Sunteți pe pagina 1din 10

Development and application of the drop number size moment modelling

to spray combustion simulations


I. Dhuchakallaya
*
, A.P. Watkins
Energy, Environment and Climate Change Research Group, School of Mechanical, Aerospace and Civil Engineering, University of Manchester, Manchester M1 7AP, United Kingdom
a r t i c l e i n f o
Article history:
Received 4 March 2009
Accepted 5 February 2010
Available online 10 February 2010
Keywords:
Spray
Combustion
Modelling
a b s t r a c t
This work presents the development and implementation of spray combustion modelling based on the
spray size distribution moments. In this spray model, the droplet size distribution of spray is character-
ised by the rst four moments related to number, radius, surface-area and volume of droplets, respec-
tively. The governing equations for gas phase and liquid phase employed are solved by the nite
volume method based on an Eulerian framework. These constructed equations and source terms are
derived based on the moment-average quantities which are the key concept for this work. The sub-model
employed for ignition and combustion is the coupling reaction rate between Arrhenius model and Eddy
Break-Up model (EBU) via a reaction progress variable. The results obtained from simulation are com-
pared with the experimental and simulation data in the literature in order to assess the accuracy of pres-
ent model. Comparing with the experimental results, present approach is capable to provide a
qualitatively reasonable prediction for auto-ignition. In addition, the ame area developed during the
combustion progress corresponds with the experimental data.
2010 Elsevier Ltd. All rights reserved.
1. Introduction
Spray combustion is important in many practical power gener-
ation devices such as industrial boilers, aircraft combustors, and
conventional engines. Understanding the details of the spray com-
bustion process including spray formation, spray structure, droplet
dynamics, heat and mass transfer, auto-ignition, and ame propa-
gation is essential in the design and performance prediction of
such applications. Spray combustion is much more complex than
the homogeneous combustion process due to variation of combus-
tible mixture at each position. Hence the combustion process de-
pends not only on fuel chemistry but also that of spray physics [1].
In the combustion process of spray in engines, the high-pres-
sure liquid fuel is injected into the combustion chamber occupied
by high-pressure and high-temperature air. The high-pressure of
injection disturbs the liquid jet surface and leads to breakup into
smaller droplets and then evaporation takes place. In the experi-
mental results, ignition generally occurs somewhere in the down-
stream portion of the vaporised cloud. This is the onset of
combustion. Normally, the models for the simulation of internal
combustion engine combustion are complicated due to the follow-
ing reasons. Firstly, the pressure present in combustion chamber is
very high and varies with time due to heat released. Secondly, the
interactions between two phases, liquid fuel and the gaseous
phase, requires closure models to perform accurate prediction. In
addition, the complexity of turbulence in the gaseous phase which
plays the essential role in mixing and combustion also raises dif-
culties. Finally, the chemical kinetic mechanisms of fuel that con-
sist of hundreds of species and thousands of chemical reactions
are the key factors in auto-ignition and the combustion model.
Hence combustion requires large computational resource to
simulate.
In models to predict the chemical kinetic mechanisms, there are
four main groups employed including detailed mechanism,
reduced mechanism, skeletal mechanism, and global mechanism.
A detailed mechanism employs all relevant reactions and species,
however, its application to engine models is limited due to lack
of veried mechanisms for the high hydrocarbon fuels and very
large computational resource requirements. A reduced mechanism
is deduced by approximation of detailed mechanisms by means of
sensitivity analysis and steady-state analysis without addition of
generic species or global reactions. One well-known reduced
mechanism model is the Shell model [2] developed in order to
predict knock in spark ignition engines. It has been adopted suc-
cessfully in CFD simulation for diesel engines [37] in which
n-heptane is selected as a representative of diesel fuel due to com-
parable octane number. However, Cox and Cole [8] and Hu and
Keck [9] argued that the Shell model does not adequately describe
the ignition phenomena which take place at low-temperatures. To
improve the prediction capability of this model, the number of
reactions should increase by introducing the chemistry controlling
1359-4311/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2010.02.002
* Corresponding author.
E-mail address: dhuchkallaya@yahoo.com (I. Dhuchakallaya).
Applied Thermal Engineering 30 (2010) 12151224
Contents lists available at ScienceDirect
Applied Thermal Engineering
j our nal homepage: www. el sevi er . com/ l ocat e/ apt her meng
the hydrocarbon oxidation process at low-temperatures. Other re-
duced mechanism models have been employed fruitfully in en-
gines, for examples, MIT model [10], and Schreibers model [11].
However, the reduced mechanism scheme also has a limitation
to describe the ignition process of higher hydrocarbons due to
complexity of their mechanisms. A skeletal mechanism employs
a combination of elementary and overall steps including both indi-
vidual and generic species. These skeletal mechanisms select only
the critical reactions and species that affect the auto-ignition pro-
duction, and they are easily adjusted to the engine model.
Although this model is less popular than reduced reaction scheme,
it has been applied successfully in many numerical modelling stud-
ies [1214]. Based on experimental data, global mechanism, the
simplest form for predicting the ignition, employs empirical rela-
tionships between fuel and physical environments to generate
the single-step chemical equation mostly presented in the Arrhe-
nius form. Due to the inexpensiveness of computer capacity and
simplicity for modelling, this scheme is extremely attractive in
CFD calculations, as appeared in many reports [1517], but it is
essential to realise that the constants have no evident physical
meaning and are usually obtained by matching experimental re-
sults. Due to representation of the combustion process by a single
overall reaction, the information about the intermediate products
which can be pollutant precursors cannot be calculated.
Both premixed and diffusion ame theories are involved in the
spray combustion. There is no distinctive boundary between pre-
mixed and diffusion combustion, and then both premixed and dif-
fusion ame theories can be applied to spray combustion problems
[18]. However, in the spray combustion processes, the reaction rate
mainly depends on diffusion rate of fuel vapour and oxidizer spe-
cies. Furthermore, the complexity of models multiplies signi-
cantly as the turbulence in the ame sheet becomes dominant.
The detailed chemical kinetics is then required to interact with
the complicated turbulent combustible gases in order to describe
a full range of reaction rates in combustion process. A crucial ques-
tion related to this complex turbulencechemistry interaction is
how to formulate and how to close the mean turbulent reaction
rates. The most rigorous approach is that of using the probability
density function (PDF) transport equation model because the
chemical reaction source terms of each species can be derived di-
rectly from kinetics theories without any modelling and they can
produce the nearly perfect closure of the reaction source terms
