Date post: | 10-Dec-2023 |
Category: |
Documents |
Upload: | independent |
View: | 0 times |
Download: | 0 times |
Introduction: process and hence are crucial from the modeling
point of view. Accurate prediction of ignition delay Application of computational fluid dynamics (CFD) requires proper chemical reaction modeling. There has recently grown very popular as a means to are four different approaches to specify chemical investigate various in-cylinder processes. Along kinetic mechanisms and they include detailed with the main combustion event, there are many mechanisms, skeletal mechanisms, reduced physical processes that need to be described mechanisms and global mechanisms [2]. The carefully to predict proper mixture formation detailed mechanisms involve all those chemical process. The mixture formation can affect the species and reactions which are pertinent but the combustion efficiency and hence the pollutant application of these detailed mechanisms in formation. One approach is to use empirical internal combustion engines is limited due to two relationships to describe different physical aspects reasons. Primarily, it is attributed absence of the of the combustion events. There is a drawback in proven mechanisms of those hydrocarbons, such as this approach as the assumptions of the steady state diesel, that involve a large number of homologous conditions are less likely to be valid for transient series . Secondly, the requirements for conditions as encountered in compression ignition computational resource for such simulations are engines [1]. The breakup of the fuel stream into large. The skeletal mechanism provides an droplets followed by the evaporation of these alternative course. In skeletal mechanism, the droplets are critical in the mixture formation
Prediction of Heat Release Rate and Engine Emissions Using an
Unsteady Flamelet Model Approach for a Compression Ignition
Engine 1, 2 1 3 1 1 1S. Imran , F. Noor *, R. Shad , A. Hussain , Z. Anwar , G. Ahmad
Abstract
Diesel Unsteady Flamelet Model (DUFM) approach along with probability density function (PDF) has been
used to investigate the effect of breakup length on the in-cylinder pressure and the rate of heat release. Two spray
breakup models, WAVE and Kelvin-Helmohltz Rayleigh Taylor (KHRT), have been investigated by varying the
magnitude of the breakup constants. The optimized set of parameters are used to predict the in-cylinder
pressure, the rate of heat release and NOx emissions in a direct injection (DI) engine. A chemical mechanism of
n-heptane with 29 species and 52 reactions has been used.
Keywords: Chemical kinetics, Computational combustion, unsteady flamelet model, spray and combustion,
compression ignition,
Submitted: 06/03/2015, Accepted: 30/03/2015, Online: 07/04/2015
1.Faculty of Mechanical Engineering, University of Engineering and Technology Lahore, Pakistan
2.School of Engineering and Materials Science, Queen Mary University of London, Mile End Road, E1 4NS, UK
3.Department of Mechanical Engineering, University of Central Punjab, Lahore , Pakistan
*Corresponding author; Fahad Noor ([email protected])
JPIChE 43 (1) 2015: 21-36
journal homepage: www.piche.org.pk/journal
Journal of Pakistan Institute of Chemical Engineers
218-985-1 RV
21
Journal ofThe Pakistan Institute of
Chemical Engineers
Vol. XXXXIII 2015 ISSN 1813-4092
reduction in the number of species and the chemical limits burns quickly and as this process is primarily
reactions and hence the elimination of the faster governed by reaction kinetics, it is named as
premixed combustion. Diffusion or non-premixed reactions by categorizing them on the basis of their
combustion describe the slow burning of the relative significance within the full mechanism.
remaining unmixed or non-evaporated fuel which This is the easiest way to achieve mechanism
burns once properly prepared and as this reduction, but still insufficient to be adopted for
preparation is strongly dependent on mixing, this heavy hydrocarbons as diesel. Reduced
stage of combustion is generally referred as mixing mechanisms can further simplify the mechanism
controlled. For any fuel sprayed in this manner into complexity. To achieve this mechanism
the compression ignition environment, the ignition simplification (reduction), the sensitivity analysis
delay is the sum of both the physical delay as well as is carried by the assuming the steady state
the chemical delay [2]. The ignition delay can be assumptions. The sensitivity analysis analysis is
measured by using equation 2.used to eliminate the species with less contribution.
The last approach to achieve mechanism reduction (2)
is through the use of the global mechanisms.
Minimization of the number of relevant variables where is the ignition delay, P and T are the
makes these global mechanism approaches very average in-cylinder pressure and temperature
attractive especially to the CFD studies. A global before ignition. E is the activation energy whereas Amechanism can be represented by the following R is the universal gas constant. The experimental uequation,constants are represented by A and n. The above
(1)relationship shows that the ignition delay is a
strong function of temperature and pressure as
decreases when the temperature and pressure are Fuel and oxidizer mass fractions are represented by
increased. This equation has been successfully [F] and [O] whereas A, n and EA are typical
applied in number of studies. Some other Arhenius constants. n-Heptane is generally used as
parameters have also been investigated for their a diesel surrogate because of the comparable octane
possible e?ects on ignition delay. They include fuel number [3].
air equivalence ratio [46], the cetane number [7] Ignition Delay: and the engine speed [8]. Following relationship [Eq
The combustion process in a diesel engine is a 3] was proposed by Livengood and Wu [9] and used
combination of pre-mixed and diffusion combustion in many CFD codes [3].
