Sei sulla pagina 1di 14

Energy Conversion and Management 51 (2010) 1835–1848

Contents lists available at ScienceDirect

Energy Conversion and Management


journal homepage: www.elsevier.com/locate/enconman

Effect of fuel injection timing and intake pressure on the performance of a DI


diesel engine – A parametric study using CFD
B. Jayashankara, V. Ganesan *
Internal Combustion Engines Laboratory, Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai 600 036, India

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents the computational fluid dynamics (CFD) modeling to study the effect of fuel injection
Received 10 June 2008 timing and intake pressure (naturally aspirated as well as supercharged condition) on the performance of
Received in revised form 10 April 2009 a direct injection (DI) diesel engine. The performance characteristics of the engine are investigated under
Accepted 7 November 2009
transient conditions. A single cylinder direct injection diesel engine with two directed intake ports whose
Available online 23 March 2010
outlet is tangential to the wall of the cylinder and two exhaust ports has been taken up for the study.
Effect of injection timing (start of injection 16, 12 and 8 CAD bTDC) and intake pressure (1.01, 1.21
Keywords:
and 1.71 bar) on the performance of the engine has been investigated for an engine speed of
Diesel engine
Computational fluid dynamics (CFD)
1000 rpm. CFD predicted results during both suction and compression strokes under motoring conditions
Intake ports have been validated with experimental results available in the literature. Magnusson’s eddy break-up
Swirl ratio (SR) model is used for combustion simulation. Predicted performance and emission characteristics such as
Injection timing pressure, temperature, heat release, NOx, and soot are presented and discussed. The predicted values
Intake pressure reveal that retarding the injection timing results in increase in-cylinder pressure, temperature, heat
Heat release release rate, cumulative heat release and NOx emissions. Decreasing trend is observed by advancing
Emission the injection timing. In case of soot emission the increasing trend is observed up to certain crank angle
then reverse trend is seen. The supercharged with inter-cooled cases show lower peak heat release rate
and maximum cumulative heat release, shorter ignition delay, higher NOx and lower soot emissions.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction temperature and fuel chemistry. In DI diesel engines, the ignition


delay is a function of compression ratio, swirl and speed of the en-
Diesel engine is well known for its higher efficiency and fuel- gine. The heat release period consists of two different modes of
economy as compared to any other type of internal combustion en- burning, viz. premixed and diffusive burning. Initially, most of
gines. Fuel injected in a diesel engine has to mix with air and form the fuel burns in the premixed mode until the fuel air mixture that
a combustible mixture. The rate of combustion, formation of was prepared and ready to burn prior to ignition is exhausted.
smoke and particulates and emission levels are controlled strongly After this point burning proceeds only in the diffusive mode until
by the mixture preparation characteristics. The combustion pro- the end of combustion [24]. The heat release rate of the fuel causes
cess in a diesel engine is an exceedingly complex phenomena. In a variation of gas pressure and temperature within the engine cyl-
diesel engines the complexity of the combustion process is com- inder, and strongly affects the fuel economy, power output and
pounded by the non-uniform fuel distribution within the combus- emissions of the engine. It provides a good insight into the com-
tion chamber and the fuel–air mixing process. The development of bustion process that takes place in the engine. Smoke level falls
fuel jet with simultaneous evaporation of fuel and mixing with air, when the injection timing is advanced. However, at too advanced
controls the rate of heat release, there by the engine power and injection timings the smoke level rises due to fall in brake thermal
efficiency. efficiency, which leads to increased fuel input at any given output.
Combustion in direct injection diesel engine starts at the dy- Advancing the injection timing leads to reduction in smoke level
namic injection point and consists of two distinct phases: the igni- due to increase in premixed combustion phase and reduction in
tion delay and the heat release period. The ignition delay is the fuel consumption on account of higher brake thermal efficiency.
time interval between the actual dynamic injection point and start The NOx formation rate has a strong exponential relationship to
of ignition. The ignition delay depends on the cylinder pressure, temperature. High temperatures in the combustion chamber dur-
ing relatively short, homogenous, premixed combustion result in
higher NOx formation rates. NOx is reduced by decreasing the
* Corresponding author. Tel.: +91 44 22574657; fax: +91 44 22574652.
E-mail address: vganesan@iitm.ac.in (V. Ganesan). pre-mixed burning rate i.e. by the suppression of initial injection

0196-8904/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enconman.2009.11.006
1836 B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848

rate. Lower the peak temperature during initial combustion or gine equipped with two intake valves. The resulting flow was
reduction in peak temperature can lower the NOx emissions. But analysed and compared for different combustion chamber geome-
lowering the peak temperature in the combustion chamber de- tries by using CFD as well as Laser Doppler Velocimetry measure-
creases thermal efficiency which is undesirable. Further, it is diffi- ments. They found that maximum swirl and turbulence velocity
cult to reduce both NOx and particulate emissions simultaneously. values were obtained at maximum valve lift, and maximum swirl
Diesel engine emission control technology is often characterized level was maintained during the second part of the intake stroke.
by its ‘‘NOx versus PM trade-off”. This trade-off refers to the control The geometry of the piston had negligible effect on the flow char-
technology so designed to reduce NOx while increasing production acteristics during the first phase of the compression stroke, and
of particulates. Diesel engine designers must balance this trade-off significant influence was found as and when piston approached
to accomplish compliance with both the NOx and PM. A simulta- TDC. Payri et al. [23] measured the ensemble averaged tangential
neous decrease in NOx and particulate emissions in the exhaust and radial velocity components and turbulence intensity to inves-
of a diesel engine can be realized through proper control of com- tigate the flow patterns inside a four valve single cylinder motored
bustion in which the premixed and the diffusive combustion diesel engine using LDV, for various engine speeds, swirl numbers
phases are optimized. More intake air mass must be forced into and piston bowl geometries, under conditions similar to those of a
the cylinder by supercharging to burn the same quantity of fuel production engine. The work focused on the near wall region of the
and to maintain original engine power density, or to make the DI axisymmetric combustion chambers where strong swirl–squish
diesel engine operate at ultra-lean mixture to reduce NOx and soot and spray–wall interactions took place. Measured values were
emissions simultaneously [9]. Enhancing the intake swirl velocity compared with those provided by a two-zone phenomenological
contributes to the reduction of soot emission in spite of the in- model. LDV measurements for the beginning of compression phase
creased NOx emission. Supercharging can favorably enhance diffu- to half way through expansion phase were available.
sion combustion resulting in improved fuel economy for retarded Han et al. [8] carried out multidimensional computations for a
injection timings and reduced emissions. Swirl and supercharging heavy-duty diesel engine with multiple injections. Different injec-
can also be major parameters influencing in-cylinder air flow pat- tion schemes were considered, and the predicted cylinder pressure,
terns [17]. Uchida et al. [18] showed that the favourable effects of heat release rate and soot and NOx emissions were compared with
intake supercharging on exhaust emissions and fuel consumption measured data. Excellent agreements between predictions and
are achieved when intake pressure is boosted. Thus further inves- measurements were achieved after improvements in the models
tigation on effects of higher intake pressure on exhaust emissions were made. The improvements include using a RNG k–e turbulence
is necessary. model, adopting a new wall heat transfer model and introducing
Sayin and Canakci [3] have carried out an experimental inves- the nozzle discharge coefficient to account for the contraction of
tigation on a single cylinder diesel engine to study the influence of fuel jet at the nozzle exit. The present computations confirm that
injection timing on the engine performance and exhaust emissions split injection allows significant soot reduction with out a NOx pen-
using ethanol blended diesel fuel. Compared to the results of stan- alty. Based on the computations, it is found that multiple injections
dard injection timing NOx and CO2 emissions have increased and have a similar NOx reduction mechanism as single injections with
unburned HC and CO emissions have decreased for retarded injec- retarded injection timings. Kong et al. [11] developed an integrated
tion timing. On the other hand, with the advanced injection timing numerical model (wave break-up model) for diesel fuel atomization
HC and CO emissions diminished and NOx and CO2 emissions and the combined laminar–turbulent time scale model for combus-
boosted. Li et al. [9] investigated experimentally to study the effect tion. The proposed integrated model was applied to simulate three
of various intake compositions and intake pressure on combustion different diesel engines under varying fuel injection timing and
and emission characteristics in a single cylinder DI diesel engine. injection pressures. The overall combustion characteristics and
The ignition delay and the premixed combustion seems to have de- NOx and soot emissions were well predicted with just one model
creased, while the diffusion combustion is slightly enhanced when constant being varied from engine to engine. The constant reflected
the intake pressure is increased. Uchida et al. [17] made an exper- the differences in the injector and the resulting spray break-up
imental investigation to study the effect of supercharging, intake characteristics. The accurate prediction of spray dynamics was also
swirl and fuel injection rate on diesel combustion to reduce ex- crucial to combustion modeling. The effects of drop distortion on
haust emissions, particularly NOx and particulates, without affect- the drop drag forces were considered. The momentum exchange
ing the performance. Supercharging with an intercooling system between the drops and the surrounding gas was then enhanced
results in improved fuel economy and reduced emissions except for distorted drops. The effect slowed down the liquid breakup
for NOx at retarded injection timings. Increasing the injection rate and consequently improved the penetration of a combusting diesel
along with retarding the injection timing reduced both particulate spray. A better spray droplet distribution was then obtained. Com-
and NOx emissions without deterioration of fuel economy. Nwafor parisons with in-cylinder spray visualisation experiments indicated
et al. [19] and Nwafor [20] have conducted combustion studies on that improvements were needed in modeling the breakup and
both diesel – natural gas as well as diesel – vegetable oil fuels, with vapourization of the injected liquid in the early stage of injection.
the standard and advanced injection timing using the same engine. However, a good agreement was found with measured cylinder
It is stated that the longer delay lead to unacceptable rates of pres- pressure and heat release rate data suggested that the overall fuel
sure rise with the result of diesel knock because too much fuel is distributions were predicted accurately. To predict the ignition de-
ready to take part in premixed combustion. Alternative fuels lay accurately, the low temperature chemistry of hydrocarbon oxi-
exhibited longer delay periods and slower burning rate especially dation was treated using multistep ignition kinetics. A reduced
at low load operating conditions. Advanced injection timing is ex- kinetics mechanism, i.e., the Shell model, was sufficient to predict
pected to compensate these effects. The engine performance was the diesel spray ignition in both combustion bombs and diesel en-
smooth especially at low load levels by advancing the injection gines over a wide range of operating conditions. The eddy break up
timing. The ignition delay was reduced through advanced injection timescale was incorporated in the combustion model to account for
but tended to incur a slight increase in fuel consumption. Moderate turbulence mixing effects. Mehta and Bhaskar [14] developed a
advance of injection timing is recommended for low speed opera- multizone phenomenological model by incorporating spray mixing
tions. The CO and CO2 emissions were reduced with advanced model during injection involving droplet evaporation and fuel spray
injection timing. Payri et al. [22] carried out in-cylinder flow anal- air swirl interactions, two equations k–e turbulence model, hetero-
ysis during the intake and compression strokes of a DI diesel en- geneous combustion rate model based on kinetics and eddy-dissi-
B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848 1837