as well. Practically, it is extremely difcult to presume the shape
of a joint PDF depending on more than two variables and the com-
plexity of problem is also multiplied as the number of involved
species increases. Hence, this approach can be employed success-
fully with the lower hydrocarbon fuels using a Monte-Carlo solving
schemes [19,20]. Based on conditional moment closure, a compu-
tationally simpler approach is to solve transport equations for
reactive species mass fractions which are conditionally averaged
on a conserved scalar. Successful investigations of CMC in a die-
sel-like environment have been reported [21,22].
In order to reduce the computational cost, the assumption that
the instantaneous thermochemical state of the uid is related to a
conserved scalar quantity known as the mixture fraction is applied.
In other words, the local mass fraction of all the species (CO
2
,
H
2
O, O
2
, etc.) is a function of mixture fraction. Hence the transport
equations for a conserved scalar are required while the solutions of
a large number of species transport equations are not required. In
order to calculate the mean value of thermochemical variables
(temperature, density and species mass fractions), the assumed
shape probability density function (PDF) approach is applied. In
addition, it allows intermediate (radical) species prediction, disso-
ciation effects, and rigorous turbulencechemistry coupling. The
chemistry can be modelled as in mixed-is-burned, chemical equi-
librium, or near chemical equilibrium (laminar amelet) model.
The assumption of mixed-is-burned model is that the chemistry
is innitely fast and irreversible. Based on work of Burke and Schu-
mann [23], fuel and oxidant species never coexist in space and
chemical reaction also convert completely to nal products with
no reaction rate or chemical equilibrium information required.
Hence, it cannot predict intermediate species formation or dissoci-
ation effects and ame temperature results are often overpre-
dicted. In an equilibrium model, the chemistry is assumed to be
Nomenclature
B
Q2
source term for Q
2
due to break-up
B
M
mass transfer number
c reaction progress variable
C
D
drag coefcient
C
Q2
source term for Q
2
due to collision
D molar diffusivity
E
a
activation energy
h
0
f
enthalpy
k turbulent kinetic energy, thermal conductivity
L latent heat of vaporisation
Le Lewis number
n(r) number size distribution
P
k
production of turbulence K.E.
Q
0
total drop number
Q1 sum of drop radii
Q
2
sum of squares of drop radii
Q
3
sum of cubes of drop radii
r radius
s stoichiometric mass ratio
S
m
source term due to mass transferred by evaporation
S
Q2
source term due to evaporation
Sc Schmidt number
Sh Sherwood number
t time
T temperature
U velocity
Y species mass fraction
Greek symbols
e dissipation rate
l
eff
effective turbulent viscosity
m kinetic viscosity
h void fraction
r turbulent Prandtl/Schmidt number
r
m
turbulence damping coefcient
_ x reaction rate
Subscripts
2 based on surface-area-averaged
3 based on volume-averaged
32 Sauter mean
b bag break-up
F fuel
g gas
j component
l liquid
rel relative
s drop surface, stripping break-up
1216 I. Dhuchakallaya, A.P. Watkins / Applied Thermal Engineering 30 (2010) 12151224
rapid enough for chemical equilibrium. The equilibrium mass frac-
tion of each species related to mixture fraction is calculated based
on the minimization of Gibbs free energy. Then it can predict infor-
mation of intermediate species without knowledge of detailed
chemical kinetics. This approach is widely adopted in spray com-
bustion models [2426] due to its inexpensive computational re-
source and providing intermediate species details. In the amelet
model, thin diffusion layers embedded in a turbulent non-reactive
ow eld are assumed. If the chemistry is fast enough, the chem-
istry is active within a thin region where the chemistry conditions
are in (or close to) stoichiometric conditions. This thin region is as-
sumed to be smaller than Kolmogorov length scale and therefore
the region is locally laminar. It describes all chemical reactions
and molecular diffusion by means of a laminar ame structure
approximation. Thus, this model is simplied to determine the
ame surface density with the detailed chemistry data obtained
from a amelet library. Hence, the mean value of thermochemical
variables is completely described by the two quantities, mixture
fraction and scalar dissipation. The advantage of this model is that
realistic chemical kinetic effects can be incorporated into turbulent
ames. Therefore this model is far more popular than the two
models above. However it consumes more computational cost.
Successful use of amelet model in a spray combustion have been
reported in [27,28]. Traditionally, engine combustion studies have
preferred to use some simplied models for the mean reaction rate
closure. The most commonly used ones are the Eddy Break-Up
(EBU) model [29]. In these models, it is assumed that the chemistry
is much faster than the mixing, thus the rate controlling phenom-
ena is the turbulent mixing. The reaction rates appearing as source
terms in the species transport equations can be obtained from the
Arrhenius rate and the eddy break-up rate. Regardless of detailed
chemical kinetics, the intermediate species information cannot
be predicted and the ame temperature results are often overpre-
dicted. However, due to inexpensive computational cost and its
simplicity, this approach has been employed successfully in turbu-
lent combustion simulations [3032].
The principle objective of present work is to develop a simula-
tion model for a spray combustion based on the spray number size
distribution moments introduced by Beck [33]. This model has
been presented in earlier publications [3439]. In order to intro-
duce the capability of this spray model to model combustion, this
paper reports numerical investigations of diffusion ames in the
main combustion and premixed ames in the ignition phase using
the Eddy Break-Up Model, and a RANS approach to simulate the
turbulence. In the developing process, this model mainly focuses
on the combustion process of diesel fuel which is widely used in
compression ignition engines. In order to verify the validity of
the model, experimental data of diesel spray combustion and pre-
diction from other models are compared with the present model
results.
2. Mathematical model
The governing equations for liquid phase of the droplet size dis-
tribution of spray are characterised by the rst four moments. Q
0
is
the total number of drops present, Q
1
is the total sum of radii of the
drops, 4pQ
2
is the total surface-area of the drops and 4pQ
3
/3 is the
total volume of the drops which, assuming that locally all drops to
have the same liquid density, also denes the mass of liquid pres-
ent per unit volume. These constructed equations are based on the
moment-average quantities which is the key concept for this work
that there is no the segregation of the droplets into groups of equal
radius. Briey, the volume and surface area moments at any point
in the spray varied in space and time are calculated by solving
transport equations both for the moments themselves and also
for the appropriate moment-averaged convection velocities. Once
values of Q
3
and Q
2
are known, the local Sauter mean radius
(SMR), r
32
is evaluated from Q
3
/Q
2
. The presumed droplet size dis-
tribution given by
nr
16r
r
2
32
exp
4r
r
32
_ _
1
is then truncated at one end to t the SMR. Values of Q
1
and Q
0
are
then evaluated from the truncated distribution.
2.1. Transport equations
The transport equation for the fourth droplet moment is effec-
tively a liquid phase continuity equation. According to the deni-
tion of the spray moments and the moment-averaged quantities,
it can be written as
4p
3
@
@t
q
l
Q
3

