university positions itself are the
following :

The primary task is the training of engineers (at Master of Science level) possessing a sound
scientific basis and depth of knowledge, as well as the necessary competences that will enable them
to develop successful careers in a wide range of fields and functions within the community. The
bachelors programs (BSc) are intended to serve as a basis for further education at Masters level.
The TU/e also trains teachers (at Master of Science level), designers (at the Professional Doctorate
in Engineering level) and researchers (at Doctor of Philosophy level). In addition, it provides
postacademic programs and courses.



In the research field the TU/e prefers to focus, within the engineering science & technology
domain, on the specific areas in which it takes or can take a significant role in the international
scientific world, and in which it can make meaningful contributions to the knowledge-intensive
industries and other sectors of the community with a high, or rapidly developing, technology
intensity. The TU/e strives to ensure that its research results are translated into successful
innovations and serve as a basis for the creation of new enterprises. It actively encourages students
and staff to opt for entrepreneurship.
Final Project : Characterization of Diesel Sprays I-3
Departments
The TU/e is divided in 9 departments composed of various sections. Within these 9
departments, there is the Mechanical Engineering Department.

Architecture, Building and Planning
- Structural design
- Physical aspects of the built
environment
- Building engineering
- Construction engineering
- Architecture
- Design systems
- Urban planning

Electrical Engineering
- Information and communication
systems
- Measurement and control systems
- Power engineering
- Telecommunication and
electromagnetism

Chemical Engineering and Chemistry
- Inorganic chemistry
- Catalysis
- Chemical reactor engineering
- Analytical chemistry
- Polymer chemistry/coating
technology
- Solid state chemistry and materials
science
- Separation processes and fluid
phenomena
- Process development
- Macromolecular and organic
chemistry
- Synthetics technology

Technology Management
- Human Performance Management
- Information Systems
- Operations, Planning, Accounting
and Control
- Organisation Science and Marketing
- Quality and Reliabilitiy Engineering

Industrial Design
Applied Physics
- Physics and applications of ion
beams
- Physics and application of
accelerators
- Physics of surfaces and boundary
layers
- Gas dynamics
- Turbulence and vortex dynamics
- Semiconductor physics
- Equilibrium and fluid dynamics in
plasmas
- Elementary processes in gas
discharges
- Physics of nano-structures
Theoretical and experimental nuclear
physics and quantum electronics
- System and control engineering
Low temperatures
- Physical informatics and clinical
physics
- Classical theoretical physics and its
foundations

Mechanical Engineering
- Computational Rheology
- Mechanics of materials
- Polymer technology
- Process technology
- Combustion technology
- Energy technology
Dynamics and Control
- Systems engineering
- Precision engineering
- Control systems technology
- Design of industrial products

Biomedical Engineering
(in co-operation with Universiteit Maastricht)

Mathematics and Computer Science
- Mathematics
- Computer science
Final Project : Characterization of Diesel Sprays I-4
Mechanical Engineering Department
The Mechanical Engineering Department is divided into three divisions each composed of
different sections. The Section Combustion Technology is part of the Division Thermo Fluids
Engineering.

Division Computational and Experimental Mechanics (CEM)
Mechanics of materials
Polymer Technology
Micro- & Nano-Scale Engineering

Division Thermo Fluids Engineering (TFE)
Process Technology
Combustion Technology
Energy Technology

Division Dynamical Systems Design (DSD)
Dynamics and Control
Systems Engineering
Control Systems Technology
Section Combustion Technology
Research

An important aim of the research carried out in the Combustion Technology group is to
support the development of efficient and clean combustion processes of organic fuels in laminar
and turbulent flows. This is related to the development of new combustion equipment for domestic
and industrial appliances, with a special focus on engine applications. Tailor-made theoretical and
numerical models are developed using state-of-the-art physical and chemical knowledge of
reacting flow systems. These models are validated with local measurements of velocities,
temperatures and concentrations by LDV, PIV and LIF. The research is funded by STW, SMO
(Stichting Motoren Onderwijs Nederland), EU, EZ and a number of companies.

Areas of interest

The research in the Combustion Technology section can be divided in five main areas of
interest:
New Combustion Models
Turbulent Combustion Modelling
Optical Measurement Techniques in Burner Applications
Internal Combustion Engine Technology
Biomass Conversion

Final Project : Characterization of Diesel Sprays I-5
Region of Eindhoven
The TU/e has been founded 50 years ago and is located in the city of Eindhoven, whose
region is characterized by :

The industrial and technological mainstay of the Netherlands : 50% of the Dutch research &
development expenditure occurs in this region and 30% of the regional employment is
related to the industrial sector.
Attractive location in a natural environment that is close to the Randstad (Amsterdam-
Rotterdam), the industrial Ruhr area in Germany, and the European business centres of
Antwerp and Brussels.
Direct connections to the European road system and the European railway network, and
direct flights to Eindhoven Airport.
Cultural facilities at a high level (including the Van Abbe Museum for Modern Art, the Frits
Philips Music Centre, and the City theatre).
A strong concentration of educational institutes (including Technische Universiteit
Eindhoven, Fontys Polytechnics, and the Design Academy).
International sports facilities (including the Philips Stadium).
An international business centre and the most important shopping centre for the south of the
Netherlands.
A large number of high-tech enterprises, including the multinational Philips Electronics
group, Oc, ASML, DAF, DSM and the Philips High-Tech Campus situated next to the
renowned Philips Research.
700,000 inhabitants and 32,000 enterprises (with many starters in the technology sector).
People and personnel

Intake Bachelor
Students
Intake Master
Students
Total number of
Students
2002/2003 1076 1 6723
2003/2004 1011 54 7032
2004/2005 983 50 7207

2001 2002 2003 2004
Number of personnel 3006 3113 3314 2996
Number of professors (excluding
part-time honorary positions)
128 124 129,6 162
Number of PhD students 493,5 513,9 566,8 591
Administrative and support staff 1339 1355 1473 1246



Final Project : Characterization of Diesel Sprays I-6
Experimental set up

The Eindhoven High Pressure Cell (known as EHPC or "spray bomb") is designed to create
realistic engine-like conditions. The study of Diesel sprays (dynamics, geometry, combustion, soot
formation) in an environment of high pressure and high temperature is performed in the EHPC
through optical diagnostic including Mie-scattering and Schlieren. The experimental set up mainly
consists of a constant-volume combustion vessel with optical access, a high-pressure common rail
fuel injection system, a gas supply device and an optical imaging acquisition system. The gas
supply, the fuel injection and the acquisition system are controlled and operated by a computer.
Combustion vessel
The vessel is a cubic combustion chamber (cf. Figure 1) with a constant volume of 1,2957
dm
3
. Depending on window material, the cell is supposed to be able to undergo a maximum pressure
of 350 bar, even if in practice pressures are limited to 130 bar for quartz windows and 300 bar for
sapphire windows. The three large windows (on entrance, exit and on a side) are all of equal size (50
mm thickness and 100 mm diameter) and permit full access to the vessel for optical diagnostic
requiring access.

Figure 1 : Combustion vessel
As the combustion chamber needs to be supplied and controlled, several devices are
connected on the combustion vessel (cf. Figure 2) :

Two intake and exhaust valves are connected to the gas supplier and set in one
interior face of the cell, so the vacuum can be settled and the vessel can be supplied
with gases (acetylene C
2
H
2
, ethylene C
2
H
4
, oxygen O
2
, nitrogen N
2
, argon Ar and
hydrogen H
2
).

A mixing fan creates turbulence and improves the pre-combustion mixture
homogeneity, which can be ignited by a spark plug mounted on a cell corner.

A fuel injector, described below, is connected to a common rail device and is also
mounted in the cell so the fuel spray is visible through the windows.

Final Project : Characterization of Diesel Sprays I-7

Figure 2 : Schematic of the combustion vessel
To follow the evolution of thermodynamic properties in the vessel during pre-
combustion, pressure transducers and thermocouples are connected to the cell. The
pressure transducers are : a Kistler 7061 piezo-electric (0-100 bar setup) able to
follow combustion processes and a Druck piezoresistive (0-70 bar setup).

The cell is surrounded by a heating system which can heat the walls to 200 C. This
aims at simulating engine wall temperatures and to prevent water condensation on
the window surfaces.

Fuel injection system
The fuel injection system (cf. Figure 3) consists of a high-pressure air-driven pump and an
electronically controlled common rail injector. This system can provide a rail pressure up to 1500
bar at high mass flows, in order to be closed to the real injection systems nowadays in use.
Furthermore, the fuel used for experiments is a regular, commercially available Diesel fuel. With
these conditions, the experiments are as realistic as possible.


Figure 3 : Fuel common rail injection system
The Bosch injector, designed for common rail, has 8 holes (183 m diameter) but only one
fuel spray is studied here, so a drainpipe covers the other ones and collects the fuel injected which is
led to the bottom of the cell. The needle lift of the injector and the injection pressure are measured
with sensors.
Final Project : Characterization of Diesel Sprays I-8
Optical diagnostic
To investigate the Diesel spray characteristics, many optical diagnostic techniques have
been used, especially Mie-scattering and Schlieren. The main principles of these techniques are
presented below.

The Schlieren technique is based on the detection of gradients across which the density of
the subject of study changes. This detection is performed by an optical set up composed by lenses
and a collecting device (cf. Figure 4). The light source must be extended and non coherent. The
light rays are deflected when passing through a fluid presenting an inhomogeneous refractive index,
and they interact with a knife edge, which changes the illumination on the image. This interaction
increases the variations in the distribution of light on the image plane depending on the degree to
which the path of the rays was deflected. The knife edge in conjunction with the image of the light
source acts as a discriminator which makes visible the deflected light beams.

Figure 4 : Schematic of the set up for Schlieren experiments
The Mie-scattering technique (cf. Figure 5) is based on the light which is scattered by
interaction between the subject of study and a sheet of light created by a set of optical devices
(lenses and mirrors). This method only records light (in a small solid angle) scattered by the liquid
phase with a optical acquisition device perpendicularly to the spray. The back-scattered signal
intensity from the spray depends on their concentration.











Figure 5 : Schematic of the set up for Mie-scattering experiments
2 mm
Pin-hole
2 mm
Pin-hole or
Pin-head

EHPC

Xe
Lamp
HS
camera
Liquid fuel
Scattered light
collected
Laser
HS camera
Final Project : Characterization of Diesel Sprays I-9
The light source chosen for Mie-scattering experiments is an Ar-Ion laser (2W, = 488nm),
while filtered light from an AEG Xe Lamp is used for Schlieren experiments. In order to follow the
very short events studied here, a high speed camera (cf. Figure 6) is used to collect the images
created by the optical techniques. The movies collected are then analysed with numerical tools.
Lenses and mirrors are also used to create the optical paths for both techniques and to collect the
data.


Figure 6 : High speed camera
Data acquisition and control
Two networked computers are used to operate the experiments and collect the data. The
main computer controls the experiment hardware (gas supply, valves, injector, spark) and the
signals acquisition and treatment, while the second one controls the high speed camera and collects
its data. For all the measurements, an in-house developed TUeDACS data acquisition system is
used : channels at 1 MHz maximum, corresponding at 2,6e5 samples. Some programs enable the
operator to control the gas supply and to determine the properties of the mixture.
Experimental procedure and conditions
In order to generate in the cell a high temperature and high pressure environment which is
representative of real engine in-cylinder conditions, a method known as pre-combustion has been
adopted. This method consists of the spark ignition and combustion of a premixed combustible gas
mixture in the vessel. This mixture is composed of acetylene C
2
H
2
, ethylene C
2
H
4
, oxygen O
2
,
nitrogen N
2
and hydrogen H
2
. The reaction is a typical combustion :

2 2 2 4 2 2 2 2 2 2 2
a C H b C H c H d O e N f CO g H O h O e N + + + + + + +

By changing the values of the coefficients a, b, c, d and e, the operator can control the
composition of the mixture (lambda, density) and determine its properties after the reaction
(composition, maximal pressure and temperature reached). With this coefficients, it is possible to
create both inert (no oxygen left : h = 0) or reacting (h 0) ambient gas environment after
combustion, so both reacting or non-reacting sprays experiments can be performed.

Final Project : Characterization of Diesel Sprays I-10
A typical pressure history for an experiment on non reacting spray is shown in Figure 7. The
vessel is filled with the gas mixture 2 ms before time zero. After this filling, a spark ignites the
mixture which then burns, causing the pressure rise. The maximal pressure is reached at the end of
the combustion. Then the hot gases produced by the combustion process cool down. The fuel is
injected at a pressure determined by the operator.

















Figure 7 : Example of the pressure history with pre-combustion


Figures

Figure 1 : Combustion vessel............................................................................................................... 6
Figure 2 : Schematic of the combustion vessel.................................................................................... 7
Figure 3 : Fuel common rail injection system...................................................................................... 7
Figure 4 : Schematic of the set up for Schlieren experiments ............................................................. 8
Figure 5 : Schematic of the set up for Mie-scattering experiments ..................................................... 8
Figure 6 : High speed camera .............................................................................................................. 9
Figure 7 : Example of the pressure history with pre-combustion..................................................... 10


0 0.1 0.2 0.3 0.4
10
20
30
40
50
60
70
P
r
e
s
s
u
r
e

i
n

t
h
e

E
H
P
C

[
b
a
r
]
Time [s]
pre-combustion fuel injection
cooling down















Part II

EHPC Experiments
Modelling of the EHPC Cooling


















Final Project : Characterization of Diesel Sprays II-2
Contents

Introduction .................................................................................................. 3
1. Modelling of the cooling ........................................................................ 3
1.1 Previous investigations ...................................................................................................... 3
1.2 Cooling processes .............................................................................................................. 4
1.2.1 Parameters........................................................................................................... 4
1.2.2 Conduction.......................................................................................................... 4
1.2.3 Convection.......................................................................................................... 5
2. Simulation results ................................................................................... 8
2.1 Conduction......................................................................................................................... 8
2.1.1 Short time tests ................................................................................................... 8
2.1.2 Long cooling tests............................................................................................. 11
2.1.3 Time delayed conduction.................................................................................. 14
2.1.4 Discussion......................................................................................................... 14
2.2 Convection ....................................................................................................................... 15
2.2.1 Short time experiments ..................................................................................... 15
2.2.2 Long time experiments ..................................................................................... 19
2.2.3 Discussion......................................................................................................... 21
2.3 Conclusion ....................................................................................................................... 21
3. Pressure measurements......................................................................... 22
3.1 Introduction...................................................................................................................... 22
3.1.1 Issue with the pressure transducer .................................................................... 22
3.1.2 Measurements ................................................................................................... 22
3.2 Issue with the pressure offset ........................................................................................... 23
3.2.1 Pre-combustion experiments ............................................................................ 23
3.2.2 Addition and removal of nitrogen..................................................................... 26
3.2.3 Discussion......................................................................................................... 28
3.3 Reliability of the cooling behaviour................................................................................. 30
3.3.1 Thermal shock .................................................................................................. 30
3.3.2 Influence of turbulence ..................................................................................... 32
3.3.3 Discussion......................................................................................................... 33
3.4 Conclusion ....................................................................................................................... 34
Tables ......................................................................................................... 35
Figures ........................................................................................................ 35
Appendix .................................................................................................... 36
Nomenclature ............................................................................................. 39
References .................................................................................................. 39







Final Project : Characterization of Diesel Sprays II-3
Introduction
For the pre-combustion experiments in the EHPC, the cooling phase of the cell is critical to
understand and model. A better knowledge of the cooling processes would improve the
understanding of the phenomenon and thus the EHPC experiments. While a uniform temperature is
usually assumed and derived from the pressure signal with the perfect gas law, we know from
literature [1] that there are internal temperature gradients, which remain unknown since the cooling
processes are still to be determined. In this report, the modelling of the main processes that are
supposed to occur, conduction and convection, is presented and compared to experimental data.

1. Modelling of the cooling
Based on previous investigations and training periods results, it seems interesting to
compute both conduction and convection processes in order to determine which one is the
dominating process for the EHPC cooling.
1.1 Previous investigations
Some training periods have already been performed on the heat issues (including
combustion and cooling) in the EHPC. A short overview of the main conclusions which are useful
for the later analysis will be presented below.

Hans Snijders has simulated the EHPC cooling with a method of effective volume which is
assumed to be adiabatic, with reference to [2]. This method, based on conduction with a thermal
diffusivity = 35.10
-6
m
2
/s, provided a good agreement with experiments, although the tuning of
is necessary (up to a factor of 2), which could lead to unrealistic values. The thermal diffusivity is
indeed a mixture property that should be calculated from the thermodynamic conditions and the
mixture composition.

Willem Verstegen analysed both conduction and convection (also for core temperature)
processes. His literature review reveals that conduction is assumed to be the leading process in
Rapid Compression Machines [2,3] and sometimes in constant volume cell like the EHPC.
References [4,5] present a thermal boundary layer as a result from conduction, although it may not
apply to the EHPC. Willem Verstegen concluded that conduction was too slow compared to the
experimental pressure fall. Furthermore, his results with convection agreed well to the experiments,
although these results are strongly dependant on the valued gas velocity in the EHPC. This velocity
could be fitted to experimental cooling measurements with the assumption of a cooling dominated
by convection and the use of Reynolds and Prandtl numbers.

Final Project : Characterization of Diesel Sprays II-4
1.2 Cooling processes
1.2.1 Parameters
The core volume at temperature T
c
has a volume V
c
and the total constant volume is
V=(0,108)
3
m
3
. The cell is at a mass averaged (or bulk) temperature T
b
while the walls of the EHPC
are at the temperature T
w
which is supposed to be constant (T
w
= 300 K).
We want to derive the core temperature T
c
temporal profile from the experimental data for
T
b
temporal profile. The conservation of mass provides a relation presented in Eq. 1 between the
volumes and temperatures, assuming that the pressure and molar mass are uniform throughout the
volume.
c bl
bl
bl bl
b c bl c bl c
V V dV V 1 1
dV
T T T T T T
| |
= + = +
|
\

( 1 )
The boundary layer is at a temperature T
bl
and has a volume V
bl
. To solve Eq. 1 and obtain a
relation between T
b
, T
c
and T
w
, the temperature profile over the boundary layer must be known
which depends on the process involved (conduction or convection).
1.2.2 Conduction
1.2.2.1 Theoretical background
In the case of conduction, the temperature T
bl
in the boundary layer at a distance x is given
by Eq. 2 as a function of time, thermal diffusivity and temperatures T
c
and T
w
. The thermal
diffusivity can be calculated with the thermal conductivity , the core density
c
and the heat
capacity C
p
as presented in Eq. 3. The thermal boundary layer thickness can also be determined
with Eq. 4.
( )
bl w
bl w c w
c w
T T x x
erf T T T T erf
T T 2 t 2 t
| | | |
= = +
| |

\ \
( 2 )
c p
C

( 3 )
w
c
2 T
1 t
T
| |
=
|

\
( 4 )
The derivation based on the calculation of the integral term in Eq. 1 is a bit complex: only
the main stages will be presented below. With Eq. 2 and 4 we can calculate the integral as a
function of boundary layer thickness , core surface A calculated from the cell dimension (A =
6(0,108)
2
m
2
), wall temperature T
w
and a parameter c, where it is assumed that A<<V.
bl
bl
bl c 0 w
1 1 A 1 erf (y)
dV dy
T T 1 c erf (y) T

| |

=
|
+
\

with
c
w
T
c 1
T
= ( 5 )

Final Project : Characterization of Diesel Sprays II-5
Then the ratio T
c
/T
b
can be derived from Eq. 1 and 5 as a function of surface A, cell volume,
boundary layer thickness and a function (c) which is defined in Eq. 6.
0
(c 1) 1 erf (y)
(c) dy
1 c erf (y)

+
=
+

( 6 )
c
b
T A
1 (c)
T V

= + ( 7 )
A first analysis allows us to notice that at the initial time t = 0 (reference for the cooling
start), the layer thickness is equal to zero, and thus T
c
is equal to T
b
.
Another possibility to calculate T
c
is using the adiabatic relation given in Eq. 8 as a function
of core density
c
and the gamma ratio between C
p
and C
v
. The core density
c
is calculated with the
boundary layer thickness as presented in Eq. 9 where m
0
is the total mass.
c c
c c
dT d
( 1)
T

( 8 )
0
c
m
V A
=
+
( 9 )
1.2.2.2 Computation
An experimental pressure signal is used to calculate the T
b
profile with the perfect gas law.
The properties of the mixture after combustion (composition, mass, molar mass) are collected from
the experimental data.
For each time step, T
b
is derived from P
b
, and then the boundary layer thickness is
determined (Eq. 4). With we can find with Eq. 7 the ratio T
b
/T
c
with T
b
as a first estimate of T
c
in
the parameter c and in (to be refined in an iterative process). These calculations require the
computation of many physical properties such as the conductivity , the heat capacities C
p
and C
v
,
gamma, the thermal diffusivity and the density
c
. Most of them can be calculated from Carlo
Luijten's physical properties files.
1.2.3 Convection
1.2.3.1 Theoretical background
For convection transfer, the temperature T
bl
in the boundary layer at a distance x is given by
Eq. 10 as a function of time and boundary layer thickness .
1
7
bl w
c w
T T x
T T

| |
=
|

\
for x , and 1 for x > ( 10 )
1
1
7
bl 1
bl
7
bl c c 0
1 1 cA 1 y
dV dy
T T T
1 cy
| |
=
|
\ +

with
c
w
T
c 1
T
= and
x
y =

( 11 )

Final Project : Characterization of Diesel Sprays II-6
Then the ratio T
c
/T
b
can be derived from Eq. 1 and 10 as a function of surface A, cell
volume, boundary layer thickness and a function ) (c which is defined in following Eq. 12 and
obtained after calculation of the above integral of Eq. 11.
7
6 5 4 3 2
7
c 1 c
(c) 60 70c 84c 105c 140c 210c 420c 420ln(c 1)
60c c 1
+
= + + + + +

+

(12)
c
b
T A
1 (c)
T V

= + ( 13 )
According to this last Eq. 13, the boundary layer thickness must be known to calculate the
ratio T
c
/T
b
. Following are all the equations and steps needed to obtain the boundary layer thickness
from the T
b
profile:
Heat transfer coefficient h :
b
c w
c v
dT hA
(T T )
dt VC
=

( 14 )
Nusselt number :
hL
Nu =

( 15 ) where L is a characteristic length (L = 0,108 m)

Reynolds number from Nusselt relation :
1 1
3 2
Nu 0.33Re Pr = ( 16 )
where Pr is considered as a constant, Pr = 0,7
Core velocity from Reynolds relation :
c
Lv
Re

( 17 )
Boundary layer :
o
1
2
5LRe

= in laminar case (Re < 5.10

5
) ( 18 )
o
1
5
LRe

= in turbulent case (Re > 5.10

5
) ( 19 )

The Mach number Ma is calculated according to the following Eq. 20 with the specific gas
constant r given by Eq. 21 where
0
R is the universal gas constant ( ) . /( 3144 , 8
0
K mol J R = ) and M
the molar mass of the gas:
v
Ma
rT
=

( 20 )
0
R
r
M
= ( 21 )

Final Project : Characterization of Diesel Sprays II-7
1.2.3.2 Computation
An experimental pressure signal is used to calculate the T
b
profile with the perfect gas law.
The properties of the mixture after combustion (composition, mass, molar mass) are collected from
the experimental data.
As a first approximation we consider that T
c
=T
b
(for the right part of Eq. 14). From step 14,
the convection heat transfer coefficient is obtained, for each time step, from the derivative of mass
averaged temperature Tb. All the physical properties (conductivity, heat capacity and viscosity) are
calculated using Carlo Luijtens files. From Eq. 15 we obtain the Nusselt number which is useful
for computing Reynolds number with Eq. 16 considering that Prandtl number is a constant (this
assumption has been well checked).
Then the boundary layer thickness is calculated in the laminar case and turbulent case which
are both defined by Reynolds number. The limit between the two cases is around a Reynolds
number of 5.10
5
which means that in reality there is a transition zone between the two behaviours.
Finally, using Eq. 13 and the boundary layer thickness value, the ratio T
c
/T
b
can be
calculated for each time step. In order to refine the result, an iterative process is made by
substituting T
b
by T
c
(just obtained above) in the right part of Eq. 14.