processes. Liquid fuel such as diesel is injected into
the hot stream of air at the end of the compression
(3)stroke. Disintegration of the diesel fuel into smaller
droplets is followed by processes like evaporation Eq. 3 can be written as leading towards the formation of fuel vapor mixture
with in the combustion chamber. The time delay (4)
between the injection and the preparation of the
fuel for burning is referred as physical ignition When the value of the above integral (Eq. 4) exceeds
delay. Some low emperature reactions have been unity, combustion mode is activated.
reported to take place during and after the physical Spray Modelsignition delay leading to an additional delay that is
In dense sprays, the radial growth of sprays can be known as chemical delay. Auto-ignition is a
affected by the following factors [10],phenomenon in which a portion of the fuel which is
mixed with air su?ciently and is within combustible - Droplet Collisions
22 Journal of the Pakistan Institute of Chemical Engineers Vol. XXXXIII
)exp(][][][
TREOFATdt
FduA
ban -=
)/exp()( TREAP uAn-=t
11
1
0
=ò=
-
dtt
tt
1)/exp(
11
0
)(=ò
=
-
-dt
TREAP
t
t uAn
- Droplet Dispersion force. The following second order differential
equation is used to represent the TAB analogy- Levels of turbulence at the spray edge
Droplet collisions are important because they can (6)serve as a mechanism for droplet growth as well as
can contribute to droplet break up. Apart from
droplet collisions and coalescence, droplet Are the terms representing initiating dispersion can be another important factor affecting restoring and damping forces divided by the mass of spray growth. Droplet dispersion (Droplet-gas the body, whereas the instantaneous displacement phase turbulence interaction) that affects of the body from the equilibrium position is turbulence kinetic energy (k) and its dissipation described by x. The corresponding counterparts of rate (å) and hence may serve as mechanism for these three terms in liquid breakup environment droplet growth. Due to high slip velocity between are represented in the following equationthe droplet and the gas phase, the turbulence level
at the edge of spray can be high. This injection
(7)related high levels of turbulence can affect the
Substituting dimensionless y=x/C r, the above bradial growth of the spray. Apart from WAVE spray equation (Eq. 6) for acceleration takes following model, Taylor Anology Breakup (TAB) and Kelvin form,Helmohltz Rayleigh Taylor (KHRT) are the other
two spray model available for use in FLUENT.
Weber number is the main parameter that is used to (8)
define the breakup physics. It is defined as Orourke and Amsden [11] analytically solved this
the above second order differential equation [8]. (5)
Total fragmentation of the parent drop into a
number of daughter dropletss is resulted in TAB injWhere W , , u , d and are weber number, e g inj model but the model does not give any information
density, uniform velocity of the undistributed jet, about their numbers and sizes. Energy balance is
nozzle diameter and surface tension of the liquid used to estimate the number daughter droplets and
droplet, respectively. An analogy of spring mass their size. Under prediction of the droplet sizes of
system is used to describe the oscillation of the full cone diesel spray while using TAB model has
distorting droplets. The restoring force, the been reported [12, 13]. Park et al [14] reported that
external force and the damping force are the TAB model underestimates the penetration
represented by liquid surface tension, droplet drag when combined with Blob-method. Reitz [15]
and liquid viscosity respectively. The Taylor presented the KH breakup model based on the
Analogy Break up model (TAB model), originally Kelvin-Helmohltz instability . The basis for this
proposed by ORourke and Amsden [11], uses a breakup model is a fact that sinusoidal waves are
forced oscillating spring-mass system to describe setup on the surface of the liuid as it passes through
the behavior of an an oscillating drop that the nozzle hole. As the liquid jet enters into the
traversing through a gaseous atmosphere. The chamber, the interaction between the liquid and gas
aerodynamic forces deforming the droplet by a force phases help the sinusoidal waves to grow as a result
F that initiates the oscillations. The liquid surface of different aerodynamic factors. The growth rate .
