Conference PaperPDF Available

Modeling ethanol spray jet flame in hot-diluted coflow with transported PDF

Authors:

Abstract and Figures

MILD Combustion, also known as flameless combustion, is attracting wide scientific interest due to its potential of high efficiency and low NOx emission. This paper focuses on the numerical modeling of one of the ethanol spray flame cases from the Delft Spray-in-Hot-Coflow (DSHC) burner, which has been used to study MILD oxidation of liquid fuels. The study has been carried out following the approach of dilute spray simulation. To properly account the turbulent two-phase flow system, a joint velocity-scalar PDF for continuous phase, and a joint PDF of droplet parameters for dispersed phase are employed respectively. Due to the high-dimensionality, the joint PDFs are solved by a Monte Carlo particle method, therefore it is refered to as 'Lagrangian-Lagrangian' approach. The evolution of gas phase composition is described by a Flamelet Generated Manifold (FGM) and the LMSE micro-mixing model. The droplet heating and evaporation processes are modeled with a parabolic temperature profile model. Validation of this modeling approach is carried out by comparison with experimental measurements. Results show that the spray behavior is successfully reproduced; the predicted droplet mean velocity components profiles for all droplet size classes are in very good agreement with the experimental data at various axial locations. Droplet Sauter Mean diameter (SMD) have been accurately predicted. Gas phase velocity also matches well with experimental data. Gas phase temperature is in reasonable agreement with the experiment, however, it is under predicted at the near axis region. Improvement of the accuracy on temperature prediction can be made by using a non-adiabatic FGM table including an enthalpy variable.
Content may be subject to copyright.
SPEIC14 Towards Sustainable Combustion
November 19-21, 2014, Lisboa, Portugal
MODELING ETHANOL SPRAY JET FLAME IN HOT-DILUTED COFLOW
WITH TRANSPORTED PDF
L. Ma*, B. Naud** and D.J.E.M. Roekaerts*
* Department of Process and Energy, Delft University of Technology, The Netherlands
** Modeling and Numerical Simulation Group, Energy Department, Ciemat, Spain
ABSTRACT
MILD Combustion, also known as flameless combustion, is attracting wide scientific interest due to its potential
of high efficiency and low NOx emission. This paper focuses on the numerical modeling of one of the ethanol spray
flame cases from the Delft Spray-in-Hot-Coflow (DSHC) burner, which has been used to study MILD oxidation of
liquid fuels. The study has been carried out following the approach of dilute spray simulation. To properly account
the turbulent two-phase flow system, a joint velocity-scalar PDF for continuous phase, and a joint PDF of droplet
parameters for dispersed phase are employed respectively. Due to the high-dimensionality, the joint PDFs are solved
by a Monte Carlo particle method, therefore it is refered to as 'Lagrangian-Lagrangian' approach. The evolution of gas
phase composition is described by a Flamelet Generated Manifold (FGM) and the LMSE micro-mixing model. The
droplet heating and evaporation processes are modeled with a parabolic temperature profile model. Validation of this
modeling approach is carried out by comparison with experimental measurements. Results show that the spray
behavior is successfully reproduced; the predicted droplet mean velocity components profiles for all droplet size
classes are in very good agreement with the experimental data at various axial locations. Droplet Sauter Mean
Diameter (SMD) have been accurately predicted. Gas phase velocity also matches well with experimental data. Gas
phase temperature is in reasonable agreement with the experiment, however, it is under predicted at the near axis
region. Improvement of the accuracy on temperature prediction can be made by using a non-adiabatic FGM table
including an enthalpy variable.
Keywords: Spray combustion, MILD combustion, FGM, Transported PDF
1. INTRODUCTION
Spray combustion is widely utilized in various engineering
applications, such as industrial furnaces and propulsion systems.
Modeling of turbulent spray combustion is particularly
challenging, because many physical and chemical processes
including turbulence, atomization, evaporation, combustion and
radiative heat transfer are involved and interact with each other
[1]. These phenomena and processes have to be modeled in a
proper way in the sense that the main physical characteristics
have to be accounted for but within a reasonable computational
cost. For simplicity, many spray combustion studies have been
carried out in the regime of dilute spray [2,3], and this is also the
case for this study.
Transported PDF method proposed by Pope [4] has been
proven to be a powerful closure method for modeling turbulent
reactive flow. The most outstanding advantage of transported
PDF method is that the mean reaction source term appears as a
closed term. However, for the sake of reducing computation cost,
the direct use of detailed chemistry is not employed in this study,
instead the Flamelet Generated Manifold model proposed by van
Oijen [5] is used. In FGM, the scalars, such as temperature,
species mass fraction, density etc., are stored in the lookup table
as a function of a few independent variables, e.g. mixture
fraction and progress variable. The influence of turbulence is
accounted for by the Probability Density Function (PDF) of the
independent variables. The PDF for mixture fraction is often
presumed as a beta function, and determined by the mean and
variance value obtained during the turbulent combustion
simulation. However, severial studies [1,6] already pointed out
that due to the presence of droplet evaporation the beta shape
PDF is no longer valid for mixture fraction in spray combustion.
