Int. J. Vehicle Design, Vol. 41, Nos.

1/2/3/4, 2006 83

In-cylinder reduction of PM and NOx emissions from

diesel combustion with advanced injection strategies

C.A. Chryssakis, J.R. Hagena, A. Knafl,

V.D. Hamosfakidis and Z.S. Filipi*
University of Michigan, 2031 W.E. Lay Automotive Lab.,
1231 Beal Ave., Ann Arbor, MI 48109, USA
Fax +1-734-7644256 E-mail: cchryssa@umich.edu
E-mail: jhagena@umich.edu E-mail: aknafl@umich.edu
E-mail: bhamosfa@umich.edu E-mail: filipi@umich.edu
*Corresponding author

Dennis N. Assanis
Department of Mechanical Engineering, University of Michigan,
2236 G.G. Brown, 2350 Hayward Street, Ann Arbor,
MI 48109, USA
Fax: +1-734-6473170 E-mail: assanis@umich.edu

Abstract: The effect of advanced injection strategies, including pilot- and

post-injections, on reducing pollutants from diesel combustion is
investigated through a synergistic approach combining experiments and
Computational Fluid Dynamics (CFD) simulations. It is shown
experimentally that pilot injections have the potential to reduce NOx and
Particulate Matter emissions simultaneously when the timing of the pilot is
selected appropriately. To gain further understanding of the combustion
and emissions formation mechanisms from multiple injection events, a CFD
analysis is performed to model in-cylinder processes. Results show that
benefits of pilot injection stem from improved fuel-air mixing and the
reduction of the amount of diffusion combustion. Furthermore, CFD
analysis demonstrates that post injection can accelerate the soot oxidation
process if the injection timing and the amount of fuel are suitably selected,
while simultaneously reducing NOx by reducing the amount of fuel in the
main event and lowering peak combustion temperatures.

Keywords: CFD; combustion; diesel; emissions; experiment; injection; NOx;

soot; strategies.

Reference to this paper should be made as follows: Chryssakis, C.A.,

Hagena, J.R., Knafl, A., Hamosfakidis, V.D., Filipi, Z.S. and Assanis, D.N.
(2006) `In-cylinder reduction of PM and NOx emissions from diesel
combustion with advanced injection strategies', Int. J. Vehicle Design,
Vol. 41, Nos. 1/2/3/4, pp.83±102.
Biographical notes: Christos Chryssakis is a post-doctoral research fellow at
the Institut FrancËais du PeÂtrole (IFP). He received his Diploma in
Mechanical Engineering from the National Technical University of
Athens (NTUA), Greece and the University of Stuttgart, Germany, in
2000, and his MSc and PhD in Mechanical Engineering from the University
of Michigan in 2002 and 2005 respectively. His research interests include

Copyright # 2006 Inderscience Enterprises Ltd.

84 C.A. Chryssakis et al.

engine systems modelling, CFD modelling of spray atomisation, diesel

combustion and emissions reduction.
Jonathan Hagena received his BSc in Mechanical Engineering from the
South Dakota State University in May 2000. In August 2004 he received his
MSc in Mechanical Engineering from the University of Michigan. He is
currently a PhD candidate at the University of Michigan. His research
interests include experimental research of diesel combustion, transient
engine emissions formation during in-vehicle operation, and low-emission
engine and hybrid vehicle control strategies.
Alexander Knafl received his Diploma in Mechanical Engineering from the
Technical University of Graz, Austria in 2002. He performed his MSc thesis
work as an exchange student at the University of Michigan in 2001. He
has worked for AVL in Austria from 1999 to 2002. He is currently a PhD
candidate at the University of Michigan. His research interests include
diesel engine emissions research, aftertreatment and engine behaviour under
transient operating conditions.
Vasileios Hamosfakidis received his Diploma in Chemical Engineering from
the National Technical University of Athens, Greece in 1998. In 2001 he
received his MSc in Mechanical Engineering from the University of
Wisconsin-Madison. Currently he is a PhD candidate at the University of
Michigan. His research interests include CFD modelling of IC engines,
turbulent combustion, turbulence and numerical methods.
Zoran Filipi is a Research Associate Professor in the Department of
Mechanical Engineering and the Assistant Director of the Automotive
Research Center at the University of Michigan. He received his MS and
PhD degrees in Mechanical Engineering from the University of Belgrade
(1987, 1992 respectively). Current interests include modelling of engine
physical processes and engine systems, low-emission engine technologies,
turbocharging, heat transfer in homogeneous charge compression ignition
engines, and hybrid propulsion systems.
Dennis Assanis is a Jon R. and Beverly S. Holt Professor of Engineering;
Director of the Automotive Research Center; and Chair of the Department
of Mechanical Engineering at the University of Michigan. He received his
BSc Marine Engineering from the University of Newcastle-upon-Tyne, UK
(1980), SM in both Naval Architecture and Mechanical Eng. from the
Massachusetts Institute of Technology (1982), SM in Management of
Technology from MIT (1986), and a PhD in Power and Propulsion from
MIT (1985). Current interests include modelling and experiments for
studies of IC engine processes and automotive systems.