4p
3
@
@x
j
q
l
Q
3
U
l3j
S
m
2
where S
m
is the source term of mass transferred by evaporation
from droplets to the gas phase.
The equations for the remaining moments can be written in a
similar manner, but the source terms are different. The transport
equation of the Q
2
moment of the spray is,
@
@t
Q
2

@
@x
j
Q
2
U
l2j
B
Q2
C
Q2
S
Q2
3
where B
Q2
, C
Q2
and S
Q2
are the source terms due to droplet breakup,
dropletdroplet collisions and droplet evaporation, respectively.
To evaluate the liquid mass-average velocity, the liquid phase
momentum equation is derived based on the work of Harlow and
Amsden [40] for particulate ows as
@
@t
q
l
1 hU
l3i

@
@x
j
q
l
1 hU
l3i
U
l3j
U
l3i
S
m

@
@x
j
q
l
1 hr
v
m
l
@U
l3i
@x
j
_ _
S
U
3j
4
where r
v
is the coefcient of Melville and Bray [41], m
l
is the turbu-
lent equivalent viscosity and S
U
3j
is the source term of momentum.
The remaining equation for the moment-average velocity is de-
rived in a similar manner as
@
@t
Q
2
U
l2j

@
@x
j
Q
2
U
l2i
U
l2j
U
l3i
U
l2i
B
Q2

@
@x
j
Q
2
r
v
m
l
@U
l2i
@x
j
_ _
S
U
2j
U
l2i
S
Q2
5
The liquid phase energy equation can be expressed as
@
@t
q
l
1 hh
l

@
@x
j
q
l
1 hU
l3j
h
l

@
@x
j
q
l
1 hr
v
m
l
@h
l
@x
j
_ _
S
E
6
where S
E
is the inter-phase energy transfer.
As in this work ows with relatively high momentum will be
analysed, the effect of gravity can be neglected. Hence the gaseous
conservation of momentum equation, including turbulence effects
can be written as,
@
@t
hq
g
U
gi

@
@x
j
hq
g
U
gj
U
gi
U
gi
@
@t
hq
g

@
@x
j
hq
g
U
gj

_ _

@
@x
j
l
eff
h
@U
gi
@x
j

@U
gj
@x
i
_ _ _ _
h
@P
@x
i

@
@x
j
2
3
hq
g
kd
ij

2
3
l
eff
@U
k
@x
k
_ _
S
m
U
l3i
U
gi
S
U
j
7
I. Dhuchakallaya, A.P. Watkins / Applied Thermal Engineering 30 (2010) 12151224 1217
where d
ij
is Kronecker delta (e.g. d
ij
= 1 if i = j; and d
ij
= 0 if i j).
The source term S
U
j
represents the momentum exchange between
the liquid and the gas phases per unit time in a control volume.
In single phase ows, the third term on the left hand side equals
to zero, as it is a multiple of the continuity equation [42]. In our
case, due to inter-phase mass transfer, this can be non-zero
and therefore it should be included. The penultimate term is de-
scribed by Mostafa and Mongia [43] as the momentum growth
term, and it results from the initial relative velocity between
the generated vapour (initially travelling at liquid velocity) and
the carrier gas.
Although in the momentum conservation equation no explicit
combustion term appears, the chemistry still has a strong inuence
on the ow behaviour. Due to the thermal energy liberated along
the ame front, the dynamic viscosity l, the density q and as a
consequence the velocity U
gi
change dramatically in comparison
with cold ows.
Similarly, the energy transport equation for the gas phase
including energy transfer due to mass transfer and heat release
due to reaction can be written as
@
@t
hq
g
h
g

@
@x
j
hq
g
U
gj
h
g

@
@x
j
l
eff
r
t
h
@h
g
@x
j
_ _ _ _
h
@
@x
j
PU
gj

S
E

_
x
F
Q
F
8
where Q
F
is the amount of heat released by the combustion of an
unit mass of fuel and _ x
F
is the mean reaction rate of fuel. The liquid
mean temperatures employed in this work developed by Watkins
[39] are the volume-averaged and surface-area-averaged tempera-
tures. In the transport equation for liquid phase energy, the liquid
energy is represented by a volume-averaged temperature. In practi-
cal, a surface-averaged-temperature which is normally higher than
volume-averaged temperature relates directly to heat and mass
transfer phenomena.
When the liquid fuel droplets evaporate, the fuel droplets will
transform in phase from liquid to gas and will be transported to-
gether with the rest of the ambient gas. Consequently, the mixture
of fuel and oxygen is consumed during combustion resulting in
heat release and product gases. Hence it is necessary to determine
the species transport equations of both reactants and products as
presented below
@
@t
hq
g
Y
k

@
@x
j
hq
g
Y
k
U
gj

@
@x
j
l
eff
r
Y
h
@Y
k
@x
j
_ _ _ _
S
m

_
x
k
9
where S
m
is the mass transfer source term for fuel vapour species
only. If species is not fuel vapour, this term is vanished. _ x
k
is the
mean reaction rate of species k.
The ke turbulence model of Launder and Spalding [44] is used
here. This model solves transport equations for the turbulence ki-
netic energy and its dissipation rate. The transport equation for the
turbulent kinetic energy is expressed as:
@
@t
hq
g
k
@
@x
j
hq
g
U
gj
k
@
@x
j
l
eff
r
k
h
@k
@x
j
_ _ _ _
hP
k
q
g
eh 10
whereas the dissipation rate transport equation is given as:
@
@t
hq
g
e
@
@x
j
hq
g
U
gj
e
@
@x
j
l
eff
r
e
h
@e
@x
j
_ _ _ _
hC
e1
P
k
e
k
hC
e2
q
g
e
2
k
hC
e3
q
g
e
@U
gj
@x
j
11
The turbulence kinetic energy production rate is given as:
P
k
q
g
C
l
k
2
e
@U
gi
@x
j

@U
gj
@x
i
_ _
@U
gj
@x
i

2
3
d
ij
q
g
k q
g
C
l
k
2
e
@U
gk
@x
k
_ _
@U
gj
@x
i
12
The constants take the values C
e1
= 1.44, C
e2
= 1.92, C
e3
= 0.373,
r
t
= 1.0, r
e
= 1.3. All the source terms are calculated by considering
the effect of the gas phase on the liquid phase in terms of the drop-
let size distribution moments. The details of source terms are de-
rived and discussed in [45], thus here the derivation of these
source terms will not be reproduced.
2.2. Spray sub-models
The evaporation of droplets is caused by heat and mass transfer
between the droplets and the surrounding gas. The mass transfer
source term in terms of the droplet moments can be derived as
S
m
2pq
g
D
g
ln1 B
M
Q
1
Sh
Q
13
where the Sherwood and Reynolds numbers employed here are
modied as
Sh
Q
2 0:6Re
0:5
Q
Sc
0:333
Re
Q

2q
g
U
l2
U
g

l
g
Q
2
Q
1
_ _
14
The source term for Q
2
due to mass evaporation is given by
S
Q
2

q
g
D
g
q
l
ln1 B
M
Sh
Q0
Q
0
15
where
Sh
Q0
2 0:6Re
0:5
Q0
Sc
0:333
Re
Q0

2q
g
U
l2
U
g

l
g
Q
1
Q
0
_ _
16
The source term for liquid energy transport equation is given by
S
E
2pQ
1
k
g
T
g
T
l2

ln1B
M

1=Le
1B
M

1=Le
1
Nu
Q
q
g
D
g
ln1B
M
Sh
Q
L
_ _
17
where the mass transfer number, B
M
, is dened in terms of fuel va-
pour mass fraction (B
M
= (Y
F,s
Y
F
)/(1 Y
F,s
)).
In most research, the types of droplet instability can be classi-
ed into two modes including bag breakup and stripping breakup
modes. These are straightforward in nature and have been widely
adopted in many Discrete Droplet Model codes. According to the
correlation of Reitz and Diwakar [46], the unstable droplets are
considered to be undergoing breakup when
We
q
g
U
2
rel
d
r
l
> 12:0 18
for bag breakup, and the critical radius can be evaluated as
r
b