pation concept and single step kinetic rate model for soot processes. by Kong et al. [11], the kinetics is almost totally neglected. The ap-
These models are compared with the experimental data and the proach followed here is based on the Abraham’s characteristic-
predicted results such as cylinder pressure, exhaust soot and NO time concept. Unlike the original approach, in the current model
emissions are in good agreement with experimental data. Patterson the chemistry is taken into account by means of a three-step mech-
et al. [21] used modified KIVA-II code to study the effects of fuel anism based on the well recognized sequential combustion man-
injection characteristics on diesel engine performance and emis- ner of hydrocarbons. The study is focused on the effects of both
sions. The soot formation was modelled using Nagle–Strickland- turbulence and kinetics on the local and overall combustion rates.
Constable oxidation formulation. Their results showed that NOx The purpose is to identify the controlling parameters under various
emissions were extremely sensitive to the factors that influence operating conditions in order to contribute to the basic under-
the cylinder gas temperature such as crevice flow. They further standing of the physics and chemistry of the combustion process.
indicated that soot emissions increased as the injection pressure The Laminar and Turbulent Characteristic Time combustion
was lowered consistent with the experimental data. However, it model (LaTCT) was implemented in StarCD. Its background is Eddy
was noted that the increase in injection pressure beyond 90 MPa Dissipation combustion model by Magnussen and Hjertager [13]
does not improve the soot and NOx emission trade off significantly, but LaTCT version of the (Eddy Dissipation Combustion) model
indicating that there was a limit to the usefulness of increased for SI engines was formulated by Abraham et al. [1]. The current
injection pressure for reducing NOx emissions. Reitz and Diwakar version is implemented by Patterson et al. [21] but without using
[25] have included drop breakup in dense spray calculation. They the thermodynamic equilibrium values for the mass fractions of
modified KIVA to address the drop breakup. They have reported the species in the reaction rate equation.
that drop breakup influences the spray penetration, vapourization The ignition process was modeled using the Shell model devel-
and mixing in high-pressure sprays. The spray droplets size was oped by Halstead et al. [7]. The Shell multistep kinetics model was
controlled by breakup and coalescence. Droplet breakup dominated used to model diesel ignition. The model introduces a simplified
in hallow cone sprays because coalescence was minimized by set of generic reactions and species to account for the multistage
expanding spray geometry. They have also analysed the flow field ignition and cool flame phenomena that have been observed with
with injected droplets that had a size equal to the nozzle orifice hydrocarbon fuels [27]. In the Shell model [7], eight generic reac-
diameter. The droplets break-up due to the interaction with the tions are used to represent fuel, intermediate species and products.
gas phase flow field yielded a core region, which contained rela- The premise of the Shell model is that degenerate branching plays
tively large droplets. The computed droplet size, droplet velocity an important role in determining the cool flame and the two-stage
and spray penetration compared well with LDV measurements in ignition phenomena that are observed during the auto-ignition of
high-pressure sprays. hydrocarbon fuels. A chain-propagation cycle is formulated to de-
The RNG turbulence model is based on the work of Yakhot and scribe the history of the branching agent together with one initia-
Orszag [29] and implemented in StarCD [30]. Kong et al. [11] and tion and two termination reactions [11]. The transition from auto-
Han et al. [8] have compared the performance of the standard ignition to the main combustion process is based on the local cell
k–e and the RNG turbulence models in diesel engine combustion. temperature: when the temperature of a cell exceeds 1000 K, it
Kong et al. [11] reported that the RNG model produced higher undergoes main (high-temperature) combustion. The emission
gas temperatures and higher turbulent viscosities than the stan- models used are the extended Zeldovich mechanism to predict
dard k–e model. Han et al. [8] showed that the RNG model gives NO formation, the Hiroyasu model for soot formation and the Na-
better results for the smaller bore DI diesel engines and also RNG gle and Strickland-Constable model for soot oxidation [16].
model yielded better emission predictions. It was found that, The combustion model was combined with the ignition model
although the standard k–e turbulence model produced adequate to simulate the overall combustion process in a diesel engine. Once
results for the larger engines, the RNG turbulence model gave bet- the cell temperature reaches 1100 K, the combustion is assumed to
ter agreement with the experimental data for all the engines. start. The ignition model was used whenever and wherever the
In diesel engines, a significant portion of combustion is thought temperature was lower than 1100 K to simulate the ignition chem-
to be mixing-controlled. Hence, interactions between turbulence istry and locate the ignition cells. If the temperature was higher
and chemical reactions have to be considered. A laminar-and-tur- than 1100 K, the combustion model was activated for describing
bulent characteristic-time combustion model for spark-ignited en- high temperature chemistry.
gines [1] was extended to model diesel combustion processes. The Soot formation and oxidation have been modeled by using the
turbulent mixing-time formulation is based on the eddy-dissipa- Hiroyasu soot formation and the Nagle and Stricland-Constable
tion model of Magnussen and Hjertager [13], who proposed that oxidation models respectively. For NO the extended Zeldovich
diffusion combustion is controlled by the mixing of the fuel with mechanism has been used. As proposed by Han et al. [8] a correc-
the oxidiser. The eddy break-up model proposed by Magnussen tion factor of 1.5 is used to convert NO predictions to allow NOx
and Hjertager [13] which takes into account the effect of turbu- estimation.
lence on the combustion process. The model relates the rate of Liquid fuel jet breakup is modeled using Reitz and Diwakar
combustion to the rate of dissipation of turbulent eddies. The rate model [25]. Droplet – wall and droplet – droplet interactions
of reaction depends on the mean concentration of a reacting spe- including splashing, rebounding and coalescence are handled using
cies, the turbulent kinetic energy and the rate of dissipation of this Weber correlations. The liquid injection is simulated using the
energy. The great advantage of this approach is that it requires very ‘blob’ injection method of Reitz and Diwakar, modified according
low computational effort, since the cells of the computational do- to Han et al. [8] in order to take into account the effective nozzle
main are conceived as well-stirred reactors with homogeneous area. The droplet drag model considers the effects of drop deforma-
composition, temperature and pressure. Assumptions about the tion. Two different mechanisms were used to characterize the
flame front propagation are not required: numerically the flame breakup of the spray. The droplet break-up due to shear flow is de-
front does not exist. Abraham et al. [1] developed the laminar- scribed by the wave break-up model of Reitz. Reitz’s model is
and-turbulent characteristic time model for spark-ignition engines based on the growth of Kelvin–Helmholtz instabilities on the drop-
based on the eddy-dissipation concept. In this model, the combus- let surface resulting from the relative velocity between gas and li-
tion rate is determined by the dominant rate between a turbulent quid phase. The droplet break-up due to the acceleration that takes
time scale and a kinetic (or laminar) time scale. In the model for- place at the interface between liquid drops and the gas phase is
mulated by Abraham and extended to diesel engine combustion modeled according to the Rayleigh–Taylor model theory.
1838 B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848