In transported PDF method, the PDF is directly solved, therefore
the turbulence-chemistry interaction is considered in a more
precise manner.
Moderate or intense low oxygen dilution (MILD)
Combustion is a promising technology that mitigates the
combustion-generated pollutants whilst meeting thermal
efficiency needs [7,8]. Delft Spray-in-Hot-Coflow (DSHC)
burner was designed to study the fundamental aspects of
flameless oxidation of light oils [9]. The present paper reports a
first numerical study of this flame with transported PDF method.
2. EXPERIMENTAL DADABASE
The schematic of the DSHC burner is shown in Figure 1. The
hot-diluted coflow is produced by the secondary burner that
operates on air and Dutch nature gas (DNG). This hot-diluted
coflow is used to mimic the mixture of air with combustion
products in MILD combustion furnace. The liquid fuel, ethanol
in this study, is injected by a pressure swirl atomizer in the
hot-diluted coflow.
The available experimental database includes the following
data: radial profiles of coflow temperature, velocity components
and O2 volume concentration; radial profiles of the droplet
diameter and velocity components at different axial locations;
and radial profiles of gas phase temperature at different axial
locations. A complete description on the DSHC burner and
corresponding database can be found in [9]. These data are
compared with the simulation results for validation purpose.
Figure 1. Schematic of DSHC burner.
3. NUMERICAL SETUP
3.1 Computation domain & boundary conditions
Figure 2 shows a picture of the DSHC flame, on which the
computational domain of this study is illustrated with yellow
rectangle. As the flame is axisymmetric, a 2D axisymmetric
simulation is conducted. The inlet boundary is chosen such that
it is sufficiently far from the atomizer tip to avoid the influence
of the dense spray region and also below the region where the
ignition starts. In this case, the axial location Z=8 mm is chosen
as the inlet boundary, this is also the first axial location where
the dispersed phase properties were measured.
The dispersed phase boundary conditions are assigned
according to the corresponding experimental data. However, the
gas phase boundary conditions cannot be directly specified from
the measurements. This is because, firstly, the existence of the
droplets prevents the use of Laser Doppler Anemometry (LDA)
for gas phase velocity measurement in the spray region,
therefore the gas phase velocity is obtained from the Phase
Doppler Anemometry (PDA) measurements, assuming that the
small droplets (D < 6 µm) strictly follow the gas phase behavior.
However the PDA result only is available at a relative small
region where the small droplets are present, and these data are
insufficient to accurately assign the whole inlet boundary
profiles. Secondly, as the FGM model is used as the combustion
model, the control variables, mixture fraction and progress
variable, have also to be assigned at the inlet boundary. But, they
are not directly available from experimental measurement.
However, the necessary properties are available at the coflow
exit (Z = 0mm). A full spray combustion simulation including
the spray atomization process is conducted with ANSYS
Fluent15 to provide reliable boundary conditions at Z=8 mm for
the present study. In the Fluent simulation, the pressure-swirl
atomizer is modeled with Linearized Instability Sheet
Atomization (LISA) model. The turbulence flow field is
described by Reynolds Stress Model (RSM). And the
turbulence-chemistry interaction, to be consistent with the
current study, is also modeled by FGM model but with presumed
shape PDF method. In this Fluent simulation, beta PDF is used
for both mixture fraction and progress variable. To examine the
reliability of the boundary condition provided by the Fluent
simulation, not only the results at Z=8 mm but also at other axial
locations are compared with experimental data. The compared
variables include: the droplet velocity components for different
droplet size classes, droplet Sauter Mean Diameter (SMD), gas
phase velocity components as well as gas phase temperature.
Reasonably good agreement with experimental data has been
achieved by this Fluent simulation. The modeling detail and
results of the Fluent simulation will be reported separately, part
of the results can be found in [10]. This gives us enough
confidence to employ Fluent simulation results as boundary
conditions for the current study. The velocity, mixture fraction
and progress variable boundary conditions for this study are
shown in Figure 3.
Figure 2 Picture of DSHC flame with indication of axial
location of experimental data.