6r
l
q
g
U
2
rel
19
In the same manner, the criterion for stripping breakup mode is
We

Re
p > 0:5 20
and the critical radius for stripping breakup type can be calculated
as
1218 I. Dhuchakallaya, A.P. Watkins / Applied Thermal Engineering 30 (2010) 12151224
r
s

r
2
l
2q
g
U
3
rel
l
g
21
In order to break up, unstable droplets are required to reach not
only the critical radius but also the critical lifetime. Corresponding
to Reitz and Diwakar [46], the critical lifetimes for an unstable drop-
let to break up as given for each breakup regime can then be eval-
uated as
s
b

p
4

q
l
d
3
r
l

s
s

d
2U
rel

q
l
q
g

22
According to Faeth et al. [47], the outcome diameter of the droplet
breakup should be taken as a resultant Sauter Mean Diameter
rather than the exact diameter of all droplets. This assumption is
fairly reasonable as discussed in Beck and Watkins [36]. The corre-
lation for stripping breakup mode can be rearranged as
r
32;out
6:2
q
l
q
g
_ _
0:25
l
l
2q
l
U
rel
_ _
0:5
r
0:5
in
23
Assuming, the volume of droplet is preserved in the breakup
process that means no change in density and liquid mass. Hence
the change in the third droplet moment due to the breakup of a
single drop is given by
Q
2;out
Q
2;in

r
3
in
6:2
q
l
q
g
_ _
0:25
l
l
2q
l
U
rel
_ _
0:5
r
0:5
in

r
3
in
r
in
24
The change in the surface area per unit time for a single droplet
is given by dividing by the appropriate lifetime for an unstable
droplet. This equation is then integrated over all drops undergoing
stripping breaking up in a control volume to give the source term
as
B
Q2;s

Q
1;s
Q
2;s

0:5
6:2
q
l
q
g
_ _
0:25
l
l
2q
l
U
rel
_ _
0:5
1
U
rel
q
l
q
g
_ _
0:5

Q
1;s
1
U
rel
q
l
q
g
_ _
0:5
25
where Q
1,s
and Q
2,s
are the total sum of radii and total sum of
squares of radii undergoing stripping breakup. In this analysis, the
non-integer moment Q
1.5,s
is approximated as Q
1;s
Q
2;s

0:5
.
In the case of bag break-up, there is no appropriate correlation
found in the literature, and the surface area is assumed to double in
the breakup process, which is equivalent to the production of eight
equally sized droplets. Thus
dQ
2
dt

r
0:5
in
p
q
l
2r
l
_ _
0:5
26
Summing this over the appropriate droplets, and geometrically
interpolating between the surrounding moments to integrate over
non-integer powers of the radius, the source term becomes
B
Q2;b

Q
0;b
Q
1;b

0:5
p
q
l
2r
l
_ _
0:5
27
where Q
1,b
and Q
0,b
represent the total sum of radii and total num-
ber of the droplets undergoing bag breakup.
These two contributions to the change in the surface-area drop-
let distribution moment are then summed to give the total change
in the moment equation source term due to breakup.
In the implementation of the droplet collisions based on a semi-
empirical model, there are three stages to accomplish. Firstly, the
number of collisions between droplets occurring in any control
volume is determined by using the collision frequency concept of
ORourke and Bracco [48]. Their expression for the probability of
a collision between two drops per unit volume per unit time, pro-
viding they are in the same control volume, is adapted to the new
model by multiplying this collision probability by the appropriate
number distributions and integrating over all drops. The second
stage of the model determines how many of these collisions result
in each of the regimes of coalescence, bounce and separation, de-
scribed by Orme [49]. The two parameters required to determine
these proportions are the Weber number and the impact parame-
ter. Chart of the different collision regimes as functions of Weber
number and impact parameter proposed by Qian and Law [50] is
used to provide the probabilities of each of the possible outcomes.
The critical Weber numbers shown on the charts are translated
into critical radii and the number size distribution is used to deter-
mine the probability that any given drop lies between adjacent
critical radii. The outcome of a collision is decided by the Weber
number of the smaller drop, according to Orme, and the impact
parameter. The nal stage of the collisions model is to determine
the effect of the predicted collisions on the drop surface area mo-
ment of the size distribution, as the liquid volume is conserved
during collisions. The surface area change is approximated by that
obtained from a collision between two drops of equal radius and to
result in either one (coalescence) or ve (separation) drops, also of
equal radius, such that the total drop volume is conserved.
The drag force is caused by the motion of a particle through
the ambient gas. Thus, droplet drag effects acceleration of droplets.
The terms of pressure, gravity, added-mass and Basset forces in the
droplet equation are not taken into account due to their complex-
ity and uncertain signicance [51]. Thus, based on the steady drag
coefcient, C
D
, given by the correlation of Wallis [52] for a solid
sphere, the liquid momentum transfer source terms due to drag
can be derived as follows
S
U
3j
6pl
g
U
rel;j
Q
1
1:8pq
g
jU
rel
jQ
2