From the above literature review it is seen that there is consid- 3. Mathematical model
erable interest in the CFD analysis of CI engine flow field. From the
review it is apparent that the effect of injection timing and intake The in-cylinder flow is simulated by solving the three equations,
pressure has not been systematically analyzed. Hence in this paper namely conservation of mass, momentum, and energy and also
it is proposed to analyze the effect of the above two parameters on species concentration in case of combustion.
the performance of a DI diesel engine using CFD. It is hoped that
the present effort will fill the gap.
3.1. Conservation equations for the flow field
The aim of this paper is to present the results of a comprehen-
sive CFD study on the performance characteristics of a heavy-duty,
The mass and momentum conservation equations solved for the
DI diesel engine. The approach followed is divided in two parts:
unsteady, compressible and incompressible, three-dimensional in-
validation of the predicted results and analysis of the performance
cylinder flow in Cartesian tensor notation are represented as [28]:
of engine under consideration. For the purpose of validation, calcu-
lations of the suction and compression strokes are performed and @q @
the predicted results are compared with available experimental þ ðquj Þ ¼ sm ð1Þ
@t @xj
data. The results obtained in terms of swirl ratio, radial velocity, @ @ @p
and turbulent intensity is compared with Laser Doppler Velocime- ðqui Þ þ ðquj ui  sij Þ ¼  þ si ð2Þ
@t @xj @xi
try (LDV) measurements [22,23]. In the second part, CFD predic-
tions of the combustion process have been discussed for different where t – time, xi – Cartesian coordinate (i = 1, 2, 3), ui – absolute
fuel injection timing and different intake pressure. The combustion fluid velocity component in direction xi, p – piezometric pressur-
prediction includes cylinder pressure, temperature, heat release,
NOx and soot emissions, so that the combustion process could be
analyzed in detail. In particular, differences observed for the differ-
ent injection timing and different intake pressure are discussed
and appropriate conclusions are drawn. These provide an insight
on the influence of injection timing and intake pressure on the
combustion characteristics.

2. Engine geometry and computational details

A single cylinder DI diesel engine having a toroidal combustion


chamber with two directed intake ports whose outlet is tangential
to the wall of the cylinder and two exhaust ports has been used.
The geometrical and fuel injection details of the engine are sum-
marized in Table 1. The pre-processor GAMBIT is used to create
the entire computational domain of the engine including intake
and exhaust ports and commercial computational fluid dynamics
code STAR-CD is used for the solution of governing equations and
post processing the results. The computational mesh employed
for the simulation is shown in Fig. 1. Hexahedral block structured
mesh is employed for the entire computational domain with
505,542 cells.

Table 1
Geometrical and fuel injection details of the engine.

Bore 130.0 mm
Stroke 150.0 mm
Connecting rod length 275.0 mm
Displacement 1991 cc
Rated speed 2000 rpm
Bowl diameter 76.0 mm
Bowl depth 29.0 mm
Bowl volume 113.5 cc
Static bumping clearance 1.55 mm
Compression ratio 15.5:1
Intake valve diameter 44.4 mm
Intake valve opening 9° bTDC
Intake valve closure 62° aBDC
Exhaust valve opening 43° bBDC
Exhaust valve closure 10° aTDC
Fuel pump Bosch P 7100
Injector nozzle Bosch DLLA
Nozzle opening pressure 250 bar
Number of holes 5
Hole diameter 0.33 mm
Cone angle 145°
Start of fuel injection 12° bTDC
Fuel used n-Dodecane
Fuel injection quantity 207 mg/cycle
Fuel injection duration 30° Fig. 1. Computational domain of the engine with varying cross sectional area intake
ports.
B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848 1839

e = ps  q0gmxm, where ps is static pressure, q0 is reference density, 3.2. Conservation equation for species or mass transfer
gm are gravitational acceleration components and xm are coordi-
nates from a datum where q0 is defined, q – density, Each constituent, m of a fluid mixture, whose local concentra-
 
tion is expressed as a mass fraction mm, is assumed to be governed
¼ RT Pp mm , where mm is the mass fraction of a constituent with
M Mm by a species conservation equation of the form:
molecular weight Mm, T is the temperature and R is the universal @ @
gas constant, sij – stress tensor components, sm – mass source, si – ðqmm Þ þ ðquj mm  F m;j Þ ¼ sm ð8Þ
@t @xj
momentum source components (assumed to be negligible).
where Fm,j – diffusional flux component ¼ qDm @m m
@xj
q
 uj m0m .
The flow being assumed to be Newtonian, the following consti-
The rightmost term representing the turbulent mass flux whose
tutive relation is specified connecting the components of the stress
model is given later, sm – rate of mass production or consumption
tensor sij to the velocity gradients:
due to chemical reaction.
2 @uk
sij ¼ 2lsij  l dij  q
 u0i u0j ð3Þ
3.3. Turbulence modeling
3 @xk
where l – molecular dynamic fluid viscosity, dij – Kronecker delta The in-cylinder flow is turbulent in nature at all speeds and
(=1 when  i = j,  and 0 otherwise), sij – rate of strain dimensions of the engine. It is necessary to model the turbulence
@u
tensor = 12 @u
@x
i
þ @xj . to capture properties of in-cylinder fluid dynamics.
j i

The rightmost terms in Eq. (3) represents the additional Rey-


nolds stresses due to turbulent motion and u0 are the fluctua- (1) The ‘standard’ model [5,12,26], in which the high (turbulent)
tions about the ensemble average velocity. The Reynolds Reynolds number forms of the k and e equations are used in
stresses are linked to the mean velocity fields via turbulence conjunction with algebraic ‘law of the wall’ representations
models. of flow, heat and mass transfer for the near wall region.
(2) The ‘Renormalization group’ (RNG) version of the k–e model
3.1.1. Conservation equations for heat transfer [29,30], denoted as RNG k–e model. This is employed in high
Heat transfer is implemented through the following general Reynolds number form in conjunction with ‘law of the wall’
forms of the enthalpy conservation equation for a fluid mixture functions.
[10] depending on the value of Eckert number:
All forms of the k–e models assume that the turbulent Reynolds
U2 stresses and scalar fluxes are linked to the ensemble averaged flow
Ec ¼ ð4Þ properties in an analogous fashion to their laminar flow counter-
cp T
parts as [12]:
The values of Eckert number of the order of 1.0 or more recom-  
2 @uk
mends the solution of a transport equation for total enthalpy and if  u0i u0j ¼ lt sij 
q lt þ qk dij ð9Þ
3 @xk
the Eckert number is less than 1.0, the solution of transport equa-
lt @h
tion for static enthalpy will suffice. The energy equation derived q u0j h0 ¼  ð10Þ
from the first law of thermodynamics, the mathematical forms rh;t @xj
are represented as follows: l @mm
q u0j m0m ¼  t ð11Þ
rm;t @xj
3.1.1.1. Static chemico-thermal enthalpy where k – turbulent kinetic energy, lt – turbulent viscosity, rh,t –
turbulent Prandtl number, 2rm,t – turbulent Schmidt number, lt –
@ @ @P @p @ui C qk
turbulent viscosity ¼ fl l e , where Cl is an empirical coefficient,
ðqhÞ þ ðqhuj  F h;j Þ ¼ þ uj þ sij þ sh ð5Þ
@t @xj @t @xj @xj usually taken as a constant and fl is the damping function, to be de-
where h is the static enthalpy, defined by: fined when the individual model variants are presented.
X X
h  cp T  c0p T 0 þ m m H m ¼ ht þ mm Hm ð6Þ 3.3.1. Standard k–e model equations