Figure 3 Boundary conditions for dilute spray simulation of
the DSHC flame
3.2 Numerical model
The turbulent two phase flow system of the DSHC flame is
described by a hybrid finite volume / transported PDF approach
implemented in the in-house code ‘PDFD’. The continuous
phase is described by a joint velocity-scalar PDF, and the
dispersed phase is described by a joint PDF of droplet position,
velocity, temperature, diameter, and the gaseous properties 'seen'
by the droplet. Due to the high-dimensionality, the joint PDFs
are solved by Monte Carlo particle method. In contrast with
more standard Eulerian-Lagrangian approach, in PDFD, both the
gas phase and the dispersed phase evolution are defined by
Lagrangian equations, therefore we refer it as
'Lagrangian-Lagrangian' approach. To overcome the bias error
due to the limited number of computational particles in the
Monte Carlo method, the mean velocities and Reynolds stresses
are calculated using a Finite-Volume (FV) method, in which the
Reynolds Averaged Navier Stokes (RANS) equations are
solved. The algorithm of PDFD code is shown in Figure 4.
Figure 4 Schematic of the hybrid FV / MC algorithm in
PDFD code
The finite volume submodel provides the mean velocity and
its gradient, mean pressure gradient, Reynolds stresses and mean
turbulent dissipation rate to the Monte Carlo part. The
fluctuating velocity increment of gas phase particle is
determined by generalized Langevin model (GLM), specifically,
the Lagrangian isotropisation of production model (LIPM). To
be consistent with LIPM, the isotropisation of production
Reynolds-stress model (LRR-IPM), is used for the modeling of
the pressure strain correlation in the finite volume part. The
evolution of gas phase composition is described by FGM and the
Linear Mean Square Estimation (LMSE) micro-mixing model
[11], also known as interaction by exchange with the mean
(IEM):
mix
d Sdt


(1)
 
1
2
mix dt
  
 
(2)
Where
is the modeled scalar variance decay frequency.
S is the source term. Since the tabulated chemistry model FGM
is used, the scalar θ in this case include only the independent
variables of the lookup table, namely, the mixture fraction and
progress variable. Unlike the pure gaseous flame, the mixture
fraction in spray combustion is no longer a conserved scalar, its
source is due to the vaporization of droplets. The source term of
progress variable is retrieved from the FGM lookup table as a
function of independent variables, as shown in Figure 5. The
influences of the spray are also represented by the extra source
terms appearing in the momentum and Reynolds stress equations
in the finite-volume, this is the so-called two-way coupling.
Figure 5 Source term of progress variable
The droplet heating and evaporation processes are modeled
with the parabolic temperature profile model which is in the
category of finite conductivity models. In contrast with the
widely used infinite conductivity model, where the temperature
distribution inside the droplet is assumed uniform, this model
takes into account the finite rate heat conduction inside the
droplet. This is important where the droplet heating process is
fast as is the case in the hot-diluted coflow condition of this
study.
In flamelet-like models, the multi-dimension flames are
considered to be a set of 1D flamelets. The 1D flamelets are
characterized by different controlling parameters to describe the
local variations of the real flame. For the FGM model, the
controlling parameters are mixture fraction and progress variable.
There are different methods existing for the construction of the
2D FGM lookup table [5]. A commonly used one is to first
calculating different steady flamelet equations with scalar
dissipation rate increasing from a zero to the extinction value.
And then mapping these steady flamelets together with the
unsteady flamelet at the extinction scalar dissipation rate to
mixture fraction and progress variable space. Another approach
is to solve the unsteady process of 1D diffusion flame in physical
space from pure mixing until the steady flame is established. The
flamelets at different time are then transformed into mixture
fraction and progress variable space. Comparing to the
"extinguishing" FGM generated by the first method, the second
method generates a "auto-igniting" FGM table. Therefore it is
more suitable to describe the auto-ignition process of the DSHC
flame. The auto-igniting FGM table in this study is generated
with the CHEM1D code developed at the Eindhoven University
of Technology [12]. The detailed ethanol high temperature
oxidation mechanism containing 57 species and 383 reactions by
Marinov [13] is employed. The ignition process is illustrated by
the temperature profiles in mixture fraction space with
increasing time, as shown in Figure 6. The progress variable in
this study is defined as:
2 2 2
2 2 2
CO H O H
CO H O H
Y Y Y
CM M M
  
(3)
Where Y and M denote the mass fraction and molecular weight.
Figure 6 Temperature profiles for auto-ignition FGM table
3. RESULTS AND DISCUSSION
3.1 Dispersion phase results
In spray combustion, the dispersed phase not only provides
the evaporated fuel vapor for combustion, but also directly
modulates the continuous phase flow field. A correct prediction
of the dispersed phase behavior is essential for successful
modeling of spray combustion. Therefore the predicted droplet
results are firstly presented and compared with the experimental
data to validate the dispersed phase sub-models. The gas phase
results will be discussed in next subsection.
Droplet mean axial and radial velocity profiles are shown in
Figure 7. The results are plotted in a matrix of subplots with each
subplot representing a certain droplet size class (columns) at a
certain axial location (rows). To save space, only five droplet
classes are shown here. The droplet size increases from left to
right of the matrix and the axial location increases from bottom
to top. It can be clearly seen that the predicted droplet velocity
matches very well with the experimental for all droplet classes at
all the axial locations compared. At higher axial location, the
simulation results have a wider radial distribution than the
corresponding measured data. This is because in the experiment,
the PDA measurement is only carried out in the region where the
droplet concentration is relatively high. At the inner and outer
spray edge where the droplet number density is low, the
experimental data is not available.