0:687
l
g
Q
1
2
_ _
0:313
U
rel;j
28
and
S
U
2j

9
2
Q
0
q
l
U
rel;j
l
g

1:35
q
l
q
g
jU
rel
jQ
1

0:687
l
g
Q
0
2
_ _0:313
U
rel;j
29
2.3. Combustion model
In detailed chemical mechanism reaction, thousands of reac-
tions and hundreds of species are involved which is an extremely
complex and costly task. Due to numerical limitations in comput-
ing time and storage capacity, the irreversible fast single-step reac-
tion is employed in order to simplify the complex combustion. In
present work, the major objective of the combustion modelling
performed is to obtain the heat release. The details of the emissions
are not sought here. The chemical reaction rate usually in the form
of an Arrhenius equation, is
x
c
Aq
ab
g
Y
a
F
Y
b
O
expE
a
=R
u
T 30
According to Westbrook and Dryer [53], the experimentally
determined constants for diesel simulations are A = 3.2 10
11
,
a = 0.25, b = 1.50 and E
a
/R
u
= 15,100.
After the onset of combustion, the local gas temperature be-
comes very high. Consequently, chemistry grows very fast, and
the combustion is generally controlled by the turbulent mixing
rate rather than by the amount of mixture. Thus the absence of
fuelair mixing does not imply zero reaction rate. This situation
is not taken into account by the chemical reaction rate as men-
tioned above. In present work, the Eddy Break-Up model (EBU)
introduced by Magnussen and Hjertager [54] is employed. It
assumes innitely fast chemistry. Hence, the turbulent mixing
reaction source term can be obtained
I. Dhuchakallaya, A.P. Watkins / Applied Thermal Engineering 30 (2010) 12151224 1219
x
t
Bq
e
k
min Y
F
;
Y
O
s
; C
Y
P
1 s
_ _
31
where B and C are modelling constants which depend on both the
structure of the ame and the reaction between the fuel and the
oxygen and s is the stoichiometric oxygen to fuel mass ratio.
In practice, the chemical reaction rate is active before the ame
occurs. The turbulent mixing rate is very high, hence the overall
reaction rate is controlled by chemical reaction rate. Once the
ame is ignited, the Eddy Break-Up rate is generally smaller than
the Arrhenius rate, and reactions are mixing-limited. Conse-
quently, the chemical reaction rate and the turbulent mixing reac-
tion rate have to be coupled in some ways that guarantees a
continuous transition between premixed and non-premixed com-
bustion. The continuity of the transition is a necessary condition
to assure the physical meaning of the model results. In the com-
mercial software such as Fluent, STAR-CD, the total reaction rate
is basically dened as
x minx
c
; x
t
32
The reaction rate which is smaller controls the combustion rate.
Alternatively, the coupling reaction rate correlation proposed by
Pires da Cruz et al. [55] giving more continuity in transition is
x x
c
1 c x
t
c 33
where c is the reaction progress variable which has value between 0
(premixed ame) and 1 (diffusion ame). It refers to how much of
combustible mixture has already burned. Here, to simplify the anal-
ysis, the denition of the reaction progress variable based on the
assumption that the Lewis number is unity and constant heat
capacity and molecular weight can be dened as
c
T T
R
T
P
T
R
34
where subscripts R and P are represented to reactants and products,
respectively.
3. Computational solution scheme
In this section, the overview of numerical procedures is pre-
sented. The nite volume method is employed to carry out the
solution of the transport equation system in an Eulerian frame-
work. The temporal difference method is performed using the Eu-
ler implicit method while the spatial discretisation is implemented
using the hybrid scheme. The spray combustion studied in this pa-
per are axisymmetric, thus here all the conservation or transport
equations for both the liquid and gas phases are solved on the same
two-dimensional ( z, r) axisymmetric orthogonal computational
grid. A staggered grid arrangement is adopted for the liquid and
gas phase velocity components. Euler implicit temporal differenc-
ing and hybrid upwind/central spatial differencing are employed
to render all the liquid and gas phase transport equations into -
nite volume forms. Discussion of these processes and the algorithm
used can be found in Beck and Watkins [36]. For clarity the major
algorithm steps are reproduced here.
The solution algorithm is based on the PISO algorithm of Issa
[56], with the liquid phase equations added into it. The PISO algo-
rithm provides an efcient non-iterative solution procedure that
couples the gas-phase pressure and velocity components by an
operator splitting technique and solves the equations of motion
for the gas phase in a predictorcorrector fashion. The current
scheme solves the liquid equations only once, at the beginning of
the time step. The use of a non-iterative scheme implies that some
effects are lagged, i.e. carried forward from one time step into the
next. In the current approach, the drop break-up and collision
effects are calculated at the end of the time step, and the amended
source terms for the Q
2
equations and for the surface-area-aver-
aged momentum equations are therefore carried forward to the
beginning of the next time step.
The solution proceeds in the following manner at each time
step:
Step 1. The transport equations for the moments Q
2
and Q
3
of
the drop size distribution, are solved. The void fraction is
updated. Moments Q
0
and Q
1
, requiring approximation from
the assumed distribution, are calculated.
Step 2. The inter-phase drag source terms are evaluated.
Step 3. The transport equations for the moment-average liquid
velocities U
l2
and U
l3
, are solved.
Step 4. The gas phase velocity components are predicted using
the gas phase momentum equations.
Step 5. The rst set of pressure correction equations are solved
(see Issa [56] for details). The gas-phase pressures, densities and
velocity components are corrected.
Step 6. The transport equations of each species mass fraction are
solved.
Step 7. The gas phase energies in mixing and reactive zone are
evaluated.
Step 8. The second set of pressure correction equations is solved.
The gas-phase pressures, densities and velocity components are
corrected again.
Step 9. Transport equations are solved for the turbulence kinetic
energy and its dissipation rate. The equations are coupled and
iterated to convergence. The turbulent viscosity is recalculated.
Step 10. Source terms are calculated for the effects of the break-
up of unstable drops and collisions between drops on the Q
2
transport equations and on the surface-area-averaged momen-
tum equations.
The stability and robustness of this scheme has been demon-
strated by the ability of the method to obtain converged solutions
of the equations for all the test cases that have been attempted to
date. The accuracy of this spray model has been partially assessed
Table 1
The physical conditions of spray injection test cases.
Fuel Diesel (CN 55)
Time step (ls) 0.5
Injection cell of largest side (mm) 0.5
Cell grid (z r cells) 109 73
Inlet SMR (lm) 10.0
Trap pressure (MPa) 2.7
Maximum injection pressure (MPa) 80.0
Trap temperature (K) 830
Liquid temperature (K) 298
Injector diameter (mm) 0.18
Injection duration (ms) 3.8
Fuel injection quantity (mg) 28.0
X
Y
0 0.05 0.1 0.15 0.2
0
0.01
0.02
0.03
Fig. 1. Grid used for test case in the spray combustion. (The domain is 200 mm long
and 30 mm in radius with 109 73 cells.)
1220 I. Dhuchakallaya, A.P. Watkins / Applied Thermal Engineering 30 (2010) 12151224
in earlier publications [3439] by comparing with the experimen-
tal results for both non-evaporating and evaporating sprays. The
comparison results are in reasonably good agreement with exper-
imental data. Numerous grid and time step dependence tests have
been also carried out in the earlier publications. The results from
those tests have been used to set these parameters in this work.
4. Results and discussion
In order to assess the potential of this model for the application
in diesel spray combustion, the auto-ignition and combustion
model are veried. The combination of a single-step Arrhenius
model and the Eddy Break-Up model (EBU) via a reaction progress
variable is adopted in calculation the ignition and combustion peri-
ods due to its simplicity and inexpensive computational cost. The
experimental results from literatures are also required to compare.
The physical conditions are given in Table 1 for the test case based
on the experiments of Akiyama et al. [57]. High-pressure diesel fuel
was injected into a high-temperature environment in a rapid com-
pression machine. The simulations are solved on a computational
grid shown in Fig. 1 with inlet SMR of 10 lm.
The heat reaction rate results of two coupling models are com-
pared as presented in Fig. 2. As discussed above, before the ame
appears, the turbulent mixing rate is much higher than the chem-
ical reaction rate. The combustion equation of Model 1 is presented
in Eqs. (32) and (33) represents for Model 2. In Model 1, the tran-
sition between Arrhenius rate and Magnussen model is sudden,
0
5000
10000
15000
20000
25000
30000
35000
0 2 4 6 8 10
Time after injection (ms)
B
u
r
n
i
n
g

r
a
t
e

(
m
i
c
r
o
g
r
a
m
/
m
s
)