and T – temperature, mm – mass fraction of mixture constituent m, The particular high Reynolds number form of the k–e model is
Hm – heat of formation of constituent m, cp – mean constant – pres- appropriate to fully turbulent, compressible or incompressible
sure specific heat at temperature T, c0p – reference specific heat at flows. The transport equations are as follows.
temperature T0, Fh,j – diffusional energy flux in direction
P
@T
xj ¼ k @x  u0j h0 þ m hm qDm @m
q m
@xj
, where the middle term represents 3.3.1.1. Turbulence energy
j

the turbulent flux of energy whose model is given later, sh – energy @ @ @ leff @k  @ui
source and ht – thermal enthalpy. ðqkÞ þ ðquj kÞ ¼ þ 2lt sij
@t @xj @xj rk @xj @xj
The static enthalpy, h is defined as the sum of the thermal and  
2 @ui @ui
chemical components, the latter being included to cater for chem-  lt þ qk  qe ð12Þ
ically reacting flows. 3 @xi @xi
where leff = l + lt.
3.1.1.2. Total chemico-thermal enthalpy. A governing equation for
total chemico-thermal enthalpy, H may be formed by summing 3.3.1.2. Turbulence dissipation rate
an equation for mechanical energy conservation Eq. (2) and the   
@ @ @ leff @ e e @ui 2
static chemico-thermal enthalpy Eq. (5): ðqeÞ þ ðqeuj Þ ¼ þ C e1 2lt sij  
@t @xj @xj re @xj k @xi 3
  
@
ðqHÞ þ
@
ðquj H  F h;j  ui sij Þ ¼
@P
þ si ui þ sh
@ui @ui @ui e2
ð7Þ  lt þ qk þ C e2 qe  C e3 q
@t @xj @t @xi @xi @xi k
ð13Þ
where H ¼ 12 ui ui þ h.
1840 B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848

where re is the turbulent Prandtl number and C e1 ; C e2 , and C e3 are Abraham et al. [1] presented a laminar-and-turbulent character-
empirical coefficients, whose values taken from [5,12,26] are given istic-time combustion model for spark-ignition engines. Patterson
in Table 2. The turbulent viscosity lt is obtained with fl set equal to et al. [21] extended this concept to model diesel engines. In diesel
unity. engines, a significant portion of the combustion is thought to be mix-
ing controlled, so Magnussen’s model should be able to describe the
3.3.2. RNG k–e model equations rate of combustion. However, it is recognised that the initiation of
The procedure in this model systematically removes the small combustion relies on laminar chemistry [21]. Turbulence starts to
scales of motion from the governing equations by expressing their have an influence only after combustion events have already been
effects in terms of large scale motions and a modified viscosity. The observed. Eventually, combustion will be dominated by turbulent
equations of the model are quoted below [30]: mixing effects in the regions of sl  st, where sl is the laminar (or
chemical) time scale and st is the turbulent (or mixing) time scale.
3.3.2.1. Turbulence energy The laminar time scale, however, is not negligible near the
 injector regions (where the high injection velocity makes the tur-
@ @ @ leff @k  @ui bulent time scale very small), in shear layers generated by the fuel
ðqkÞ þ ðquj kÞ ¼ þ 2lt sij
@t @xj @xj rk @xj @xj sprays and in near-wall regions. For this reason, the laminar-and-
  turbulent characteristic-time combustion model has been used
2 @ui @ui
 lt þ qk  qe ð14Þ for the present work in preference to the EBU combustion model.
3 @xi @xi
According to Magnussen et al.,
 
qA Y O BY p
3.3.2.2. Turbulence dissipation rate Rf ¼ min Y f ; 2 ; kg m3 s1 ð16Þ
st rf 1 þ rf
     
leff @ e e @ui 2 @u @ui
þ C e1 2lt sij  lt i þ qk where st ¼ k=e, the turbulent time scale.
re @xj k @xj 3 @xi @xi As mentioned earlier, Patterson et al. [21] recognised that the
@ui e2 C l g3 ð1  g=g0 Þ e2 laminar time scale cannot be assumed to be negligible in a number
þ C e2 q e  C e3 q  q ð15Þ
@xi k 1 þ bg3 k of important regions of the diesel engine combustion chamber. It is
for this reason that the turbulent time scale in Eq. (16) is to be re-
where g  S ke S  ð2sij sij Þ1=2 . placed by a characteristic time, sc to give:
 
Comparison of Eqs. (14) and (15) with their counterparts Eqs. qA Y O BY p
Rf ¼ min Y f ; 2 ; ð17Þ
(12) and (13) in the standard model reveals that the obvious dis- sc rf 1 þ rf
tinctive feature of the RNG k–e version is the additional and last
term in the dissipation equation, which represents the effect of where
mean flow distortion on e. The coefficients of the model are fixed sc ¼ Characteristic combustion time scale ¼ sl þ f st ð18Þ
empirically as in [26] and are recorded in Table 3. The fl coefficient
is set to unity. A and B are modal coefficients, A = 4, B = 0.5, Yf, Y O2 , and YP are
the fuel, oxygen and product mass fractions respectively and rf is
3.4. Combustion and ignition models the oxygen to
nO2 M O2
The simulation of diesel engines involves modelling several pro- fuel mass fraction ¼ ð19Þ
nf M f
cesses such as spray dynamics, auto-ignition, chemistry, turbu-
lence, etc. as well as the interactions between them such as The laminar time scale, sl is derived from an Arrhenius type
turbulence–chemistry interactions, etc. Ignition and combustion reaction rate [14] with the pre-exponential constant,
are represented as two separate processes. A = 2.0  1010 s1 as:
 
E
3.4.1. The laminar-and-turbulent characteristic-time model
sl ¼ A1 ½C 7 H16 0:25 ½O2 1:5 exp ð20Þ
Rm T
This combustion model is based on the works of Magnussen and
Hjertager [13], Abraham et al. [1] and Patterson et al. [21]. where [C7H16] is the fuel concentration and E the activation energy.
Magnussen and Hjertager [13] presented a model based on an The turbulent time scale st is proportional to the eddy turnover
eddy break-up (EBU) concept, which relates the rate of combustion time
 
to the rate of dissipation of eddies and expresses the rate of reac- k
tion by the mean concentration of a reacting species, the turbulent st ¼ C 2 ð21Þ
e
kinetic energy and the rate of dissipation of this energy.
where C2 = 0.142 [21].
The delay coefficient, f, present in Eq. (18) is chosen to simulate
Table 2 the increasing influence of turbulence on combustion after ignition
Values assigned to standard k–e turbulence model coefficients.
has occurred and has the form [21]:
Cl rk re rh rm Ce1 Ce2 Ce3
1  ec
0.09 1.0 1.22 0.9 0.9 1.44 0.33 1.92 f ¼ ð22Þ
0:632
where the parameter, c is a reaction progress variable given by
Y ft  Y f
Table 3 c¼ ð23Þ
Y ft  Y res
Values assigned to RNG k–e turbulence model coefficients.

Cl rk re rh rm Ce1 Ce2 Ce3 g0 b Yft and Yres are the total fuel mass fraction and the amount of fuel
left at the end of combustion respectively. The characteristic com-
0.085 0.179 0.179 0.9 0.9 1.42 0.387 1.68 4.38 0.012
bustion time scale is given by
B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848 1841


sl at the limit of no combustion The Eqs. (26)–(30) give the local reaction rates for the ignition
sc ¼ ð24Þ
species, which are general functions of the rates of transport by
sl þ st at the limit of complete combustion
convection and diffusion from other parts of the flow. Thus, for
3.4.2. Ignition model each of the species U (U  R*, B, or Q), a transport equation of
Shell auto-ignition model was originally developed for the the form
study of knock in gasoline engines [7]. It has been extended to pre-
dict ignition for fuels of higher molecular weight such as diesel
@ qY U
þ rðqUY U  CrY U Þ ¼ SU ð31Þ
fuels [11]. The model complements the standard numerical com- @t
bustion model. is solved in conjunction with the other species Eq. (8), with the
The chain-propagation cycle provides the framework for the appropriate SU term inserted from one of the Eqs. (26)–(30). A
mathematical model. It is assumed that the reaction chain may new enthalpy source related to temperature rise caused by the igni-
be propagated by any number of radical reactions, but each prop- tion model Eq. (31) is also calculated at this stage and used as a
agation step shows a first-order dependence on the radicals in- source term in the thermal enthalpy transport equation.
volved. It is further assumed that the radical concentrations are
in a steady state relation to each other. 3.5. NOx Modeling
In this model, generic molecular species are introduced. Species
which play a similar role in the ignition chemistry, but which may Nitrogen oxides are important air pollutants produced mainly
be of varying detailed structure are combined and accounted for as by diesel engines. The NOx concentration has a little influence on
a single entity. The model is capable of simulating the cool flame the flow field. The time scale for NOx reactions is larger than the
phenomenon, a sharp well defined two-stage ignition, a rapid time scales for the turbulent mixing process and the combustion
acceleration of the reaction rate after the onset of ignition and a of hydrocarbons that control the heat-releasing reactions. Hence,
transition from two-stage to single-stage ignition with increasing the computations of NOx can be decoupled from the main reacting
temperature [27]. flow field predictions. Two mechanisms have been identified for
The Shell auto-ignition model makes use of the following gener- the formation of nitric oxide during the combustion of hydrocar-
ic molecules: bons, namely: Thermal NOx and Prompt NOx.