Droplet Sauter Mean Diameter (SMD) is shown in Figure 8.
In the region where the experimental data are available, both the
trend and magnitude of SMD are correctly predicted. The slight
under-prediction at the spray outer edge (large radial position)
may be attributed to the inaccurate specification of droplet size
distribution at the inlet boundary. In this simulation, droplet are
considered only in the diameter range of [0, 70] µm, respecting
the known total mass flux. This was done because the number
density of the droplets larger than 70 µm is quite low in this
flame, due to both the quick evaporation and changed
atomization mechanism by the hot coflow as presented in the
experimental paper by Rodrigues et al [9]. However, exclusion
of the big droplets may have a non-negligible influence on the
mass flux of large droplet at the spray outer edge because of their
ballistic trajectories and the large amount of mass they contain.
Figure 7 Droplet velocity: solid lines - simulation results,
circular dots - experimental data, red - axial velocity, blue -
axial velocity.
3.2 Gas phase results
From flame picture Figure 2, it can be clearly observed that
the DHSC flame is a lifted flame. Since the spray is issued into a
hot-diluted coflow which temperature much higher than the
auto-ignition temperature of ethanol, the spray flame is mainly
stabilized by the growth of auto-ignition kernels. The lift-off
height is therefore a balance between the auto-ignition delay
time, the evaporation time scale and the flow time scale. As
demonstrated in [10], the steady flamelet model is unable to
predict the lift-off nature of this flame, and therefore leads to a
wrong prediction of the flame structure. However, the
auto-ignition FGM model used in this study, can capture the
flame lift-off. Although there is no quantitative measurement on
the lift-off height of this flame, by comparing flame picture
Figure 2 and OH concentration and temperature field in Figure
10, we could conclude that the flame lift-off has been rather
accurately predicted.
Figure 8 Droplet Sauter Mean Diameter: solid lines
simulation results, circular dots experimental data.
Figure 9 Gas phase velocity: solid lines - simulation results,
circular dots - experimental data, red - axial velocity, blue -
axial velocity.
Gas phase velocity and temperature are shown in Figure 9
and Figure 11 respectively. Good agreement with the available
measurement data is observed for both mean axial and radial
velocity profiles at all the axial locations compared. A slight
under-prediction of the gas phase radial velocity at the lower
axial locations can be explained from the following two aspects:
first, the gas phase velocity boundary conditions obtained from
Fluent simulation have some discrepancies with the experimental
data, as shown in Figure 3. The reason for these discrepancies
are explained in [10]. Second, the experimental gas phase
velocity is obtained by PDA with small droplets as tracer. The
small droplets need time to fully follow the gas phase, therefore
the PDA results at the lower axial location maybe cannot been
exactly interpreted as gas phase velocity. Further downstream,
where the PDA tracer fully followed the gas phase flow, the
match between the simulation and experiment becomes better.
From axial location Z=45 mm onward, not enough tracer
droplets are available due to the evaporation, therefore no
experimental data are available thereafter.
The gas phase temperature in the spray region has been
measured with Coherent Anti-Stokes Raman Scattering (CARS)
technique for spray combustion [9]. A reasonable good
agreement with the experimental data is obtained in this study.
The flame peak temperature as well as the flame width are
correctly predicted. The radial position of the peak temperature
is a little shifted towards the center. This may mainly be caused
by the mixture fraction profile specified at the inlet boundary. As
explained in [10], the Fluent simulation used for providing inlet
boundary information predicts smaller spray dispersion angle
comparing with experiment. Consequently the distribution of the
vaporized vapor is also narrower than that in reality. Close to the
spray axis, an opposite temperature trend is predicted. The
simulation shows a small temperature peak in the center, while
the temperature progressively decreases toward the center in the
experiment. This is because near the spray axis many small
droplets exist, considerable gas phase enthalpy loss happens due
to the fast evaporation of small droplets. The enthalpy loss,
however, cannot be accounted by the 2D adiabatic FGM table
used in the current study. As a consequence, the temperature has
been over-predicted in this region. This problem can be solved
by including enthalpy loss as another independent variable of the
FGM table, namely, using a non-adiabatic FGM table.
Figure 10 Contour plots from simulation results, left: OH
mass fraction, right: gas phase temperature.
Figure 11 Gas phase mean and RMS temperature: solid line
mean temperature simulation , dashed line RMS
temperature simulation, circular dots - mean temperature
experiment, triangular dots RMS temperature experiment.