Model 1; B=4
Model 1; B=8
Model 1; B=12
Model 1; B=inf
Model 2; B=4
Model 2; B=5
Model 2; B=6
Fig. 2. Time history of burning rate at different model for test case. (Model 1:
x = min(x
c
, x
t
); Model 2: x = x
c
(1 c) + x
t
c; where xc 3:2 10
11
qY
0:25
F
Y
1:50
O
exp15; 100=T; xt Bq
e
k
min YF ;
YO
s
;
YP
1s
_ _
).
Time 1.8 ms 3.0 ms 4.2 ms 5.2 ms
Gas temperature
Fuel vapour
mass fraction
Flame
Luminosity [57]
Fig. 3. The contour of gas temperature and fuel vapour mass fraction comparing with the experiment results of Akiyama et al. [57].
I. Dhuchakallaya, A.P. Watkins / Applied Thermal Engineering 30 (2010) 12151224 1221
while this change is gradual in Model 2. Although the gas temper-
ature is high enough, in Model 1, the reaction mode cannot switch
until Arrhenius rate is higher than the mixing rate. As shown in
Fig. 2, the ignition delay time for Model 1 is then longer than for
Model 2 because Model 2 can switch the mode increasingly with
the higher temperature resulting in predicting auto-ignition rea-
sonably. Furthermore, the burning rate from this coupling mode
is comparable with the experimental data of Akiyama et al. [57].
This coupling reaction rate can also perform successfully in auto-
ignition prediction as shown by Pires da Cruz et al. [55]. Hence,
the transition between self-ignition and high-temperature com-
bustion in all this work employs this coupling mode.
The auto-ignition delay time in present work is based on the
denition of Kang et al. [17] that the ignition delay time is the mo-
ment that the gas temperature increases rapidly. The gas temper-
ature starts to increase sharply at around 1.8 ms as presented in
Fig. 3 which shows the development of gas temperature and fuel
mass fraction. Thus the auto-ignition delay time for this test case
is approximately 1.8 ms while the rst appearance of the luminous
ame in experiment of Akiyama et al. [57] is faster around 1.4 ms.
The location of ignition occurs in the vapour region on the periph-
ery of the liquid spray at around 30 mm from nozzle and 5 mm
from centre line. Moreover, this ignition is located in the lean mix-
ture zone, whereas the corresponding stoichiometric ratio with the
fuel mass fraction is around 0.074. At the auto-ignition point, the
fuel mass fraction is only approximately 0.04 related to a lean
equivalence ratio of 0.54. This result corresponds with the simula-
tion result of Tao et al. [58] and the DNS data of Rutland and Wang
[59]. The possible reason that the ignition occurs at lean mixture
locations is that these locations have experienced less evaporative
cooling due to lesser liquid mass. The droplet temperature is then
able to increase easily. The fast increase of gas temperature during
auto-ignition is caused by heat released resulting in switching
from the chemical reaction control to the turbulent mixing reac-
tion control. After ignition took place, ame subsequently spreads
around the liquid core and moves quickly downstream along the
vapour region inwards to the axis of the fuel jet as presented in
Fig. 3. The increased gas temperature enhances the evaporation
rate of liquid fuel, so that large values of fuel mass fraction appear
consequently. The ame then propagates rapidly downstream until
the fuel vapour downstream is all consumed resulting in the high
reaction rate. During the consumption of fuel vapour, the ame be-
comes shorter and wider, and then it appears nally like a reball.
This simulation also corresponds with experimental results of [60
62]. Additionally, the simulated process is quite similar to the sim-
ulation results of [58,63].
At a fuel injection period of 3.0 ms, the predicted gas tempera-
ture is rather higher than of experimental results as shown in
Fig. 3. However, the ame shapes at 4.2 and 5.2 ms are reasonably
comparable. In addition, ame distributions are also compared in
term of ame area as shown in Fig. 4. Occurrence of predicted
ame is earlier than experimental result, and then the ame occur-
rence becomes closer later on. The detailed comparison of ame
temperature represented in term of temperature histogram is also
illustrated as in Fig. 5. Comparing ame area at 4.0 and 5.0 ms after
start of injection, the predicted total ame areas at specic period
is nearly the same as the experimental data as shown in Fig. 4, and
the means of these distributions are located nearly at the adiabatic
ame temperature and are also reasonably good in agreement with
experimental results as presented in Fig. 5. As seen, the results pre-
dicted by this model present good agreement with the experimen-
tal results at time of 4.0 and 5.0 ms after start of injection,
especially the mean temperatures. The predicted average ame
temperature is 1950 K at 4.0 ms, and the ame temperature con-
tinues increase until 2100 K at 5.0 ms later on. These correspond
to the experimental results. However, their amplitudes of these
ame distributions are quite lower than the peak values of ame
area obtained from the experimental data. Although the ame
areas are quite similar, large ame area appearing in these simula-
tions present in low-temperature range, especially in spray tails.
Fig. 6 shows how accurately present model can capture heat re-
lease rate. In the early stage, heat release rate slightly decreases be-
low zero due to heating up the liquid droplets. It then rapidly
increases after ignition occurs resulting in high reaction rate. Later
on, the diffusion ame becomes dominant to control the reaction.
The ame approaches combustion chamber wall at time of 5.2 ms
in the experiment. Hence the results after time of 5.2 ms are not
considered. Comparing with the experimental results, the peak of
reaction rate obtained from the experiment is higher than the peak
0
500
1000
1500
2000
2500
3000
2 2.5 3 3.5 4 4.5 5 5.5
Time after injection (ms)
F
l
a
m
e

a
r
e
a

(
m
m
2
)
Experimental data
Predicted data
Fig. 4. Development of ame area for the experiment of Akiyama et al. [57].
0
200
400
600
800
1000
1200
1400
1600
1800
2000
1400 1600 1800 2000 2200 2400 2600 2800
Flame temperature (K)
F
l
a
m
e

a
r
e
a

(
m
m
2
)
Exp. (5.0 ms ASI)
Sim. (5.0 ms ASI)
Exp.(4.0 ms ASI)
Sim. (4.0 ms ASI)
Fig. 5. Comparing ame temperature histogram for experiment of Akiyama et al.
[57] at 4.0 and 5.0 ms after start of injection.
-100
0
100
200
300
400
500
600
700
800
900
0 2 4 6 8 10
Time after injection (ms)
H
e
a
t

r
e
l
e
a
s
e
d

r
a
t
e

(
J
/
m
s
)