RH Hydrocarbon fuel (CmH2n+2) 3.5.1. Thermal NOx


R* Radical formed from the fuel Thermal NOx is strongly temperature dependent and it is pro-
B Branching agent duced by the reaction of atmospheric nitrogen with oxygen at ele-
Q Intermediate species vated temperatures. Thermal NOx is formed by high temperature
P Products consists of CO, CO2, and H2O oxidation of atmospheric nitrogen. For thermal nitric oxide, the
principle reactions are generally recognised to be those proposed
The chemical model consists of the following set of equations: by the following three extended Zeldovich mechanisms:
K1
Primary initiation RH+O2 ? 2R* N2 þ O () NO þ N ð32Þ
K 1
Main propagation R* ? R* + P K2
R* ? R* + B N þ O2 () NO þ O ð33Þ
K 2
R* ? R* + Q
K3
R* + Q ? R* + B N þ OH () NO þ H ð34Þ
K 3
Branching B ? 2R*
Linear termination R* ? Inert products The rate constants for these reactions have been measured in
Quadratic termination 2R* ? Inert products numerous experimental studies [6,15], and the data obtained from
these studies have been critically evaluated by Baulch et al. [2]. The
expressions for rate coefficients for reactions (32)–(34) based on
The non-reactive species (inert products) in the termination equa-
Baulch et al. [2] are given below:
tions are assumed to be equivalent to nitrogen.
 
The rate coefficients of the above reactions take the Arrhenius 38370
form:
K 1 ¼ 1:8  1011 exp m3 ðkg moleÞ1 s1 ð35Þ
T
ðE=RTÞ  
k ¼ Ae ð25Þ 425
K 1 ¼ 3:8  1010 exp ð36Þ
Since generic species are used in the model, the above reactions T
 
as written would not conserve mass. To use the reaction set, the 4680
K 2 ¼ 1:8  107 T exp ð37Þ
equations are mass balanced. This is achieved by appropriately T
defining the molecular weights of the generic species R*, B, Q and P.
 
6 20820
The mathematical model for the above simultaneous reaction K 2 ¼ 3:8  10 T exp ð38Þ
T
comprises the following set of differential equations [27]:  
10 450
1 dnR n o K 3 ¼ 7:1  10 exp ð39Þ
SR  ¼ 2 kQ ½RH½O2  þ kB ½B  kT ½R2  f3 kP ½R ð26Þ T
V dt  
11 24560
1 dnB K 3 ¼ 1:7  10 exp ð40Þ
SB  ¼ f1 kP ½R þ f2 kp ½Q ½R  kB ½B ð27Þ T
V dt
1 dnQ where K1, K2 and K3 are the forward and K1, K2 and K3 are the
SQ  ¼ f4 kP ½R þ f2 kp ½Q ½R ð28Þ
V dt backward rate constants for reactions (32)–(34) respectively.
1 dnO2 The rate of formation of NOx is significant only at high temper-
SO  ¼ pkP ½R ð29Þ
V dt atures since the thermal fixation of nitrogen requires the breaking
ðnO2  nO2 ðt ¼ 0ÞÞ of a strong N2 bond. This effect is represented by the high activa-
nRH ¼ þ nRH ðt ¼ 0Þ ð30Þ
pm tion energy of reaction Eq. (32), which makes this reaction the rate
1842 B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848

 
limiting step of the Zeldovich mechanism. The activation energy K A P O2
RT ¼ x þ K B P O2 ð1  xÞ ð48Þ
for oxidation of N is small, hence a quasi-steady state can be estab- 1 þ K Z P O2
lished. Based on this assumption, the instantaneous rate of forma-
tion of NO is: where

q2 P O2

R  m  x¼ ð49Þ
mNO
1 þ K 1 M K
O2 mOH PO2 þ KK TB
= 2 M þ K3 M
NO O2 OH
2 3
mNO   PO2 is the partial pressure of oxygen and the rate constants in g/
m O m N
2K 1 M mO mNO mH mNO
 42K 1 2
 NO
K 2 þ K 3 5 cm2 s are:
MO M N2 K 2 mO2 þ K 3 mOH M O MNO MH M NO
MO
2
M OH  
30; 000
 m3 ðkg moleÞ1 s1 ð41Þ K A ¼ 20 exp ð50Þ
RT
 
Assumptions have to be made to obtain the concentration of 3 15; 200
K B ¼ 4:46  10 exp ð51Þ
radicals O, OH and H, if these are not calculated by the combustion RT
model.  
97; 000
K T ¼ 1:51  105 exp ð52Þ
RT
3.5.2. Prompt NOx  
4100
Prompt NOx is formed from the reaction of hydrocarbon frag- K Z ¼ 21:3 exp ð53Þ
RT
ments and molecular nitrogen. It has weak temperature depen-
dence and a lifetime of only several microseconds. The
contribution to NOx emission in diesel engines from this mecha- 3.7. Droplet break-up model
nism can be significant from regions of rich mixtures in the com-
bustion chamber such as in the proximity of the fuel spray. The injected fuel droplets become unstable under the action of
The production rate of prompt NOx is given by De Soete [4] as, the interfacial forces induced by their motion relative to the in-cyl-
inder air. The model employed to determine the size rate of change
RNO;pt ¼ kpt M NO kg m3 s1 ð42Þ
of the droplets is that of Reitz and Diwakar [25]. According to this
where kpt is a rate constant and is defined as: model, droplet break-up due to aerodynamic forces occurs in one
  of the following modes:
Ea
kpt ¼ kr ½O2 b ½N2 ½Fuel exp kg mole m3 s1 ð43Þ
RT 1. ‘Bag break-up’, in which the non-uniform pressure field around
[O2], [N2] and [Fuel] are species concentrations in the droplet causes it to expand in the low-pressure wake region
kg mole m3 s1. The oxygen reaction order, b depends on experi- and eventually disintegrate when surface tension forces are
mental conditions. According to De Soete [4], b varies with the oxy- overcome.
2. ‘Stripping break-up’, a process in which liquid is sheared or
gen mole fraction X O2 as follows:
stripped from the droplet surface.
b ¼ 1:0; X O2 6 0:0041
b ¼ 3:95  0:9 ln X O2 ; 0:0041 6 X O2 6 0:0111 In each case, theoretical studies have provided a criterion for
the onset of break-up and concurrently an estimate of the stable
b ¼ 0:35  ln X O2 ; 0:0111 6 X O2 6 0:03
droplet diameter, Dd,stable, and the characteristic time scale, sb of
b ¼ 0:0; X O2 P 0:03 the break-up process. This allows the break-up rate to be calcu-
lated from

3.6. Soot modeling dDd ðDd  Dd;stable Þ


¼ ð54Þ
dt sb
The soot emission model written in the Arrhenius single step
form considers the rate of change of soot mass to be equal to the where Dd is the instantaneous droplet diameter. The criteria and
rate of formation less the rate of oxidation [8]: time scales are as follows.

