Академический Документы
Профессиональный Документы
Культура Документы
Horsalsvagen,
7
412 96, Goteborg - Sweden
http://www.tfd.chalmers.se/ valeri
phone: +46-31-772-1396 e-mail: valeri@tdf.chalmers.se fax: +46-31-18-09-76
ABSTRACT
The new eddy dissipation concept (EDC) model based on the operator-splitting procedure applied
to the mass conservation equations for species participating in reversible chemical reactions representing combustion in a PaSR partially stirred reactor (PaSR) volume is analyzed. The model
has been implemented in the KIVA-3V code together with a new droplet collision procedure, and
its application to spray combustion simulation is illustrated by the results of the 3-D modeling of
the DI Diesel Volvo D12C engine. The chemical mechanism of diesel oil surrogate consisted of 68
species (including soot aromatic precursors) participating in 285 reversible reactions has been applied to simulation of MK (Modulated Kinetics) diesel spray combustion regime.Mean features of
MK combustion under conditions of delayed injection when auto-ignition has much in common with
the HCCI process are predicted in accordance with experimental data.
INTRODUCTION
The MK combustion process based on a lowtemperature premixed mode was suggested for simultaneous reduction of NOx and particulate matter. To combat emissions formation, the injection
timing is retarded extending the ignition delay and
premixed combustion mode contribution. Optimization of the process requires modeling based on the
detailed chemistry. Since the time scales for kinetic
and turbulent micro-mixing are comparable during
the long ignition period characteristic of the MK
mode, turbulence/chemistry interaction must be accounted for. The Eddy Dissipation Concept (EDC)
is a promising technique to facilitate such combustion analysis. However, the EDC is the model which
was not yet analyzed in details, albeit the model is
said to be implemented in practically all commercial
codes available. For example, in the latest review by
Veynante and Vervisch published recently in [1], the
model has been referred as belonging to the family
of infinitely fast chemistry models and summarized
in only 15 half-lines of the following text: The eddy
dissipation model (EDC) is a direct extension to nonpremixed flame of the eddy break up (EBU) closure
initially devoted to turbulent premixed combustion...
The fuel burning rate is calculated according to:
Yo
Yp
),
wf =
min(Yf , ,
k
s
1+s
(1)
where and are adjustable parameters of the closure. In the above equation, ... the reaction rate is
limited by a deficient species. To account for the existence of burnt gases bringing the energy to ignite
the fresh reactants, this species may be the reaction
product (when is non-zero)... Actually, for large ,
the model is difficult to justify.
In fact, the above review attributes the EDC model
only to its first, Magnussen and Hjertager, M-H, formulation [2]. In this model, the reaction product
term (when is non-zero) is introduced into the rate
expression to account for finite-rate chemistry effects (ignition/quenching) that makes it incorrect to
place the EDC approach in the family of infinitely
fast chemistry models. Some other versions of so
called EDC model, e.g. [3], were formulated without a proper connection with the original EDC approach. Also, Fluent [4] recently claimed the implementation of a new version of the EDC model ...for
the modeling of turbulent finite-rate chemistry. It is
completely different from the eddy dissipation model
that has long been available. This paper is aimed to
fill a gap in the EDC based models analysis starting
with an outline of the classic EDC model features.
Since the retarded injection was simulating on a
relatively coarse mesh generated during the expansion stroke, it required a modification of the collision
model because some default collision models introduce computational artifacts into the spray simula-
rewritten as
dc1
c1 c o
c c1
=
= fr (c) =
d
mix
Turbulence/Chemistry Interaction
The new model formulation is based on the
operator-splitting procedure applied to the reactive
species conservation equations. In terms of this approach, the time differencing was performed in three
steps: the first step was assumed to be the convection
contribution, the second one - diffusion effect without a contribution of micro-mixing, and the third step
was the chemical kinetics effect coupled with micromixing. The last step of such a mass balance can be
interpreted as representing combustion in a constant
volume partially stirred reactor (PaSR) of a computational cell size, where reactions occur in a fraction of
its volume. Since the reaction zone parameters, as a
rule, can not be resolved on a computational grid, the
sub-grid diffusion term due to micro-mixing was approximated with the help of introduction of a micromixing time as suggested in the interaction with the
mean approach [5]. Then, assuming that chemical
processes proceed in such a way that the shortest
chemical time associated with the particular (reference) species was constrained by the micro-mixing
time, a system of PaSR equations consisting of two
sets can obtained
dc1
c
c1 c o
c c c1
= ,
=
= ,
dt
c
mix
c
(2)
c
2
c mix
(4)
(3)
Ns
Y
(cs )
0
sr
s=1
kbr (T )
Ns
Y
(cs )
00
sr
s=1
where kfr and kbr are the rate coefficients for the forward and backward stages of the r-reaction, T is the
temperature, and Ns is the total number of species
in the mixture.