The MATLAB
TM
scripts written for the computation of conduction and convection can be
found in Appendix A.

Final Project : Characterization of Diesel Sprays II-8
2. Simulation results
The cooling temperature signals have been extracted from the experimental signals : only
the temperature decrease after the maximal temperature is presented on the following figures.
2.1 Conduction
Since conduction is expected to be dominated by convection on the first phase of cooling,
the model of conductive cooling has been tested also starting after this time. Core temperature T
c

derived from both T
c
/T
b
ratio (Eq. 7) and adiabatic relation (Eq. 8) have been compared to the bulk
temperature T
b
. Short time experiments (2,6s) and long time ones (26s) have been performed with
the mixing fan switched off to measure the natural processes.
2.1.1 Short time tests
2.1.1.1 Experimental results
In this part, the models have been tested on the beginning of the cooling (samples of 2,6
seconds). On Figure 1, the core temperature results derived from both T
c
/T
b
ratio ("zeta") and
adiabatic relation are plotted for a mixture density of = 3,8 kg/m
3
. The temporal evolution of
temperature difference between "zeta" T
c
and T
b
is represented in Figure 2.

0 0.5 1 1.5 2 2.5 3
500
1000
1500
2000
2500
Evolution of Temperature for the Cooling
Cooling Time [s]
T
e
m
p
e
r
a
t
u
r
e

i
n

t
h
e

E
H
P
C

[
K
]
Tb
Tc zeta
Tc adiabatic

Figure 1 : Comparison of T
b
and T
c

At first sight we can notice that the adiabatic assumption is not representative for this kind
of cooling : the final core temperature obtained with Eq. 8 is 2200 K instead of around 600 K. This
model is much too slow for such a cooling with a fast decrease at the beginning.

Final Project : Characterization of Diesel Sprays II-9

0 0.5 1 1.5 2 2.5
0
20
40
60
80
100
120
Difference of Temperature Conduction-Exp.Data for the Cooling
Cooling Time [s]
T
e
m
p
e
r
a
t
u
r
e

i
n

t
h
e

E
H
P
C

[
K
]

Figure 2 : Evolution of temperature difference between "zeta" T
c
and T
b
After an initial increase in the difference between "zeta" T
c
and T
b
, due to the initialised
value of T
c
, the two signals go closer with the time. However, the first increasing difference could
result from both a wrong assumption of conduction.

The evolution of the boundary layer thickness in Figure 3 does not seem realistic with an
asymptote after approximately 1,5 s and a maximal value of 3 mm. The evolution of density in
Figure 4 is realistic : the density reaches a minimal value at the same time that the boundary layer
thickness reaches its maximum, as predicted by Eq. 9.

0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
2
2.5
3
3.5
x 10
-3
Boundary Layer Thickness delta
Cooling Time [s]
B
o
u
n
d
a
r
y

L
a
y
e
r

T
h
i
c
k
n
e
s
s

[
m
]

Figure 3 : Evolution of boundary layer thickness

Final Project : Characterization of Diesel Sprays II-10
0 0.5 1 1.5 2 2.5 3
3.2
3.3
3.4
3.5
3.6
3.7
3.8
3.9
4
Evolution of Density
D
e
n
s
i
t
y

[
k
g
/
m
3
]
Cooling Time [s]

Figure 4 : Evolution of density
2.1.1.2 Discussion
For short time experiments (approximately 2,6s), the results seem to be quite satisfying, as
the calculated "zeta" core temperature get closer to the bulk temperature with a decreasing
difference after initialisation. Furthermore, some parameters (boundary layer thickness and density)
present an unrealistic behaviour with possible values. The adiabatic calculation method seems to be
much too slow to be a good assumption in this case.
However, this could also mean that the cooling is not conductive at all, which could explain
that the adiabatic relation provides very far results from the bulk temperature. The "zeta" relation
would be false in this case.


Final Project : Characterization of Diesel Sprays II-11
2.1.2 Long cooling tests
2.1.2.1 Experimental results
In this part, the models have been tested on the whole cooling (periods of 26 seconds, which
is thought to be long enough to be relevant). On Figure 5, the core temperature results derived from
both T
c
/T
b
ratio ("zeta") and adiabatic relation are plotted for a mixture density of = 3,8 kg/m
3
.
The temporal evolution of temperature difference between "zeta" T
c
and T
b
is represented in Figure
6.
0 5 10 15 20 25 30
400
600
800
1000
1200
1400
1600
1800
Evolution of Temperature for the Cooling
Cooling Time [s]
T
e
m
p
e
r
a
t
u
r
e

i
n

t
h
e

E
H
P
C

[
K
]
Tb
Tc zeta
Tc adiabatic

Figure 5 : Comparison of T
b
and T
c
for a whole cooling
5 10 15 20 25
0
20
40
60
80
Difference of Temperature Conduction-Exp.Data for the Cooling
Cooling Time [s]
T
e
m
p
e
r
a
t
u
r
e

i
n

t
h
e

E
H
P
C

[
K
]

Figure 6 : Evolution of temperature difference between "zeta" T
c
and T
b
The difference is first increasing (the initial equality is due to the initial value of thermal
boundary thickness ) and then decreasing until approximately t = 4s, and then increase again. The
initial increase and then decrease could indicate that conduction is the leading phenomenon after a
certain time (around t = 0,5s). However, the final increase after t = 4s does not confirm this
assumption. Furthermore, the core temperature is even increasing a bit, which is not acceptable.

Final Project : Characterization of Diesel Sprays II-12
2.1.2.2 Comparison with Hans Snijders modelling
Since the modelling presented above does not provide good results on long cooling tests, we
can compare these results with the modelling done by Hans Snijders (index H.S.) with a constant
thermal diffusivity of 35.10
-6
, 60.10
-6
or 70.10
-6
m
2
/s depending on experimental results fitting.
The results are presented for
HS
= 35.10
-6
m
2
/s in Figure 7 and Figure 8.

0 5 10 15 20 25 30
400
600
800
1000
1200
1400
1600
1800
Evolution of Temperature for the Cooling
Cooling Time [s]
T
e
m
p
e
r
a
t
u
r
e

i
n

t
h
e

E
H
P
C

[
K
]
Tb
Tc zeta
Tc zeta H.S
Tc adiabatic
Tc adiabatic H.S.

Figure 7 : Comparison of T
b
and T
c
for calculated and H.S. diffusivity values

0 5 10 15 20 25 30
0
50
100
150
200
Difference of Temperature Conduction-Exp.Data for the Cooling
Cooling Time [s]
T
e
m
p
e
r
a
t
u
r
e

i
n

t
h
e

E
H
P
C

[
K
]
Tc cond - Tb
Tc cond H.S. - Tb

Figure 8 : Difference between "zeta" T
c
and T
b
for calculated and H.S. diffusivity values
The behaviour of "zeta" T
c
is almost the same than with the calculated thermal diffusivity ,
although the difference with T
b
is higher. With the adiabatic relation, the cooling is faster with the
H.S. value, although it is still much too slow. For higher values of , the difference is higher. The
maximum value calculated is around 20.10
-6
m
2
/s.

Final Project : Characterization of Diesel Sprays II-13
2.1.2.3 Discussion
From these long cooling tests, we know that the adiabatic relation does not provide a good
core temperature for the whole cooling. The "zeta" core temperature behaviour is more satisfying,
although we obtain a rise in temperature, probably due to the evolution of (Eq. 4) which is
increasing all the time (except a slight plateau) due to the influence of the factor in (t)
1/2
which is
stronger than the factor (1-T
w
/T
c
).
The evolutions of and deriving and are strange, because they seem to indicate that the
model is only valid for a certain range of time (not more than 3-4 seconds) as suggested by the
change in behaviour illustrated by Figure 9 and Figure 10. However, no other model of conduction
has been found to be better appropriate to this problem, although there are probably.
0 5 10 15 20 25 30
2.8
3
3.2
3.4
3.6
3.8
Evolution of Density
D
e
n
s
i
t
y

[
k
g
/
m
3
]
Cooling Time [s]

Figure 9 : Evolution of density
0 5 10 15 20 25 30
0
1
2
3
4
5
6
7
x 10
-3
Boundary Layer Thickness delta
Cooling Time [s]
B
o
u
n
d
a
r
y

L
a
y
e
r

T
h
i
c
k
n
e
s
s

[
m
]

Figure 10 : Evolution of boundary layer thickness
Both and evolutions present a plateau at the same time (approximately between 1,5 and
2,5 seconds after the start of cooling). This behaviour is expected (the boundary layer is not
supposed to grow infinitely due to geometrical limits and the density has a minimal value) but not
so early if conduction is dominating the whole cooling.

Final Project : Characterization of Diesel Sprays II-14
2.1.3 Time delayed conduction
As "zeta" conduction and adiabatic relations seem to provide no good results on whole
cooling, we have tested them after a few seconds when convection is assumed to be neglected (no
more efficient velocity). Figure 11 show the results for an initial temperature T
c
= 530 K after t = 4s
(derived from convection modelling presented in further stages).

5 10 15 20 25
450
500
550
600
Evolution of Temperature for the Cooling
Cooling Time [s]
T
e
m
p
e
r
a
t
u
r
e

i
n

t
h
e

E
H
P
C

[
K
]
Tb
Tc zeta
Tc adiabatic

Figure 11 : Comparison of T
b
and T
c
for time delayed conductive cooling
We can see that the "zeta" relation is not influenced by the initial temperature, and it
provides the same results than previously. However, the adiabatic relation is much more appropriate
for this time delayed cooling than for a whole cooling or the first cooling phase. The shape of the
curve agrees quite well with the T
b
signal, although we should obtain a decreasing difference, which
is not totally the case.
It might be possible that the adiabatic relation is really representing the cooling after a first
convective cooling phase : it would explain that at the beginning (cf. short time experiments) the
adiabatic relation is much too slow. However, the core density is derived from with Eq. 9, and
calculated from Eq. 4 is assumed to be dubious.
2.1.4 Discussion
From these experiments with conduction, we can assume that the "zeta" relation is not
appropriate to simulate the cooling phase, as it provides results close to the reference signal when it
should not (convection is dominating at the beginning) and unrealistic results when conduction is
dominating. For the modelling of convection, the adiabatic relation seems to be more realistic, as
we can see with time delayed simulations with initial conditions provided by convection results
(developed in further stages).



Final Project : Characterization of Diesel Sprays II-15
2.2 Convection
2.2.1 Short time experiments
The results presented here are obtained with a pressure signal of 2,6s at 100 kHz which
presents no singularities (no thermo shock for example). The gas density is 10,6 kg/m3. According
to the modelling explained in the previous section and to the Matlab scripts, the most important
results are plotted just below and then explained.

In Figure 12 is plotted Reynolds number which is obtained by the Nusselt relation (Eq. 16).
This number is helpful to determine whether the flow is turbulent or laminar.

0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
2
2.5
x 10
6
Reynolds number
Time [s]

Figure 12 : Reynolds number (experiment time: 2,6s)
First we can observe that Reynolds dimensionless number decreases from 2.10
6
to 2.10
5
.
Because the limit between turbulent and laminar is around 5.10
5
, the curve crosses this envelope
limit. In fact, this is not a precise point but more a transition zone which is situated somewhere
between 10
5
and 10
6
.

If we consider the limit at 5.10
5
, this leads to a discontinuous and unphysical boundary layer
thickness: at the beginning, when Re > 5.10
5
, we are in the turbulent case so Eq. 19 is used and then
Eq. 18 is used for the laminar part of the curve. Thats why the two cases are studied separately.
The two boundary layer thicknesses, which are obtained using the Reynolds number, are given in
Figure 13 and Figure 14.







Final Project : Characterization of Diesel Sprays II-16
0 0.5 1 1.5 2 2.5 3
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
x 10
-3
Boundary layer thickness in laminar case[m]
Time [s]

Figure 13 : Boundary layer thickness in laminar case (experiment time : 2,6s)
In laminar case, the boundary layer is increasing from 0,4 mm to 1,4 mm. These low values
and this kind of behaviour seem to be both realistic. In the turbulent case, the boundary layer
thickness is much higher and increasing from 6 mm to 1 cm as depicted in Figure 14.

0 0.5 1 1.5 2 2.5 3
5
6
7
8
9
10
11
x 10
-3
Boundary layer thickness in turbulent case[m]
Time [s]

Figure 14 : Boundary layer thickness in turbulent case (experiment time: 2,6s)

For the turbulent case, the boundary layer thickness is around 1 cm at the end of the cooling
which seems to be a little too high [6] than expected because the characteristic length of the EHPC
is 10,8 cm.


Final Project : Characterization of Diesel Sprays II-17
With the boundary layer thickness, we can now obtain the core temperature T
c,
according to
Eq.13, and compare it with the bulk temperature T
b
. The result is plotted in Figure 15 and the
difference between the two curves is given in Figure 16.
0 0.5 1 1.5 2 2.5 3
600
800
1000
1200
1400
1600
1800
2000
2200
2400
2600
Temperature cooling modelling [K]
Time [s]
Tc laminar
Tc turbulent
Tb

Figure 15 : Cooling temperature (experiment time: 2,6s)
We can denote that the laminar profile, depicted in Figure 16, is really close to T
b
because
the difference is less than 10 K and decreasing all along the cooling process.
0 0.5 1 1.5 2 2.5 3
0
20
40
60
80
100
120
Difference between Tc and Tb
Time [s]
Laminar
Turbulent

Figure 16 : Difference between Tc and Tb (experiment time: 2,6s)
For the turbulent case, the difference between T
c
and T
b
is higher but still decreasing as
expected. It is also important to notice that the shape of the curve is nearly the same than the one of
T
b
which means that the convection modelling explains quite well the cooling for this study
(described at the beginning of this sub-section).

Final Project : Characterization of Diesel Sprays II-18
Another important parameter for the convection is of course the velocity. For each time step,
the velocity is calculated from the Reynolds relation (Eq. 17) and the result is given just below in
Figure 17.
0 0.5 1 1.5 2 2.5 3
0
20
40
60
80
100
120
140
Velocity [m/s]
Time [s]

Figure 17 : Velocity (experiment time: 2,6s)
The velocity is decreasing from 130 m/s to 5m/s according to the fact that Reynolds number
decreases a lot during the cooling time. The velocity average is in fact around 15m/s and the start
velocity is not so high than it seems if we calculate the Mach number (Figure 18) by using Eqs. 20
and 21 and if we compare to [7].
The Mach number is at its maximum at the beginning of the cooling process when the
velocity is the highest. Then it is decreasing from 0,14 to less than 0,01 at the end. This proves that
we are not in a supersonic case and that there is no shock wave. We can conclude that the Mach
number is not too high and that a velocity of 130 m/s at the beginning is realistic even if it seems
high at a first approach.
0 0.5 1 1.5 2 2.5 3
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Mach number

Figure 18 : Mach number (experiment time: 2,6s)
Then other experiments were performed and analysed with this model. They were short time
signals (2,6 s or less) at 100 kHz too with the mixing fan switched off. The results are very good too
and comparable to what we obtained just above.

Final Project : Characterization of Diesel Sprays II-19
2.2.2 Long time experiments
The signal length is this time of 26s. The fan is turned off and the fuel density is 3,8 kg/m
3
.
These experiments are analysed by the same model than at the previous sub-section. These
experiments present the particularity that at a certain moment the velocity tends to 0 m/s because of
a very flat T
b
curve after that moment (the T
b
derivative is almost zero in this case).

The consequence of that is that almost all the parameters tend to 0. This leads to a T
c
profile
that does not approach very well the T
b
one with a difference of around 100 K. These last results are
plotted in Figure 19 and Figure 20. The average velocity for this cooling is around 0.8 m/s which
much lower than for the short time experiments.
0 5 10 15 20 25 30
0
5
10
15
20
25
30
Velocity [m/s]
Time [s]

Figure 19 : Velocity (experiment: 26s @10kHz, fan off)
5 10 15 20 25
500
600
700
800
900
1000
Temperature cooling modelling [K]
Time [s]
Tc laminar
Tc turbulent
Tb

Figure 20 : Temperature cooling (experiment: 26s @10kHz, fan off)

Final Project : Characterization of Diesel Sprays II-20
First we can denote the computation difficulties due to a derivative of T
b
which is almost
zero and creates in all the steps of the modelling a lot of irregularities in the curves.
These plots show that after 4s of cooling, the T
b
signal is very flat and leads to a velocity
almost equal to zero. All the parameters are also very low and the consequence is a cooling
temperature modelled quite far from the original signal.
But finally we can say that it proves that the model is quite nice because it becomes false as
soon as the velocity is too low which means that convection is not the leading cooling process at
that moment. It could be conduction and the temperature is too low for providing radiation.
Furthermore, the boundary layer thicknesses obtained for both laminar case and turbulent
one are too high compared to the cell size. For the laminar flow, its value increases and then
decreases to stabilize around 10 cm. For the turbulent flow, the boundary layer thickness has the
same behaviour with a value at the end of the cooling process of 6 cm. These values are unrealistic
for our case because the EHPC length is typically 10,8 cm. They are respectively plotted in Figure
21 and Figure 22.
5 10 15 20 25
0
0.05
0.1
0.15
0.2
0.25
0.3
Boundary layer thickness in laminar case[m]
Time [s]

Figure 21: Boundary layer in laminar case (experiment: 26s @10kHz, fan off)
0 5 10 15 20 25
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
Boundary layer thickness in turbulent case[m]
Time [s]

Figure 22 : Boundary layer in turbulent case (experiment: 26s @10kHz, fan off)

Final Project : Characterization of Diesel Sprays II-21
2.2.3 Discussion
To conclude on all this study of the convection, we can say that the convection modelling
gives some good results. The shape of temperature cooling curve is well reproduced and all the
parameters seem to have physical values (velocity, dimensionless numbers). The difference
between T
c
and T
b
decreases during the cooling process which is in accordance with what was
expected.
As soon as the velocity is zero, it seems that the convection modelling is not enough to
predict the behaviour of the cooling. At that moment conduction (and maybe radiation) is certainly
the leading cooling process because the velocity is too low to predict well the cooling process.
2.3 Conclusion
With these simulations, we are able to confirm that convection is the leading process on the
first cooling phase which is much too fast to be conduction. The parameters of convection derived
from this simulation agree well with realistic and experimental values. However, when convection
starts to be inefficient due to the low velocity, it seems that a conductive process then appears,
although only the adiabatic modelling seems to be realistic. Gas radiation could also be tested to
complete these investigations.



Final Project : Characterization of Diesel Sprays II-22
3. Pressure measurements
3.1 Introduction
3.1.1 Issue with the pressure transducer
According to Willem Verstegen who worked on it, the final pressure value after the cooling
of pre-combustion is significantly different from the initial pressure value in the EHPC, even when
the number of molecules is preserved during the pre-combustion. If the cooling curve obtained with
these measurements is not correct, every parameter value calculated from that pressure, and
especially the temperature in the EHPC, is also wrong.
It is very important to have a good estimate of pressure and temperature in the combustion
chamber for the experiments performed on the EHPC. Indeed, the Diesel spray has to be injected at
the right pressure and temperature values in order to simulate the engine conditions.
These measurement differences observed on experiments may come from various causes.
First, the pressure sensors could give deviated results : two kinds of sensors were used, on whom a
thermal shock due to combustion might produce an offset. It could be also a consequence of the
chemical mix which of course changes during the combustion process and could produce a normal
pressure increase that could be predicted.
This report aims at analysing the issues with the pressure measurement : find out why there
is such a pressure difference, when it happens, what it is based on Previous acquisition data have
been analysed and other measurements have been performed.
3.1.2 Measurements
Two Kistler 7061 piezoelectric pressure transducers (0-100 bar setup) are used to measure
the combustion process in the EHPC. The signal of the Kistler transducers can be measured at
various measurement frequencies MF, though here the experiments are mostly performed at 10
kHz. Due to data acquisition specifications, this means that maximal time of measurement lies
around 26 s. Different channel connections in the cell and signal converter settings of the Kistlers
were also tested.
A Druck piezoresistive pressure transducer (0-70 bar) is also connected to the EHPC. Due to
a valve, this transducer cannot measure the pressure during combustion, but if the operator runs
quickly into the lab to open the valve, it is possible to measure with this sensor the pressure inside
the EHPC just after combustion. It is also possible to leave the inlet valve open during combustion,
but the results might not be good, as the Druck is not designed for such high pressure gradients
created by combustion.
Other pressure measurements can be performed on pressure events due to N
2
addition or
removal from the EHPC in order to see whether or not the problem had to deal with the heat release
caused by the combustion process in the cell, and to check if the two kinds of sensors had the same
response to a variable pressure without combustion.
For all the measurements a TUeDACS data acquisition system is used : 4 channels at 1 MHz
max (about 2.6e5 samples max). The initial pressure and the final pressure are measured, and the
ratio f i P P is calculated.
All the figures are plotted on Matlab from the experimental data. However, as these data
often show a noisy instable behaviour due to the sensibility of sensors, a Savitzky - Golay filter
(polynomial order of 2 and frame size of 1001) has been used in order to smooth the signals and
their derivatives. With this filter, it is easier to analyse the signals which may not be too much
transformed.