Predicted data
Experimental data
Fig. 6. Time history of heat release rate for the experiment of Akiyama et al. [57].
1222 I. Dhuchakallaya, A.P. Watkins / Applied Thermal Engineering 30 (2010) 12151224
of the simulation result. This is possible that the evaporation model
employed here performs improperly at low-temperature. This
model can produce less fuel vapour for burning than the experi-
mental result at low-temperature leading to a lower heat release
rate in early period of combustion. Later on, the simulation can
generate fuel vapour properly at high-temperature leading to com-
parable heat release rate with the experimental result in main
combustion period. Principally, this combustion model is capable
to capture this spray combustion experiment fairly well.
5. Conclusions
The aim of this work is to develop and implement a simulation
model for a spray combustion based on the spray size distribution
moments introduced by Beck [33], suitable for diesel-like compres-
sion ignition engines. In the present spray model, the droplet size
distribution of spray is characterised by the rst four moments re-
lated to number, radius, surface-area and volume of droplets,
respectively. The governing equations for gas phase and liquid
phase employed here are solved by the nite volume method
based on an Eulerian framework. These constructed equations
are based on the moment-average quantities which is the key con-
cept for this work. The source terms of sub-models including drop-
let breakup, collision, evaporation, and the interactions between
the liquid phase and the gas phase are derived in term of droplet
moments.
In the main section, the sub-model employed for ignition and
combustion is the coupling reaction rate between Arrhenius model
and Eddy Break-Up model (EBU) via a reaction progress variable
which is widely used in combustion modelling. In addition, auto-
ignition which is the onset of combustion is also investigated.
The auto-ignition takes place in the lean fuel vapour region on
the periphery of the liquid spray and subsequently spreads around
the liquid core and then moves quickly downstream along the
vapour region. Comparing with the experimental results, this
approach is capable qualitatively to reasonable prediction auto-
ignition. In addition, the ame area developed during combustion
progressed is also corresponding with the experimental data. How-
ever, the predicted results of this model seemingly present over-
prediction in ame temperature distributions. This might be that
the sub-model of turbulence/chemistry interaction employed here
is based on innitely fast chemistry assumption. However, overall
of this model is capable to predict moderately the phenomena of
spray combustion.
Acknowledgement
The nancial support from Thammasat University (Thailand) is
gratefully acknowledged.
References
[1] J.B. Heywood, Internal Combustion Engine Fundamentals, McGraw-Hill
International, New York, 1988.
[2] M.P. Halstead, A. Kirsch, C.P. Quinn, The autoignition of hydrocarbon fuels at
high temperature and pressure tting of a mathematical model, Combustion
and Flame 30 (1977) 4560.
[3] H. Schpertns, W. Lee, Multidimensional Modelling of Knocking Combustion
in S.I. Engines, SAE Technical Paper Series No. 850502, 1985.
[4] M.A. Theobald, W.K. Cheng, A Numerical Study of Diesel Ignition, ASME Paper
No. 87-FE-2, 1987.
[5] S.C. Kong, R.D. Reitz, Multidimensional Modelling of Diesel Ignition and
Combustion using a Multistep Kinetics Model, ASME Paper No. 93-ICE-22,
1993.
[6] R.D. Reitz, C.J. Rutland, Development and testing of diesel engine CFD models,
Progress in Energy and Combustion Science 21 (1995) 173196.
[7] P. Belardini, C. Bertoli, C. Beatrice, A. DAnna, N. DelGiacomo, Application of a
reduced kinetic model for soot formation and burntout in three dimensional
diesel combustion computations, in: Proceedings of the 26th Symposium
(International) on Combustion, Italy, 1996.
[8] R.A. Cox, J.A. Cole, Chemical aspects of the autoignition of hydrocarbonair
mixtures, Combustion and Flame 60 (1985) 109123.
[9] H. Hu, J.C. Keck, Autoignition of Adiabatically Compressed Combustible Gas
Mixtures, SAE Technical Paper Series No. 872110, 1987.
[10] J.S. Cowart, J.C. Keck, J.B. Heywood, C.K. Westbrook, W.J. Pitz, Engine knock
predictions using fully-detailed and a reduced chemical kinetic mechanism,
in: 23rd Symposium (International) on Combustion, The Combustion Institute,
Pittsburgh, Pennsylvania, 1990.
[11] M. Schreiber, A.S. Sakak, A. Lingens, J.F. Grifths, A reduced thermokinetic
model for the autoignition of fuels with variable octane ratings, in: 25th
Symposium (International) on Combustion, Irvine, 1994.
[12] Z. Zhao, J. Li, A. Kazakov, S.P. Zeppieri, F.L. Dryer, Burning velocities and a high
temperature skeletal kinetic model for n-decane, Combustion Science and
Technology 177 (2005) 89106.
[13] H.S. Soyhan, J. Andrae, Evaluation of kinetic models for autoignition of
automotive reference fuels in HCCI applications, Turkish Journal of
Engineering and Environmental Sciences 31 (2007) 18.
[14] M. Valorani, F. Creta, F. Donato, H.N. Najm, D.A. Goussis, Skeletal mechanism
generation and analysis for n-heptane with CSP, Proceedings of the
Combustion Institute 31 (2007) 483490.
[15] P.M. Najt, D.E. Foster, Compression-ignited Chemical Kinetic Mechanism for
Hydrocarbon Fuels Homogeneous Charge Combustion, SAE Technical Paper
Series No. 830264, 1983.
[16] T. Takagi, C.Y. Fang, T. Kaminoto, T. Okamoto, Numerical simulation of
evaporation, ignition and combustion of transient sprays, Combustion Science
and Technology 75 (1991) 112.
[17] S.H. Kang, S.W. Baek, J.H. Choi, Autoignition of sprays in a cylindrical
combustor, International Journal of Heat and Mass Transfer 44 (2001) 2413
2422.
[18] K.K. Kuo, Principles of Combustion, John Wiley & Sons, New York, 1986.
[19] Q. Tang, W. Zhao, M. Bockelie, R.O. Fox, Multi-environment probability density
function method for modelling turbulent combustion using realistic chemical
kinetics, Combustion Theory and Modelling 11 (2007) 889907.
[20] T.S. Kuan, R.P. Lindstedt, Transported probability density function modeling of
a bluff body stabilized turbulent ame, Proceedings of the Combustion
Institute 30 (2005) 767774.
[21] Y.M. Wright, G. De Paola, K. Boulouchos, E. Mastorakos, Simulations of spray
autoignition and ame establishment with two-dimensional CMC,
Combustion and Flame 143 (2005) 402419.
[22] A.Y. Klimenko, R.W. Bilger, Conditional moment closure for turbulent
combustion, Progress in Energy and Combustion Science 25 (1999) 595687.
[23] S.P. Burke, T.W. Schumann, Diffusion ames, in: Proceedings of the First
Symposium on Combustion, Pittsburgh, Swampscott, MA, 1928 (Reprint of
Proceedings Published by The Combustion institute in 1965).
[24] C.D. Rakopoulos, K.A. Antonopoulos, D.C. Rakopoulos, Development and
application of multi-zone model for combustion and pollutants formation in