tension responsible to keep the spherical shape of At which a wave grows on the droplet surface and
the droplet intact and resist the droplet deformation wave length of these waves can be expressed as
is represented by restoring force of the spring. The
viscosity effects resulting in the friction forces (9)
inside the droplet are represented by the damping
232015
23
2
,,lr
Cm
d
lrC
m
K
lr
guC
m
F ldk
relF
r
m
r
s
r
r==
sr inj
injge
duW 2
)(=
xm
dx
m
K
m
F&,,
ylr
Cylr
Cr
u
lC
Cy l
dkrelg
b
F &&&232
2
r
m
r
s
r
r--=
)4.11)(1(
38.034.0][
7.0
5.15.0
30
TZ
Wer gl
++
+=W
s
r
S. Imran, F. Noor, R. Shad, A. Hussain, Z. Anwar, G. Ahmad
24 Journal of the Pakistan Institute of Chemical Engineers
(10) generally used in combination with Rayleigh-
Taylor (RT) method to represent secondary breakup
of the droplet. Instability can be induced along the Where Z and T are the Ohnesorge number and the
normal to the interface between the two fluid of Taylor number respectively whereas r is the radius 0
different densities if the interface experience any of the jet before distribution. These two numbers
acceleration or deceleration [17]. Rayleigh-Taylor are functions of Reynolds number Re and wave
(RT) model is based on the Taylor's work of number W can be expressed ase instability. On a liquid-gas interface, if the
acceleration is directed into liquid, the interface is
(11) stable but disturbances (waves) can be setup if the
acceleration is directed into the gaseous medium. The breakup of a parent droplet of radius r can also The growth rate at which these waves grow and be modeled by applying the wave theory as proposed the wave length of these waves can be expressed asby Reitz [15]. Due to the assumption that child
droplet is formed by shearing of the surface of the
parent droplet, the radius of the child droplet is (14)
proportional to the wave length of the parent
droplet (15)
(12) The droplet is only allowed to breakup if its is
small than its diameter. Similar to the adjustable The KH model of spray breakup differs from the
constant B1 [Eq. 13] for the KH breakup model, RT TAB model in way that unlike TAB model, the
model has an adjustable constant in the form of C3. former approach does not allow complete breakup of
Both of these constants serve the same purpose and the droplet. In this approach, the droplet loses its
take into account the effect of initial conditions like mass continuously as it is penetrating into the
turbulence and cavitation on the secondary droplet gaseous medium. The rate at which the droplet
breakup. In order to overcome the deficiencies radius reduces is function of the original (parent)
reported in the two spray models (KH and RT), droplet radius, child droplet radius and
these two are generally applied in combined form as characteristic breakup time .bu
KH-RT spray model. In KH-RT model, both KH and
RT models compete are allowed to calculate (13)
unstable waves. If the breakup is predicted by RT B0 in equation [12] is a model constant whose
model with actual time step, the whole droplet values is fixed at 0.61 whereas B1 in equation [13] is
breakup occurs according to RT model. Otherwise an adjustable parameter whose value can range
the child droplets and reduction in diameter is between 1.73 [15] and 60 [16]. B1 takes into account
obtained through KH the influence of nozzle geometry (any turbulence
created) on the spray breakup. It has been
suggested that a higher value of B1 shall reduce the
droplet breakup and increase the penetration
length whereas a smaller value can increase the
spray disintegration and result in better fuel mixing
and also reduced penetration. One of the reported
drawback of this method of spray is that it results in
droplets with bigger than the KH radius in areas
near the nozzle tip. This is why this model is
model. If RT breakup model is
applied to the spray portion immediately after
nozzle exit, the diameter reduction is too fast.
Therefore it is only applied after a certain length
and only KH breakup is allowed near the nozzle exit
[18].
Droplet Drag Modeling:
When the droplet interacts with the gas inside the
combustion chamber, they exchange their
momentum. As a result the droplet decelerates and
Vol. XXXXIII
g
cl
WeTR
WeZ == ,
)(
)]([
33
22/3
gl
gla
rr
rr
s -
-=W
)(
323
glaC
rr
sp
+=L
6.067.1
7.05.0
0 )865.01(
)4.11)(45.01(02.9
gWe
TZ
r +
++=
L
L= 0Brnew
25
the gas phase accelerates due to difference in their
velocities before the exchange. The droplet (18)experiences a drag force. A spherical droplet of
radius r moving with a relative velocity u shall rel When y=0, it behaves as a sphere and when y=1 it experience a drag FD given by the following behaves as a disk.equation
Modeling NOX
(16) NO is a generic name given to a three different X
compounds; Nitric oxide (NO, the major
contributor), nitrogen dioxide (NO ) and nitrous 2A is the frontal area of the sphere. The drag f
oxide (N O). The NO formation can be attributed to 2 Xcoefficient CD for the sphere depends upon following chemical kinetics processesReynolds number for the gas phase Reg for low
- Thermal NO formation Reynolds number (Re<1000). X
Prompt NO formation X
(17) Fuel NO formationX
For Re>1000, a constant value of 0.424 is used [19]. Thermal NO X
The assumption that the droplet maintains its The oxidation of nitrogen present in the air is shape as a sphere does not hold when it interacts responsible for thermal NO formation. Extended Xwith gas with a sufficiently large Weber number. It Zeldovich mechanism is generally used to is deformed. The drag coefficient becomes a function determine the formation of thermal NO . These Xof Re as given in equation 17 as well as of its
reactions are highly temperature dependent.amplitude of oscillation. The drag coefficient of a
distorting drop can be modeled as lying between the
CD of a sphere, the lower limit, and the CD of a of (19)
disk.