Final Project : Characterization of Diesel Sprays II-23
3.2 Issue with the pressure offset
The first problem to solve is to check the supposed offset between initial and final pressure.
Indeed, if there is really a difference between the two values, all the pressure measurements are
wrong, which is critical for the reliability of experiments.
3.2.1 Pre-combustion experiments
First of all, pre-combustion experiments have been performed with the different sensors that
could be used, in order to compare their signals. The interest was focused on the initial and final
values, with the calculation of the pressure ratio.
3.2.1.1 Experiments with Kistlers
A first set of experiments was performed with the Kistlers #18 and #19. For one
measurement, the two Kistlers have been plugged in the other channel (measurement EHPC-003).
The results are shown in the following in the Table 1 :

reference
030806
pressure
transducer
MF
signal
converter
settings
P_initial P_final
ratio
f i P P
P
EHPC-001 KISTLER #19 100 kHz LONG 3,08 4,96 1,61 1,88
EHPC-002 KISTLER #19 10kHz LONG 3,10 3,39 1,09 0,29
EHPC-003 KISTLER #19 10kHz LONG 3,07 3,41 1,11 0,34
EHPC-004 KISTLER #19 10kHz LONG 3,34 3,70 1,11 0,37
EHPC-005 KISTLER #19 10kHz LONG 3,09 3,45 1,12 0,36
EHPC-006 KISTLER #19 10kHz LONG 3,06 3,42 1,12 0,37
EHPC-007 KISTLER #19 10kHz LONG 5,42 6,29 1,16 0,88
EHPC-008 KISTLER #19 10kHz LONG 6,84 7,60 1,11 0,76
1,12 0,48

EHPC-001 KISTLER #18 100 kHz LONG 3,08 4,69 1,52 1,61
EHPC-002 KISTLER #18 10kHz LONG 3,10 3,52 1,14 0,42
EHPC-003 KISTLER #18 10kHz LONG 3,09 3,53 1,14 0,44
EHPC-004 KISTLER #18 10kHz LONG 3,32 3,79 1,14 0,47
EHPC-005 KISTLER #18 10kHz SHORT 3,07 3,04 0,99 -0,02
EHPC-006 KISTLER #18 10kHz MEDIUM 3,05 2,93 0,96 -0,12
EHPC-007 KISTLER #18 10kHz LONG 5,42 6,59 1,22 1,17
EHPC-008 KISTLER #18 10kHz LONG 6,85 7,80 1,14 0,96
1,10 0,47
Table 1 : Comparison between the two Kistlers sensors
It seems that the experiment EHPC-001 performed at 100 kHz is too short to get a steady
final pressure value, and we see that the ratio f i P P is very high (above 1,5). These results will not
be taken into account to calculate the mean ratio which is around 1,11. As the EHPC-005 and the
EHPC-006 are performed with SHORT and MEDIUM signal converter settings, the ratio values
obtained form that experiments cannot be fully trusted. The relevant experimental mean value of the
pressure ratio without these experiments is around 1,14 with a standard deviation of 0,03.

Final Project : Characterization of Diesel Sprays II-24
The curves obtained with the data from the measurements show the predictable evolution of
pressure as a function of time. Only for measurement EHPC-005, the curve is showing a depression
before the end of cooling caused by the SHORT signal converter setting (for Kistler #18) which
should no longer be used.


Figure 23 : Pressure measurement with Kistlers for a combustion event
On Figure 23, we can see that the two Kistler sensors show similar evolutions (fast pressure
rise and exponential decrease) and values, which let us use either the #18 or #19 to get the pressure
values, at least at the beginning and the end of the experiments.

Final Project : Characterization of Diesel Sprays II-25
3.2.1.2 Experiments with Kistler and Druck
A second set of experiments was performed with the Kistler #19 and the Druck sensor. The
goal of this set of experiments is to determine whether or not the two kinds of sensors could both
give correct values and evolution of pressure. Furthermore, the measurement frequency was tried to
be lowered to 5 kHz in order to get a larger time sample, and other measurements were made
separately a few time after the first sample (typically 5 s between measurements). Unfortunately,
measurements at 5 kHz were not possible to perform, and a measurement frequency of 10 kHz
(about 26 s of data acquisition) was kept. The results are shown in Table 2 :

reference
040506
pressure
transducer
measurement
settings
P_initial P_final
ratio
f i P P
P
EHPC-001 KISTLER #19 3,17 4,22 1,33 1,06
EHPC-002 KISTLER #19 3,17 3,62 1,14 0,46
EHPC-003 KISTLER #19 time delay 3,17* 2,83 0,89 -0,34
EHPC-004 KISTLER #19 3,14 3,60 1,15 0,46
EHPC-005 KISTLER #19 time delay 3,14* 2,92 0,93 -0,23
EHPC-006 KISTLER #19 time delay 3,14* 2,88 0,92 -0,27
EHPC-007 KISTLER #19 time delay 3,14* 2,88 0,92 -0,26
EHPC-008 KISTLER #19 3,16 3,52 1,11 0,36
EHPC-009 KISTLER #19 time delay 3,16* 2,85 0,90 -0,31
EHPC-010 KISTLER #19 time delay 3,16* 2,89 0,91 -0,28
EHPC-011 KISTLER #19 time delay 3,16* 2,88 0,91 -0,28
0,98 -0,07

EHPC-001 DRUCK 3,16 3,22 1,02 0,06
EHPC-002 DRUCK 3,17 3,05 0,96 -0,12
EHPC-003 DRUCK time delay 3,17* 2,92 0,92 -0,25
EHPC-004 DRUCK inlet valve open 3,14 3,00 0,96 -0,14
EHPC-005 DRUCK time delay 3,14* 2,84 0,90 -0,31
EHPC-006 DRUCK time delay 3,14* 2,91 0,93 -0,23
EHPC-007 DRUCK time delay 3,14* 2,87 0,91 -0,27
EHPC-008 DRUCK inlet valve open 3,16 3,02 0,96 -0,14
EHPC-009 DRUCK time delay 3,16* 2,92 0,92 -0,25
EHPC-010 DRUCK time delay 3,16* 2,92 0,92 -0,25
EHPC-011 DRUCK time delay 3,16* 2,89 0,92 -0,27
0,93 -0,22
(
*
)
The initial pressure values are the initial values before combustion and not at the beginning of the measure itself.
Table 2 : Pressure ratio for long time measurements with Druck and Kistler sensors
It seems obvious that the pressure is still decreasing at the end of the first time sample, even
with a measurement frequency which would provide a 26 seconds acquisition duration. For each
experiment, the pressure ratio at the end of the first sample was significantly higher than the ratio
after a few more seconds. However, waiting too long is not necessary, as we can notice from the
samples EHPC-005 to EHPC-007 or EHPC-009 to EHPC-0011, where the final pressure values
dont significantly change with a longer time. The amount of time to wait may probably depend on
the experimental conditions, which are here quite smooth (maximum pressure lower than 40 bar).

Final Project : Characterization of Diesel Sprays II-26
The average pressure ratio between final and initial pressure is about 0,98 for the Kistler and
0,93 for the Druck if the first experiment, which was performed to warm the set up, is not taken into
account. But if we only compute the average ratio from the values of the time delayed experiments
(up to 1 min after the first sample), there is almost one common ration of about 0,915 with a
standard deviation of 0,01.
3.2.2 Addition and removal of nitrogen
This set of measurements aimed at comparing the static and dynamic behaviour of the two
kinds of sensors (Kistler and Druck). As nitrogen is not reacting in the EHPC, we can check if the
sensors give similar results, which cannot be verified with combustion due to the valve protecting
the Druck. These experiments consisted in static and dynamic measurements. The following results
seem to show that the Kistlers and the Druck sensors have the same behaviour.
3.2.2.1 Static experiments
A medium pressure level of nitrogen (about 20 bar) was established in the EHPC, and the
inlet valve was open during the measurement, in order to detect an eventual deviation on Kistler or
Druck signals. The curves on Figure 24 show the typical results which can be obtained.


Figure 24 : Static pressure measurement with Kistler and Druck
We can notice a deviation in the Kistlers signal, which is decreasing. This behaviour could
be explained by the piezoelectric nature of this sensor, which tends to deviate. Anyway, on the time
sample of the measurements and for the application this sensor is dedicated to, a deviation of a
magnitude of a few percents is not significant.

Final Project : Characterization of Diesel Sprays II-27
3.2.2.2 Dynamic experiments
From a medium pressure level of nitrogen, the operator alternatively opens and closes the
outlet valve during the measurement acquisition in order to create a dynamic order. Hence we can
check if the two sensors signals show the same behaviour and the same values, which is almost the
case on the typical curves on Figure 25.


Figure 25 : Dynamic N
2
pressure measurement with Kistler and Druck
However, there is a slight difference between the two signals. According to the statistics
showed in Table 3, the difference between the two signals shows a standard deviation of around
0,06 bar for a range of pressures from 10 to 22 bar, which is not significant.

reference 052306 EHPC-004 EHPC-005
max 0,072 0,108
min -0,134 -0,074
mean -0,042 0,008
P between
Kistler and
Druck
[bar] standard deviation 0,060 0,050
Table 3 : Statistics about the difference between Kistlers and Drucks signals
Both sensors must also be able to give dynamic values of pressure, at least with a non
reacting mix, even though a combustion event is too critical (high gradients) to be measured with
the Druck. The pressure values given by the two sensors are very close, which leads us to think that
these values can be trusted, at least for non reacting phenomenon (inert gas) which introduce
smoother pressure gradients compared to combustion events.


Final Project : Characterization of Diesel Sprays II-28
3.2.3 Discussion
It seems that the pressure ratio f i P P of around 2 which was observed before is quite wrong.
It might come from a too short measurement time which leads the operator to measure a false final
pressure on the cooling curve. Actually, it would be surprising that the Kistlers pressure sensors are
affected by such a strong thermal shock that it could create such an offset, as they are specially
designed for internal combustion engine in thermodynamic conditions close to those in the EHPC.
A tolerance of a few percents is acceptable, but not 50 % as it was first observed.
Although this cooling is fast and mainly obtained in 5 seconds, it is not completed until a lot
more time, as we can see from his exponential evolution. If we look at the slope of the curve, we
notice that pressure seems to be still decreasing even after the 26 seconds of data acquisition. The
cooling may not be finished yet, which could explain the ratio value almost close to 1 but a little bit
over.
A theoretical approach of the final mix may help us to understand the experimental results.
The combustion equation that has to be considered in the EHPC-003 experiment (reference 030806)
is the combustion of acetylene C
2
H
2
:
2 2 2 2 2 2
5 5
C H + O 2CO +H 0+(-1) O
2 2

We can extract the needed values from the experiment data and the theoretical results of
Carlo Luijten concerning the parameters after combustion for one relevant experiment. Initially we
have n
0
= 0,16788 moles of gas mix ( = 1,10), which is calculated and sent to the Mass Flow
Control system. This value is said to be not so precise, so we will keep in mind that a tolerance
should be adopted on the MFC results, as well as on the theoretical predictions. With this number of
initial molecules, we should have n
f
= 0,16165 moles of gas after the combustion, which means that
there is a ration of around 0,96 between final and initial number of gas molecules (between 0,87 and
1,07 with a 10 % deviation on initial number of molecules).

Theoretically, we should find a final pressure lower than the initial one for the same
temperature. For the Kistler #18, we have an initial pressure of P
0
= 3,085 bar and a final pressure
of P
f
= 3,525 bar (ratio of 1,14) for an acquisition sample at 10 kHz. The initial temperature is T
0
=
295,7 K (22,7C). With the constant R = 8,31441 and the constant cell volume V = 1,2957 dm
3
,
with the perfect gas law, which is precise enough for the involved pressures, we should find :
0 0
0 0
n .R.T
P = P = 3,277 bar
V

This calculated value is 6 % higher than the measured pressure, which may come from the
poor reliability of both theory and MFC which does not provide the precise wanted number of
molecules.
For a final pressure of P
f
= 3,525 bar and the calculated final number of molecules after the
combustion, we should find a final temperature of :
f
f f
f.
P V
T = T = 330,4 K (57,4C)
n R

Final Project : Characterization of Diesel Sprays II-29
This relatively high temperature lead us to believe that at the time we make the final
pressure measurement, the cooling is not finished yet, and even if it is mainly done, we have to wait
again to measure the real pressure after cooling. At this point, we should have T
f
= T
0
, which means
that the corresponding pressure should be :
f
f 0 f
0
n
P = P P = 2,961 bar
n

As there are less molecules after combustion, it is obvious that the final pressure after
having reached the initial temperature is lower than in the initial mix. Since we assume the thermal
equilibrium after cooling, the pressure ratio is equal to the molecules ratio.

The experimental results gives a pressure ratio between 0,91 and 0,92 for a given set of
parameters which should in theory lead to a 0,96 ratio. If we take into account the accuracy of the
MFC, the theory assumptions and the measurements precision, we can conclude that both kinds of
sensors provide a good measure of static pressure in the cell, at least with a precision less than 5 %.
It seems obvious that there is no thermal shock affecting the sensors, at least not definitively. If the
operator wants a relevant value of the final pressure, however, he will have to wait for about 1 min
after the first sample so the cooling could be considered as finished.


Final Project : Characterization of Diesel Sprays II-30
3.3 Reliability of the cooling behaviour
The whole cooling curve is very important in the EHPC experiment, because the injection
settings are based on the pressure and temperature values from the sensors signals. Now that we
have checked that the pressure sensors came back to the initial pressure value after a combustion
event, we have to be sure that the pressure profile and values are reliable.
3.3.1 Thermal shock
When the inlet valve is open, the Druck gives a feedback from the combustion but it is not
fast enough to follow the huge gradient of pressure due to combustion. That is the reason why we
cannot trust the values during combustion given by the Druck, although the pressure values could
be correct after this event. The Figure 26 shows these results :


Figure 26 : Inability for the Druck sensor to cope with the combustions pressure gradient
On the main cooling phase just after the combustion, there is no noticeable behaviour that
could lead us to think that the pressure profile and values of Kistler are not correct. However, the
cooling signal may seem good although it is not.


Final Project : Characterization of Diesel Sprays II-31
However, when the main decrease of pressure due to cooling is done, we can sometimes also
notice a difference between the signals given by the Kistler and the Druck on Figure 27. Indeed, the
Kistlers signal is showing a slight but visible rise of pressure and finally the end of the cooling, as
we can notice on the zoom on Figure 28. With the Druck, no such behaviour can be detected.


Figure 27 : Difference between Drucks
and Kistlers signals

Figure 28 : Rise of pressure due to a
thermal shock on Kistler sensor
This cooling behaviour of the Kistler might be typical from a thermal shock that affects its
signal. Some mixtures are more likely to create such shocks, which are more pronounced when the
mixture is near stoichiometric and burns at a higher speed (hence higher gradients). As the Druck is
a piezoresistive sensor and is located behind a valve in a tube, the thermal shock does not affect its
signal.

Final Project : Characterization of Diesel Sprays II-32
3.3.2 Influence of turbulence
As the pressure signal is not always showing a thermal shock, experiments were performed
in order to understand the influence of the combustion phenomenon on the appearance of a thermal
shock. The mixing fan in the cell was switched on and then off for the same experimental settings.
Figure 29 is showing the typical results which can be obtained :


Figure 29 : Influence of mixing fan on the cooling behaviour
As the thermal shock may not be always very pronounced, it is useful to detect it on the first
order derivative of the signal, plotted on Figure 30, which should remain negative if there is no
pressure rise but only fall, and should go over zero if there were a rise.


Figure 30 : Comparison of the first order derivative signals

Final Project : Characterization of Diesel Sprays II-33
It seems that the thermal shock is no longer affecting the Kistlers pressure signal when the
mixing fan is off. Indeed, the shock is caused by pressure and temperature gradients, which are
higher with mixing because of the higher turbulence created by the fan. This difference in rise of
pressure and in amplitude of the pressure gradients are easy to see on the following figures : we can
notice that the rise of pressure is earlier (Figure 31) and shows a higher gradient with the fan
creating the turbulent conditions in the cell (Figure 32). These conditions are more likely to cause
thermal shocks.


Figure 31 : Influence of mixing fan
on the combustion start

Figure 32 : Influence of mixing fan
on the amplitude of pressure gradients

3.3.3 Discussion
It has been shown that there were some thermal shocks affecting the cooling behaviour
given by the Kistler sensor. However, although we only see the consequences of this shock (slight
rise of pressure) after the main cooling of the EHPC, it is actually affecting the whole curve,
because the cause of the thermal shock occurs at the beginning of the combustion event with the
high gradients.
In order to avoid too fast combustion, the amount of nitrogen can be increased in the pre-
combustion mixture, which smoothes the process. This could make the pressure and temperature
gradients lower and prevent the Kistler from thermal shocks. At a given amount of nitrogen, using
mixtures with a lambda value which provides lower burning velocities could prevent the Kistler
sensor from thermal shocks.


Final Project : Characterization of Diesel Sprays II-34
3.4 Conclusion
Through this analysis of data acquisition with the Kistlers and the Druck pressure
transducers, we have not found that the combustion affects the sensors by creating an offset which
leads to an unexpected ratio between initial and final pressure. This wrong hypothesis is probably
due to a too short time of acquisition which does not allow the operator to analyse the whole
cooling phase in the EHPC. It seems that the two kinds of sensors provide good pressure values for
static experiments, which is the case at the initial and final states. For dynamic behaviour, the
combustion process introduces a thermal shock, due to high pressure and temperature gradients,
which affects the cooling signal of Kistler sensors (the Druck cannot measure combustion events),
and it must still have an influence on pressure measurement at the Diesel spray injection.





Final Project : Characterization of Diesel Sprays II-35
Tables
Table 1 : Comparison between the two Kistlers sensors ................................................................... 23
Table 2 : Pressure ratio for long time measurements with Druck and Kistler sensors ...................... 25
Table 3 : Statistics about the difference between Kistlers and Drucks signals ............................... 27

Figures
Figure 1 : Comparison of T
b
and T
c
..................................................................................................... 8
Figure 2 : Evolution of temperature difference between "zeta" T
c
and T
b
.......................................... 9
Figure 3 : Evolution of boundary layer thickness ............................................................................. 9
Figure 4 : Evolution of density ....................................................................................................... 10
Figure 5 : Comparison of T
b
and T
c
for a whole cooling................................................................... 11
Figure 6 : Evolution of temperature difference between "zeta" T
c
and T
b
........................................ 11
Figure 7 : Comparison of T
b
and T
c
for calculated and H.S. diffusivity values ................................ 12
Figure 8 : Difference between "zeta" T
c
and T
b
for calculated and H.S. diffusivity values .............. 12
Figure 9 : Evolution of density ....................................................................................................... 13
Figure 10 : Evolution of boundary layer thickness ......................................................................... 13
Figure 11 : Comparison of T
b
and T
c
for time delayed conductive cooling ...................................... 14
Figure 12 : Reynolds number (experiment time: 2,6s) ...................................................................... 15
Figure 13 : Boundary layer thickness in laminar case (experiment time : 2,6s)................................ 16
Figure 14 : Boundary layer thickness in turbulent case (experiment time: 2,6s)............................... 16
Figure 15 : Cooling temperature (experiment time: 2,6s).................................................................. 17
Figure 16 : Difference between Tc and Tb (experiment time: 2,6s).................................................. 17
Figure 17 : Velocity (experiment time: 2,6s)..................................................................................... 18
Figure 18 : Mach number (experiment time: 2,6s) ............................................................................ 18
Figure 19 : Velocity (experiment: 26s @10kHz, fan off).................................................................. 19
Figure 20 : Temperature cooling (experiment: 26s @10kHz, fan off) .............................................. 19
Figure 21: Boundary layer in laminar case (experiment: 26s @10kHz, fan off)............................... 20
Figure 22: Boundary layer in turbulent case (experiment: 26s @10kHz, fan off) ............................ 20
Figure 23 : Pressure measurement with Kistlers for a combustion event .......................................... 24
Figure 24 : Static pressure measurement with Kistler and Druck...................................................... 26
Figure 25 : Dynamic N
2
pressure measurement with Kistler and Druck........................................... 27
Figure 26 : Inability for the Druck sensor to cope with the combustions pressure gradient ............ 30
Figure 27 : Difference between Drucks and Kistlers signals ......................................................... 31
Figure 28 : Rise of pressure due to a thermal shock on Kistler sensor ............................................. 31
Figure 29 : Influence of mixing fan on the cooling behaviour .......................................................... 32
Figure 30 : Comparison of the first order derivative signals ............................................................. 32
Figure 31 : Influence of mixing fan on the combustion start............................................................ 33
Figure 32 : Influence of mixing fan on the amplitude of pressure gradients .................................... 33


Final Project : Characterization of Diesel Sprays II-36
Appendix
Appendix A : MATLAB
TM
scripts for modelling
temperature_cooling.m

This is the main program which does the acquisition (call to functions derived from
plotdrukbomPF.m and druksjekkuh.m) of the experimental data (pressure signals, mixture
properties, various settings) and run the conduction and convection files to obtain the temperature
signals.