There is a substantial difference between Eqs (2)
and (4) - the first model has the analytical solution,
while the application of the second one requires numerical integration. Combination of the terms in the
model (4) leads to a definition of the reactive volume
as introduced in [7]:
c1 = c + co (1 ),
(5)
where the multiplier = /( + mix ) is wedged between 0 and 1, while is the integration step.
Rewriting the first relation of (4) in terms of the
reactor exit parameters, one can get
c1 c o
c c1
c
(6)
= fr o /(cs o + term
r ),
c
c + mix
cs o f r o
,
cs o + term
r + termr mix
= fr (c1 )
=
(7)
containing no reaction zone parameters c which cannot be resolved on a grid, but replacing their effect
with the rate multiplier . The superscript o and the
subscript s denotes the value at a start of the integration step and the index of the reference species
defined below, respectively.
This expression looks different from a finite-rate
formulation of the original EDC model [8] known as
ws =
M
c c1s
s
,
1 M
mix
(8)
cos fro
termr + term
r mix
c1s
=
c + mix
cos
+ c + mix
1
c1 c1
= H( s , s )
2
c mix
which is perfectly consistent with the PaSR analytical solution (3). In a limiting case of chemical equilibrium, the model approaches the M-H model.
cos
= ( 00 sr 0 sr )[kfr (T )1 kbr (T )2 ] =
term
r
where
1 =
Ns
Y
n=1
(c1n )
0
nr
, 2 =
Ns
Y
n=1
(c1n )
00
nr
1
cs | =0
= (term
r
c1s
c
)
c c + mix
(9)
= cos ,
cos for
,
cos + term
r
As shown in [7], introduction of reference species assures the equilibrium conditions implementation at
for each reaction in the mechanism.
(10)
(11)
(12)
(13)
lo l 2 + 3D
( )3 3
v o lo
(14)
d = lo Re 1+D ,
(15)
= ( ) [( )1/2 ](1)
1
k
= ( ) (c /Ret ) 2 ,
(16)
where = 3(D3)
1+D , and k and are the turbulent
kinetic energy and energy dissipation rate, respectively. From experiments reported in [10] follows
that the fractal dimension for turbulent dissipation
is about D=2.7, i.e
k
mix = ( ) (c /Ret )0.621
In particular, the eddy break-up time k/ corresponds to D=5, and Kolmogorov micro-scale time
k = (/)1/2 - to D=3.
Another approach developed is based on the usage
of Kolmogorov expression for the micro-mixing time
with the molecular viscosity replaced by a fraction of
the effective viscosity assumed to be responsible for
micro-mixing. If, for example, the RNG k model
is employed, the turbulent viscosity related to k, the
turbulent kinetic energy, and , the dissipation rate
of k, is given by the general expression.
s
c k 2 / 2
t = l (1 +
)
(17)
l
= (l + s t + 2 l s t ),
where s t is the standard k- kinematic viscosity.
Provided that the first two terms in the expression
(17) contribute to conventional diffusion transport,
the geometrical mean term can be assumed to determine the characteristic time scale for micro-mixing
in the turbulence/chemistry interaction model, if
written in a form similar to the Kolmogorov time definition, k = (l /)1/2 replacing the molecular viscosity by its turbulent counterpart, i.e.,
mix = (2
p
k
l ts /)1/2 = (2 c k )1/2 ,
(18)
Mole fraction
0.02
0.01
0.0002
0.0004
0.0006
Time, s
0.0008
The simulation of a spray liquid core fragmentation starts by injecting droplets of the size of the
nozzle orifice. The followed droplet breakup is modeled by providing an initial drop size distribution in
a form of the power law as proposed in [16]
n+1 r n
if 0 r ro
ro ( ro )
f (r) =
,
(19)
0
otherwise
where n=0.5, and r and ro are the drop and injector
nozzle radii, respectively.
Since in the KIVA code the droplets are injected as
parcels of the identical mass, the distribution r3 f(r)
normalized to unit total mass is used to describe the
parcel size distribution
n+4 r n+3
if 0 r ro
ro ( ro )
(20)
g(r) =
0
otherwise
C14H28
C7H16
C7H8
0.03
0.001
Figure 1: Calculated fuel surrogate and its component time histories in the course of combustion development in the premixed mixture (=0.5) at po =41.0
bar, To =900 K.