(20)
(21)
The original Zeldovich mechanism consists of only
two reactions (Eqs 19 and 20). Once the third
reaction (Eq. 21) is included, it becomes an extended
Zeldovich mechanism. The third reactions has been
reported to contribute towards thermal NOX
formation when operating in rich mixtures [3]. The
net rate of reaction for the three reactions involved
in the Zeldovich extended mechanism can be
expressed as
-
-
2015
Fig. 1: Illustration of dynamic drag on a fuel droplet
)632.21(, yCC sphereDD +=
]][[]][[]][[]][[]][[]][[ 3,2,1,3,22,21, HNOKONOKNNOKOHNKONKNOKdt
NOfrrfff ---++=
(22)
S. Imran, F. Noor, R. Shad, A. Hussain, Z. Anwar, G. Ahmad
2
232
3
4
2 dt
xdrACuF lfDrel
gD pr
r==
]6
Re1[
Re
243/2
,g
g
sphereDC +=
NONNO +Û+ 2
NNOON +Û+2
NOHOHN +Û+
26 Journal of the Pakistan Institute of Chemical Engineers
where K and K represent the forward and reason why NO formation is highly temperature f r X
backward reaction rate constants respectively. The dependent. Nitrogen atom has small activation
first reaction in Zeldovich mechanism involves an energy and in fuel-lean flame (when there is enough
N molecule. The two nitrogen molecules are bonded oxygen), the rate of nitrogen consumption becomes 2
equal to its formation.together through a triple bond with a dissociation Hence a quasi-steady state can be assumed. Eq. 22 energy of 941 KJ/gmol. This step is a rate limiting can be written as
step in Zeldovich mechanism. This is the main
The dependence of the rate of NO formation on the (25)X
oxygen concentration and the temperature is clear
from the Eq 23. The concentrations of two [O] and For partial equilibrium approach, the eq. 25 can be
[OH] radicals are required apart from the written as
concentration of the two stable species O2, N . 2
(26)Various modeling strategies are adopted to
The partial equilibrium approach has generally determine the concentration of [O] and [OH] been reported to result in higher NO formation rate. radicals [3]. Equilibrium, Partial equilibrium and The predicted local concentration approach uses an local concentration are the three approaches that advanced chemistry model and takes [O] from the can be applied to determine the concentration of [O] local O species mass fraction. radical.
Conservation Equations Slower kinetics of the NO formation when X
The continuity equation can be written ascompared to the rate hydrocarbon oxidation
suggests that the thermal NO are formed after the X
(32) completion of the main combustion event. The
Using Einstein's notation, the continuity equation equilibrium approach calculates the formation of can expressed as the thermal NO after decoupling it from the main X,
(33) combustion process, by assuming that the
combustion reaction has attained equilibrium. The
following equation provides the equilibrium where D , Dt is substantial derivative of . For
concentration of the oxygen atom.incompressible flows, is constant. The continuity
equation reduces to(24)
One problem with the equilibrium approach for the (34)
determination of [O] radical concentration is the Using the same notations as for the continuity assumption itself. The concentration of [O] radical equation, the momentum equation can be expressed is usually more abundant than the equilibrium aslevels. The remedy to this problem is to assume the
concentration of [O] radical in partial equilibrium. (35)
As an improvement to the equilibrium approach, a
third body reaction can be included in to the The energy equation can be expressed as:
dissociation-recombination of oxygen
Vol. XXXXIII
)./(
][][
][1
][][
][1
]][[2 3
3,22,
1,
22,21,
22,1,
21, smg
OHkOk
NOk
OkNk
NOkk
NOKdt
NO
ff
r
ff
rr
f
+
-
=(23)
)/(][64.36][ 3/271232/12
2/1 mgmoleOTO T-=
0)( =¶
¶+
¶
¶u
xt i
rr
0=¶
¶+
i
i
x
u
Dt
Dr
r
0)(. =¶
¶==Ñ
i
i
x
uudivu
j
i
ij
ji
ji
jj fxx
p
x
uu
t
u
Dt
Du+
¶
¶+
¶
¶-=
¶
¶+
¶
¶=
trr ][
2/12 ][][ OkO p=
MOOMO ++Û+
27
(36) at a suitable time. A probability density function
table is generated at every fractional time step. This
table provides the flow field properties for he next Methodology
fractional time step [3].Unsteady Flamelet Approach
Pdf Approach:In the finite rate chemistry approach, the chemical
Being highly turbulent in nature, the combustion in source term is calculated by applying Arhenius
IC engines is governed by Navier Stokes equations. equation which does not take into account the
Various methods are employed to solve these eddies produced as a result of turbulence. The finite
Navier Stokes equations as the direct solution of rate chemistry model is more suitable for laminar
these equations computationally expensive. flames but it can also be applied to the situations
Reynolds averaging of these equations is one way to where there is small chemistry turbulence
solve these equations. For turbulent chemistry in interactions. The unsteady flamelet approach
IC engines, the Reynolds averaging approach achieves significant reduction in calculation time
results in two unknown terms for turbulent scalar when compared to the finite rate chemistry
flux and mean reaction rate. ANSYS FLUENT uses approach. This model solves the flamelet species
gradient diffusion to model turbulent scalar flux and energy equations at the same time it solves the
and treats turbulent convection as enhanced flow equation. The solution of the flamelet and the
diffusion. Laminar Eddy Dissipation Concept flow equations is advanced simultaneously. For
(EDC) and finite rate chemistry models are used to every fractional step, the properties of the flow are
model the mean reaction rate [3].used to calculate the properties of the flamelet