% file to load
path = 'location of the experimental data file' ;
filename = 'EHPC';
mn = 3 ;
MF = 10e3 ;

% constant values
R = 8.3144; % [J/(mol.K)] Universal gas constant
S = 6*(108e-3)^2; % [m2] Surface
Vbomb = (108e-3)^3; % [m3] Volume
L = 108; % [mm] Characteristic length for Nusselt and Reynolds
Tw = 300 ; % [K] Wall temperature

load pureparameters; % load NasaThermo
iO2=1; iCH4=2; iH2=3; iH2O=4; iCO2=6; iN2=7; iC2H4=8; iC2H2=9;

% collect the cooling pressure signal
A = pressure_signal_acquisition(path, filename, mn, MF);
cooling_pressure = A(:,1);
cooling_time = A(:,2);

% collect the settings
B = settings_acquisition(path, filename, mn, MF);
density = B(1); % [kg/m3]
molar_mass = B(2); % [g/mol]
total_mass = B(3)*1e-3; % [g]
composition = [0, 0, 0, B(5), 0, B(6), B(7), 0, 0, 0];

% calculating the temperature T_b from the pressure signal with perfect gas law
cooling_temperature = (cooling_pressure*1e5)*(molar_mass*1e-3)/(density*R);

Tb = cooling_temperature;
Pb = cooling_pressure;
t = cooling_time;

% reduce the numbers of elements
for i = 1 : 256
Tb_bis(i) = Tb(1000*i);
Pb_bis(i) = Pb(1000*i);
t_bis(i) = t(1000*i);
end

% calculating Tc from conduction and convection
Tc_conduction = conduction(Tb_bis,t_bis,Tw,composition,total_mass,molar_mass,density);
Tc_convection = convection(Tb_bis,Pb_bis,t_bis,Tw,composition,molar_mass,density)


Final Project : Characterization of Diesel Sprays II-37
conduction.m

This is a function which calculates T
c
with the assumption of conductive cooling (adiabatic
or "zeta"). The input arguments are the temperature experimental signal and the properties of the
mixture and the output is the core temperature signal. It can also be used to test conduction after a
certain time (as shown below).

function [Answer] = conduction(Tb,t,Tw,composition,m0,M0,rho)

% initialization
N = length(t);
Tcc = Tb;
m = 40; % start of conductive cooling
Tc(m) = 530; % [K]
Tc2(m) = 530; % [K]

for i = m+1 : N
Tcc(m+1) = 530;
for k = 1 : 2 % iterative process
% calculating the zeta function for conduction
c(i) = Tcc(i)/Tw-1;
sum = 0;
step = 0.01;
integral(i)=0;
for j=[1:501]
x = (j-1).*step;
integrand(j)=(1-erf(x))./(1+c(i).*erf(x));
integral(i) = integral(i)+integrand(j).*step;
end
zetaconduction(i) = (c(i)+1)./sqrt(pi).*integral(i);

% calculating Tc
Pb(i) = 1e-5*Tb(i)*rho_C(i)*R/(M0*1e-3); % [bar]
lambda_C(i) = conductivity(Pb(i),Tcc(i),composition,pr); % [W/m/K]
Cp(i) = Cprealmix(Pb(i),Tcc(i),composition,pr)/(M0*1e-3); % [J/kg/K]
Cv(i) = Cvrealmix(Pb(i),Tcc(i),composition,pr)/(M0*1e-3); % [J/kg/K]
gamma(i) = Cp(i)/Cv(i); % [-]
alpha_C(i) = lambda_C(i)/(rho_C(i)*Cp(i)); % [m2/s]
delta_T(i) = (2/sqrt(pi))*(1-Tw/Tcc(i))*sqrt(alpha_C(i)*(t(i)-t(m))); % [m]
rho_C(i) = m0*1e-3/(Vbomb+S*delta_T(i)); % [kg/m3]

Tc(i) = abs(Tb(i)*(1+(S*delta_T(i)/Vbomb)*zetaconduction(i))); % "zeta" Tc
Tc2(i) = Tc2(m)*(rho_C(i)/rho)^(gamma(i)-1); % adiabatic Tc
Tcc(i) = Tc(i);
end
end
[Answer] = [Tc, Tc2];


Final Project : Characterization of Diesel Sprays II-38
convection.m

function [Answer] = convection(Tb,Pb,t,Tw,composition,M0,rho)

N= length(Tb) ;
% Temperature first order derivative
for k = 2 : N-1
diff_Tb(k) = (Tb(k+1)-Tb(k-1))./(t(k+1)-t(k-1)) ;
end
diff_Tb(1) = diff_Tb(2) ; diff_Tb(N) = diff_Tb(N-1) ;

% Convection heat transfer coefficient % [W/m2.K]
for k = 1 : N
hc(k) =-(rho.*Vbomb.*Cvrealmix(Pb(k),Tb(k),composition,pr).*diff_Tb(k))/((M0.*1e-3).*S.*(Tb(k)-Tw));
end

% Nusselt number
for k=1:N
Nu(k) = (hc(k).*L.*1e-3)./conductivity(Pb(k),Tb(k),composition,pr) ;
end

% Prandtl number
Pr = 0.7 ; % Considered as a constant

% Reynolds number
for k = 1:N
Re(k) = (Nu(k)./(0.33.*Pr.^(1/3))).^2 ;
end

% Velocity
for k = 1 : N
v(k) = (viscosity(Pb(k),Tb(k),composition,pr).*Re(k))./(rho.*L.*1e-3) ;
end
v_average = sum(v)/N

% Boundary layer
for k = 1 : N
delta_lam(k) = 5.*L.*1e-3.*Re(k).^(-1/2) ; % Boundary layer laminar case [m]
delta_tur(k) = L.*1e-3.*Re(k).^(-1/5) ; % Boundary layer turbulent case [m]
end

% Comparison between Tc and Tb
for k = 1 : N
c(k) = (Tb(k)-Tw)./Tw ;
zetaconvection(k) = (c(k)+1)./(60.*c(k).^7).*(-60.*c(k).^7./(c(k)+1)+70.*c(k).^6-84.*c(k).^5+105.*c(k).^4-
140.*c(k).^3+210.*c(k).^2-420.*c(k)+420.*log(c(k)+1)) ;
temp_comp_lam(k) = 1 + (S.*delta_lam(k)./Vbomb).*zetaconvection(k) ;
temp_comp_tur(k) = 1 + (S.*delta_tur(k)./Vbomb).*zetaconvection(k) ;
Tc_lam(k) = Tb(k).*temp_comp_lam(k) ; % Cooling temperature in laminar case
Tc_tur(k) = Tb(k).*temp_comp_tur(k) ; % Cooling temperature in turbulent case
end
ecart_lam = Tc_lam - Tb ;
ecart_tur = Tc_tur - Tb ;

% Pressure cooling Pc
for k = 1 : N
Pc_tur(k) = (rho.*(R./(M0*1e-3)).*Tc_tur(k))./(1e5) ;
Pc_lam(k) = (rho.*(R./(M0*1e-3)).*Tc_lam(k))./(1e5) ;
end

% Then two loops (convergence on velocity) or more are performed with the Tc obtained instead of Tb.
[Answer] = [Tc_lam, Tc_tur];

Final Project : Characterization of Diesel Sprays II-39
Nomenclature
A surface (m)
Cv heat capacity at constant volume (J/kg.K)
Cp heat capacity at constant pressure (J/kg.K)
h heat transfer coefficient (W/m.K)
L characteristic length (m)
m
c
core mass (kg)
Nu Nusselt dimensionless number
Pr Prandtl dimensionless number
Re Reynolds dimensionless number
T
b
mass averaged temperature (K)
T
bl
boundary layer temperature (K)
T
c
core temperature (K)
T
w
wall temperature (K)
v core velocity (m/s)
V cell volume (m
3
)
V
bl
boundary layer volume (m
3
)
V
c
core volume (m
3
)
thermal diffusivity (m/s)
heat capacities ratio
boundary layer thickness (m)
dynamic viscosity (N.s/m2)
thermal conductivity (W/m.K)

c
core density (kg/m
3
)

References

[1] Siebers D.L., Ignition Delay Characteristics of Alternative Diesel Fuels : Implications on
Cetane Number, SAE Paper 852102, 1985.

[2] Tanaka S., Ayala F., Keck J.C., A reduced chemical kinetic model for HCCI combustion or
primary reference fuels in a rapid compression machine, Combustion and Flame 133, pp.
467-481, 2003.

[3] Tanaka S., Ayala F., Keck J.C., Heywood J.B., Two-stage ignition in HCCI combustion and
HCCI control by fuels and additives, Combustion and Flame 132, pp. 219-239, 2003.

[4] Keck J.C., Thermal boundary layer in a gas subject to a time dependent pressure,
Characters in heat and mass transfer 8, pp. 313-319, 1981.

[5] Xisheng Luo, Unsteady flows with phase transition, PhD thesis TU/e, ISBN90-3861915-4,
2004.

[6] Adrian Bejan, Heat Transfer, New York, Wiley, 1993.

[7] Durett R.P., Oren D.C., and Ferguson C.R., A Multidimensional Data Set for Diesel
Combustion Model Validation: I. Initial Conditions, Pressure History and Spray Shapes,
Society of Automotive Engineers, 1987.
















Part III

EHPC Experiments
Modelling of the Liquid Length
of Diesel Sprays

























Final Project : Characterization of Diesel Sprays III-2
Contents

Introduction .................................................................................................. 3
1. Siebers model for liquid length .............................................................. 4
1.1 Liquid length scaling law................................................................................................... 4
1.1.1 Experimental and literature background............................................................. 4
1.1.2 Spray model and assumptions ............................................................................ 4
1.1.3 Derivation of the liquid length scaling law......................................................... 5
1.2 Engineering correlation...................................................................................................... 7
1.3 Hiroyasu model of vaporizing spray.................................................................................. 8
1.3.1 Spray model ........................................................................................................ 8
1.3.2 Derivation of the model ...................................................................................... 8
1.4 Numerical solution........................................................................................................... 10
1.4.1 Physical properties............................................................................................ 10
1.4.2 Assumptions ..................................................................................................... 10
1.4.3 MATLAB
TM
Program....................................................................................... 11
2. Simulation results ................................................................................. 12
2.1 Results trends ................................................................................................................... 12
2.1.1 Siebers scaling law ........................................................................................... 12
2.1.2 Engineering correlation .................................................................................... 18
2.1.3 Hiroyasu model of vaporizing spray................................................................. 20
2.2 Experiments ..................................................................................................................... 25
2.2.1 Test matrix........................................................................................................ 25
2.2.2 Results analysis................................................................................................. 25
2.3 Comparison with experiments ......................................................................................... 26
2.4 Discussion........................................................................................................................ 29
3. Conclusion............................................................................................ 30
Tables ......................................................................................................... 31
Figures ........................................................................................................ 31
Appendix .................................................................................................... 32
Nomenclature ............................................................................................. 39
References .................................................................................................. 40









Final Project : Characterization of Diesel Sprays III-3
Introduction
Improvement of the in-cylinder combustion processes in Diesel engines is one of the main
goals in research to meet more and more constraining emissions regulations. The liquid phase of
fuel is one of the most important aspects to focus on. Indeed, the liquid fuel penetration promotes
the fuel-air mixing but must remain below the distance to the piston bowl wall in order to prevent
incomplete combustion. The influence of various parameters regarding fuel injection and in-
cylinder conditions has to be better understood. As a consequence, some models have been
developed to predict the geometry of the Diesel spray : maximum axial penetration of the liquid
phase (also defined as the liquid length in the following) and its angle.
In this report, two main models of the liquid phase fuel penetration are presented, showing
two different physical approaches of the Diesel spray penetration. The first one has been presented
by Siebers [1,2]. He developed a scaling law for the liquid length of fuel in a Diesel spray mainly
based on an idealized geometry and the following assumption : the rate of evaporation is
determined by the mixing rate of fuel and hot air (so-called mixing-limited evaporation). The
second one has been developed by Hiroyasu [3] and is mainly based on fuel droplets vaporization.
Both models will be explained in the following.
The goal of this study is to analyse and compute the Siebers model and compare it with the
Hiroyasu model that has already been computed by M. Boot and with the experiments performed by
P. Frijters and R. Peters in the Eindhoven High Pressure Cell (EHPC) using a Mie-scattering set up.
This comparative study will focus on the influence of various parameters on liquid length : fuel
injection parameters (pressure and temperature) and in-cylinder thermodynamic conditions
(temperature, pressure and density). These parameters will be tested on a wide range of values,
especially for real engine conditions.


Final Project : Characterization of Diesel Sprays III-4
1. Siebers model for liquid length
1.1 Liquid length scaling law
The liquid length scaling law has been fully explained by Siebers [1,2]. The following is
only a detailed summary of the main ideas and assumptions developed in this model. Furthermore,
all equations useful for the model understanding or computing will appear in this report.
1.1.1 Experimental and literature background
The liquid length is strongly linked to fuel vaporization. Espey and Dec [4] showed that
temperature and gas density have strong effects on liquid length, such as have fuel properties, as
presented by Browne et al. [5]. Furthermore, they found that the primary source of energy used to
vaporize the fuel is hot air entrained into the Diesel spray and not energy from combustion, since
the liquid length was established before ignition. Although the orifice diameter of the injector seems
to have an important influence on it [5], Kamimoto et al. [6] and Yeh et al. [7] have shown that the
liquid length is not affected by injection pressure. These trends have been confirmed by Siebers [1].
All these references indicate that vaporization in a Diesel spray is controlled by mixing
processes including entrainment of hot air (i.e., by turbulent mixing) and the overall transport and
mixing of fuel and air throughout the spray cross-section [1,2]. This conclusion is the basis of
Siebers model for liquid length, assuming that local interphase transport processes (such as droplet
vaporization) do not limit the rate of vaporization because these are fast relative to turbulent mixing
processes.
1.1.2 Spray model and assumptions
Siebers developed his model from an idealized spray model, with a simple geometry in cone
(Figure 1). Mass, momentum and energy balances are applied on a integral control surface. The
main assumptions about this idealized spray, justified in details in [2], are the following ones :
- quasisteady flow with a uniform growth rate (i.e., constant spray angle )
- uniform velocity, fuel concentration and temperatures profiles (i.e., perfect mixing
inside the spray boundaries) over a cross-section (fixed x)
- no velocity slip between the injected fuel and the entrained ambient gas


Figure 1 : Schematic of the idealized Siebers spray model

Final Project : Characterization of Diesel Sprays III-5
In Figure 1, the thermodynamic properties of the entrained ambient gas are P
a
, T
a
and
a
,
while fuel is injected at x = 0 in this gas at P
a
, with a temperature T
f
and a density
f
. At x = L
(downstream control surface), the fuel is supposed to be just completely vaporized and the vapor
phase is at a saturated condition in equilibrium with the ambient gas and the liquid phase of fuel.
1.1.3 Derivation of the liquid length scaling law
The liquid length scaling law is derived by solving mass, momentum and energy balances
for the axial location x = L (complete vaporization). This derivation is done in three main steps that
are summarized in the following.
1.1.3.1 Complete fuel vaporization
In this first step, the mixing ratio (in terms of mass flow rates of fuel and ambient gas) is
determined from the mass (Eq. 1 and 2) and energy (Eq. 3) balances. From the conservation of mass
for fuel (indexes f) and ambient gas (indexes a) in Eq. 1-2, the mass flow rate of fuel is constant and
equal to its value at x = L. The total entrained ambient gas mass flow rate is equal to the ambient
gas mass flow rate at x = L.
f f f f f f
m A U m (L) (L) A(L) U(L) = = = & & ( 1 )
a a
m (L) (L) A(L) U(L) = & ( 2 )
From the conservation of energy for the mixture fuel/ambient gas in Eq. 3, we have a
relation between initial specific enthalpies for liquid fuel (temperature T
f
and pressure P
a
) and
ambient gas (temperature T
a
and pressure P
a
) and specific enthalpies for saturated fuel vapor
(temperature T
s
and vapor pressure P
s
defined by T
s
) and ambient gas (temperature T
s
and pressure
P
a
-P
s
) after mixing and complete vaporization of the fuel. We can obtain Eq. 4 by rearranging Eq. 3
to find the mass flow rate ratio.
f f f a a a a a f f s a a s a s
m (L) h (T , P ) m (L) h (T , P ) m (L) h (T ) m (L) h (T , P P ) + = + & & & & ( 3 )
f a a a a s a s
a f s f f a
m (L) h (T , P ) h (T , P P )
m (L) h (T ) h (T , P )

=

&
&
( 4 )
The enthalpy difference in the numerator in Eq. 4 is the specific energy brought by the
entrained ambient gas to heat and vaporize the fuel. The enthalpy difference in the denominator is
the energy required to heat and vaporize the fuel. Using the real gas equation of state of Eq. 5 and
substituting it in Eq. 4, we obtain Eq. 6 where the mass flow rate ratio, now defined as B, is a
function of the fuel and ambient gas properties (temperatures T
f
and T
a
, pressure P
a
, molecular
weight M
f
and M
a
, compressibilities Z
f
and Z
a
) and the unknown T
s
.
P Z R T M = ( 5 )
a s a s s f a a a a s a s
f s s a s a f s f f a
Z (T , P P ) P M h (T , P ) h (T , P P )
B
Z (T , P ) (P P ) M h (T ) h (T , P )

= =

( 6 )
The method used to determine T
s
is developed in later stages and is based on an iteration
process until convergence to solve Eq. 6.

Final Project : Characterization of Diesel Sprays III-6
1.1.3.2 Axial variation of the fuel/ambient gas ratio
An additional assumption is needed to derive the axial variation of the mass flow rate ratio.
According to Kuo [8], the axial variation of this ratio in a vaporizing spray is similar to that in a
non-vaporizing isothermal spray, which means that the flow has locally homogeneous flow. Naber
and Siebers [9] showed that time and length scales developed for non-vaporizing and isothermal
sprays correlated also vaporizing spray penetration data.
[ ]
2
A(x) x tan( / 2) = ( 7 )
2
f f
A d
4

= ( 8 )
a
f
d C d = ( 9 )
Eq. 7-9 give the geometrical properties of the spray : the cross-section A(x), the effective
orifice diameter d
f
determined by the orifice coefficient of area-contraction C
a
, and the effective
fuel initial cross-section A
f
.
tan( / 2) a tan( / 2) = ( 10 )
f f
a
d
x
tan( / 2)
+

=

( 11 )
x x x
+
= % ( 12 )
A relationship between idealized spray angle and real spray angle is presented in Eq. 10.
The value of the constant a = 0,66 was determined by Naber and Siebers [9] to fit their experimental
results, and this value will be discussed in later stages. In Eq. 11 is x
+
defined as the spray
penetration length scale introduced in [9]. It is used in Eq. 12 to normalize the axial distance in the
spray x% .
With the additional non-vaporizing isothermal spray assumption, we can write Eq. 13-15
from the mass and momentum balances applied to the spray. With Eq. 7-15 we can rearrange them
and find Eq. 16 which provides the required relationship (in dimensionless form) for axial variation
of fuel and ambient gas mass flow rates in the spray.
f f f f f
m A U m (x) = = & & ( 13 )
a a
m (x) A(x) U(x) = & ( 14 )
f f f a
m U m (x) U(x) m (x) U(x) = + & & & ( 15 )
f
2
a
m (L) 2
m (L)
1 16x 1
=
+
&
&
%
( 16 )

Final Project : Characterization of Diesel Sprays III-7
1.1.3.3 Liquid length scaling law
The liquid length scaling law can now be derived by substituting B from Eq. 6 into Eq. 16
and solve for L
%
which is the value of x% corresponding to the maximal penetration of liquid phase
fuel (i.e., the dimensionless liquid length) to obtain Eq. 17. The dimensional form in Eq. 18 is
obtained by the combination of Eq. 17 and Eq. 11-12.
2
a a f
2
L b 1 1
B(T , P , T )
| |
= +
|
\
%
( 17 )
2
a
f
a a a f
C d
b 2
L 1 1
a tan( / 2) B(T , P , T )
| |

= +
|

\
( 18 )
The liquid length scaling law is formed by Eq. 17 or Eq. 18 coupled with Eq. 6. The value of
the constant b = 0,41 (instead of 0,25) is derived from a best fit of the liquid length scaling law with
the experiments of Naber and Siebers [1]. This value will be also discussed in later stages.
1.2 Engineering correlation
Siebers has also developed an engineering correlation [10] extended from his work on the
liquid length scaling law [2]. As detailed fuel composition and properties are typically not known,
this correlation focuses on a simple liquid length model based on properties that might be easy to
know. The equations are based on the analysis presented in the derivation of liquid length scaling
law. So the theoretical background will not be detailed once again.

A first ratio between fuel density
f
and ambient gas density
a
has to be calculated. Note
that the fuel density used in Eq. 19 is evaluated for atmospheric conditions, which is easier to
determine.
f
a
A

( 19 )
A second ratio B', also known as the specific energy ratio, is derived from the conservation
of energy in Eq. 20. All the properties such as the liquid heat capacity of fuel c
p,f,l
, the heat capacity
of ambient gas c
p,a,v
, the boiling point temperature T
b
and the latent heat of vaporization of the fuel
L
vap
are estimated at atmospheric conditions.
p,f ,l b f vap
p,a,v a b
c (T T ) L
B'
c (T T )
+
=

( 20 )
Then the correlation takes the form of a power law described in Eq. 21, depending on the
orifice diameter d and three parameters k, and . These parameters have been determined with
experimental data from Siebers [10] : k = 10,5 , = 0,58 and = 0,59. These values may be
changed in later stages in order to fit to the EHPC experimental results.
L d k A B'

= ( 21 )
This engineering correlation seems very simple to use, especially if we take into account the
need of physical properties which is much reduced compared to the Siebers scaling law.

Final Project : Characterization of Diesel Sprays III-8
1.3 Hiroyasu model of vaporizing spray
Hiroyasu has developed his model in [3]. In this overview are only presented the main basis
and equations of this model which could be interesting for the comparison with Siebers model. The
main idea of this model is to focus on the vaporization (or evaporation) of the fuel droplets created
by the jet.
1.3.1 Spray model
This model is time dependent, and there are three major events : injection time t
SOI
, time for
the jet to break-up t
bu
and time for the droplets to evaporate t
evap
. From this time scale, the geometry
of the spray can be designed in two parts : a liquid jet (before break-up) and five radial slices (index
k) containing fuel droplets until their vaporization. This geometry is illustrated by Figure 2.
Figure 2 : Schematic of the Hiroyasu spray model
The model focuses on a mean droplet which is assumed to be representative for the droplet
distribution, which might not be homogeneous.
1.3.2 Derivation of the model
We can separate the model into two main chronological parts : the first part before the
droplets formation and the second one before the total vaporization of these droplets, as represented
in Figure 3.








Figure 3 : Time scale of the spray in Hiroyasu model

1.3.2.1 Pre-breakup liquid spray penetration : t < t
SOI
+ t
bu,k

Before break-up, the length of the uniform liquid jet is controlled by the injection parameters :
injector discharge coefficient C
D
, the orifice pressure drop between injector nozzle and ambient gas
p
noz
, and by the fuel density
f
. Eq. 22 gives the uniform liquid length as a function of time until
the formation of droplets [11].
( ) [ ]
noz
D SOI
f ,0
2 p
L t C t t

( 22 )
The break-up time is different for each radial slide k. It also depends on the nozzle diameter
d of the injector. Eq. 23 gives a formula for the break-up time of radial slide k t
bu,k
.
time
t
SOI

t
bu,k

t
evap,k

x
y
Pre-Breakup
Post-Breakup
L(t)

k = 1..5

Final Project : Characterization of Diesel Sprays III-9
f ,0
bu,k
2
a
D noz
d
6 k
t 4,351
5 C p

=

( 23 )
1.3.2.2 Post-breakup vaporization : t
SOI
+ t
bu,k
t t
SOI
+ t
bu,k
+ t
evap

The spray angle can be determined at this time with an empirical formula by Hiroyasu [3].
As it is shown in Eq. 24, the spray angle depends on the properties of the ambient gas (pressure P
a
,
density
a
and dynamic viscosity
a
) and the injection pressure.
( ) ( )
( )
1 4
2
a SOI bu,k 5 noz SOI bu,k 5
2
a SOI bu,k 5
t t p t t d
0, 05
t t
= =
=

+ +
=

+


( 24 )
At t = t
SOI
+ t
bu,k
, droplets of liquid fuel are supposed to be created from the initial uniform
liquid jet. These droplets have a diameter (Sauter Mean Diameter) which depends on the properties
of the fuel (kinematic viscosity
f
, surface tension
f
, density
f
) and their radial position. Eq. 25
gives a correlation for this diameter [12].
( ) ( )
0.385 0.737 0.737 0.06 -0.54
k f,0 f,0 f,0 a SOI bu,k noz SOI bu,k
SMD 3,08 t t p t t = + + ( 25 )
For each radial slide k, we can calculate with Eq. 26 the penetration as a function of time [3].
Note that in this report we only focus on the axial liquid length, which is the maximum penetration
(k = 1).
( )
( )
( )
1 4
noz noz
k SOI bu,k D bu,k
a f ,0
p t
2 p
L t 2,95 d t t t C t
t
| |

= +
|


\
( 26 )
To calculate the maximum liquid penetration (i.e., the liquid length), we need to know the
time t
evap
. This time can be determined with the analysis of the heat transfers for a fuel droplet. This
analysis is complex and requires a lot a physical properties for the fuel in the droplet and at its
surface : vapor pressure, latent heat of vaporization L
f
, heat capacities for liquid phase c
p,f,l
and
vapor phase c
p,f,v
, heat conductivity k
f
For the ambient gas, we also need the heat capacity c
p,a,v

and the heat conductivity k
a
.
With these properties, we can calculate the heat and mass transfer numbers B
T
and B
M
, and
then compute the evolution of the droplet diameter D, its mass m
d
and the surface temperature T
f,s

(Eq. 27 29). This part of the analysis which has been computed by M. Boot is based on the droplet
evaporation model by Meyer and Heywood [13].
( ) ( ) ( )
( )
a M
SOI
f p,g
4k t ln 1 B t
dD
dt (t ) c t D
+
=

( 27 )
( )
( )
( )
( ) ( )
g
d
M
p,g
k t
dm
2 D t ln 1 B t
dt c t
= + ( 28 )
( ) ( )
( )
vap,f ,s f ,s T d
p,f ,0 d M
L t dT B t
dm
1
dt dt c m B t
| |
=
|
\
( 29 )
When the droplet is totally vaporized, its diameter D and its mass m
d
are equal to zero. At this
point, there is no liquid fuel left, and the time is t = t
SOI
+ t
bu,k
+ t
evap
. Then the liquid length can be
calculated.