4. CONCLUSIONS
In this paper, we reported a first numerical investigation of
Delft Spray-in-Hot-Coflow flame with transported PDF method
and FGM model. The in-house hybrid finite-volume/transported
PDF code "PDFD" is used for the simulation. The mean gas
phase flow field is calculated by the finite volume part with
Reynolds Stress turbulence model. The gas phase fluctuation,
the turbulence-chemistry interaction as well as the droplet
evolution are represented in the Monte Carlo part. These two
parts are coupled in the way that the finite volume part provides
the gas mean properties that are required for the Monte Carlo
part calculation, and the latter one feeds back the gas phase
density to the finite volume part. The continuous phase is
described by a joint velocity-scalar PDF, and the dispersed
phase is described by a joint PDF of droplet position, velocity,
temperature, diameter, and the gaseous properties 'seen' by the
droplet. A parabolic temperature profile model is used to
describe the droplet heating and evaporation processes.
The current modeling approach is validated by comparing
the predicted droplet and gas phase properties with available
experimental data. In general, very good agreement is obtained.
Droplet velocity, Sauter Mean Diameter are all in good
agreement with measured data, showing that the spray
sub-models including the evaporation and dispersion models
used in this study are suitable for modeling the DSHC flame.
The lift-off characteristics of this flame have been correctly
captured by the auto-ignition FGM momdel used in this study.
Gas phase velocity also matches very well with the available
experimental data. Gas phase temperatrue are in reasonable
agreement with experimental data, showing the capability of the
current modeling approach. However, it is also realized that the
2D adiabatic FGM table used in this study is insufficient to
account the enthalpy loss due to the droplet evaporation, which
resulted in a over-prediction of the gas phas temperature at the
near axis region.
5. ACKNOWLEDGEMENT
The authors would like to thank the China Scholarship
Council (CSC) for financial support for the first author. Part of
this work is also supported by the Comunidad de Madrid
through Project HYSYCOMB P2009/ENE-1597 and by the
Spanish Ministry of Economy and Competitiveness under
Projects ENE2008-06515-C04-02 and CSD2010-00011. We
thank H. Correia Rodrigues, M.J. Tummers and E.H. van Veen,
for creating the experimental dataset. G. Sarras is thanked for
the help on running the PDFD code.
REFERENCES
1. P. Jenny, D. Roekaerts, and N. Beishuizen, Modeling of
turbulent dilute spray combustion, Prog. Energy Combust.
Sci., vol. 38, pp. 846887, 2012.
2. M. Chrigui, a. R. Masri, a. Sadiki, and J. Janicka, “Large
Eddy Simulation of a Polydisperse Ethanol Spray Flame,”
Flow, Turbul. Combust., vol. 90, pp. 813832, 2013.
3. C. Heye, V. Raman, and A. R. Masri, LES/probability
density function approach for the simulation of an ethanol
spray flame, Proc. Combust. Inst., vol. 34, pp. 16331641,
2013.
4. S. B. Pope, PDF methods for turbulent reactive flows, Prog.
Energy Combust. Sci., vol. 11, pp. 119192, 1985.
5. J. A. Van Oijen and L. P. H. De Goey, Modelling of
Premixed Laminar Flames using Flamelet-Generated
Manifolds, Combust. Sci. Technol., vol. 161, pp. 113137,
2000.
6. H. Watanabe, R. Kurose, S. Hwang, and F. Akamatsu,
Characteristics of flamelets in spray flames formed in a
laminar counterflow, Combust. Flame, vol. 148, pp. 234
248, 2007.
7. A. Cavaliere and M. de Joannon, Mild Combustion, Prog.
Energy Combust. Sci., vol. 30, pp. 329366, 2004.
8. J. A. Wunning and J. G. Wunning, Flameless oxidation to
reduce thermal no-formation, Prog. Energy Combust. Sci.,
vol. 23, pp. 8194, 1997.
9. H. Correia Rodrigues, M. J. Tummers, E. H. van Veen, and
D. J. E. M. Roekaerts, Spray flame structure in
conventional and hot-diluted combustion regime, Combust.
Flame, Sep. 2014, in press.
10. L. Ma, S. Zhu, M. J. Tummers, T. H. Van Der Meer and
D.J.E.M. Roekaerts, Numerical investigation of ethanol
Spray-in-Hot-Coflow flame using steady flamelet model, in
Proc. Eight Mediterranean Combustion Symposium, 2013.
11. C. Dopazo, Non-isothermal turbulent reactive flows:
stochastic approaches. PhD thesis, State University of New
York at Stony Brook, 1973.
12. CHEM1D, A one dimension laminar flame code,
Eindhoven University of Technology,
URL:http://www.combustion.tue.nl/chem1d/.
13. N. M. Marinov, A Detailed Chemical Kinetic Model for
High Temperature Ethanol Oxidation, Int. J. Chem. Kinet.,
vol. 31, pp. 183220, 1998.