A transport equation for single-point, joint which in turn provide the input to calculate the flow Probability Density Function (PDF) can properties for that fractional. The initial flamelet alternatively be derived instead of Reynolds distribution is mixed but unburnt. Different averaging of the species and energy equations. The variables including temperature, pressure as well pressure and temperature of all the species present as scalar dissipation are volume-averaged scalar. at every time step are reflected in this PDF. For N The values of these variables for both the fuel as species the PDF has N+2 dimensions (N arises due well as the oxidizer are first calculated in a a flow to the number of species and +2 arises from the solver. The output of the flow solver serves as input temperature and pressure). Based on Pope's work to flame solver. This is repeated for every fractional [23], the following transport equation for PDF has step. The compression results in the flamelet been derived [3].temperature increase leading towards the ignition
Where is density, P and u are Favre averaged PDF expectation of X occurring provided that Y has i
occurred. The three terms on the right hand side of composition and mean fluid velocity, Ø and S are k
the eq. 37 represent unsteady rate of change of PDF, the composition space vector and rate of reaction for change of PDF due to convection arising from the species k respectively. and Ji. k are vectors
mean velocity field and the change of PDF due to representing fluctuation in fluid velocity and chemical reactions. All the terms are closed and molecular diffusion flux respectively. The two terms hence do not require any modeling. The closure of on the right hand side involves expectation and this heavily nonlinear term representing chemical conditional probability and can be summarized in reactions is a great advantage. Changes in PDF due the following analogy: < X | Y > represents the
2015
321
)().(dxdxdx
QuyuT
Dt
De s
xi
j&
&& +¶
¶+Ñ-ÑÑ= tlr
]1
[]/[)()()( , Px
J
pPu
xPSPu
xP
t i
ki
ki
k
k
i
i ¶
¶
Y¶
¶+Y¢¢
¶
¶-=
Y¶
¶+
¶
¶+
¶
¶rrrr
(36)
S. Imran, F. Noor, R. Shad, A. Hussain, Z. Anwar, G. Ahmad
28 Journal of the Pakistan Institute of Chemical Engineers
to turbulent scalar flux and molecular diffusion are solution becomes time expensive. Theoretically, the
represented by the terms on the right hand side of pre-tabulation of chemical results can help to avoid
the eq. 37. The two terms on right hand side are not integration via lookup tables but practically the
closed and hence require modeling. The composition scheme has many pitfalls. Any table constructed in
vector which is a function of mass fraction of this manner would have N+3 dimensions, where N
different species, temperature and pressure can be stands for the number of species and 3 is for
expressed as follows pressure, temperature and time. If each dimension
has X points, the total number of data points will be (38) XN+3. For 10 points in each direction and 10
where YN is the mass fraction of the nth species. species, the table would have 1013 data entries.
Integrating the reaction source term through Another problem with this pre-tabulation is that it
reaction fraction step results in following equation. stores data on some reactions that are at unrealistic
conditions, i.e., a data entry for OH radical at 300K.(39)
Computational Models
oA 90 sector mesh is used to take advantage of the Where S is the chemical source term. Some symmetry of the four equally spaced injector nozzle chemical reactions are fast and has a time scale of holes. ICEM-CFD, a state of the art mesh 10-10 seconds whereas other reactions may be generation package, was used to generate the mesh. considered slow and has a time scale of 10-3 With piston at the bottom dead center (BDC), the seconds. The difference in these time scales of mesh consists of 26788 cells. The mesh is moved results in numerical stifness. Direct or pre-from BDC to the crank angle where intake valve integration of chemical results are others options. closes. This is done through the dynamic mesh An earlier study [4] in the group has used pre-utility.integration approach to map the combustion
response in IC engines. This study used SENKIN, a Numerical Scheme and Selection of Models
sub-program in CHEMKIN-II to calculate the Unsteady flamlet model (UFM) has been used to detailed chemical kinetic reactions of air/fuel predict in-cylinder pressure, the rate of heat release mixtures at different temperatures, pressure and and different emissions. Two different spray compositions. These calculations were done prior to breakup models have been used in the UFM the engine simulations. The reaction results were approach; WAVE breakup model and KHRT. Three decoupled from their chemical time scales (order of different values of each breakup constant were used
-10about 10 and then integrated and saved in ) for the parametric study conducted to assess the 5physical time scale (order of about 10 ) in a database effect of the breakup constants on ignition delay,
file. The reactions results of different initial rate of pressure rise and the maximum in-cylinder conditions (temperature, pressure and composition) pressure. A bigger time step is used during are stored in different zones. The zones were compression of the in cylinder mixture. When
oindexed using their respective reaction conditions. injection starts, the time step is reduced to 0.1 CA These reaction results were retrieved from this for injection and combustion events. All reactions database file when needed during the simulation. It are switched on. The different physical aspects of took 29 days to generate the database file. the fluid are modelled as follows:
For a simulation containing 40,000 cells and 15
particles per cell, if the convergence is achieved in 81500 iterations then 10 stiff ordinary differential
equations are required to be solved. If each
integration is done in a few milli seconds, the
Vol. XXXXIII
),,......,,( 321 pTYYYY N=f
ò+=dt
Sdt0
01 ff
29
-
-
-
-
-
-
-
-
been used due to 3-D nature of the
computational domain. It provides general
shape of the spray.