Final Project : Characterization of Diesel Sprays III-10
1.4 Numerical solution
1.4.1 Physical properties
In order to compute the Siebers model, a lot of physical properties at various temperatures
are needed for both fuel and ambient gas resulting from the pre-combustion. As these properties are
not analytically available, we needed to use some correlations. For common required properties
with Hiroyasu model, the same correlations have been used to make the comparison honest between
the two models. Mostly, correlations are limited to a relatively short range of temperature and for
specific fuels, but it is the only one solution which was available.
Ambient gas enthalpies were calculated with programs written by C. Luijten and B. Somers
based on NASA polynomials, while a correlation [14] was used for gas viscosity. For fuel,
correlations were found : for density and liquid and vapor fuel heat capacities in the Chalmers
University database [15], for vapor fuel enthalpy the Heywood correlation [16], for latent heat of
vaporization and vapor pressure in the Hiroyasu modelling computed by M. Boot, and for
compressibility a correlation derived from Reid [17]. All these correlations for ambient gas and fuel
properties for both Siebers and Hiroyasu models are presented in Appendix A.
1.4.2 Assumptions
As the modelling of the spray angle is complex and not yet well understood, Siebers has
decided to use some experimental measured spray angles. In this study we will use the "real" spray
angle :
( )
1/4
2
a inj a
2
a
P P d
0,05

=

( 30 )
The formula in Eq. 30 has been developed by Hiroyasu [3]. This decision was made to allow
a better and easier comparison focused on liquid length with the Hiroyasu model (spray angle is not
that important in this study) and also because we would not have enough experimental values of
spray angles from the EHPC in the time of the project. However, this choice of a calculated could
lead us to think that this angle is actually the idealized spray angle . The comparison with
experiments should determine whether or not it is the case, and if we should take a = 1 for the
computing.

As the composition of the Diesel fuel used for the experiments in the EHPC is almost
unknown, a fuel composed of 70 % n-decane C
10
H
22
and 30 % 1-methylnaphtalene C
11
H
10
was
chosen to make the comparison. Anyway, Diesel fuels are composed of many species and it is hard
to represent it. As the Siebers model only fits with single component fuels, this multi components
fuel will be considered as a single component fuel with linear approximated properties. This
approximation is detailed in Appendix A. However, the molar weight and the atmospheric boiling
point temperature are known and these values will be used (T
b
= 611 K and M
f
= 179,3 g/mol).

The modelling of the enthalpy needed to heat and vaporize the fuel can be done in two
different ways, because enthalpy is a function of state and thus does only depend on the initial and
final states. The following ways should give similar results :
- heating from T
f
to T
s
and vaporizing at T
s

- vaporizing at T
f
and heating from T
f
to T
s


Final Project : Characterization of Diesel Sprays III-11
For the second way, the enthalpy required for heating could be calculated through two
methods : directly with Heywood correlation for fuel enthalpy or by integrating the correlation for
heat capacity between T
f
and T
s
. As correlations for the heat of vaporization don't seem very
accurate, it would be better to use it just once, which is the case with the second way. Furthermore,
the method used to calculate the heating of fuel from the heat capacity is not precise at all, and it
would be better to use Heywood correlations. The results of comparison between the three
possibilities for enthalpy difference are in Appendix A.

For the computing, some injector properties used for the experiments are needed : the nozzle
diameter d which is 130 m or 180 m and the constant value C
a
(orifice coefficient of area-
contraction) equal to 0,81.
1.4.3 MATLAB
TM
Program
The Siebers model for liquid length has been computed in MATLAB
TM
. A program has
been written, using some previously computed physical properties files (mostly by C. Luijten and
B. Somers) and new ones resulting from correlations research.

For a given fuel, whose required properties are calculated with the correlations, the input
parameters needed to run the program are :
- injection conditions (injection pressure P
inj
and fuel initial temperature T
f
)
- composition of the mixture used for pre-combustion
- temperature of ambient gas T
a
when fuel injection occurs

If there is no divergence in the execution (the calculation of T
s
detailed below is the critical
point), the program returns :
- liquid length L
- "real" spray angle

In the following, the main steps of the program execution are summarized. It can be divided
into three main parts :
- calculation of the initial fuel and ambient gas properties at the injection (density,
molar mass, enthalpy, gas pressure, fuel heat of vaporization)
- iterative method to solve Eq. 6 : T
s
is initialised and incremented in a loop (steps of
1C) until the convergence of the two sides of the equation (2% tolerance). Fuel
properties (vapor pressure, enthalpy, compressibility) and ambient gas properties
(enthalpy and partial pressure) are calculated for each iteration of T
s
.
- calculation of the liquid length from T
s
value, and spray angle

The program computed for the engineering correlation of Siebers is very simple. It
calculates the initial atmospheric values of fuel and ambient gas properties (density, molar mass,
heat capacity, heat of vaporization). Then the coefficients A and B' are calculated and from these
values the liquid length can be determined.

The program for the Hiroyasu model had already been written by M. Boot. After a few
modifications, it is using the same input parameters and returns the same answers.

The MATLAB
TM
scripts written for the computation of the Siebers model and the
engineering correlation can be found in Appendix B.



Final Project : Characterization of Diesel Sprays III-12
2. Simulation results
2.1 Results trends
Before comparing the model results with experiments, we have to check that the main trends
observed by Siebers in his experiments [1] and his scaling law [2] can also be visible on our
simulations, both for Siebers and Hiroyasu models. Only the trends are valuable in the following.
2.1.1 Siebers scaling law
Below are presented the first results obtained with the Siebers model. For these results,
experimental parameters have been chosen in order first to get a good overview of the results trends
for the influence of different parameters and second to fit with the experimental results for the later
comparison. However, the experimental conditions present a wide range of possibilities, and
especially the ambient gas composition for pre-combustion. As the initial mixture composition is
the basis for the calculations of ambient gas enthalpy and other properties, it could be sensitive to
this parameter, and the values on the following figures might not be correct for all parameters.
2.1.1.1 Influence of ambient gas temperature and density
500 600 700 800 900 1000 1100 1200 1300 1400 1500
0
50
100
150
200
250
300
Ambient Gas Temperature T
a
[K]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Ambient Gas Temperature and Density on Liquid Length
rho
a
= 4,7 kg/m3
rho
a
= 9,0 kg/m3
rho
a
= 13,3 kg/m3
rho
a
= 17,6 kg/m3
rho
a
= 21,8 kg/m3
rho
a
= 26,1 kg/m3

Figure 4 : Siebers - Influence of Ambient gas temperature for different densities

Final Project : Characterization of Diesel Sprays III-13
0 5 10 15 20 25 30 35 40
0
50
100
150
200
250
Ambient Gas Density rho
a
[kg/m3]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Ambient Gas Temperature and Density on Liquid Length
T = 500 K
T = 650 K
T = 800 K
T = 950 K
T = 1100 K
T = 1250 K

Figure 5 : Siebers - Influence of Ambient gas density for different temperatures
As expected from Siebers results [1], ambient gas density and temperature both have a
strong influence on the liquid length. We have a decreasing exponential behaviour when ambient
gas temperature is increasing at a given ambient gas density, and vice versa (Figure 4 and Figure 5).
For higher densities, the influence is not as pronounced as for lower densities. Density seems to
have a more non-linear influence than temperature. Note that for low temperatures and densities, the
model may be instable and anyway give unrealistic values.

Final Project : Characterization of Diesel Sprays III-14
2.1.1.2 Influence of fuel temperature
240 260 280 300 320 340 360 380 400 420 440
0
20
40
60
80
100
120
Fuel Temperature [K]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Fuel Temperature on Liquid Length
650 K - 9,1 kg/m3
800 K - 13,8 kg/m3
950 K - 18,6 kg/m3
1100 K - 23,3 kg/m3
1250 K - 28,1 kg/m3
1400 K - 32,9 kg/m3

Figure 6 : Siebers - Influence of Fuel temperature for different ambient gas conditions
Liquid length decreases linearly with increasing fuel temperature as expected, and this trend
is more visible for low temperatures and densities. Indeed, for a higher fuel temperature, the energy
required to vaporize the fuel is lower and easier reached, resulting in a lower liquid length [1]. The
influence of fuel temperature may be neglected for "normal" engine like ambient gas conditions.
Note that the fuel temperature which should be used in the model is the temperature reached by the
fuel at its entrance in the EHPC. Unfortunately, we can only measure the temperature in the fuel
tank which could be a bit lower.

Final Project : Characterization of Diesel Sprays III-15
2.1.1.3 Influence of orifice diameter
0 50 100 150 200 250 300
0
50
100
150
Orifice Diameter [m]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Orifice Diameter on Liquid Length
650 K - 9,1 kg/m3
800 K - 13,8 kg/m3
950 K - 18,6 kg/m3
1100 K - 23,3 kg/m3
1250 K - 28,1 kg/m3
1400 K - 32,9 kg/m3

Figure 7 : Siebers - Influence of Orifice diameter for different ambient gas conditions
The dependence of liquid length on orifice diameter should be linear according to Eq. 18,
which is not really the case in Figure 7. However, this can be explained by the choice of the formula
in Eq. 19 for , which is dependant on the orifice diameter d. Anyway, the dependence of liquid
length on orifice diameter can be considered as almost linear, at least for relatively high temperature
and density. Furthermore, the length tends to zero when the diameter also tends to zero. It illustrates
the fact that liquid length can be controlled through orifice size selection [1]. Note that the liquid
length is increasing faster as both temperature and density decrease.


Final Project : Characterization of Diesel Sprays III-16
2.1.1.4 Influence of orifice pressure drop
200 400 600 800 1000 1200 1400 1600
0
20
40
60
80
100
120
Orifice Pressure Drop [bar]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Orifice Pressure Drop on Liquid Length
650 K - 9,0 kg/m3
800 K - 13,3 kg/m3
950 K - 17,6 kg/m3
1100 K - 21,8 kg/m3
1250 K - 26,1 kg/m3
1400 K - 30,4 kg/m3

Figure 8 : Siebers - Influence of Orifice pressure drop for different ambient gas conditions
For the more realistic range of injection pressure values (not too low orifice pressure drop),
the simulation confirms that the pressure drop has no real influence on liquid length. This is less
visible for lower temperatures and densities, where we can notice a more pronounced decrease with
an increasing pressure drop. This behaviour may also be explained by the choice of a spray angle
from Eq. 19, because this angle creates a decreasing dependence for the liquid length on the orifice
pressure drop.

Final Project : Characterization of Diesel Sprays III-17
2.1.1.5 Influence of the ambient gas composition
The ambient gas composition defines its properties, and especially its enthalpy. As in the
Siebers model this ambient gas enthalpy is used in the computation, a change in the initial mixture
at constant lambda and density could lead to different final ambient gas enthalpies.

700 800 900 1000 1100 1200 1300
20
25
30
35
40
45
50
55
60
Ambient Gas Temperature T
a
[K]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Mixture Composition on Liquid Length
75% N2
69% N2
63% N2

Figure 9 : Siebers - Influence of Mixture composition
Different amounts of nitrogen have been tested (from 63 % to 75 %) with a constant density
= 9,85 kg/m3 and a constant lambda = 0,81 to obtain Figure 9. It seems that the initial mixture
composition has a significant effect on results for Siebers model : a difference of about 10 % is
observed. It seems to be necessary to know the initial mixture composition as input parameter for
the Siebers computation.

We can see that our simulation results agree with the Siebers experimental and scaling law
results for the main trends. However, our simulations provide very high results which do not seem
realistic but this could be related to our choice in mixture composition for the simulations to
observe some relevant trends. It will be discussed for the comparison with the EHPC experimental
results.

Final Project : Characterization of Diesel Sprays III-18
2.1.2 Engineering correlation
The Siebers engineering correlation has also been tested on the same ranges of parameters as
the scaling law. For ambient gas temperatures T
a
below the fuel boiling point temperature T
b
, the
model is not defined, which reduces the range of temperatures.
Since this engineering correlation is directly derived from the theoretical background used
for the liquid length scaling law, it would not be surprising that the results are very close for both
models, at least concerning the trends in the influences of parameters.

As we can see from Figure 10, the ambient gas temperature and it density have a similar
influence on liquid length : an exponential decrease.

400 600 800 1000 1200 1400 1600
0
10
20
30
40
50
60
70
80
Ambient Gas Temperature T
a
[K]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Ambient Gas Temperature and Density on Liquid Length
rho
a
= 4,7 kg/m3
rho
a
= 9,0 kg/m3
rho
a
= 13,3 kg/m3
rho
a
= 17,6 kg/m3
rho
a
= 21,8 kg/m3
rho
a
= 26,1 kg/m3

Figure 10 : Engineering correlation Influence of Ambient gas temperature


Final Project : Characterization of Diesel Sprays III-19
According to the linear dependence in d in the model, we see that the liquid length is
proportional to the orifice diameter as shown in Figure 11. We can check in Figure 12 that for high
enough ambient gas temperatures, the fuel temperature has no real influence on the liquid length.
These results correspond to the main trends observed with the liquid length scaling law.

0 100 200 300
0
20
40
60
80
100
120
Orifice Diameter [m]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Orifice Diameter on Liquid Length
650 K - 9,1 kg/m3
800 K - 13,8 kg/m3
950 K - 18,6 kg/m3
1100 K - 23,3 kg/m3
1250 K - 28,1 kg/m3
1400 K - 32,9 kg/m3

Figure 11 : Engineering correlation
Influence of Orifice diameter
250 300 350 400 450
0
20
40
60
80
100
120
Fuel Temperature [K]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Fuel Temperature on Liquid Length
800 K - 13,8 kg/m3
950 K - 18,6 kg/m3
1100 K - 23,3 kg/m3
1250 K - 28,1 kg/m3
1400 K - 32,9 kg/m3

Figure 12 : Engineering correlation
Influence of Fuel temperature
Generally, this engineering correlation seems to give much lower results than the Siebers
scaling law, which could be closer to realistic values. However, this could be due to the fact that the
Siebers scaling law is very sensitive to the composition of pre-combustion mixture.



Final Project : Characterization of Diesel Sprays III-20
2.1.3 Hiroyasu model of vaporizing spray
The same ranges of parameters have been used to observe the main trends in the results of
the Hiroyasu model. Only for matters of scale the ambient gas temperature range has been shifted,
because of deviations in the model for low temperatures.
2.1.3.1 Influence of ambient gas temperature and density
500 600 700 800 900 1000 1100 1200 1300 1400 1500
20
40
60
80
100
120
140
Ambient Gas Temperature T
a
[K]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Ambient Gas Temperature and Density on Liquid Length
rho
a
= 4,7 kg/m3
rho
a
= 9,0 kg/m3
rho
a
= 13,3 kg/m3
rho
a
= 17,6 kg/m3
rho
a
= 21,8 kg/m3
rho
a
= 26,1 kg/m3

Figure 13 : Hiroyasu - Influence of Ambient gas temperature for different densities
The liquid length is decreasing with an increasing ambient gas temperature following an
exponential behaviour like for Siebers model. However, ambient gas temperature seems to have
more influence than the ambient gas density since all the curves for different densities converge to a
common liquid length value for lower temperatures. It might also indicate that the model is not
valid for this very low ambient gas temperatures because below a certain temperature, curves cross
and we find higher lengths for higher densities, which is not physically acceptable.


Final Project : Characterization of Diesel Sprays III-21
0 5 10 15 20 25 30 35 40
30
40
50
60
70
80
90
100
110
Ambient Gas Density rho
a
[kg/m3]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Ambient Gas Temperature and Density on Liquid Length
T = 550 K
T = 700 K
T = 850 K
T = 1000 K
T = 1150 K
T = 1300 K

Figure 14 : Hiroyasu - Influence of Ambient gas density for different temperatures
As expected, the liquid length decreases with an increasing ambient gas density, even if the
influence of this parameter seems to be lower than the influence of ambient gas temperature.
However, for low temperatures, the model provides non-physical results : at a given ambient
temperature, the liquid length first decreases with an increasing density and then increases, which is
not realistic. Furthermore, for decreasing temperatures, the influence is higher and the differences in
liquid length are increasing.

Final Project : Characterization of Diesel Sprays III-22
2.1.3.2 Influence of fuel temperature
240 260 280 300 320 340 360 380 400 420 440
0
20
40
60
80
100
120
140
Fuel Temperature [K]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Fuel Temperature on Liquid Length
650 K - 9,1 kg/m3
800 K - 13,8 kg/m3
950 K - 18,6 kg/m3
1100 K - 23,3 kg/m3
1250 K - 28,1 kg/m3
1400 K - 32,9 kg/m3

Figure 15 : Hiroyasu - Influence of Fuel temperature for different ambient gas conditions
We can see that the liquid length is decreasing with an increasing fuel temperature as
expected. The results show that fuel temperature is a really influencing parameter in Hiroyasu
model. This influence is increasing while the ambient gas conditions (temperature and density) are
decreasing. This importance of fuel temperature is critical because the real fuel temperature is not
really known and that could have a real influence on the results for this model. If the injected fuel
temperature is under-estimated, the model could provide over-estimated values of liquid length.

Final Project : Characterization of Diesel Sprays III-23
2.1.3.3 Influence of orifice diameter
0 50 100 150 200 250 300
0
20
40
60
80
100
120
140
Orifice Diameter [m]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Orifice Diameter on Liquid Length
650 K - 9,1 kg/m3
800 K - 13,8 kg/m3
950 K - 18,6 kg/m3
1100 K - 23,3 kg/m3
1250 K - 28,1 kg/m3
1400 K - 32,9 kg/m3

Figure 16 : Hiroyasu - Influence of Orifice diameter for different ambient gas conditions
The behaviour of liquid length regarding orifice diameter is very similar to the behaviour
noticed with the Siebers model : almost linear for higher ambient gas conditions and with a more
pronounced slope for lower conditions. The liquid length is also tending towards zero when the
orifice is close to zero, which is unrealistic considering the filtered fuel properties but may be a
good indicator of an eventual deviation.

Final Project : Characterization of Diesel Sprays III-24
2.1.3.4 Influence of orifice pressure drop
0 500 1000 1500 2000 2500 3000
20
30
40
50
60
70
80
90
Orifice Pressure Drop [bar]
L
i
q
u
i
d

L
e
n
g
t
h

[
m
m
]
Influence of Orifice Pressure Drop on Liquid Length
650 K - 9,1 kg/m3
800 K - 13,8 kg/m3
950 K - 18,6 kg/m3
1100 K - 23,3 kg/m3
1250 K - 28,1 kg/m3
1400 K - 32,9 kg/m3

Figure 17 : Hiroyasu - Influence of Orifice pressure drop for different ambient gas conditions
On this plot we can notice that the injection pressure has an influence on the liquid length in
Hiroyasu model : it decreases with an increasing orifice pressure drop. However, this influence is
not so pronounced for higher ambient gas conditions.


The main trends observed for Siebers model and experiments are visible for the simulations
performed with the Hiroyasu model. However, some influences are more pronounced, especially
the fuel temperature and the injection pressure. But for low ambient gas conditions, the model gives
some unrealistic trends, which lead us to think that it is not valid over a certain range of conditions.
Furthermore, the liquid length values seem very high and unrealistic.


Final Project : Characterization of Diesel Sprays III-25
2.2 Experiments
2.2.1 Test matrix
A test matrix of interesting parameters has been determined and is summarized in Table 1.
The variable parameters are : composition of the fuel, injection pressure P
inj
, ambient gas density
a

and ambient gas temperature T
a
. The ambient gas density is related to the pre-combustion mixture,
whose composition is also known.

PCCI "Normal" Engine Post-injection
Temperature T
a
[K]
475 - 800 800 - 1200 1200 - 1500
Density
a
[kg/m
3
] 6 - 42
Injection pressure P
inj
[bar] 750 - 1500
Table 1 : Test matrix for the EHPC experiments
The temperature range is composed of three parts : the low temperatures simulate pre-
injection/HCCI like conditions, the middle temperatures are representative of a "normal" engine and
the high temperatures simulate post-injection like conditions.
2.2.2 Results analysis
The liquid length and the spray angle have been determined from the movies caught by the
HS camera during the EHPC experiments with Mie-scattering and Schlieren. These experiments
and their analysis have been performed on a wide range of temperatures and densities by P. Frijters
R. Peters and F. David. A typical result of Mie-scattering experiment is shown on Figure 18. It has
been artificially coloured for a better view.


Figure 18 : Typical result of Mie-scattering technique
The liquid length, which is the main subject of this study, has been determined from the
difference of intensity in a movie between the liquid core and the gas in the background. This
analysis is difficult because of the instability of the spray and the threshold of intensity chosen to
separate liquid from gas. All this analysing work has been computed in MATLAB
TM
files by R.
Peters. The results obtained with this technique may be under-estimated because the very thin
droplets of liquid fuel do not probably reflect enough the laser to provide a sufficient intensity on
the movies.

Final Project : Characterization of Diesel Sprays III-26
2.3 Comparison with experiments
This comparison will be performed for only one fuel, since the models are only computed
for one ideal Diesel fuel. Furthermore, due to experimental issues and time limits, it has been
impossible to perform enough experiments to test the models on the whole test matrix. The
following experiments used for the validation and comparison of the models have been performed
for a density = 9,5 kg/m
3
, an injection pressure P
inj
= 800 bar and a fuel temperature T
f
= 300 K.