... Furthermore, the progress of chemical reaction in inhomogeneous mixtures in droplet combustion can be characterized by reaction progress variable c, which can be defined in terms of a suitable species mass fraction so that it assumes a value equal to zero in the unburned gas and monotonically increases to assume a value equal to unity in the fully burned gas. The interdependence of mixture fraction and reaction progress variable c plays a key role in the modeling of spray combustion using flamelet generated manifold (FGM) (Ma et al., 2014;Sadiki et al., 2012). Furthermore, the characterization of probability density functions (PDFs) of mixture fraction and reaction progress variable c and their respective gradients (i.e., Ñ and Ñc) in terms of presumed functions play pivotal roles in flamelet and conditional moment closures (CMC; Borghesi et al., 2011;Ge and Gutheil, 2006;Sadiki et al., 2012;Tyliszczak et al., 2014). ...
... These modeling issues will also be valid for droplet combustion because the evaporation of droplets will lead to partially premixed combustion in gaseous phase. Although different presumed PDF approaches have been used in the past in the context of spray combustion (Ge and Gutheil, 2006;Ma et al., 2014;Sadiki et al., 2012;Tyliszczak et al., 2014), the statistics of the interdependence of and reaction progress variable c and their gradients (i.e., Ñand Ñc) have received limited attention (Luo et al., 2011;Wandel, 2013Wandel, , 2014Xia and Luo, 2010) in existing literature. This deficit has been addressed here by analyzing the statistical behavior of and c, and their respective gradients (i.e., Ñand Ñc) based on 3D DNS data of freely propagating statistically planar turbulent flames propagating into droplet mist for different values of root-mean-square velocity fluctuation u 0 ; droplet diameter a d , and droplet equivalence ratio ...
... Indeed, the agreement of the β-function PDF withP ð Þ is consistent with the modeling assumptions made by Sadiki et al. (2012) and Ma et al. (2014) in the context of FGMbased simulations of turbulent spray combustion. However, it must be emphasized that the use of the β-function PDF to modelP ð Þ may not be suitable in all cases of droplet combustion, as has already been shown by Ge and Gutheil (2006) and Luo et al. (2011). ...
Article
Full-text available
The statistics of reaction progress variable, , and mixture fraction, , and their gradients (i.e., and ) in flames propagating in droplet mist, where the fuel was supplied in the form of monodisperse droplets, have been analyzed for different values of turbulent velocity fluctuations ( ), droplet equivalence ratios ( , and droplet diameters ( based on three-dimensional direct numerical simulations (DNS) in a canonical configuration under decaying turbulence. The combustion process in the gaseous phase has been found to take place predominantly in fuel-lean mode, even for . The probability of finding fuel-lean mixture increases with increasing initial droplet diameter due to slower evaporation of larger droplets. It has been shown that the joint probability density function (i.e., joint PDF) of and (i.e., ), cannot be approximated in terms of discrete delta functions throughout the flame brush for the cases considered here. Furthermore, the magnitude of cannot be adequately approximated by the product of marginal PDFs of , and variable, (i.e., ). The statistical properties of the Favre probability density functions (Favre-PDFs) of the mixture fraction, , and oxidizer-based reaction progress variable, , have been analyzed at several locations across the flame brush and a -function distribution has been found to capture the Favre-PDFs of and obtained from the DNS data. Furthermore, a log-normal distribution has been shown to capture the qualitative behaviors of the PDFs of the gradient of the mixture fraction and the gradient of the reaction progress variable, and , respectively, but discrepancies between the log-normal distribution and the DNS data were observed at the tails of PDFs. In addition, the interrelation between and was examined in terms of the PDFs of the cosine of the angle between them (i.e., and it was observed that most droplet cases exhibited much greater likelihood of positive values of than negative values. Finally, the joint PDF of and , , has been compared with that of P( (i.e., assuming statistical independence of and ) and a good level of agreement has been obtained. The bivariate log-normal distribution has been considered both assuming correlation between and and assuming no correlation for the purpose of modeling , and the variant with no correlation has been found to be more successful in capturing qualitative behavior of although quantitative discrepancies have been observed due to inaccuracies involved in parameterizing P( and P( by log-normal distributions.