Secondary breakup model: Both WAVE and
KHRT536 breakup models have been utilized
as described in the following section.
Droplet deformation: Dynamic drag model has
been used due to its ability to capture droplet
deformation540 due to aerodynamic forces.
Droplet collision and coalescence: Droplet
collision and coalescence models have been
considered. These models capture head-on
collision, side collision and multi-body collision
and coalescence. If these models are not used,
the prediction of particle breakup due to
collision is not captured.
Wall-film: This model is used to predict the
formation of thin fuel film at contact regions.
This model takes into account the particle
splashing, convective heat transfer from wall
to fuel film and the evaporation of the fuel film. Turbulence: RNG k-å has been used. It takes
into account swirling flow. Its computational NO : Extended Zeldovich mechanism has been X
cost is lower than LES model. used to model the formation of NO . Thermal X
and prompt NO have been predicted. The Discrete Phase: Langrangian approach has X
effect of turbulence on the formation of NO has been used to track individual droplets. It X
requires smaller computing resource and been considered. Temporal fluctuation of
simulation runtime. temperature and species concentration result
from the turbulence.Primary breakup: Solid cone injection type has
2015
Fig. 2: Computational Mesh used in the study
Fig. 3: General numerical scheme used in the study
S. Imran, F. Noor, R. Shad, A. Hussain, Z. Anwar, G. Ahmad
30 Journal of the Pakistan Institute of Chemical Engineers
ignition delay for the two lower values of the Results and discussion
breakup constant B and a longer ignition delay 1
Parametric Study when B was set at 50 can be explained on this basis. 1
Three values (10, 25 and 50) of the breakup model All three cases of UFM with different values of B 1
constant B are used. Three values (10, 30 and 50) of 1 have captured the ignition delay trend. The peak
the KHRT breakup model constant C are used. pressure occurs at the same crank angle position for 3
When unsteady flamlet model (UFM ) uses WAVE all three cases. This can explain the higher rate of
model for the secondary break, the effect of pressure rise when the ignition delay is maximum
variation in B is significant. Ignition, rate of at B =50. With shorter breakup length, the fuel is 1 1
pressure rise as well the maximum in-cylinder atomised and combustion commences earlier but
pressure are all affected by any change in this the dispersion of the fuel may not cover the whole
breakup constant. As the value of this breakup volume of high temperature air available at that
constant is increased, ignition is slightly delayed particular crank angle (reduced air utilization).
and a higher rate of pressure rise as well as higher With longer breakup length, the fuel atomisation
peak in-cylinder pressures are observed. The and hence ignition is delayed but air utilization is
minimum value of B produces the best agreement improved. Hence a larger proportion of the fuel is 1
consumed at the same time leading to higher rate of with the experimental data. Theoretically, a higher
pressure rise and hence increased heat release rate. value of B shall reduce the breakup and hence 1
When UFM uses KHRT model to simulate the increase the penetration length. On the other hand,
secondary breakup model, a very good agreement is a smaller value of B results in increased spray 1
noted between the experimental and the atomisation. The increased spray atomisation computational results.results in better fuel/air mixing. Relatively shorter
Vol. XXXXIII
Initial and Boundary Conditions
Table 1 summarizes the initial boundary conditions used in the simulations.
Table 1: Initial and boundary conditions used in the simulation
Sr No. Initial boundary conditionso1 Start crank angle -140 BTDC
o2 Stop crank angle 130 ATDCo3 Time step during compression stroke 0.5o4 Time step during injection, ignition and combustion 0.1
5 Temperature of cylinder wall 545 K
6 Temperature of cylinder head 610 K
7 Temperature of cylinder piston 650 K
8 Initial pressure 0.16 MPa
9 Fuel inlet temperature 370 K
10 Pressure discretization scheme Standardnd11 Turbulent dissipation rate discretization scheme second 2 Order Upwind
312 Density for diesel 840 kg/m
13 Saturation vapour pressure scheme Piecewise Linear
14 Droplet surface tension calculation Piecewise Polynomial
31
The effect of change in C on ignition delay, rate of value of C . Larger diameter droplets shall take 3 3
pressure rise and the peak in-cylinder pressure is longer to evaporate and hence result in slower rate
less pronounced when compared to the effect of of pressure rise. For the largest C value, the 3
change in WAVE breakup model constant B . For pressure drops more rapidly in the expansion stroke 1
the two smaller values of C , no difference was of the cycle.3
observed in terms of ignition delay, rate of pressure In-Cylinder Pressure and Rate of Energy rise and the peak in-cylinder pressure. When C was 3 Releaseset a very higher number, smaller rate of pressure
Figures 4 and 5 show computationally predicted in-rise and lower peak in-cylinder pressure were
cylinder pressure using UFM approach utilizing the obtained. According to the model, the breakup will
WAVE and KHRT breakup models respectively. An only occur if the wavelength of the wave growing on
overall better agreement is achieved with UFM the surface of the liquid droplet is smaller than the
approach when KHRT622 breakup model is used. droplet diameter.