On Figure 19, we can see the models compared with the experimental results. At first sight,
we can notice that the Hiroyasu model is providing too much high values and that the Siebers model
and the Engineering correlation are quite close to the experimental results, although for lower
temperatures it seems to deviate a bit.
700 800 900 1000 1100 1200 1300
20
30
40
50
60
70
80
90
100
Ambient Gas Temperature T
a
[K]
L
i
q
u
i
d

L
e
n
t
g
h

[
m
m
]
Comparison of Siebers & Hiroyasu models with EHPC Experiments
Siebers
Engineering Correlation
Hiroyasu
Experiments

Figure 19 : Comparison between models and experimental results


Final Project : Characterization of Diesel Sprays III-27
In Siebers model, there are two empirical coefficients a and b than could be tuned to fit to
the experimental results just as Siebers does himself. The ratio b/a can be reduced to 95 % of its
initial value : this value has been found to be the most efficient for a better fitting to the
experimental results, as illustrated by Figure 20. For the Engineering correlation the empirical
parameters k, and can be tuned. A simple change with = 0,49 seems to improve the results as
we can see in Figure 21.
700 800 900 1000 1100 1200 1300
20
25
30
35
40
45
50
55
60
Ambient Gas Temperature T
a
[K]
L
i
q
u
i
d

L
e
n
t
g
h

[
m
m
]
Tuning of Siebers model
Siebers
Siebers x 0,95
Experiments

Figure 20 : Tuning of Siebers model
700 800 900 1000 1100 1200 1300
20
30
40
50
60
70
80
Ambient Gas Temperature T
a
[K]
L
i
q
u
i
d

L
e
n
t
g
h

[
m
m
]
Tuning of the Engineering Correlation
Siebers
Siebers x 0,95
Experiments

Figure 21 : Tuning of the Engineering Correlation

Final Project : Characterization of Diesel Sprays III-28
Concerning the Hiroyasu model, some tunings of the heat transfers inside the derivation
have been tested without significant results. The reduction to 45 % of its initial value has been
found to be the best tuning (Figure 22), although the shape of the curve is not the same than the
other models and the experimental results.
700 800 900 1000 1100 1200 1300
20
30
40
50
60
70
80
90
100
Ambient Gas Temperature T
a
[K]
L
i
q
u
i
d

L
e
n
t
g
h

[
m
m
]
Tuning of Hiroyasu model
Hiroyasu
Hiroyasu x 0,45
Experiments

Figure 22 : Tuning of the Hiroyasu model
The results of the comparison between tuned models, with the above mentioned tuned
parameters, and the experimental results are illustrated in Figure 23.
700 800 900 1000 1100 1200 1300
20
25
30
35
40
45
50
55
60
Ambient Gas Temperature T
a
[K]
L
i
q
u
i
d

L
e
n
t
g
h

[
m
m
]
Comparison of Siebers & Hiroyasu models with EHPC Experiments
Siebers x0,95
Engineering Correlation beta = 0.49
Hiroyasu x0,45
Experiments

Figure 23 : Comparison between tuned models and experimental results

Final Project : Characterization of Diesel Sprays III-29
2.4 Discussion
The Siebers model, even with the initial empirical coefficient values, seems to provide liquid
length values close to the experimental ones. This is not surprising since the experimental set up
used by Siebers for his investigations and to develop his model is very similar to the EHPC, with
almost the same settings. The slight tuning added to improve the results is not significant towards
the relevance of the model.
The Engineering correlation, since it is directly derived from the Siebers model, gives very
similar results. However, the results for lower temperatures may be overestimated, which is
improved with a slight tuning. Furthermore, this tuning seems to be natural for such a simple
engineering correlation and it does not denature the model.
The tuning of Hiroyasu model is very strong, and does not even provide very good results.
The overestimation of liquid length is directly related to the model assumptions of droplet
evaporation, which is probably not the limiting process according to the results. Although the
Hiroyasu model is more sensitive to the fuel temperature, the error between the measured fuel
temperature and the real fuel temperature on the nozzle might not improve radically the results.

Final Project : Characterization of Diesel Sprays III-30
Conclusion
We can conclude that the Siebers model matches very well the few experiments that have
been performed to validate it. However, this model requires a lot of physical properties correlations,
which could limit its range of use, since it may be difficult to find so many correlations for different
fuels or wide ranges of temperatures and pressures. It can be used to predict the fuel liquid length in
the EHPC.
The Engineering correlation gives also very good results compared to these experiments
with a slight tuning. This tuning is easy to perform and the model is probably requiring it, since it is
very simple. Another advantage is the few physical properties correlations required : only a few
properties at atmospheric pressure are needed, which makes its use much more simple than the
Siebers model. It also can be used to predict the liquid length in the EHPC, and for the comparison
of different fuels it seems to be easier to use than the Siebers model.
The Hiroyasu model seems to be useless in this case, as it provides very high values for
liquid length and requires a strong tuning to give results closer to experiments, although the
influence of ambient gas temperature is not well simulated. It is recommended not to use it if
realistic values are to be found.

All these conclusions are valid only for the few experimental conditions that have been
tested by now. With higher densities the results may change, although the Siebers model and the
Engineering correlation are more likely to provide a good estimate of liquid length than the
Hiroyasu model.


Final Project : Characterization of Diesel Sprays III-31
Tables
Table 1 : Test matrix for the EHPC experiments............................................................................... 25

Figures
Figure 1 : Schematic of the idealized Siebers spray model ................................................................. 4
Figure 2 : Schematic of the Hiroyasu spray model .............................................................................. 8
Figure 3 : Time scale of the spray in Hiroyasu model ......................................................................... 8
Figure 4 : Siebers - Influence of Ambient gas temperature for different densities............................ 12
Figure 5 : Siebers - Influence of Ambient gas density for different temperatures............................. 13
Figure 6 : Siebers - Influence of Fuel temperature for different ambient gas conditions .................. 14
Figure 7 : Siebers - Influence of Orifice diameter for different ambient gas conditions................... 15
Figure 8 : Siebers - Influence of Orifice pressure drop for different ambient gas conditions ........... 16
Figure 9 : Siebers - Influence of Mixture composition...................................................................... 17
Figure 10 : Engineering correlation Influence of Ambient gas temperature .................................. 18
Figure 11 : Engineering correlation Influence of Orifice diameter ............................................... 19
Figure 12 : Engineering correlation Influence of Fuel temperature............................................... 19
Figure 13 : Hiroyasu - Influence of Ambient gas temperature for different densities....................... 20
Figure 14 : Hiroyasu - Influence of Ambient gas density for different temperatures........................ 21
Figure 15 : Hiroyasu - Influence of Fuel temperature for different ambient gas conditions ............. 22
Figure 16 : Hiroyasu - Influence of Orifice diameter for different ambient gas conditions .............. 23
Figure 17 : Hiroyasu - Influence of Orifice pressure drop for different ambient gas conditions ...... 24
Figure 18 : Typical result of Mie-scattering technique...................................................................... 25
Figure 19 : Comparison between models and experimental results................................................... 26
Figure 20 : Tuning of Siebers model ................................................................................................. 27
Figure 21 : Tuning of the Engineering Correlation............................................................................ 27
Figure 22 : Tuning of the Hiroyasu model......................................................................................... 28
Figure 23 : Comparison between tuned models and experimental results......................................... 28



Final Project : Characterization of Diesel Sprays III-32
Appendix
Appendix A : Correlations for physical properties
Ambient gas

Enthalpy

Ambient gas enthalpy has been determined with the programs written by C. Luijten and Bart
Somers based on NASA polynomials.



From these figures and the range of temperature and pressure expected for the ambient gas
for the EHPC experiments, it seems that pressure is not really affecting the enthalpy.

Heat capacity (M. Boot)

C
p,a
= 1,9327e-10 T
a
4
7,9999e-7 T
a
3
+ 1,1407e-3 T
a
2
4,4890e-1 T
a
+ 1,0575e3 [J/(kg.K)]

Thermal conductivity (M. Boot)

k
a
= 1,5207e-11 T
a
3
4,8574e-8 T
a
2
+ 1,0184e-4 T
a
3,9333e-4 [W/(mK)]

Dynamic viscosity (M. Boot)

a
= -1,3257e-17 T
a
4
+ 4,7194e-14 T
a
3
7,1799e-11 T
a
2
+ 8,078e-8 T
a
2,227e-7 [Pa.s]



Final Project : Characterization of Diesel Sprays III-33
Diesel fuel

Composition

A linear interpolation between the properties of n-decane C
10
H
22
(70 %) and 1-
methylnaphtalene C
11
H
10
(30 %) has been used to determine the critical pressure P
c
, the critical
temperature T
c
and the Pitzer's acentric factor .

C
11
H
10
C
10
H
22

Diesel fuel
(70 % C
10
H
22
+ 30 % C
11
H
10
)
Critical temperature T
c
[K] 772 617,7 664
Critical pressure P
c
[bar] 36 21,2 25,6
Pitzer's acentric factor [-] 0,310 0,489 0,435

Density (M. Boot)

f
= -6,3973e-9.T
f
4
+ 9,396e-6 T
f

3
- 5,3483e-3 T
f

2
+ 0,66441 T
f
+ 909,54 [kg/m^3]

Compressibility (fitted from Reid)

for : 0,6 < T
r
= T/T
c
< 1,024 Z = -11,102 T
r
3
+ 22,877 T
r
2
16,248 T
r
+ 4,9016 [-]
for : 1,024 < T
r
= T/T
c
< 1,3 Z = 102,75 T
r
3
- 347,36 T
r
2
+ 393,16 T
r
+ - 148,48 [-]


This correlation was derived from an interpolation in tables from [17], and it may not be
used outside the recommended range of use : 0,6 < T
r
= T/T
c
< 1,3.

Final Project : Characterization of Diesel Sprays III-34
Heat of vaporization (M. Boot)

L
vap
= -1,6084e-3 T
f
3
+ 1,5215 T
f
2
801,69 T
f
+ 4,9549e5 [J/kg]


Different correlations for heat of vaporization have been tested and compared with the one
used in the Hiroyasu model : from C
10
H
22
and C
12
H
26
data and by Pitzer. As the interesting values
are for low temperatures (choice made in numerical resolution), it seems that the correlation used by
M. Boot is precise enough.

Vapor pressure (M. Boot)

P
s
= 9,9609e-8 T
f
5
- 1,2876e-4 T
f
4
+ 0,060736 T
f
3
12,414 T
f
2
+ 1000,7 T
f
20518 [Pa]




Final Project : Characterization of Diesel Sprays III-35
Vapor fuel enthalpy (Heywood)

T
in
= T
f
/1000 ; A = 4.1868e3
h
f
= A [-9,1063 T
in
+ 123,485 T
in
2
- 47.9133 T
in
3
+ 8.0823 T
in
4
- 0,0518/(T
in
-50,128)] [J/mole]



Heat capacities (M. Boot)

C
p,vap
= 4,1569e-9 T
f
4
- 1,0237e-5 T
f
3
+ 68091e-3 T
f
2
+ 2,3341 T
f
+ 507,82 [J/(kg.K)]

C
p,liq
= 3,1754e-18 T
f
4
+ 1,5891e-6 T
f
3
- 2,1011e-3 T
f
2
+ 4,8215 T
f
+ 763,14 [J/(kg.K)]

These correlations are used to calculate the energy received by the liquid fuel from the
ambient gas to reach the temperature T
s
. However, this integration is very basic and it may not be
very accurate. The Heywood correlation coupled to the correlation for heat of vaporization at the
initial temperature will be used in the computation, even if the three possibilities are not that
different.




Final Project : Characterization of Diesel Sprays III-36
Dynamic viscosity (M. Boot)

f
= 1,0646e-11 T
f
4
1,733e-8 T
f
3
+ 1,0611e-5 T
f
2
- 2,9025e-3 T
f
+ 0,30034 [Pa.s]

Surface tension (M. Boot)

f
= 4e-11 T
f
3
1,5e-8 T
f
2
- 8e-5 T
f
+ 0,05 [N/m]

Vapor thermal conductivity (M. Boot)

k
f
= 5,8005e-15 T
a
4
3,572e-11 T
a
3
+ 8,6428e-8 T
a
2
+ 4,5135e-5 T
a
0,013969 [W/(mK)]





Final Project : Characterization of Diesel Sprays III-37
Appendix B : MATLAB
TM
scripts for Siebers models
% load NasaThermo
load pureparameters;
iO2=1; iCH4=2; iH2=3; iH2O=4; iCO2=6; iN2=7; iC2H4=8; iC2H2=9;

% Input experimental conditions
T_a = 760; % [K]
T_f = 300; % [K]
p_inj = 1e5*760; % [Pa]
d_noz = 1e-6*130; % [m]
Ca = 0.81; % [-]
mO2 = 2160; % [mg]
mN2 = 8985; % [mg]
mC2H2 = 864; % [mg]
molO2 = mO2*1e-3/pr(iO2).Mass; % [mol]
molN2 = mN2*1e-3/pr(iN2).Mass; % [mol]
molC2H2 = mC2H2*1e-3/pr(iC2H2).Mass; % [mol]
compmol = [molO2 0 0 0 0 0 molN2 0 molC2H2 0]

% calculate mixture properties after combustion
A = druksjekkuh_tom(0, mC2H2, mN2, 0, mO2, 295, 0, 0, 0); % derived from druksjekkuh.m
lambda = A(1); % [-]
density = A(2); % [kg/m3]
if A(7) > 0
compmol2 = [A(7) 0 0 A(5) 0 A(4) A(6) 0 0 0];
else
compmol2 = [0 0 0 A(5) 0 A(4) A(6) 0 0 0];
end

% Ambient gas ("air") thermodynamic properties
R = 8.3143; % [J/(mol.K)] Universal Gas Constant
M_a = 28.96; % [g/mol] Molar masses of air
rho_a = density; % [kg/m^3]
p_a = R*T_a*density/(M_a*1e-3); % [Pa]
p_a_bar = p_a*1e-5; % [bar]
dp_noz = p_inj-p_a; % [Pa] Pressure drop over the nozzle
mu_a = - 1.3257e-017*T_a^4 + 4.7194e-14*T_a^3 - 7.1799e-11*T_a^2 + 8.078e-8*T_a -
2.227e-7; % [Pa.s] Air dynamic viscosity

% Computation half cone angle
theta = 0.025*[rho_a*dp_noz*d_noz^2/mu_a^2]^0.25; % [degrees]

% Diesel fuel thermodynamic properties
M_f = 179.3; % [g/mol] Molar masses of Diesel fuel
T_c = 663.99; % [K] Critical temperature of Diesel fuel
P_c = 25.64; % [bar] Critical pressure of Diesel fuel
w = 0.4353; % [-] Pitzer accentric factor
rho_f = - 6.3973e-9*T_f^4 + 9.396e-6*T_f^3 - 0.0053483*T_f^2 + 0.66441*T_f + 909.54;
% [kg/m^3] Density
c_p_f = 3.1754e-18*T_f^4 + 1.5891e-6*T_f^3 - 0.0021011*T_f^2 + 4.8215*T_f + 763.14;
% [J/(kg.K)] Liquid heat capacity

Final Project : Characterization of Diesel Sprays III-38
L_vap = -0.0016084*T_f^3 + 1.5215*T_f^2 - 801.69*T_f + 4.9549e+005;
% [J/kg] Latent heat of vaporization
p_s = 9.9609e-8*T_f^5 - 0.00012876*T_f^4 + 0.060736*T_f^3 - 12.414*T_f^2 + 1000.7*T_f -
20518; % [Pa] Vapor pressure

% Empirical coefficients of Siebers
a = 0.66;
b = 0.41;

% Initialization
T_s = T_f;
p_s = corr_Boot_Psat(T_s);

h_a1 = Hrealmix(p_a_bar,T_a,compmol2,pr)/(M_a*1e-3); % [J/kg] air enthalpy
h_a2 = Hrealmix(p_a_bar-p_s,T_s,compmol2,pr)/(M_a*1e-3); % [J/kg] air enthalpy
delta_ha = h_a1-h_a2; % [J/kg]
rho_a = (M_a*1e-3)*(p_a_bar-p_s)*1e5/(R*T_a); % [kg/m3]

h_f_Ts = corr_Heywood_hf(T_s); % [J/kg] fuel enthalpy
h_f_Tf = corr_Heywood_hf(T_f); % [J/kg] fuel enthalpy
L_vap_Tf = corr_Boot_Lvap(T_f)/(M_f*1e-3);
delta_hf = h_f_Ts - h_f_Tf + L_vap_Tf;
Z_f_Ts = corr_compressibility(T_s); % [-] fuel compressibility
rho_f_vap = (M_f*1e-3)*(p_s*1e5)/(Z_f_Ts*R*T_s); % [kg/m3] fuel vapor density

B_left = rho_f_vap/rho_a;
B_right = delta_ha / delta_hf;
ratio = (B_right - B_left)/ B_right;

% Iterative process until covergence: B_left = B_right
step = 1;
precision = 2e-2;
while abs(ratio) > precision
T_s = T_s + step;
p_s = corr_Boot_Psat(T_s);
Z_f_Ts = corr_compressibility(T_s);
rho_f_vap = (M_f*1e-3)*(p_s*1e5)/(Z_f_Ts*R*T_s);
B_left = rho_f_vap/rho_a;
h_a2 = Hrealmix(p_a_bar-p_s,T_s,compmol2,pr)/(M_a*1e-3);
delta_ha = h_a1-h_a2;
h_f_Ts = corr_Heywood_hf(T_s);
delta_hf = h_f_Ts - h_f_Tf + L_vap_Tf;
B_right = delta_ha / delta_hf;
ratio = (B_right - B_left)/ B_right;
end

B = (B_left + B_right)/2;

% Liquid Length Scaling Law
L = (b/a)*(rho_f/rho_a)^0.5*((Ca)^0.5*d_noz/abs(tan(theta*pi/180)))*((2/B+1)^2-1)^0.5; % [m]

Final Project : Characterization of Diesel Sprays III-39
Nomenclature
a empirical coefficient determined by Siebers
A cross-sectional area
b empirical coefficient determined by Siebers
B
M
mass transfer number
B
T
heat transfer number
C
a
orifice area-contraction coefficient
C
D
discharge coefficient
c
p
heat capacity
d orifice diameter
D droplet diameter
h enthalpy
k heat conductivity
L liquid length
L
%
dimensionless liquid length
L
vap
latent heat of vaporization
M molecular weight
m
d
droplet mass
m& mass flow rate
P pressure
R universal gas constant
T temperature
U velocity
x axial coordinate of the spray
x
+
characteristic length scale
x% dimensionless characteristic length scale
Z compressibility

idealized spray angle
experimental spray angle
viscosity
density
surface tension
cinematic viscosity






Final Project : Characterization of Diesel Sprays III-40
References
[1] Siebers D.L., Liquid-Phase Fuel Penetration in Diesel Sprays, SAE Paper 980809, 1998.

[2] Siebers D.L., Scaling Liquid-Phase Fuel Penetration in Diesel Sprays, Sprays Based on
Mixing-Limited Vaporization, SAE Paper 1999-01-0528, 1999.

[3] Hiroyasu H. and Arai M., Fuel Spray Penetration and Spray Angle in Diesel Engines,
Transactions of SAE, Vol. 21, Sect. 5, 1980.

[4] Espey C. and Dec J.E., The Effect of TDC Temperature and Density on the Liquid-Phase
Fuel Penetration in a D.I. Diesel Engine, Transactions of the SAE, Vol. 104, Sect. 4, pp.
1400-1414, 1995.

[5] Browne K.R., Partridge I.M. and Greeves G., Fuel Properties Effects on Fuel/Air Mixing in
an Experimental Diesel Engine, SAE Paper 860223, 1986.

[6] Kamimoto T., Yokota H. and Kobayashi H., Effect of High Pressure Injection on Soot
Formation Processes in a Rapid Compression Machine to Simulate Diesel Flames,
Transactions of the SAE, Vol. 96, Sect. 4, pp. 4783-4791, 1987.

[7] Yeh C.-N., Kamimoto T., Kobori S. and Kosaka H, 2-D Imaging of Fuel Vapor
Concentration in a Diesel Spray via Exciplex-Based Fluorescence Technique, SAE Paper
932652, 1993.

[8] Kuo K.K., Principles of Combustion, John Wiley and Sons, New York, NY, 1986.

[9] Naber J.D. and Siebers D.L., Effects of Gas Density and Vaporization on Penetration and
Dispersion of Diesel Sprays, Transactions of the SAE, Vol. 105, Sect. 3, pp. 82-111, 1996.

[10] Higgins B.S., Mueller C.J. and Siebers D.L., Measurements of Fuel Effects on Liquid-Phase
Penetration in DI Sprays, SAE Paper 1999-01-0519, 1999.

[11] Jung D. and Assanis D.N., Multi-Zone DI Diesel Spray Combustion Model for Cycle
Simulation Studies of Engine Performance and Emissions, SAE Paper 2001-01-1246, 2001.

[12] Elkotb M.M., Fuel Atomisation for Spray Modelling, Prog. Energy Combustion. Sci., 8,
1982, pp 61.

[13] Meyer R. and Heywood J.B., Evaporation of In-Cylinder Liquid Fuel Droplets in an SI
Engine: A diagnostic-Based Modelling Study, SAE Paper 1999-01-0567, 1999.

[14] SFPE Handbook of Fire Protection Engineering, 2nd Edition Table B-2, 1995.

[15] Chalmers Liquid Fuel Database: http://tfdpc67.tfd.chalmers.se/cgi-bin/chem.cgi

[16] Heywood J.B., Internal Combustion Engine Fundamentals, Mc Graw Hill, 1988.

[17] Reid R.C., Prausnitz J.M. and Poling B.E., The properties of gases & liquids, 4th Edition,
pp. 34-39, 1987.
















Part IV

EHPC Experiments
Literature Review : Advanced Laser
Diagnostics for Soot Formation


















Final Project : Characterization of Diesel Sprays IV-2
Contents

Introduction .................................................................................................. 3
1. Soot characteristics................................................................................. 4
1.1 Generalities ........................................................................................................................ 4
1.2 Important parameters ......................................................................................................... 4
2. Thermal radiation : two-colour method ................................................. 5
2.1 Theoretical background...................................................................................................... 5
2.2 Experimental set up............................................................................................................ 6
2.2.1 Determination of parameters .............................................................................. 6
2.2.2 Experimental apparatus ...................................................................................... 7
2.3 Results................................................................................................................................ 8
3. Light extinction....................................................................................... 9
3.1 Theoretical background...................................................................................................... 9
3.2 Experimental set up............................................................................................................ 9
3.2.1 Determination of parameters .............................................................................. 9
3.2.2 Experimental apparatus .................................................................................... 10
3.3 Results.............................................................................................................................. 12
4. Laser Elastic Scattering ........................................................................ 13
4.1 Theoretical background.................................................................................................... 13
4.2 Combination with other techniques ................................................................................. 13
4.2.1 Combination with light extinction.................................................................... 13
4.2.2 Combination with thermal radiation................................................................. 14
4.3 Experimental set up.......................................................................................................... 14
4.3.1 Determination of parameters ............................................................................ 14
4.3.2 Experimental apparatus .................................................................................... 15
4.4 Results.............................................................................................................................. 16
5. Laser Induced Incandescence............................................................... 17
5.1 Theoretical background.................................................................................................... 17
5.2 Combination with other techniques ................................................................................. 17
5.2.1 Combination with LOS techniques................................................................... 17
5.2.2 Time Resolved LII............................................................................................ 18
5.3 Experimental set up.......................................................................................................... 19
5.3.1 Determination of parameters ............................................................................ 19
5.3.2 Experimental apparatus .................................................................................... 20
5.4 Results.............................................................................................................................. 21
6. Discussion............................................................................................. 22
6.1 Summary.......................................................................................................................... 22
6.2 Recommendations............................................................................................................ 23
Tables ......................................................................................................... 24
Figures ........................................................................................................ 24
References .................................................................................................. 25

Final Project : Characterization of Diesel Sprays IV-3
Introduction
The formation of soot during the incomplete combustion of hydrocarbon fuels remains one
of the least solved problems of combustion in diesel engines and is a major component of the
exhaust emissions. Therefore it is necessary to be able to experimentally study the soot
characteristics in the actual or future experimental TU/e set up (EHPC, optical engine, burner).
Optical techniques seem to be appropriate for this kind of investigation, as they are most of the time
non intrusive methods and can be easily set up. There are several different optical techniques for
soot diagnostic, and each of them present advantages and drawbacks depending on the
characteristics studied and the kind of set up (high pressure engine or burner, atmospheric flame).
This report is an overview of the main optical techniques used for soot diagnostic. It is not
specific to one TU/e set up, though an eventual future application will be under high pressure,
which has to be taken into account. These techniques include thermal radiation, light extinction,
light scattering, laser-induced incandescence. All of them are used in high pressure experimental set
up, such as fundamental burners, high pressure combustion vessels or optically accessible engines.
A comparison between these methods will be performed with diagnostic criteria defined in the
following section.