dM soot dM form dM oxid


¼  ð44Þ 3.7.1. Bag break-up
dt dt dt
dM form Here, instability is determined by a critical value of the Weber
¼ K f Mf v ð45Þ number, We, thus:
dt
where qj~ ud j2 Dd
u ~
We  P C bl ð55Þ
  2r d
Ef
K f ¼ Af P0:5 exp ð46Þ
RT where rd is the surface tension coefficient and Cbl is an empirical
coefficient having a value in the range of 3.6–8.4 [25]. A value of
Mfv is the fuel vapour mass, P is the pressure in bar, Af = 300,
Cbl = 6 is chosen. The stable droplet size is that which satisfies the
Ef = 12500 cal/mole.
equality in Eq. (55).
The oxidation rate is adopted from the Nagle Strickland-Consta-
The associated characteristic time is
ble oxidation model [16] and is given by,

dM oxid 6M c C b2 q1=2
d Dd
3=2

¼ M R ð47Þ sb ¼ ð56Þ
dt qs Ds form T 4r1=2
d

where Mc is the molecular weight of carbon, qs is the soot density where Cb2 = empirical coefficient = p, rd = surface tension coeffi-
(2 g/cm3), Ds is the soot diameter (30 nm) and, cient, Dd = instantaneous droplet diameter.
B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848 1843

3.7.2. Stripping break-up ser Doppler Velocimetry (LDV) measurements of the same engine
The criterion for the onset of this regime is equipped with toroidal combustion chamber at different locations
inside the combustion chamber at a radial distance of 34 mm from
We
pffiffiffiffiffiffiffiffi P C s1 ð57Þ the axis of the cylinder and at a distance of 4 and 16 mm respec-
Red
tively from the cylinder head which is shown in Fig. 2. One of
where Cs1 is a coefficient with the value 0.5 [25]. The characteristic the exhaust valves of the standard cylinder is blocked, and a thick
time scale for this regime is quartz window is inserted in the space for this purpose.
 1=2 Fig. 3 shows the variation of non-dimensional radial velocity
C s2 qd Dd with crank angle near TDC position at two measuring locations.
sb ¼ ð58Þ
2 q j~
u ~ud j The predicted radial velocity has similar trend and almost close
to the experimental results during the expansion stroke at all loca-
where Cs2 is the empirical coefficient, is in the range of 2–20 [25],
tions except the location nearer to the cylinder head during com-
and is presently taken to be 20, ~
u and ~ ud are instantaneous fluid
pression stroke. The predicted results show a centripetal flow,
and droplet velocity.
starting from velocity values near zero to higher values when the
piston approaches TDC and the flow moves into the bowl. The
4. Methodology
experimental results show positive values for radial velocities dur-
ing the compression stroke which implies that the flow in this
The computational fluid dynamics code, STAR-CD has been used
stage is centrifugal and this contradicts the mass conservation cri-
to solve the discretized Navier–Stokes equations. The RNG k–e
teria. Near the cylinder head at crank angle around TDC, there is a
model with standard wall function has been employed for physical
strong squish effect evident in the large gradients of radial velocity.
modeling. The program is based on the pressure-correction meth-
Up to TDC, the piston moves up and this provokes a centripetal mo-
od and uses the PISO algorithm. The first order upwind differencing
tion of the flow which is maintained at the beginning of the expan-
scheme (UD) is used for the momentum, energy and turbulence
sion stroke, the direction of the radial velocities is then reversed.
equations and the temporal discretization is implicit. Spatial dis-
The radial motion of the flow loses significantly its strength in
cretisation is done using the upwind differencing scheme for
the points located far away from the cylinder head. In this case,
momentum, turbulent kinetic energy/dissipation and temperature,
CFD predictions seem to be more reasonable.
and central differencing scheme for density. Temporal discretisa-
Fig. 4 shows the variation of swirl ratio with crank angle near TDC
tion is done using the implicit scheme. Combined laminar and
position at two measuring locations. Decreasing trend in swirl ratio
turbulent characteristic time combustion model with Shell auto-
is observed during compression stroke due to friction at the wall.
ignition model is used to characterise ignition and combustion.
However, while approaching TDC, swirl is enhanced as the flow
The Reitz–Diwaker model [25] is used to characterise droplet
accelerates in preserving its angular momentum within the smaller
break-up, with rebound boundary condition at the walls. The mod-
diameter piston bowl. During the expansion stroke, reverse squish
el employed to determine the rate of change of size of droplets is
as the flow exits from the piston bowl and wall friction contributes
that of Reitz and Diwakar, which mathematically represents insta-
to the sudden fall of the swirl ratio. The predicted results are gener-
bility of droplets by a critical value of the Weber number. Mono-
ally in good agreement with experimental results at all locations.
tone Advection and Reconstruction Scheme (MARS) [STAR-CD
The results show faster decay of swirl during the expansion stroke
Methodology Volume] is used to discretise species scalars. NOx
at locations near the cylinder head due to reverse squish–swirl
emission is modelled by using extended Zeldovich mechanism.
interaction.
The soot emission model written in the Arrhenius single step form
Fig. 5 shows the variation of non-dimensional turbulent inten-
which considers the rate of change of soot mass is used to model
sity with crank angle near TDC position at two measuring loca-
soot emissions.
tions. The turbulent intensity decays almost linearly as it

5. Boundary and initial conditions

Constant pressure boundary conditions are assigned to both in-


take and exhaust ports, so the dynamic effects are neglected. Fric-
tional effects at the walls are not taken into account, i.e., the
smooth wall option for turbulent flow boundary condition is used.
Adiabatic wall for heat transfer has been assumed. The initial val-
ues of pressure and temperature are considered as homogeneous
in the whole domain. As the residual swirl of the flow in the cylin-
der at the end of the exhaust stroke is not taken into account, the
flow is supposed to be quiescent initially. The initial turbulent
intensity is set at 5% of the mean flow, and the integral length scale
is estimated with the mixing length model of Prandtl. The walls of
the intake ports, the lateral walls of the valves and the cylinder
head, the cylinder wall and the piston crown that form the walls
of the combustion chamber are considered adiabatic.

6. Validation of the predicted results

The CFD predicted results of the engine having varying cross


sectional area intake ports with a 20° bend angle is used for valida-
tion with the experimental results obtained by Payri et al. [22,23].
The results obtained for radial velocity, swirl ratio and turbulent
intensity variations at different crank angles are compared with La- Fig. 2. Measuring locations.
1844 B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848

Location z1 = 4 mm, Expt. Payri et al. Location z2 = 16 mm, Expt. Payri et al.
r = 34 mm Predicted r = 34 mm Predicted
2 2
1 1
u/Vp

u/Vp
0 0
-1 -1
-2 -2
-3 -3
280 300 320 340 360 380 400 280 300 320 340 360 380 400
Crank angle, deg Crank angle, deg

Fig. 3. Non-dimensional radial velocity versus crank angle at various locations compared with experimental LDV data.

Location Z1 = 4 mm, Expt. Payri et al. Location Z2 = 16 mm, Expt. Payri et al.
r = 34 mm Predicted r = 34 mm
4 4 Predicted

3 3
SR

SR
2 2

1 1
280 300 320 340 360 380 400 280 300 320 340 360 380 400
Crank Angle, deg. Crank Angle, deg.

Fig. 4. Swirl ratio versus crank angle at various locations compared with experimental LDV data.

approaches TDC and starts its downward motion during the expan- 1000 rpm. The start of injection (SOI) 16°, 12° and 8° bTDC and
sion stroke. Since the bowl geometry is open, the radial motion is the step profile is considered for all the cases in the present study.
weak and the squish effect is small, particularly at the measure- Figs. 6 and 7 show the cylinder averaged pressure and temper-
ment points which are located relatively far away from the walls. ature profiles obtained for three injection timings (SOI). The peak
It is clear that turbulence dissipation rate is faster than the turbu- cylinder pressure and temperature obtained for the three injection
lence generation rate. The predicted results are reasonably in good timings are summarized in Table 4. The advanced injection timing
agreement with experimental results. shows higher peak pressure and temperature and retarded injec-
tion timing shows lower peak pressure and temperature with ref-
7. Results and discussion erence to the base case. As the injection timing is advanced,
pressure and temperature inside the cylinder is not sufficient to
Based on the confidence gained from validation, the study is ex- ignite the fuel as a result a large amount of evaporated fuel is accu-
tended to evaluate the effect of fuel injection timing and intake mulated during the ignition delay period. However, in the case of
pressure on diesel engine performance and emissions. A compre- retarded injection timing, pressure and temperature inside the cyl-
hensive study of the effect of injection timing and intake pressure inder is sufficient to ignite the fuel and a relatively small amount of
with respect to performance and emissions are presented below. evaporated fuel is accumulated during the ignition delay period.
Figs. 8–10 show the cylinder averaged instantaneous heat re-
7.1. Effect of fuel injection timing lease rate, exploded view of heat release rate and cumulative heat
release respectively. The ignition delay, combustion duration, peak
The simulation is carried out to study the effect of fuel injection heat release rate and cumulative heat release obtained for different
timing on the performance of the engine operating at a speed of injection timing is summarized in Table 5. The advanced injection

Location Z1 = 4 mm, Expt. Payri et al. Location Z2 = 16 mm, Expt. Payri et al.
r = 34 mm r = 34 mm Predicted
0.8 Predicted 0.8

0.6 0.6
u'/Vp

u'/Vp

0.4 0.4

0.2 0.2
280 300 320 340 360 380 400 280 300 320 340 360 380 400
Crank Angle, deg. Crank Angle, deg.