Figures. 6 and 7 show computationally predicted A larger value of C shall reduce the breakup and 3 rate of energy release using UFM approach result in larger droplet diameter. This can explain utilizing the WAVE and KHRT breakup models the smaller rate of pressure rise with increasing respectively.
2015
Table 2: Effect of Wave breakup constant B on peak combustion pressure1
Chosen value of B Peak Combustion Pressure/ Mpa1
10 6.085
25 6.261
50 6.47
Fig. 4: Comparison of In-cylinder pressure for different values of breakup constant B using wave breakup model1
S. Imran, F. Noor, R. Shad, A. Hussain, Z. Anwar, G. Ahmad
32 Journal of the Pakistan Institute of Chemical Engineers Vol. XXXXIII
Figure 5: Comparison of In-cylinder pressure for different values of breakup constant C using KHRT breakup model3
oFrom Figures 6 and 7, it is clear that the occurs 1 CA before the corresponding peak for the
experimental rate of energy release rises first experimentally obtained rate of energy release
compared to the computational trend line which curve. All predicted results produced higher second
refrects that the model does not predict the ignition rate of energy release peak when compared to the
behavior well but the start of the main combustion experimental results. For the different values of
event is well predicted. This may be due to limited breakup constants tested within the different
knowledge of the injection mass profile, differences models used, B = 10 (WAVE model) and C = 30 1 3
between fuel surrogate and the real diesel, (KHRT model) give better prediction of in-cylinder
cavitation in the injector, pressure changes in the pressures and the rate of energy release. Hence
injector, fuel residuals and injector dribble. With these two values have been chosen to be used for the
UFM, two rate of energy release peaks are observed. FRC model.
The first predicted rate of energy release peak
Fig. 6: Comparison of rate of energy release for different values of breakup constant B using wave breakup model1
33
Prediction of Emissions used. Due to this reason this combination has been
chosen to predict three of the in cylinder emissions. Figures 8, 9 and 10 show comparison of The computational results for three emissions were experimentally and numerically obtained lower than the levels collected experimentally but emissions of specific NO , CO and CO respectively. X 2
the UFM approach has captured the trends. The Of all combinations of combustion and spray models agreement between the computational and the tried with different breakup constants, the UFM experimental results improved as the equivalence produces the best estimate of the in-cylinder ratio was increased.pressure and the rate of energy release when KHRT
spray model with a breakup constant (C ) set at 30 is 3
2015
Fig. 7: Comparison of rate of energy release for different values of breakup constant C using KHRT breakup model3
Fig. 8: Comparison of experimentally and numerically obtained specific NOX emissions
S. Imran, F. Noor, R. Shad, A. Hussain, Z. Anwar, G. Ahmad
34 Journal of the Pakistan Institute of Chemical Engineers Vol. XXXXIII
Fig. 9: Comparison of experimentally and numerically obtained specific CO emissions2
Fig. 10: Comparison of experimentally and numerically obtained specific CO emissions
The computationally predicted specific NO was experimentally obtained specific CO emissions X 2
12% lower than the experimentally obtained values range between 8 to 10 % across a range of
at an equivalence ratio of 0.5. The agreement equivalence ratios. The CO emissions were the least
between the computationally and experimentally well predicted. The agreement between the
obtained specific NO was improved and lie within computationally and experimentally obtained X
specific CO is 15%, 7% and 12% at the three 10% of each other with an equivalence ratio of 0.68.
equivalence ratios tested.The agreement between the computationally and
Conclusions: the rate of energy release.
The study investigated a set spray and combustion Computationally, the results of different
models to obtain an optimized set of these models. emissions are under-predicted for the cases
The optimized set was used to predict in cylinder tested but the trends have been well captured.
pressure, rate of heat release and various engine Nomenclatureemissions. The conclusions of the study are
Abbreviations summarized as follows,
CAD crank angle degrees When unsteady flamelet model is used in CI compression ignition combination with WAVE breakup model, the
DI direct injectionignition, the rate of pressure rise and the
References:maximum in-cylinder pressure are all affected
1. C. Fieberg, L. Reichelt, D. Martin, U. Renz, by any change in the breakup constant. With and R. Kneer. Experimental and numerical an increase in breakup constant (B ), the 1
investigation of droplet evaporation under ignition is slightly delayed and a higher rate of diesel engine conditions. International Journal pressure rise is observed. A higher value of the of Heat and Mass Transfer, 52(1516):3738 breakup constant gives shorter breakup and 3746, 2009.hence reduced penetration length.