Final Project : Characterization of Diesel Sprays IV-4
1. Soot characteristics
In this section, the main soot characteristics are shortly presented. The main parameters used
for soot characterization and usually experimentally determined are listed.
1.1 Generalities
Soot is produced during the combustion of hydrocarbon fuels. The formation of soot is the
most complex chemical system in flames. Its mechanism is normally described in four steps:
nucleation, surface growth, coalescence and then aggregation. Soot particles emitted due to
combustion consist of chainlike aggregates composed by spherical units having hexagonal
structures similar to graphite. Individual particles appear in the early stages of soot formation, just
after nucleation, while these particles more likely tend to form aggregates afterwards. Soot particles
contain several thousands of carbon atoms formed from simple fuel molecules within few
microseconds. Their sizes and morphology are characteristic of the nature of original fuel and its
oxidation products. For example, the sizes of soot individual particles emitted by Diesel combustion
in current engines vary on a range of 10 to 70 mm. At a particular flame location, aggregate size
distribution is expected to be broad (polydisperse) while particle spheres would have nearly uniform
diameters (monodisperse).
1.2 Important parameters
For the characterization of soot due to combustion, optical techniques should allow some
parameters to be experimentally determined. These important parameters are :
Size (primary particle diameter D and aggregate size)
Distribution (soot volume fraction f
v
, number density C
n
, KL factor)
Temperature T

Although it has been common to focus on soot concentration, especially emission
regulations, size is possibly the most important parameter to govern the interaction of a particle with
its surroundings.
Furthermore, the nature of the results (quantitative or qualitative), the kind of visualization
(point, line, 2D) and the ease of the set up will be taken into account in this comparative
overview of optical techniques for soot diagnostic.

Final Project : Characterization of Diesel Sprays IV-5
2. Thermal radiation : two-colour method
Incomplete combustion is characterized by its high luminosity caused by the thermal
radiation of soot particles at high temperature. The two-colour method, also known as the two-
wavelengths or radiation method, is based on the use of two different wavelengths and is performed
mostly to determine the soot temperature T, but it can also provide a value of the KL factor.
2.1 Theoretical background
The whole theory of thermal radiation will not be detailed here. Only the main useful
assumptions and equations from [1] will appear below. The idea of the two-colour method is to
detect the thermal radiation signals at two different wavelengths and to determine the flame
temperature from their ratio. Note that the flame temperature is assumed to be close enough to the
soot temperature T.
The radiation emissive power (spectral irradiance) of a black body E
b,
depends on the
wavelength and the temperature T of this black body as described by Eq. 1, where C
1
= 3,7418.10
-
6
W.m
2
and C
2
= 1,4388.10
-2
m.K are respectively the first and the second Planck' constants.
( )
2
1
b,
C T 5
C
E
e 1

=



( 22 )
The monochromatic emissivity of a non-black body

(i.e., the fraction of black body

radiation emitted by a surface at wavelength ) is defined in Eq. 2, where I

(T) and I
b,
(T) are the
monochromatic emissive power of respectively a non-black and a black body at temperature T and
wavelength . The apparent temperature T
a
is defined as the temperature of a black body which
emits the same radiation intensity at wavelength as a non-black body at a temperature T.
b, a
b, b,
I (T )
I (T)
I (T) I (T)

= = ( 23 )
Combining Eq. 1 and 2, the monochromatic emissivity

can be determined theoretically in

Eq. 3. In practice, for soot particles, there is a widely used correlation by Hottel and Broughton [2]
for

described in Eq. 4 which enables the two-colour method to be derived.

( )
( )
2
2 a
C T
C T
e 1
e 1

( 24 )
( )
KL
1 e

= ( 25 )
In Eq. 4, K is the absorption coefficient, which is proportional to the number density C
n
of
the particles, and L is the geometric optical thickness of the flame along the optical axis of the
detection system. The parameter depends on the optical and physical properties of the soot.

Combining Eq. 3 and 4, we can obtain the KL factor as described in Eq. 5. This KL factor
can be eliminated by rewriting the equations for two distinct wavelengths
1
and
2
. This
combination results in Eq. 6 which can be solved if the apparent temperatures T
a1
and T
a2
are
known.
( )
( )
2
2 a
C T
C T
e 1
KL ln 1
e 1

| |

=
|
|

\
( 26 )

Final Project : Characterization of Diesel Sprays IV-6
( )
( )
( )
( )
1 2
2 1 2 2
2 1 a1 2 2 a2
C T C T 1 2
C T C T
e 1 e 1
1 1
e 1 e 1




| | | |

=
| |
| |

\ \
( 27 )

The KL factor can be determined with the value of T from Eq. 6. This factor, which is a
criteria by itself, is proportional to the soot volume fraction f
v
, which can be determined with a few
assumptions (Rayleigh approximation, path length L, parameter and complex refractive index m
known) in Eq. 7.
v
2
2
1 K
f
m 1
6 Im
m 2

=
| |

|
+
\
( 28 )
The soot refractive index function E(m), which is used in many optical diagnostic
techniques and is discussed in later stages, is defined in Eq. 8.
2
2
m 1
E(m) Im
m 2
| |
=
|
+
\
( 29 )
2.2 Experimental set up
2.2.1 Determination of parameters
Some parameters have to be determined in order to calculate the temperature T and the soot
volume fraction f
v
. The two wavelengths have to be chosen carefully, and it influences the values of
the parameter . Furthermore, the apparent temperatures T
a1
and T
a2
have to be measured for these
wavelengths.
2.2.1.1 Choice of wavelengths
The studies using two-colour methods have been performed with infra-red and visible
wavelengths. According to Zhao and Ladommatos [&], visible wavelengths, as separated as
possible, are preferred for the following reasons : greater sensitivity to changes in flame
temperature, greater difference in signal output (lower signal-to-noise ratio) and lower sensitivity to
the choice of for the estimation of the temperature and the KL factor. However, the need to avoid
emissions in visible from reacting radicals may limit the range of appropriate wavelengths. Typical
values for from literature [1] are from 460 nm to 750 nm for visible and from 750 to 4000 for
infra-red.
2.2.1.2 Choice of optical and physical properties
From the literature review reported by Zhao and Ladommatos [1], a recommended value of
this parameter for most fuels is = 1,39 when using visible wavelengths, and close to unity for
infra-red wavelengths (range from 0,8 m to 7 m). However, it seems that the choice of values for
is a critical point for the accuracy of the method to obtain T.
The value of the soot refractive index function E(m) is critical, because the estimation of the
refractive index m has a huge influence on the accuracy of the method to convert the KL factor into
the soot volume fraction f
v
. A widely used value for m has been determined in [3] and seems very
common : m = 1,57 0,56i, which provides of value of E(m) = 0,260. This value seems to fit for all
optical techniques since it is a characteristic of soot and wavelength. Experimental results from
Chang and Charalampopoulos [4] are summarized in Table 4.

Final Project : Characterization of Diesel Sprays IV-7

Wavelength
(m)
Refractive index
m = n - ki
Refractive index
function E(m)
0,337 1,48 0,77i 0,377
0,532 1,75 0,61i 0,241
0,63 1,80 0,58i 0,221
0,810 1,85 0,57i 0,208
1,064 1,91 0,59i 0,204
Table 4 : Refractive index variation with wavelength [4]
2.2.1.3 Determination of apparent temperatures
The apparent temperatures T
a1
and T
a2
can be measured with a calibrated two-colour optical
pyrometer system. However, for this measurement a calibrated emission source is required if
accurate results of temperature are required, which is difficult to find [1,5]. Though a black body
would be ideal, a tungsten lamp is often used for the calibration as described in [1].
2.2.2 Experimental apparatus
A typical schematic from a thermal radiation / two-colour set up is represented in Figure 1
and Figure 2, which are extracted the work from Bento et al. [6] on co-flow propaneair laminar
diffusion flames in a relatively high pressure combustion chamber (up to 8 bar stable flames). It
represents the two dimensional imaging system for temperature and soot. Typical optical devices
(lens, shutter, aperture) are used to guide the beam, while a digital high speed camera is used to
collect the signals.





Figure 1 : Schematic of high pressure combustion
chamber for the two-colour method [6]

Figure 2 : Schematic of the optical set up for the two-
colour method [6]
Note that a two-dimensional visualization should be theoretically possible, but it seems very
difficult to perform it and other techniques are much more used for that.

Final Project : Characterization of Diesel Sprays IV-8
2.3 Results
A series of emission detections at a given height in a flame can be inverted (Abel inversion)
to obtain radially resolved values of emission from which temperature T and soot volume fraction f
v

can be determined, as shown in Figure 3 and Figure 4 extracted from Bento et al. [5].


Figure 3 : Results of soot volume fraction
measurements with the two-colour method [6]

Figure 4 : Results of temperature measurements
with the two-colour method [6]
For engines applications of two-colour diagnostic there are large variations in KL factor
results (and in a lower scale temperature results) due to the irregularity of the process (random
location of ignition and combustion, motion) : only cycle averaged values would be relevant for
KL factor [1]. This problem should not occur on stable flames in a burner set up.

As the measurement of temperature with this method is based on the soot thermal radiation,
there will be no information in zones where the soot concentration is not high enough to provide a
reasonable radiation, which is a limit for this method [6].

The two-colour method is widely used for the determination of soot temperature, sometimes
in order to complete a diagnostic performed with another optical technique. Temperature
measurements seems to be reliable, with a reasonable accuracy of 3,5 % in the radial region of the
flames, while in the core of the flame the uncertainty is too high [6,7]. The soot volume fraction
measurements suffer from a high uncertainty around 40 % [6]. This could be explained by the poor
estimation of the refractive index function E(m) (14 to 24 % uncertainty according to Krishnan et
al. [8]), because the soot volume fraction f
v
is very sensitive to this value.

Final Project : Characterization of Diesel Sprays IV-9
3. Light extinction
When a beam of light passes through a cloud of soot particles, its power is attenuated. The
light extinction method is based on the measurement of this light extinction. It has been a popular
and accurate way to measure soot concentration due to its simplicity and accuracy.
3.1 Theoretical background
An overview of the extinction theory is presented below. For clarity, the equations may have
been simplified under usual assumptions that will not always be justified here. The attenuation of
the light is caused by both scattering and absorption by the soot particles. The Beer-Lambert law in
Eq. 9 gives the KL factor, also known as the "optical thickness", as a function of the incident light
intensity I
0
and the light intensity I after passing through the soot cloud. The parameter K is the
extinction coefficient of the cloud of particles, including both scattering and absorption, and L is the
path length of the beam. The ratio between collected and incident intensities is the transmittance
of the cloud of particles.
( )
0
KL ln I I ln = = ( 30 )
Assuming the Rayleigh scattering approximation (typical diameter D much smaller
compared to laser wavelength ), the absorption dominates the scattering in terms of efficiency, as
described in Eq. 10.
2 2
4
abs scat 2 2
m 1 8 m 1 D
Q 4xIm Q x Re , x 1
m 2 3 m 2
| | | |
= >> = = <<
| |
+ +
\ \
( 31 )
The soot volume fraction f
v
can be determined by Eq. 11 (or Eq. 12), with a few simplifying
assumptions, as a function of the extinction coefficient K (or the path length L) and the soot
refractive index function E(m). Values of l
abs
used in Eq. 12 have been determined by Chang and
Charalampopoulos [4].
v
sa
K
f
6 (1 )E(m)

=
+
( 32 )
0 abs
v abs
sa
I l
f ln , l
I L 6 (1 )E(m)

| |
= =
|
+
\
( 33 )
3.2 Experimental set up
3.2.1 Determination of parameters
The coefficient
sa
is the scattering to absorption ratio, as scattering has to be taken into
account in some cases. Soot particles can indeed form aggregates which are more likely to scatter
light due to their size, which could lead to approximated values of
sa
= 0,2 - 0,4 according to some
studies [9,10,11]. The soot refractive index function E(m) is the same as previously discussed, with
an approximated value of E(m) = 0,260 corresponding to the common value m = 1,57 0,56i. Most
studies using light extinction are performed with wavelengths in visible [1,5,12,13].


Final Project : Characterization of Diesel Sprays IV-10
The determination of the path length L is required to obtain quantitative results. The natural
soot luminosity measurement could provide a value of path length, and therefore a path-length-
averaged soot volume fraction. Pickett and Siebers [14] found that the path length could be defined
as distance between radial positions on the wings of the luminosity profile where the luminosity
level equals approximately 5 % of the peak, as illustrated by Figure 5. This method fits for many
chamber conditions, and axial profiles from path length L similar to Figure 6 can be determined.





Figure 5 : KL and luminosity radial profiles at a
constant axial distance from the injector [14]

Figure 6 : Soot path length L axial profile [14]
3.2.2 Experimental apparatus
Although the light extinction method is mostly used as a Line Of Sight technique (LOS)
which provides measurements along a line of points through the soot cloud in the combustion
chamber, it is possible to obtain a two-dimensional imaging of soot.
3.2.2.1 Experimental difficulties
The main difficulties to perform light extinction measurement are the interferences with
other physical phenomenon due to the behaviour of soot particles crossed by a laser beam. These
difficulties are often eliminated or at least reduced with the use of proper optical devices.

Refractive index gradients due to gradients in temperature, density or mixture composition
can deflect the laser beam as it passes through the combustion chamber : this is the principle of the
Schlieren technique. This phenomenon could disperse the transmitted beam by as much as 100
mrad. As a consequence, the detection system has to be wide enough with a focusing lens to collect
all the transmitted light, and show an uniform sensitivity [1,13,14]. This beam steering is significant
and has to be addressed, especially in high pressure environments, where the increased density
creates greater refractive index gradients which exacerbates the problem [13]. Practical solutions
used by searchers have been a diffuser plane [14], or most likely an integrating sphere removing
any beam orientation information from the transmitted light [13,14], though almost no information
about the collecting angle is available on literature.

The natural soot incandescence due to combustion creates interferences with the light
extinction. Its influence on the measured light intensities can be reduced by the use of a very narrow
band-pass interference filter (1 - 10 nm full-width at half-maximum FWHM) centred at the laser
wavelength in front of the photodiode [1,13,14]. Aperture can also be used to restrict the
combustion luminosity reaching the diffuser when such a device is used [13]. The transmitted light
can also be corrected by subtracting the natural soot incandescence collected by the filter system
with the laser blocked, which gives a correction up to 4 % of the signal for the highest ambient gas
pressure and temperature [14].

Final Project : Characterization of Diesel Sprays IV-11

Parasite interferences could also come from the absorption or the scattering of ambient gas,
dust and fuel droplets when working on an engine [1]. Eventually, soot deposition on windows
could also attenuate the light collected [13].
3.2.2.2 Line Of Sight Imaging
The light extinction technique is a LOS technique, thus it can only provide discrete
measurements along a line of points through the soot cloud. The experimental apparatus shown
schematically in Figure 7 is very simple and consists of a laser beam whose transmitted signal after
having crossed the flame is collected with several optical devices.

Figure 7 : Schematic of the optical set up for both light extinction and 2-D visualization [14]
In order to plot radial or axial profiles of results, it is necessary to change the geometrical
settings and to perform a new measurement. For the visualization of soot distribution, a matrix of
measurement points are required, which is tedious and time-consuming and would only provide
ensemble-averaged soot distributions due to the different measurements required [1].
3.2.2.3 Two-dimensional imaging
For a two-dimensional imaging, an expanded light source has to be used. In order to cope
with the refractive index issues previously explained, a diffuser and a filter are used which suppress
the Schlieren effect with diffused light [1].



Figure 8 : Schematic of a light extinction set up for two-dimensional visualization [1]
However, the light extinction method is not widely used for two-dimensional soot
visualization in the literature due to other more performing techniques.

Final Project : Characterization of Diesel Sprays IV-12
3.3 Results
The critical point to determine the soot volume fraction f
v
is that the measurement of I and I
0

provides only a value of the KL factor, where the path length L is not always constant during
combustion events and anyway difficult to measure [13]. Without this value of L, a tomographic
reconstruction (three-points Abel inversion) [14] can be used to extract the extinction coefficient K
from the KL radial profile (Figure ) and then determine the soot volume fraction f
v
with Eq. 11, as
illustrated by Figure 9. The approach of Pickett and Siebers [14] to determine the path length L
seems to be interesting and provides axial profiles of path-length-averaged soot volume fraction
v
f ,
as illustrated in Figure 10 for different ambient conditions.




Figure 9 : Soot volume fraction f
v
radial profile
obtained from a three-point Abel inversion of the
KL data [14]

Figure 10 : Path-length-averaged soot volume
fraction
v
f axial profile [14]
Note that a qualitative diagnostic can be performed directly with the optical thickness KL.
The results of soot volume fraction may also be used for the calibration of another optical
diagnostic technique combined with light extinction [15], as these results are accurate enough.

According to the widespread use of this technique, the results of soot volume fraction may
be considered as accurate enough for further diagnostic. The results accuracy is affected mostly by
the uncertainty of the refractive index function E(m), as discussed previously. The effect of
scattering may also be a problem regarding uncertainties of the light extinction method. Further
information about accuracy issues for light extinction method can be found in [16]. For accuracy
considerations, regions with a too high or too low extinction are not taken into account within
acquired data, using thresholds in the ratio I/I
0
(for example between 2 % [14] and 95 % [12]).

Final Project : Characterization of Diesel Sprays IV-13
4. Laser Elastic Scattering
Particles excited by a laser beam scatter light in different directions, and this scattered light
depends on the properties of these particles, especially their size and concentration. Laser Elastic
Scattering is a widely used technique to study the soot formation process in flames.
4.1 Theoretical background
The scattering method hereby described is valid for the Rayleigh approximation, which
deals with particles with a diameter D which are much smaller compared to the incident laser
wavelength , which is the case for soot particles. The elastic scattering means that the energy, and
thus the wavelength, of the scattered light is the same as for the incident laser light. The scattering
theory below is simplified for clarity and is mostly derived from [1,17]. Here we will only take care
of scattering light collected at a given angle (mostly perpendicularly) to the incident laser sheet.

If we assume the particles as isotropic spheres in a monodisperse system (uniform
diameters), the light is scattered by the particles with a scattering coefficient K
scat
given by Eq. 13.
2
5 2
6 6
scat n scat n 4 2
2 m 1
K C D K C D
3 m 2

=
+
( 34 )
We can see that the scattering coefficient K
scat
is proportional to the number density C
n
and
the particle diameter D, which is a distribution averaged size, at the sixth power. However, none of
these parameters can be determined separately from the scattering signal without other
informations.
4.2 Combination with other techniques
We cannot determine soot parameters from the scattering signal itself : it has to be
complemented with another signal based on other optical techniques. This has been performed
mostly with light extinction, though it is also possible to use thermal radiation. Tree and Foster [18]
have performed experiments and compared both combinations, whose results are presented and
discussed in [1].
4.2.1 Combination with light extinction
Scattering technique has been commonly combined with light extinction [7,12,18]. Indeed,
the extinction coefficient K
ext
, given by Eq. 14, is proportional to the number density C
n
and the
particle diameter D at the third power. This is true for the Rayleigh approximation, where forward
scattering is dominated by extinction.
2
2 2
3 3
ext abs n ext n 2
m 1
K K Im C D K C D
m 2
| |
= =
|
+
\
( 35 )
The unknown number density C
n
can be eliminated by taking the ratio between the
scattering and the extinction coefficients as presented by Eq. 15, which provides a signal
proportional to the particle diameter D at the third power.

Final Project : Characterization of Diesel Sprays IV-14
2
2
3
2
3 3
scat scat
2
ext ext
2
m 1
m 2 K 2 K
D D
K 3 K m 1
Im
m 2

+
| |
=
|
| | \
|
+
\
( 36 )
Therefore the primary particle diameter D can be determined through Eq. 15 with the
measurement of both scattered light and light extinction. However, D is an distribution averaged
diameter, and for polydisperse clouds of particles, it may be biased by the size distribution which
remains unknown. From the value of particle diameter D and the knowledge of the calibration
constants, the number density C
n
can be calculated from the scattering or the light extinction signal,
although differences between the two results may occur [1].
However, since individual particles tend to form aggregates whose size may not be valid for
the Rayleigh approximation. Therefore, the particle diameter D is more a "volume equivalent sphere
diameter" neither than a real particle size nor an aggregate size in aggregation regions [12,19].
An approach with scattering at two different angles has been performed by Yang and Koylu
[12] which could distinguish primary particles from soot aggregates.
4.2.2 Combination with thermal radiation
As shown by Tree and Foster work [18] presented and discussed in [1], the scattering
diagnostic could be also completed with thermal radiation measurements using the two-colour
method previously presented. Indeed, with this technique, it is possible to obtain the value of the
K
ext
L factor in addition to the soot temperature T. This K
ext
L factor, according to the extinction
theory with the Rayleigh approximation in Eq. 14, is proportional to the product C
n
D
3
L. If the
optical length L is known, the same determination method used for scattering/extinction provides
the values of particle diameter D and number density C
n
. However, the advantage of combining
scattering with thermal radiation is that the soot temperature can be simultaneously determined,
which can be useful.
4.3 Experimental set up
4.3.1 Determination of parameters
As extinction technique is combined with scattering, the same parameters have to be
determined compared to extinction alone. The extinction coefficient K
ext
has to be extracted from
the KL factor measured with the incident and attenuated light intensities. The path length can be
determined with the same technique than previously explained.
The laser wavelength is chosen in visible. Note that the power used for scattering
measurements may be higher than for extinction measurements (1 W against 0,15 W for an argon-
ion laser at 514,5 nm in [12]).
When thermal radiation is combined with scattering, the path length L has to be determined
for further calculations. Tree and Foster approximated it by matching the peak of soot volume
fraction [18].