Conference Paper
Full-text available
The flameless combustion is a promising approach to increase the combustion efficiency and decrease the pollutant emission. In contrast with the wide range of investigations of its gaseous counterpart, the flameless combustion of liquid fuel is still in its infant stage. Numerical simulation provides an efficient way for the understanding of the flameless spray combustion in detail. In the present study, the Delft Sray-in-Hot-Coflow flame, which wasdesigned to mimic the flameless oxidation of light oils, was modeled using ANSYS Fluent and validated by the experimental data. Steady laminar flamelet model was employed with the detailed ethanol high temperature oxidation mechanism of Marinov. The turbulent two-phase flow field was predicted by the standard k-$epsilon$ turbulence model taking into account the turbulence two-way coupling. The Linearized Instability Sheet Atomization (LISA) model was used to simulate the injection and atomization processes of the liquid fuel. Turbulence dispersion of the droplets was calculated by a Random Walk Model. The predicted droplet SMD matches well with the experimental data. The calculated droplet velocities at different downstream elevations for different droplet size classes show good agreement with the measured results. Predicted gas phase velocities also display reasonably good agreement. The main failure of the present model is its inability to predict absence of combustion in the high strain region, in the vicinity of the injector exit. The presence of the hot coflow accelerates the evaporation of droplets and increases the scalar dissipation rate at the near injector region, compared to combustion in cold air. Comparing with the conventional spray combustion, more efforts have to be made on the precise evaporation model and the auto-ignition and local extinction processes in order to accurately predict the flameless spray combustion.
Article
Full-text available
In order to reduce the computational cost of flame simulations, several methods have been developed during the last decades, which simplify the description of the reaction kinetics. Most of these methods are based on partial-equilibrium and steady-state assumptions, assuming that most chemical processes have a much smaller time scale than the flow time scale. These assumptions, however, give poor approximations in the ‘colder’ regions of a flame, where transport processes are also important.The method presented here, can be considered as a combination of two approaches to simplify flame calculations, i.e. a flamelet and a manifold approach. The method, to which we will refer as the Flamelet-Generated Manifold (FGM) method, shares the idea with flamelet approaches that a multi-dimensional flame may be considered as an ensemble of one-dimensional flames. The implementation, however, is typical for manifold methods: a low-dimensional manifold in composition space is constructed, and the thermo-chemical variables are stored in a database which can be used in subsequent flame simulations. In the FGM method a manifold is constructed using one-dimensional flamelets. Like in other manifold methods, the dimension of the manifold can be increased to satisfy a desired accuracy. Although the method can be applied to different kinds of flames, only laminar premixed flames are considered here.Since the major parts of convection and diffusion processes are present in one-dimensional flamelets, the FGM is more accurate in the ‘colder’ zones of premixed flames than methods based on local chemical equilibria. Therefore, less controlling variables are sufficient to represent the combustion process. Test results of one and two-dimensional premixed methane/air flames show that detailed computations are reproduced very well with a FGM consisting of only one progress variable apart from the enthalpy to account for energy losses.
Article
A detailed chemical kinetic model for ethanol oxidation has been developed and validated against a variety of experimental data sets. Laminar flame speed data (obtained from a constant volume bomb and counterflow twin-flame), ignition delay data behind a reflected shock wave, and ethanol oxidation product profiles from a jet-stirred and turbulent flow reactor were used in this computational study. Good agreement was found in modeling of the data sets obtained from the five different experimental systems. The computational results show that high temperature ethanol oxidation exhibits strong sensitivity to the fall-off kinetics of ethanol decomposition, branching ratio selection for C2H5OH + OH ↔ Products, and reactions involving the hydroperoxyl (HO2) radical. The multichanneled ethanol decomposition process is analyzed by RRKM/Master Equation theory, and the results are compared with those obtained from earlier studies. The ten-parameter Troe form is used to define the C2H5OH(+ M) ↔ CH3 + CH2OH(+ M) rate expression as k∝ = 5.94E23 T-1 68 exp(- 45880 K/T) (s-1) k° = 2.88E85 T-18.9 exp(- 55317 K/T) (cm3/mol/sec) Fcent = 0.5 exp(- T/200 K) + 0.5 exp(- T/890 K) + exp(- 4600 K/T) and the C2H5OH(+ M) ↔ C2H4 + H2O(+ M) rate expression as k∝ = 2.79E13 T0 09 exp(- 33284 K/T) (s-1) k° = 2.57E83 T-18 85 exp(- 43509 K/T) (cm3/mol/sec) Fcent = 0.3 exp(-T/350 K) + 0.7 exp(-T/800 K) + exp(-3800 K/T) with an applied energy transfer per collision value of 〈ΔEdown〉 = 500 cm-1. An empirical branching ratio estimation procedure is presented which determines the temperature dependent branching ratios of the three distinct sites of hydrogen abstraction from ethanol. The calculated branching ratios for C2H5OH + OH, C2H5OH + O, C2H5OH + H, and C2H5OH + CH3 are compared to experimental data.
Article
A consistent probability density function (PDF) based combustion modeling approach for large eddy simulation of turbulent spray combustion is formulated. A Lagrangian Monte-Carlo approach for solving the PDF transport equation in the context of low-Mach number LES formulations is developed. The LES/PDF approach is used to simulate an experimental pilot-stabilized ethanol spray flame. In this particular flame, droplet evaporation occurs away from the flame front, leading to separation of the two processes. A high-temperature pre-flame zone with a stratified mixture is found to exist throughout the length of the flame. The high flow rate of the droplet-laden air prevents the flame front from propagating through this evaporated, but well-mixed fuel/air mixture. Comparison with experiments displayed large discrepancies concerning the droplet inflow conditions. In particular, the inflow turbulence is significantly modified by droplet-turbulence interactions in the injection pipe.