Fig. 5. Non-dimensional turbulent intensity versus crank angle at various locations compared with experimental LDV data.
B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848 1845

90 3000
80
2500

Heat Release Rate, J/deg


70
60
Pressure, bar

2000
50
40 1500
30
1000
20
10 500
0
300 320 340 360 380 400 420 440 0
340 360 380 400 420 440
Crank Angle , de g.
Crank Angle, deg
SOI 16 CAD bTDC SOI 12 CAD bTDC
SOI 8 CAD bTDC SOI 16 CAD bTDC SOI 12 CAD bTDC
SOI 8 CAD bTDC
Fig. 6. Cylinder averaged pressure.
Fig. 8. Average heat release rate.

2500
200

2000

Heat Release Rate, J/deg


Temperature, K

150
1500

100
1000

500 50

0
300 320 340 360 380 400 420 440 0
340 360 380 400 420 440
Crank Angle, deg.
Crank Angle , de g
SOI 16 CAD bTDC SOI 12 CAD bTDC
SOI 16 CAD bTDC SOI 12 CAD bTDC
SOI 8 CAD bTDC
SOI 8 CAD bTDC
Fig. 7. Cylinder averaged temperature.
Fig. 9. Exploded view of average heat release rate.

timing shows maximum peak heat release rate and maximum


cumulative heat release and retarded timing shows lower peak 6000
heat release rate and lower cumulative heat release compared to
base case. In the case of advanced injection timing, a large amount 5000
Cum_Heat Release, J

of evaporated fuel is accumulated resulting in longer ignition de-


lay. The longer ignition delay leads to rapid burning rate and the 4000
pressure and temperature inside the cylinder rises suddenly.
3000
Hence, most of the fuel burns in premixed mode causes maximum
peak heat release rate, maximum cumulative heat release and 2000
shorter combustion duration. In the case of retarded injection tim-
ing, the accumulation of evaporated fuel is relatively less resulting 1000
in shorter ignition delay. The shorter ignition delay leads to slow
burning rate and slow rise in pressure and temperature. Hence, 0
most of the fuel burns in diffusion mode rather than premixed 340 360 380 400 420 440
mode resulting in lower peak heat release rate, lower cumulative Crank Angle, deg
heat release and longer combustion duration. SOI 16 CAD bTDC SOI 12 CAD bTDC
SOI 8 CAD bTDC

Fig. 10. Average cumulative heat release.


Table 4
Comparison of fuel injection timing.

Start of injection, CAD Peak pressure, Peak temperature, Figs. 11 and 12 show the cylinder averaged NOx and soot emis-
bTDC bar @ CAD K @ CAD
sions. The NOx and soot emission obtained for different injection
16 84.36 @ 367.2 2006.80 @ 383.4 timing is summarized in Table 6. The advanced injection timing
12 76.47 @ 368.6 1918.25 @ 387.0
shows higher NOx (6.88%) through out the predicted range. The
8 67.45 @ 370.4 1817.29 @ 391.7
soot emissions are higher in the beginning around up to 390 CAD
1846 B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848

Table 5 Table 6
Comparison of fuel injection timing. Comparison of fuel injection timing.

Start of injection, Ignition Combustion Peak Cum_heat Start of injection, CAD NOx, g/kg fuel @ 450 Soot, g/kg fuel @ 450
CAD bTDC delay, CAD duration, CAD HRR, J/ release, J bTDC CAD CAD
deg
16 37.75 0.68
16 6.9 45.1 2780 5580 12 35.32 0.68
12 5.9 45.5 2460 5400 8 32.90 0.70
8 5.4 49.5 2120 5300

50 120
45
100
40
NOx, g/kg fuel

35
80

Pressure, bar
30
25
60
20
15 40
10
5 20
0
340 370 400 430 0
Crank Angle , de g. 300 320 340 360 380 400 420 440
SOI 16 CAD bTDC SOI 12 CAD bTDC Crank Angle, deg.
SOI 8 CAD bTDC
1.01 bar 1.21 bar 1.71 bar
Fig. 11. Cylinder averaged NOx emission.
Fig. 13. Cylinder averaged pressure.

7.2. Effect of intake pressure


0.8
0.7 The simulation is carried out to study the effect of intake pres-
0.6 sure (naturally aspirated and supercharged with inter-cooled con-
Soot, g/ kg fuel

0.5 ditions) on the performance of the engine. The operating speed of


0.4 the engine is 1000 rpm, intake pressure of 1.01, 1.21 and 1.71 bar
and an injection timing of 12° bTDC with step profile are consid-
0.3
ered for the simulation. Figs. 13 and 14 show the cylinder averaged
0.2 pressure and instantaneous temperature profiles obtained for
0.1 three intake pressure cases. The peak cylinder pressure and tem-
0.0 perature obtained for the three intake pressure cases are summa-
340 370 400 430 rized in Table 7. The supercharged with inter-cooled cases show
Crank Angle, de g. higher value of peak pressure and lower temperature compared
to base case. Out of these two cases, the highly supercharged case
SOI 16 CAD bTDC SOI 12 CAD bTDC
(1.71 bar) shows higher peak pressure and lower temperature. The
SOI 8 CAD bTDC
supercharging increases the amount of air inside the cylinder
Fig. 12. Average cumulative soot emission.

2500

and reduced during rest of the period compared to base case. The
higher NOx and reduction in soot emission is due to high temper- 2000
Temperature, K

ature caused by pre-mixed burning and high rate of oxidation of


soot. The retarded injection timing shows lower value of NOx 1500
(6.85%) throughout the predicted range, whereas soot emissions
are lower in the beginning around up to 390 CAD and an increase 1000
in soot emissions is observed during rest of the period compared to
base case. At the end of combustion the retarded injection timing 500
shows 2.94% increase in soot emission. The lower NOx and lower
soot emission is due to lower temperature of diffusion burn and in-
0
crease in soot emission is due to lower rate of oxidation of soot at 300 320 340 360 380 400 420 440
low temperature. The reduction trend in NOx emission is observed
Crank Angle, deg.
in all the cases and it is due to reduction in-cylinder temperature.
Based on overall performance and emission characteristics of the
1.01 bar 1.21 bar 1.71 bar
engine the optimum injection timing (start of injection) is 12°
bTDC for the engine under consideration. Fig. 14. Cylinder averaged temperature.
B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848 1847

Table 7 8000
Comparison of intake pressure.
7000
Intake pressure, bar Peak pressure, bar @ CAD Peak temperature, K @ CAD

Cum_Heat Release, J
6000
1.01 76.47 @ 368.6 1918.25 @ 387.0
1.21 86.97 @ 369.2 1887.70 @ 384.6 5000
1.71 111.94 @ 368.4 1778.26 @ 387.8
4000
3000
which is not taking part in combustion which results in decreasing 2000
of temperature inside the cylinder.
1000
Figs. 15–17 show the cylinder averaged instantaneous heat re-
lease rate, exploded view of heat release rate and cumulative heat 0
release respectively. The ignition delay, combustion duration, peak 340 360 380 400 420 440
heat release rate and cumulative heat release for different intake Crank Angle , de g
pressure cases are given in Table 8. The supercharged with inter-
cooled cases show lower peak heat release rate and maximum 1.01 bar 1.21 bar 1.71 bar
cumulative heat release compared to base case. In case of super- Fig. 17. Average cumulative heat release.
charged with inter-cooled cases, pressure and temperature inside
the cylinder is initially higher resulting in shorter ignition delay.
The air fuel mixture inside the cylinder is overall lean and leads
to slow burning rate and slowly raises the pressure and tempera- Table 8
Comparison of intake pressure.
ture inside the cylinder. Hence, with the availability of air for com-
plete combustion and longer combustion duration, the fuel burns Intake Ignition Combustion Peak HRR, Cum_heat
in diffusion mode with lower peak heat release rate and maximum pressure, bar delay, CAD duration, CAD J/deg release, J
cumulative heat release. 1.01 5.9 45.5 2460 5400
1.21 5.4 49.2 1980 5960
1.71 4.6 57.8 1230 7510

3000

2500
Heat Release Rate, J/deg

70

2000 60
50
NOx, g/kg fuel

1500
40
1000 30
20
500
10
0 0
340 360 380 400 340 360 380 400 420 440
Crank Angle, de g
Crank Angle, deg.
1.01 bar 1.21 bar 1.71 bar 1.01 bar 1.21 bar 1.71 bar
Fig. 15. Average heat release rate.
Fig. 18. Cylinder averaged NOx emission.