Final Project : Characterization of Diesel Sprays IV-15
4.3.2 Experimental apparatus
As scattering is mostly combined with extinction and the set up may be similar, the
experimental set up of thermal radiation will not be presented here.
4.3.2.1 Experimental difficulties
The scattering measurement set up has to be calibrated so that absolute scattering
coefficients of soot can be distinguished from other sources of scattered light, especially the
components of the set up (walls and windows). This can be done with well documented scattering
gas (nitrogen and methane in [12]). Furthermore, the soot scattering intensity must be corrected for
the attenuation of the laser light as it goes through the flame to the measurement location and vice
versa, which is done with extinction measurement.
An additional difficulty to use elastic scattering in a diesel engine is that the liquid fuel
droplets would scatter light as well as soot particles, creating a significant interference. This
problem could be solved by using two different polarizations of laser light, although the experiment
is significantly more complicated [1].
The natural flame luminosity can be suppressed at the collection of the signal by using a
short camera shutter time [1].
4.3.2.2 Set up
A typical set up for combined scattering and extinction measurements is presented as a
schematic in Figure 11. The scattering signal is collected by a photo-multiplier tube PMT (or a
CCD camera which is less sensitive), while the extinction signal is collected by a photodiode
collector. Note that the PMT is only used for point measurements. A polarizer is used to only
collect the scattered light with the same polarization than the laser source. Diaphragms are placed at
various locations along the optical path to eliminate possible light reflections from the components
and apparatus.

Figure 11 : Schematic of a combined scattering/extinction set up [12]
This kind of experimental apparatus seems to be easy to set up, with one single laser source.
However, the main difficulties are the interfering issues which have to be addressed mainly due to
the use of extinction technique.
Note that a two-dimensional imaging of the soot clouds with scattering is possible and has
been successfully used to map the soot distribution : a laser sheet (as shown in Figure ) has to be
produced from the laser beam with optical devices and the plane scattering image can be collected
with a camera (HS, CCD or PMT).

Final Project : Characterization of Diesel Sprays IV-16
4.4 Results
Typical results from both scattering/extinction and scattering/radiation techniques can be
extracted from Tree and Foster [18] as presented in [1]. It includes time-resolved values of number
density C
n
, particle diameter D, soot volume fraction f
v
and soot temperature T (for combination
with thermal radiation). Both techniques provide almost similar behaviours. Figure 13 is the result
of single scattering experiments on a running engine extracted from [20], which is almost
impossible to interpret nor use to determine soot concentration or size.


Figure 82 : Temporal results from scattering/extinction and
scattering/radiation techniques [18]










Figure 13 : Spatial result from scattering
experiment (16,5 ATDC) [20]
The scattering/extinction technique is thought to be more quantitative than the
scattering/radiation technique because the later involves the approximation of the path length L and
possibilities for radiations from soot outside the scattering measurement volume to be detected [1].

Concerning the scattering/extinction technique, the uncertainty level in the determination of
particle diameter D is estimated to be about 30-35 % [1,12], without taking into account the
uncertainty on refractive index. The scattering/radiation technique suffers from an additional source
of error with the determination of path length L [1]. Furthermore, it seems that the interpretation of
scattering signal is difficult, since it could be due to a higher concentration or larger particles or a
combination of both [1].

Final Project : Characterization of Diesel Sprays IV-17
5. Laser Induced Incandescence
Laser-induced incandescence (LII) is generated when a cloud of soot is irradiated with an
intense laser light. The soot particles are heated to a temperature (~ 4000 K) well above the
surrounding gas temperature due to the absorption of laser energy. Then the heated soot particles
emit black body radiation corresponding to the elevated soot temperature. The LII signal should be
able provide measurements of soot volume fraction f
v
, particle size D and a two-dimensional
imaging, alone or in combination with other optical techniques. This technique has been recently
implemented and is more widely used nowadays.
5.1 Theoretical background
The theory of LII is mainly based on energy and mass balances equations, which are
described in detail in [1] and will not be only presented here. From Melton work [21], one of the
results presented in Eq. 16 is that the LII signal is roughly proportional to the soot volume fraction
due to its dependence on the particle number density C
n
and diameter D at the power (3 + 0,154/),
where is the detection wavelength in m.
(3 + 0,154 )
LII n LII v
I C D I f

( 37 )
This relation of proportionality is only true for relatively a high laser power. If the
calibration constants can be determined (with extinction for example), quantitative measurements of
soot volume fraction can be performed. Note that for a detection wavelength between 0,4 and 0,7
m, the power of D varies between 3,38 and 3,22 and confirms the approximation previously used
in Eq. 16.
5.2 Combination with other techniques
Although the LII signal is already proportional to the soot volume fraction f
v
, it can be
combined with other techniques to provide more quantitative results and other soot characteristics.
This can be done with LOS techniques (light extinction and elastic scattering) and through the study
of the heated soot cooling.
5.2.1 Combination with LOS techniques
5.2.1.1 Calibration with extinction
Light extinction measurements can be used in order to calibrate the LII signal which is
proportional to the soot volume fraction but is a absolute measurement of incandescent light. With
this calibration, it should be possible to obtain quantitative results. The soot volume fraction is
measured at a single spatial point of a flame with extinction and is used to calibrate the LII signal
[1,15,19]. However, the uncertainty of light extinction results could lead to uncertainties on the
order of a factor of 2 when used for LII calibration [15,17].

Final Project : Characterization of Diesel Sprays IV-18
5.2.1.2 Combination with elastic scattering
Combining LII with elastic scattering allows the determination of other soot characteristics
such as the number density C
n
and the primary particle diameter D [1,22]. As previously described,
the scattering signal I
scat
is proportional to the C
n
D
6
factor while the LII signal I
LII
is approximately
proportional to the C
n
D
3
factor. From these two signals, the number density C
n
and the primary
particle diameter D can be determined as presented in Eq. 17 and Eq. 18.
3
scat
LII
I
D
I
( 38 )
2
LII
n
scat
I
C
I
( 39 )
The use of scattering provides semi quantitative results for the number density and the
particle diameter, which is not possible with one of these techniques alone.
5.2.2 Time Resolved LII
Melton [21] introduced the idea that the temporal profile of the LII signal after the laser
pulse could be used to determine the particle size. The main assumption is that the LII signal decay
of a larger soot particle is longer than for a smaller soot particle, as illustrated by Figure 14 for
different ambient temperatures. Using this size dependence of the cooling, the particle diameter D
can be determined by analysing the curve directly or by taking the ratio of the LII signal at two
delay times after the laser pulse [5,19] which allows a two dimensional measurement of size. Note
that the particle diameter D determined by this technique is a real particle diameter and is not
exactly the same than the diameter D (volume equivalent sphere diameter) previously discussed.


Figure 14 : Influence of primary soot particle size D on the LII signal decay time [5]
However, since ambient temperature (almost the same than soot temperature) is needed, the
TIRE LII requires modelling of the cooling to interpret the temporal profile of the LII signal. Note
that the ambient temperature can be determined experimentally by using the thermal radiation
technique described in Section 2.

Final Project : Characterization of Diesel Sprays IV-19
The modelling has been developed by Will et al. and is presented in [5] but it will be shortly
described below. The energy balance in Eq. 19 is composed by the absorption of laser energy
(efficiency Qabs, irradiance Ei), heat conduction to the surroundings (heat transfer coefficient h,
particle temperature T, surrounding gas temperature T
0
), vaporization (heat of vaporization H
v
,
molar mass of solid carbon M, particle mass loss due to vaporization dm/dt), thermal radiation
(emission coefficient , blackbody spectral irradiance E

b
) and change of internal energy (density ,
specific heat C).
( )
2
2
v
abs i 0
3
2 b
D H dm
Q E h T T D
4 M dt
D dT
D (D, )E (T, )d C 0
6 dt

( 40 )
From Eq. 19, the temporal behaviour of particle temperature and size can be obtained by
numerical integration. The size dependence of the LII signal can also be determined : the relative
LII signal intensity is proportional to the particle surface, i.e. to its square diameter D
2
as shown in
Eq. 20 where R() is the spectral characteristic of the detection path.
2 b
LII
I D R( ) (D, )E (T, )d

( 41 )
By calculating the ratio of integrated LII signals at two delay times and compare it with the
signal ratio from the model for different particle sizes, it is possible to obtain a two-dimensional
measurement for the particle diameter D.
5.3 Experimental set up
5.3.1 Determination of parameters
5.3.1.1 Determination of laser power
The LII technique requires the use of a high power laser. For the LII combined with
extinction, it is important not to use a too high laser power, because extinction measurements could
be affected, but high enough to obtain the incandescent signal required [17,19]. Indeed, if the laser
power is too low, the soot particles would not be heated enough. Furthermore, with a too high laser
power, there would be too much vaporization from polyaromatic hydrocarbons and C
2
and C
3

radicals which would create Laser-Induced Fluorescence (LIF) potentially affecting the LII signal
[1,5,17]. Accordingly, for single LII measurements, there is an optimal zone for laser power just
above vaporization limit where the LII signal is stable for small variations in laser power [17].
5.3.1.2 Determination of wavelengths
The LII signal presents a low signal-to-noise ratio which requires the used of a gated camera
with very short gating down to nanoseconds and a clever selection of excitation and detection
wavelengths to optimise the collected signal [1]. A short excitation wavelength (for example
Nd:YAG at 532 nm) would be more effective in the soot particles heating, improving the signal.
The recommended use of a short detection wavelength (preferentially 400 nm) would better
separate heated soot from non-heated soot but lower the overall intensity collected [1]. However, a
long detection wavelength could improve the proportionality between the LII signal and the soot
volume fraction since this signal is proportional to the diameter D at the power (3 + 0,154/) [23].

Final Project : Characterization of Diesel Sprays IV-20
5.3.2 Experimental apparatus
As the experimental apparatus and the associated issues for the other techniques have
already been presented above, only the specific difficulties for LII will be presented here.
5.3.2.1 Experimental difficulties
The natural flame luminosity can be suppressed at the collection of the signal by using a
gated intensified camera and interference filters, which would also allow the low signal-to-noise
ratio LII signal to be better analysed. Furthermore, there will be no interference from liquid droplets
scattering, as the LII signal is derived from the heated soot particles [1] and can be detected just
after the laser pulse [24]. To overcome the severe beam attenuation due to the use of some fuels
such as standard diesel fuels, it is recommended to use only low-sooting fuels [1,14]. When used on
running engines, the rate of window fouling and the overheating can be reduced by firing only once
every 10
th
or 20
th
cycle and collect these images [1]. When using the extinction calibration, some
vaporization could affect the extinction properties of soot, which can be avoided by delaying the
extinction beam from the LII laser sheet [19].
5.3.2.2 Set up
The typical set up for LII experiments on a burner (almost similar for high pressure burner)
combined with extinction for calibration is illustrated in Figure 15. A single laser source can be
used to produce the extinction beam and the LII sheet. A CCD camera is used to collect the LII
signal, while the collection of the attenuated beam is similar to the system presented in Section 3.

Figure 15 : Schematic of the set up for LII combined with extinction [19]
For experiments combining LII and scattering on a running engine, Figure 16 shows a
typical set up. It consists of a single laser source and an optical device to collect both signals and
then separate them. The collection is performed with two CCD cameras using filters at two different
wavelengths as the detection wavelengths are different for LII and scattering signals detection. A
similar set up is possible on other experimental apparatus such as EHPC or high pressure burners.

Figure 16 : Schematic of the set up for LII combined with scattering [1]

Final Project : Characterization of Diesel Sprays IV-21
5.4 Results
As previously presented, the LII technique can provide two dimensional measurements of
soot volume fraction f
v
(quantitative when combined with extinction measurements), primary
particle diameter D (with TIRE technique or combined with scattering) and number density C
n

(with scattering). Typical results for these different techniques are shown in Figure 17, Figure 18
and Figure 19. However, LII cannot provide information on aggregate size distribution, which are
far more important quantities, although some studies have been performed on this subject [25].



Figure 17 : Spatial distributions for f
v
, D and C
n

results with combined LII and scattering [1]

Figure 18 : Spatial distribution for soot volume
fraction fv results with combined LII
and extinction [19]

Figure 19 : Spatial distribution for particle size D
results with TIRE LII [19]
Although acceptable quantitative soot volume fraction results were obtained in the premixed
burn phase in diesel combustion, it seems that the results of LII become irregular in the diffusion
combustion phase [14,24]. Furthermore, since the LII and extinction signals depend on the
unknown refractive index m, the results accuracy might be influenced by the choice of m, which
anyway changes with the experimental conditions. Another factor creating inaccuracy is that the LII
signal is not strictly proportional to the soot volume fraction because of the dependence on D at a
power slightly superior to 3.

The TIRE LII technique requires a good knowledge of ambient temperature, which is
experimentally obtained only with a very careful calibration of the thermal radiation results [5,22].
An overestimation of temperature could lead to an underestimation of D (for example up to 20% for
a 200 K difference). When using a cooling model, uncertainties and imperfections in this model
would also affect the results. Furthermore, if the as long as possible decay time is not taken into
account in the decay curve, the decay time is underestimated and an overestimation of D up to 25%
could occur [19]. These effects of estimated temperature and decay time on the determination of
particle diameter D are illustrated by Figure .

Final Project : Characterization of Diesel Sprays IV-22
6. Discussion
6.1 Summary
The two-colour method seems to be the best technique to quantitatively determine the soot
temperature field with an acceptable accuracy depending on the knowledge of optical parameters
(m, , L). The results of soot volume fraction f
v
are not accurate enough to provide reliable values
due to the uncertainties on optical parameters, though the KL factor can be determined with the
same order of uncertainty than temperature. Non uniformities in soot concentration and
temperatures affect slightly the measured temperature and severely the KL factor.

The light extinction method is a widely used technique to obtain average soot volume
fraction f
v
(or directly the KL factor) for soot diagnostic. A two-dimensional visualization of the
soot spatial distribution is also possible. The basic experimental set up and its calibration are
simple, but some optical devices have to be added in order to deal with several interferences
(Shadowgraph and Schlieren effects) in the transmitted signal, which are sometimes neglected for
ease of use but could be accentuated with high pressure.

The combination of light scattering and extinction has been commonly used to determine the
particle diameter D as well as soot volume fraction f
v
. However, its use in diesel engines seems to
be limited and would be more appropriate for combustion vessels. The combination of light
scattering and thermal radiation does not seem to be widely used, though it provides the same
informations and the soot temperature T in addition. However, it suffers from higher uncertainties
than the first combination due to the required path length determination.

The recent LII technique has been proven to be the most effective way to visualize soot
particles in two dimensions. It is a useful diagnostic tool for spatially and temporally resolved
measurement of soot volume fraction f
v
and primary particle size D (when combined with scattering
or with the TIRE method) in a wide range of applications. When performing TIRE, the temperature
T can also be determined with thermal radiation or by using a cooling model. The LII technique is
widely used for two dimensional imaging of soot volume fraction. The set up is not too
complicated, even in combination with scattering or extinction.



Final Project : Characterization of Diesel Sprays IV-23
6.2 Recommendations
It should be reminded that for the soot diagnostic optical methods presented in this report,
the range of use (pressure and temperature, ambient density, type of set up, type of fuel) may be
relatively small, and complications in the set up or in the analysis of results could occur.
For applications in high pressure such as engines, high pressure burners or EHPC, the choice
of a technique strongly depends on which parameter is investigated. The results presented in this
report are summarized in Table 5.

Size D Concentration f
v
Temperature T 2-D visualization
Thermal Radiation (2 colours)


Light Extinction




LES



+ extinction


Light Elastic
Scattering
(LES)
+ radiation


LII



+ extinction



+ scattering



Laser Induced
Incandescence
(LII)
TIRE

= possible and used, = difficult or not used
Table 5 : Comparative overview of optical techniques for soot diagnostic
The thermal radiation and the light extinction techniques seem to be appropriate only for a
first approach, because it is not used for two dimensional imaging, and cannot provide any result for
soot particle size. However, for temperature measurements, the thermal radiation is the best
technique despite its relative inaccuracy.
For particle sizing, combining scattering and extinction or LII and scattering seems both
appropriate, although the use of extinction with scattering does not allow a two dimensional
visualization. However, no information is available on aggregate sizes.

It seems that the LII technique, combined with scattering or calibrated with extinction, might
be the more complete technique for soot diagnostic regarding a two dimensional analysis of particle
size and distribution. However, for a simple global diagnostic, light extinction may be enough and
would provide informations about soot distribution.


Final Project : Characterization of Diesel Sprays IV-24
Tables
Table 1 : Refractive index variation with wavelength ......................................................................... 7
Table 2 : Comparative overview of optical techniques for soot diagnostic....................................... 23

Figures
Figure 1 : Schematic of high pressure combustion chamber for the two-colour method.................... 7
Figure 2 : Schematic of the optical set up for the two-colour method................................................. 7
Figure 3 : Results of soot volume fraction measurements with the two-colour method ..................... 8
Figure 4 : Results of temperature measurements with the two-colour method ................................... 8
Figure 5 : KL and luminosity radial profiles at a constant axial distance from the injector.............. 10
Figure 6 : Soot path length L axial profile......................................................................................... 10
Figure 7 : Schematic of the optical set up for both light extinction and 2-D visualization ............... 11
Figure 8 : Schematic of a light extinction set up for two-dimensional visualization......................... 11
Figure 9 : Soot volume fraction f
v
radial profile obtained from a three-point Abel inversion .......... 12
Figure 10 : Path-length-averaged soot volume fraction
v
f axial profile............................................ 12
Figure 11 : Schematic of a combined scattering/extinction set up .................................................... 15
Figure 12 : Temporal results from scattering/extinction and scattering/radiation techniques........... 16
Figure 13 : Spatial result from scattering experiment (16,5 ATDC) ................................................. 16
Figure 14 : Influence of primary soot particle size D on the LII signal decay time .......................... 18
Figure 15 : Schematic of the set up for LII combined with extinction.............................................. 20
Figure 16 : Schematic of the set up for LII combined with scattering .............................................. 20
Figure 17 : Spatial distributions for f
v
, D and C
n
results with combined LII and scattering ............. 21
Figure 18 : Spatial distribution for soot volume fraction fv with combined LII and extinction....... 21
Figure 19 : Spatial distribution for particle size D results with TIRE LII ......................................... 21












Final Project : Characterization of Diesel Sprays IV-25
References
[1] Zhao H. and Ladommatos N., Optical Diagnostics for Soot and Temperature Measurement
in Diesel Engines, Prog. Energy Sci., Vol. 24, pp. 221-225, 1998.

[2] Hottel H. and Broughton F., Ing. and Eng. Chem., p. 166, 1932.

[3] Shaddix C.R. and Smyth K.C., Combustion and Flame 107, pp. 314320, 1996.

[4] Chang H. and Charalampopoulos T., Proc. R. Soc. Lond. A 430, pp. 577-591, 1990.

[5] Schraml S., Dankers S., Bader K., Will S. and Leipertz A., Soot Temperature Measurements
and Implications for Time-Resolved Laser-Induced Incandescence (TIRE-LII), Combustion
and Flame 120, pp. 439-450, 2000.

[6] Bento D.S., Thomson K.A., Glder .L., Soot formation and temperature field structure in
laminar propaneair diffusion flames at elevated pressures, Combustion and Flame, 2006.

[7] Thomson K.A., Glder .L., Weckman E.J., Roydon A. Fraser a, Greg J. Smallwood G.J.
and Snelling D.R., Soot concentration and temperature measurements in co-annular,
nonpremixed CH4/air laminar flames at pressures up to 4 MPa, Combustion and Flame
140, pp. 222232, 2005.

[8] Krishnan S.S., Lin K.-C. and Faeth G.M., Journal of Heat Transfer 122, pp. 517524, 2000.

[9] Kee S.-S., Mohammadi A., Hirano H., Kidoguchi Y. and Miwa K.,
SAE Paper 2003-01-1792, 2003.

[10] Krishnan S.S. and Faeth G.M., ASME Journal of Heat Transfer 123, pp. 331339, 2001.

[11] Koylu U.O. and Faeth G.M., Journal of Heat Transfer, p. 152, 1994.

[12] Yang B. and Koylu U.O., Detailed soot field in a turbulent non-premixed ethylene/air flame
from laser scattering and extinction experiments, Combustion and Flame 141, pp. 5565,
2005.

[13] Musculus M.P.B. and Pickett L.M., Diagnostic considerations for optical laser-extinction
measurements of soot in high-pressure transient combustion environments, Combustion and
Flame 141, pp. 371391, 2005.

[14] Pickett L.M. and Siebers D.L., Soot in diesel fuel jets : effects of ambient temperature,
ambient density and injection pressure, Combustion and Flame 138, pp. 114135, 2004.

[15] Smooke M.D., Long M.B, Connelly B.C., Colket M.B. and Hall R.J., Soot formation in
laminar diffusion flames, Combustion and Flame 143, pp. 613628, 2005.





Final Project : Characterization of Diesel Sprays IV-26
[16] Zhou Z. and Choi M.Y., Measurement of Dimensionless Extinction Constant of Soot
Generated Using Various Fuels, Proceedings Combustion Fundamentals and Applications,
April 23-26, 1995

[17] Collin R., Laser Diagnostics for Applications to In-Cylinder Engine Investigations, Doctoral
Thesis, Lund Institute of Technology, 2004.

[18] Tree D. and Foster D., SAE Paper 940270, 1994.

[19] Axelsson B., Collin R. and Bengtsson P-E., Laser-induced incandescence for soot particle
size measurements using on-line extinction calibration, Appl. Opt. B 72, pp. 367-372, 2001.

[20] Juhlin G., Development and Application of Laser Diagnostics for Studies of Phenomena
Related to IC Engine Combustion, Doctoral Thesis, Lund Institute of Technology, 2003.

[21] Melton L.A., Soot Diagnostics based on Laser Heating, Appl. Opt. 23, pp. 2201-2208, 1984.

[22] Green R.M. and Witze P.O., Laser-induced incandescence and elastic-scattering
measurements of particulate-matter volume fraction changes during passage through a
dilution tunnel, Proc. 10th Int. Symposium. on Appl. of Laser Tech. to Fluid Mech., 2000.

[23] Axelsson B., Collin R. and Bengtsson P-E., Laser-induced incandescence for soot particle
size measurements in premixed flat flames, Appl. Opt. 21, Vol. 39, pp. 367-372, 2000.

[24] Choi D., Iwamuro M., Shima Y., Senda J. and Fujimoto H., The Effect of Fuel-Vapor
Concentration on the Process of Initial Combustion and Soot Formation in a DI Diesel
Engine Using LII and LIEF, SAE Paper 2001-01-1255, 2001.

[25] Liu F., Smallwood G.J., Snelling D.R., Effects of primary particle diameter and aggregate
size distribution on the temperature of soot particles heated by pulsed lasers, Journal of
Quantitative Spectroscopy & Radiative Transfer 93, pp. 301312, 2005.