direct injection diesel engine running with vegetable oil or its bio-diesel,
Energy Conversion and Management (2007) 48 18811901.
[25] M. Lapuerta, O. Armas, R. Ballesteros, J. Fernandez, Diesel emissions from
biofuels derived fromSpanish potential vegetable oils, Fuel 84 (2005) 773780.
[26] Y. Cui, K. Deng, J. Wu, A direct injection diesel combustion model for use in
transient condition analysis, Proceedings of the Institution of Mechanical
Engineers 215 (2001) 9951004.
[27] H. Lehtiniemi, F. Mauss, M. Balthasar, I. Magnusson, Modeling diesel spray
ignition using detailed chemistry with a progress variable approach,
Combustion Science and Technology 178 (2006) 19771997.
[28] M. Hossain, W. Malalasekera, Numerical study of bluff-body non-premixed
ame structures using laminar amelet model, Proceedings of the Institution
of Mechanical Engineers 219 (2005) 361370.
[29] D.B. Spalding, Mixing and chemical reaction in steady conned turbulent
ames, in: The 13th Symposium (International) on Combustion, The
Combustion Institute, Pittsburgh, Salt Lake City, 1970.
[30] A.D. Gosman, P.S. Harvey, Computer Analysis of FuelAir Mixing and
Combustion in an Axisymmetric DI Diesel Engine, SAE Technical Paper Series
No. 820036, 1982.
[31] D.M. Wang, Modelling Spray Wall Impaction and Combustion Processes of
Diesel Engines, Ph.D. Thesis, UMIST, Manchester, 1992.
[32] A. Yoong, A.P. Watkins, Modelling of LPG spray development, evaporation and
combustion, International Journal of Engine Research 5 (2004) 469497.
[33] J.C. Beck, Computational Modelling of Polydisperse Sprays without Segregation
into Droplet Size Classes, UMIST, Manchester, 2000.
[34] J.C. Beck, A.P. Watkins, The droplet number moments approach to spray
modelling: the development of heat and mass transfer sub-models,
International Journal of Heat and Fluid Flow 24 (2003) 242259.
[35] J.C. Beck, A.P. Watkins, On the development of a spray model based on drop-size
moments, Proceedings of the Royal Society of London A 459 (2003) 13651394.
[36] J.C. Beck, A.P. Watkins, On the development of spray sub-models based on
droplet size moments, Journal of Computational Physics 182 (2002) 586621.
[37] J.C. Beck, A.P. Watkins, The simulation of fuel sprays using the moments of the
drop number size distribution, International Journal of Engine Research 5
(2004) 121.
[38] J.C. Beck, A.P. Watkins, The simulation of water and other non-fuel sprays
using a new spray model, Atomization and Sprays 13 (2003) 126.
[39] A.P. Watkins, Modelling of mean temperatures used for calculating heat and
mass transfer in sprays, International Journal of Heat and Fluid Flow 28 (2007)
388406.
I. Dhuchakallaya, A.P. Watkins / Applied Thermal Engineering 30 (2010) 12151224 1223
[40] F.H. Harlow, A.A. Amsden, Numerical calculation of multiphase uid ow,
Journal of Computational Physics 17 (1975) 1952.
[41] W.K. Melville, K.N.C. Bray, A model of the two-phase turbulent jet,
International Journal of Heat and Mass Transfer 22 (1979) 647656.
[42] H.K. Versteeg, W. Malalasekera, An Introduction to Computational Fluid
Dynamics: The Finite Volume Method, Longman Scientic & Technical,
Malaysia, 1995.
[43] A.A. Mostafa, H.C. Mongia, On the modelling of turbulent evaporating sprays:
Eulerian versus Lagrangian approach, International Journal of Heat and Mass
Transfer 30 (1987) 25832593.
[44] B.E. Launder, D.B. Spalding, Lectures in Mathematical Models of Turbulence,
Academic Press, London, 1972.
[45] J.C. Beck. Computational Modelling of Polydisperse Sprays without Segregation
into Droplet Size Classes, Ph.D. Thesis, UMIST, Manchester, 2000.
[46] R.D. Reitz, R. Diwakar, Structure of High Pressure Fuel Sprays, SAE Technical
Paper Series No. 870598, 1987.
[47] G.M. Faeth, L.P. Hsiang, P.K. Wu, Structure and break-up properties of sprays,
International Journal of Multiphase Flow 21 (1995) 99127.
[48] P.J. ORourke, F.V. Bracco, Modelling of drop interactions in thick sprays and a
comparison with experiments, in: Conference on Stratied Charge Automotive
Engines, No. C404/80, Institute of Mechanical Engineers (IMechE), 1980.
[49] M. Orme, Experiments on droplet collisions, bounce, coalescence and
disruption, Progress in Energy Combustion Science 23 (1997) 6579.
[50] J. Qian, C.K. Law, Regimes of coalescence and separation in droplet collision,
Journal of Fluid Mechanics 331 (1997) 5980.
[51] B.R. White, Particle dynamics in two phase ows, Encyclopedia of Fluid
Dynamics 3 (1986) 240282.
[52] G.B. Wallis, One-dimensional Two-phase Flows, McGraw-Hill International,
New York, 1969.
[53] C.K. Westbrook, F.L. Dryer, Chemical kinetic modelling of hydrocarbon
combustion, Progress in Energy and Combustion Science 10 (1984) 157.
[54] B.F. Magnussen, B.H. Hjertager, On mathematical modelling of turbulent
combustion with special emphasis on soot formation and combustion, in: 16th
Symposium (International) on Combustion, The Combustion Institute,
Pittsburgh, 1976.
[55] A. Pires da Cruz, T.A. Baritaud, T.J. Poinsot, Self-ignition and combustion
modeling of initially nonpremixed turbulent systems, Combustion and Flame
124 (2001) 6581.
[56] R.I. Issa, Solution of the implicitly discretised uid ow equations by operator-
splitting, Journal of Computational Physics 62 (1986) 4065.
[57] H. Akiyama, H. Nishimura, Y. Ibaraki, N. Iida, Study of diesel spray
combustion and ignition using high-pressure fuel injection and a micro-
hole nozzle with a rapid compression machine: improvement of combustion
using low cetane number fuel, Society of Automotive Engineers of Japan 19
(1998) 319327.
[58] F. Tao, V.I. Golovitchev, J. Chomiak, Application of complex chemistry to
investigate the combustion zone structure of DI diesel sprays under engine-
like conditions, in: The Fifth International Symposium on Diagnostics and
Modeling of Combustion in Internal Combustion Engines (COMODIA 2001),
Nagoya, 2001.
[59] C.J. Rutland, Y. Wang, Turbulent liquid spray mixing and combustion
fundamental simulations, Journal of Physics 46 (2006) 2837.
[60] C. Crua, D.A. Kennaird, S.S. Sazhin, M.R. Heikal, Diesel autoignition at elevated
in-cylinder pressures, International Journal of Engine Research 5 (2004) 365
374.
[61] G. Bruneaux, M. Aug, C. Lemenand, A study of combustion structure in high
pressure single hole common rail direct diesel injection using laser induced
uorescence of radicals, in: The Sixth International Symposium on Diagnostics
and Modeling of Combustion in Internal Combustion Engines, (COMODIA
2004), Yokohama, Japan, 2004.
[62] L.C. Ganippa, S. Andersson, J. Chomiak, Combustion characteristics of diesel
sprays from equivalent nozzles with sharp and rounded inlet geometries,
Combustion Science and Technology 175 (2003) 10151032.
[63] A. Larsson, Optical Studies in a DI Diesel Engine, SAE Technical Paper Series No.
1999-01-3650, 1999.
1224 I. Dhuchakallaya, A.P. Watkins / Applied Thermal Engineering 30 (2010) 12151224

S-ar putea să vă placă și