Detailed reaction mechanisms for n-heptane and
toluene oxidation have been developed, validated using shock-tube and constant volume auto-ignition experimental data, reduced (using SENKIN code and
reaction path analysis software of the Chemkin-2
[12] package) to the mechanism consisting of 68
species participating in 285 reactions and implemented into the KIVA-3V code [13] to simulate diesel
spray combustion. This mechanism was proven (see
[14]) to predict the fuel auto-ignition, combustion development and formation of polycyclic aromatic hydrocarbons (PAHs) till gaseous soot precursors based
on acenaphthylene, A2R5, then converted directly
into soot, the substance assumed to have graphite
properties. The mechanism for both aliphatic and
aromatic components includes principal fragments
of C1 -C7 and O/H chemistry taken from different
sources, but mainly from LLNL and MIT databases
posted in the Internet. The fuel surrogate decomposition into constituent components is treated as
two global stages of oxidative pyrolysis. The reaction
mechanism and species thermodynamic properties
are provided by [15]. Fig. 1 illustrates the dynamics of the diesel fuel surrogate decomposition into
constituent components in chemical kinetics simulations. The physical properties of the fuel surrogate
were taken as properties of real diesel fuel compiled
in the DI model of the KIVA-3V fuel library with the
original fuel chemical formula replaced by C14 H28 .
n
exp(),
n!
(21)
where = t is the mean expected number of binary droplet collisions during a computational time
increment, t, and the collision frequency between
the droplets reads
=
n2
(R1 + R2 )2 |~u12 |
V ol
R1 + R2 C2
t ,
)e
12
(22)
c
Solving the above equation, one can get
c
cs = c1s
+ c
2 Usage of a PSR solver of the Chemkin-2 (or Cantera) package
with the residence time equal to mix seems more reasonable.
c + mix
2
c mix
f (c) =
131 103 m
150 103 m
1.0 103 m
260 103 m
0.235 103 m
1500-1700 rpm
0-5.8 ATDC
8.4 CA
velocity table
0.062 g
300 - 500 m/s
1.31 bar
340-350 K
153.0 deg
9.5 - 12.5 deg
350 K
80
60
40
Calculation, SOI=0. Ca
Experimental data
20
0
30
20
10
10
20
Ca ATDC
30
40
50
60
(a)
Volvo D12C Engine Modeling
(Retarded injection)
300
250
SOI=0.0 Ca
SOI=3.0 Ca
SOI=6.0 Ca
RoHR, J/Ca
200
150
100
50
0
10
10
20
30
such conditions, the ignition can occur at several regions simultaneously and then rapidly spreads over
the chamber volume similar to the HCCI (homogeneous charge compression ignition) or MK (modulated kinetics) regime. In this case the entire heat release occurs in the premixed mode, although again
it is apparent that the mixture is not well premixed.
This regime of diesel spray combustion is reproduced
in the modeling as illustrated by Figs 2 a-b) by predicted in-cylinder pressure and RoHR vs Ca histories
and temperature distributions in Fig. 3. The temperature plot illustrates that the combustion zone is
more distributed compared with the classic diffusion
flame structure. From the plot one can also conclude
that with the help of the new collision model a realistic spray pattern has been predicted. The corresponding in-cylinder pressure vs time histories for
40
Ca ATDC
(Retarded injection)
100
Pressure, atm
80
60
40
SOI=0.0 Ca
SOI=3.0 Ca
SOI=6.0 Ca
20
0
10
10
20
30
40
Ca ATDC
Figure 4: Calculated n-cylinder pressure vs time histories at different SOI illustrating MK combustion.
References
[1] Veynante, D., Vervisch, L., Turbulence Combustion Modeling, Progress in Energy and Combustion Science, 28: 193-266 (2002)
[2] Magnussen, B.F., Hjertager, B.H., On the Numerical Modeling of Turbulent Combustion
with Special Emphasis on Soot Formation and
Combustion, Sixteenth Symposium (International) on Combustion, Pittsburgh: The Combustion Institute, p. 719-729 (1976)
[3] Kong, S.F., Marriot, C.D.,Reitz, R.D., and Christensen, M., Modeling and Experiments of HCCI
Engine Combustion Using Detailed Chemical
Kinetics with Multidimensional CFD, SAE Paper 2001-01-1026 (2001)
[4] Fluent-6:http://www.fluent.com/software/fluent/
focus/reacting.htm (2002)
[5] Aubry, C., and Villermaux, J., J. Chemical Engineering Science, 30, p. 457 (1975)