Article
A laboratory scale experimental setup was built to study ethanol pressure-swirl spray flames in a coflow of either air or hot-diluted oxidant. The latter case resembles conditions similar to those found in practical combustion systems of liquid fuels operating in MILD conditions. First, experiments have been performed to investigate the phenomena associated with the atomization process. High-speed visualizations show that in the presence of a hot-diluted coflow, an almost immediate disruption of the liquid jet takes place, indicating significant changes in the atomization mechanism, compared to the case with air coflow. Secondly, a comprehensive set of measurements was obtained by complementary single-point measurement techniques to reveal the gas and droplets flow fields as well as temperature fields. Measurements of droplet size and velocity components in the spray region were obtained by phase Doppler anemometry. Gas temperature was measured using coherent anti-Stokes Raman spectroscopy even in regions with droplet density as high as 10 5 cm - 3 . It has been observed that in a reacting spray in air coflow, an inner and an outer flame-front are present. For a reacting spray with similar injection pressure in hot-diluted coflow, weakening of the inner flame-front is attributed to the fact that the gaseous mixture becomes increasingly rich towards the center region. Consequently, a significant reduction of occurrence of temperature samples above 2000 K is observed throughout most of the spray region.
Article
Ethanol is identified as an interesting alternative fuel. In this regards, the predictive capability of combustion Large Eddy Simulation approach coupled to Lagrangian droplet dynamic model to retrieve the turbulent droplet dispersion, droplet size distribution, spray evolution and combustion properties is investigated in this paper for an ethanol spray flame. Following the Eulerian-Lagrangian approach with a fully two way coupling, the Favre-filtered low Mach number Navier-Stokes equations are solved on structured grids with dynamic sub-grid scale models to describe the turbulent carrier gas phase. Droplets are injected in polydisperse manner and generated in time dependent boundary conditions. They evaporate to form an air-fuel mixture that yields spray flame. Part of the ethanol droplets evaporates within the prevaporization area before reaching the combustion zone, making the flame to burn in a partially premixed regime. The chemistry is described by a tabulated detailed chemistry based on the flamelet generated manifold approach. The fuel, ethanol, is modeled by a detailed reaction mechanism consisting of 56 species and 351 reversible reactions. The simulation results including excess gas temperature, droplet velocities and corresponding fluctuations, droplet mean diameters and spray volume flux at different distances from the exit plane show good agreement with experimental data. Analysis of combustion spray features allows gaining a deep insight into the two-phase flow process ongoing.
Article
A detailed chemical kinetic model for ethanol oxidation has been developed and validated against a variety of experimental data sets. Laminar flame speed data (obtained from a constant volume bomb and counterflow twin-flame), ignition delay data behind a reflected shock wave, and ethanol oxidation product profiles from a jet-stirred and turbulent flow reactor were used in this computational study. Good agreement was found in modeling of the data sets obtained from the five different experimental systems. The computational results show that high temperature ethanol oxidation exhibits strong sensitivity to the fall-off kinetics of ethanol decomposition, branching ratio selection for C2H5OH + OH ↔ Products, and reactions involving the hydroperoxyl (HO2) radical.The multichanneled ethanol decomposition process is analyzed by RRKM/Master Equation theory, and the results are compared with those obtained from earlier studies. The ten-parameter Troe form is used to define the C2H5OH(+M) ↔ CH3 + CH2OH(+M) rate expression as k∞ = 5.94E23 T−1.68 exp(−45880 K/T) (s−1)ko = 2.88E85 T−18.9 exp(−55317 K/T) (cm3/mol/sec)Fcent = 0.5 exp(−T/200 K) + 0.5 exp(−T/890 K) + exp(−4600 K/T) and the C2H5OH(+M) ↔ C2H4 + H2O(+M) rate expression as k∞ = 2.79E13 T0.09 exp(−33284 K/T) (s−1)ko = 2.57E83 T−18.85 exp(−43509 K/T) (cm3/mol/sec)F cent = 0.3 exp(−T/350 K) + 0.7 exp(−T/800 K) + exp(−3800 K/T) with an applied energy transfer per collision value of = 500 cm−1.An empirical branching ratio estimation procedure is presented which determines the temperature dependent branching ratios of the three distinct sites of hydrogen abstraction from ethanol. The calculated branching ratios for C2H5OH + OH, C2H5OH + O, C2H5OH + H, and C2H5OH + CH3 are compared to experimental data. © 1999 John Wiley & Sons, Inc. Int J Chem Kinet 31: 183–220, 1999