250
0.8
Heat Release Rate, J/deg

200 0.7
0.6
Soot, g/ kg fuel

150
0.5
0.4
100
0.3

50 0.2
0.1
0 0
340 360 380 400 420 440 340 360 380 400 420 440
Crank Angle, deg Crank Angle, deg.

1.01 bar 1.21 bar 1.71 bar 1.01 bar 1.21 bar 1.71 bar

Fig. 16. Exploded view of average heat release rate. Fig. 19. Average cumulative soot emission.
1848 B. Jayashankara, V. Ganesan / Energy Conversion and Management 51 (2010) 1835–1848

Table 9 [6] Flowers WL, Manson RK, Kruger CH. In: 15th symposium on combustion, The
Comparison of intake pressure. Combustion Institute; 1975, p. 823.
[7] Halstead MP, Kirsch LJ, Prothero A, Quinn CP. A mathematical model for
Intake pressure, bar NOx, g/kg fuel @ 450 CAD Soot, g/kg fuel @ 450 CAD hydrocarbon auto-ignition at high pressures. Proc Roy Soc Lond
1975;A346:515–38.
1.01 35.32 0.68
[8] Han Z, Uludogan A, Hampson GJ, Reitz, RD. Mechanism of soot and NOx
1.21 40.63 0.62
emission reduction using multiple-injection in a diesel engine. SAE paper no.
1.71 56.05 0.33 960633; 1996.
[9] Li Jianwen, Ou Chae Jae, Park SB, Paik HJ, Park JK, Jeong YS, et al. Effect of intake
composition on combustion and emission characteristics of DI diesel engine at
high intake pressure. SAE 970322; 1997.
Figs. 18 and 19 show the cylinder averaged NOx and soot emis- [10] Jones WP. Models for turbulent flows with variable density and combustion.
sions. The NOx and soot emission obtained for different intake In: Kollmann W, editor. Prediction methods for turbulent flows. Washington
(DC, USA): Hemisphere; 1980. p. 423–58.
pressure cases is summarized in Table 9. Supercharged with in- [11] Kong SC, Han Z, Reitz RD. The development and application of a diesel ignition
ter-cooled cases show higher NOx and lower soot emissions com- and combustion model for multidimensional engine simulation. SAE paper no.
pared to base case, which is due to excess air. Reduction in soot 950278; 1995.
[12] Launder BE, Spalding DB. The numerical computation of turbulent flow.
emission in case of supercharging at an intake pressure of
Comput Methods Appl Mech Eng 1974;3:269–89.
1.71 bar is due to higher rate of soot oxidation and increase in [13] Magnussen BF, Hjertager BH. On mathematical modelling of turbulent
NOx is due to availability of excess air. combustion with special emphasis on soot formation and combustion. In:
Proceedings of the 16th symposium (international) on combustion,
combustion institute; 1976. p. 719–29.
8. Conclusions [14] Mehta PS, Bhaskar T. Prediction of combustion and in-cylinder emissions in a
direct injection diesel engine using multi-process models. In: The 5th
international symposium on diagnostics and modeling of combustion in IC
Computational fluid dynamic investigation is carried out to engines, COMODIA 2001; July 2001.
study the effect of fuel injection timing and intake pressure on [15] Monat JP, Hanson RK, Kruger CH. Shock tube determination of the rate
the performance of a DI diesel engine with toroidal combustion coefficient for the reaction N2 + O (NO + O. In: 17th Symposium (international)
on combustion, The Combustion Institute; 1979. p. 543–52.
chamber configuration operating at a speed 1000 rpm. Based on [16] Nagle J, Strickland-Constable RF. Oxidation of carbon between 1000–2000 C.
the overall performance and emission characteristics, the following In: Proceedings of 5th carbon conference, vol. 1; 1962. p. 154.
conclusions are drawn from the present investigation: Advanced [17] Uchida Noboru, Daisho Yasuhiro, Saito Takeshi. The control of diesel emissions
by supercharging and varying fuel-injection parameters. SAE 920117; 1992.
injection timing results in increase in-cylinder pressure, tempera-
[18] Uchida Noboru, Daisho Yasuhiro, Saito Takeshi, Sugano Hideaki. Combined
ture, heat release rate, cumulative heat release and NOx emissions effects of EGR and supercharging on diesel combustion and emissions. SAE
(6.88%) and retarded injection timing results in reverse trend (NOx 93060l; 1993.
emission 6.85%). In case of advanced injection timing the soot [19] Nwafor OMI, Rice G, Ogbonna AI. Effect of advanced injection timing on the
performance of rapeseed oil in diesel engines. Renew Energy 2000;21:433–44.
emissions show increasing trend up to certain crank angle then re- [20] Nwafor OMI. Effect of advanced injection timing on emission characteristics of
verse trend whereas in case of retarded injection timing soot emis- diesel engine running on natural gas. Renew Energy 2007;32:2361–8.
sions show the reverse trend. The supercharged with inter-cooled [21] Patterson MA, Kong SC, Mampson GJ, Reitz RD. Modeling the effects of fuel
injection characteristics on diesel engine soot and NOx emissions. SAE paper
cases show lower peak heat release rate and maximum cumulative no. 940523; 1994.
heat release, shorter ignition delay, higher NOx (15.03% and 58.69% [22] Payri F, Benajes J, Margot X, Gil A. CFD modeling of the in-cylinder flow in
at 1.21 bar and 1.71 bar respectively) and lower soot emissions direct-injection diesel engines. Comput Fluids 2004;33:995–1021.
[23] Payri F, Desantes JM, Pastor JV. LDV measurements of the flow inside the
(8.82% and 51.47% at 1.21 bar and 1.71 bar respectively). The opti- combustion chamber of a 4-valve DI diesel engine with axisymmetric piston
mum injection timing (start of injection) is 12° bTDC. The intake bowls. Exp Fluids 1996;22:118–28.
pressure boosting up to 1.21 bar is the best option. [24] Ramos JI. Internal combustion engine modeling. New York: Hemisphere
Publishing Corporation; 1989.
[25] Reitz RD, Diwakar R. Effect of droplet breakup on fuel sprays. SAE paper no.
References 860469; 1986.
[26] Rodi W. Influence of Buoyancy and rotation on equations for turbulent length
[1] Abraham J, Bracco FV, Reitz RD. Comparison of computed and measured scale. In: Proceedings of 2nd symposium on turbulent shear flows; 1979.
premixed charge engine combustion. Combust Flame 1985;60:309–22. [27] Theobald MA, Cheng WK. A numerical study of diesel ignition. ASME report no.
[2] Baulch DL, Drysdall DD, Horne DG, Lloyd AC. Evaluated kinetic data for high 87-FE-2; 1987.
temperature reactions. Butterworth; 1973. [28] Warsi ZVA. Conservation form of the Navier–Stokes equations in general non-
[3] Cenk Sayin, Mustafa Canakci. Effects of injection timing on the engine steady coordinates. AIAA J 1981;19:240–2.
performance and exhaust emissions of a dual-fuel diesel engine. Energy [29] Yakhot V, Orszag SA. Renormalization group analysis of turbulence – I: basic
Convers Manage, vol. 50. Elsevier; 2009. pp. 203–13. theory. J Sci Comput 1986;1:1–51.
[4] De Soete GG. Overall reaction rates of NO and N2 formation from fuel nitrogen. [30] Yakhot V, Orszag SA, Thangam S, Gatski TB, Speziale CG. Development of
In: 15th Symposium on combustion, The Combustion Institute; 1975. turbulence models for shear flows by a double expansion technique. Phys
[5] El Tahry SH. k–e Equations for compressible reciprocating engine flows. AIAA J Fluids 1992;A4(7):1510–20.
Energy 1983;7(4):345–53.

Potrebbero piacerti anche