2. I. Dhuchakallaya and A.P. Watkins. Auto-A larger value of C shall reduce the breakup 3
ignition of diesel spray using the pdf-eddy and results in larger droplet diameter. This can break-up model. Applied Mathematical explain the smaller rate of pressure rise with Modelling, 34(7):1732 - 1745, 2010. increasing value of C . 3
3. Fluent Inc., New hamshire,USA. Fluent 12.0 Larger diameter droplets shall take longer to
User guide, 2006. evaporate and hence results in slower rate of
4. D.N.Assanis, Z.S.Filipi, S.B.Fiveland, and pressure rise. For the largest C value, the 3
M.Symiris. A predictive ignition delay pressure drops more rapidly in the expansion
correlation under steady state and transient stroke of the cycle.
operation of a direct injection diesel engine. For the different values of breakup constants
Transaction of ASME , Journal of Engineering tested within the different models used, B = 10 1 for Gas Turbines and Power, 125:450- 457, (WAVE model) and C = 30 (KHRT model) give 3 2003. better prediction of in-cylinder pressures and
-
-
-
-
-
Table 3: Percentage variation in the predicted emissions when compared to the experimental data across three different equivalence ratios using KHRT breakup model (C = 30) and unsteady3
flamlet approach
Equivalence ratio Type of emissions % g difference in prediction
0.50 NO -12X
0.55 NO -8X
0.68 NO -5X
0.50 CO -82
0.55 CO -102
0.68 CO -92
0.50 CO -15
0.55 CO -7
0.68 CO -12
352015 S. Imran, F. Noor, R. Shad, A. Hussain, Z. Anwar, G. Ahmad
5. H. Hiroyasu, T. Kadota, and M. Arai. atomization characteristics of gasoline injector
Development and use of spray combustion for direct injection engine. In 15th Annual
model to predict diesel engine e?ciency and Conference on720 Liquid Atomization and
pollutant emissions. Bulletin of JSME, Spray Systems, pp4751, 2002.
26:569575, 1983 15. R.D. Reitz. Modeling atomization processes in
6. S. Ikura, T. Kdota, and T. Hiroyasu. Ignition high-pressure722 vaporizing sprays.
delay of fuel sprays in a constant volume bomb. Atomization and Spray Technology,
Transaction of Japan Society of Mechanical 3:309337,1987.
Enegineers, 41:1559, 1975 16. Harun Mohamed Ismail, Hoon Kiat Ng, and
7. M.L. McMillan, N. Seigla, D.C.Srinivasan, Suyin Gan. Evaluation of non-premixed
and A.D. Tuteja. Review of gm investigations of combustion and fuel spray models for in-
the effects of fuel characteristics on diesel cylinder diesel engine simulation. Applied
engine combusti on and emiss ions . In Energy, 90(1):271279, 2012. ¡ce:title¿ Energy
Coordinating Research Council, Diesel Fuel Solutions for a Sustainable World, Special
Combustion Workshop, Atlanta, Georgia, Issue of International Conference of Applied
1983. Energy, ICA2010, April 21-23, 2010,
Singapore¡/ce:title¿.8. J.H. Shipinski, P.S. Myrers, and O.A.
Uyehara. A spray droplet model for diesel 17. G.I. Taylor. The instability of liquid surfaces
combustion. Symposium on Diesel Engine when accelerated in a direction perpendicular
Combustion Inst itut ion of Mechanical to their planes. In Batchelor732 GK (1963) The
engineers, 184-3J:2835, 1970. Scienti?c Papers of Sir Geo?ery Ingram Taylor,
3:532536 Cambridge University Press, 1963. 9. C.J. Livengood and C.P. Wu. Correlation of
auto ignition phenomena in internal 18. M. Chan, S. Das, and R.D. Reitz. Modeling
combustion engines and rapid compression703 multiple injection and egr e?ects on diesel
ma ch in es . In Th e 5t h In te rn at io na l engine emissions. In SAE paper 972864,1997.
symposium on combustion, Pittsburgh, 19. A.A. Amsden, J.D. Ramshaw, P.J. ORourke, 1955.705 and J.K. Dukowicz. KIVA, A Computer
10. J.I. Ramos. INTERNAL COMBUSTION Program for Two and Three Dimensional Fluid ENGINE MODELING. HEMISPHERE Flows with Chemical Reactions andFuel PUBLISHING CORPORATION, 1989. Sprays. Los Alamos National Labs, Rept LA-
11. P.J O'Rourke and A.A. Amsden. The tab 10245-MS, 1985.
method for numerical calculations of spray 20. D. Mewes and F. Mayinger, editors. Mixture droplet break-up. SAE International Fuels and Formation in Internal Combustion Engines. Lubricants Meeting and Exposition, Springer-Verlag Berlin Heidelberg, 2006.Toronto,Ont., canada, SAE paper No.:872089, 21. A.A. Westenberg. Kinetics of no and co in lean , 1987. premixed hydro carboair ?ames. Comb.Sci. and
12. A.B. Liu, D. Mather, and R.D. Reitz. Modelling Tech, 4:5964, 1971.
the effects of drop drag and break-up on fuel 22. C. Westbrook and F. Dryer. Chemical kinetic
sprays. In SAE Paper no.930072, 1993. modeling of hydrocarbon combustion. Prog.
Energy Comb. Tech, 1984.13. F.X. Tanner. Liquid jet atomizationand 23. S.B. Pope. Pdf methods for turbulent reactive dropletbreakupmodeling of non-evaporating
?ows. Progress Energy Combustion Science,, diesel fuel sprays. In SAE paper 970050, 1997. 11:119192, 1985.
14. S.W. Park, H.J. Kim, and C.S. Lee. An
experimental and numerical study on
36 Journal of the Pakistan Institute of Chemical Engineers Vol. XXXXIII