1 Introduction
Stringent emissions regulations for diesel engines in the USA, Europe and Japan
require employing innovative technologies in order to reduce in-cylinder nitrogen
oxide (NOx) and particulate matter (PM) emissions. Advanced fuel injection
strategies, including multiple pilot and post injections can be utilised to reach the
limits of emissions reduction independently, or in conjunction with exhaust gas
recirculation (EGR) and after-treatment systems. Optical diagnostic studies of
multiple injection strategies can offer valuable insight on the in-cylinder processes as
 In-cylinder reduction of PM and NOx emissions from diesel combustion 85

a first step towards understanding the physical mechanisms that lead to NOx and
soot formation (Park et al., 2004). However, a systematic experimental investigation
on a production-type engine is needed to quantify the benefits under realistic
operating conditions. This comprises the first part of the study presented here. In
order to gain further understanding of the combustion and emissions formation
mechanisms from multiple injection events, a computational fluid dynamics (CFD)
analysis is also performed to model in-cylinder processes. Furthermore, the CFD
tool is utilised to expand the study through adding one more degree of freedom, i.e.
one more injection event, otherwise not attainable on the hardware set-up. The
overall study presented here is performed without employing EGR in order to
analyse only the effects of the injection strategies.
Results from several contemporary investigations (Benajes et al., 2001; Ishikawa
et al., 2004; Payri et al., 2002) show that pilot injections have the potential to reduce
both NOx and PM emissions due to improved fuel±air mixing and the reduction of
the amount of diffusion combustion. Furthermore, post-injection can accelerate the
soot oxidation process if the injection timing and the amount of fuel are suitably
selected. The mechanisms leading to these results will be further analysed here in
order to provide guidance for further reduction of NOx and soot emissions without
compromising engine performance and fuel economy.
In the following sections, the experimental setup and the computational models are
briefly described. The effect of pilot injection is investigated through a comprehensive
experimental study on a V8 6L International diesel engine and the main effects are
explained using detailed computational results. The effect of post-injections and their
potential for further improving emissions trade-offs is studied computationally.
Finally, the combined effects of pilot- and post-injections are studied and the insight
is used to provide clear guidelines for optimising three-stage injection strategies for
low emissions.

2 Experimental set-up
The engine used in this investigation is a 6.0 l V-8 direct-injection diesel engine
manufactured by the International Truck and Engine Corporation. Primary engine
specifications are shown in Table 1. The engine is fitted with AVL GU12P uncooled
cylinder pressure transducers. The glow plugs have been removed and replaced by
glow plug adapters, which hold the cylinder pressure transducers. Crank angle
resolved in-cylinder pressure data is acquired with a 0.1 crank angle (CA) degree
resolution through use of an AVL 365C crank angle encoder. All crank-angle
resolved signals are measured using a multi-channel high-speed data acquisition
system AVL Indimaster 671.

2.1 Fuel injection system

The engine is equipped with a hydraulic rail injection system, where the pressurised
oil is used for actuating the injectors and creating high injection pressures in excess of
2000 bar. Combination of this hardware and sophisticated electronics permit precise
control of injection timing, pressure, quantity and scheme. The Siemens fuel injectors
are electronically controlled by the fuel injection control module (FICM).
86 C.A. Chryssakis et al.

Table 1 Engine specifications

Engine type Diesel, 4 stroke

Configuration 4 overhead valves, cam-in-crankcase 90 V-8
Displacement 6.0 L (365 in3)
Bore and stroke 95  105 mm (3.74  4.134 in)
Compression ratio 18.0:1
Aspiration Variable geometry turbocharger/charge air cooler
Rated power @ rpm 242 kW (325 hp) @ 3300 rpm
Rated torque @ rpm 760 N-m (560 ft-lb) @ 2000 rpm
Combustion system HEUI G2 direct injection

To obtain a better understanding of the injection process, two fuel injectors have
been instrumented to measure needle lift and fuel injection pressure. The fuel
injection pressure is measured with a strain-gauge sensor. Needle lift is measured
using a Hall Effect sensor mounted inside the nozzle needle valve spring. A small
magnet is brazed to the spring guide and moves up and down with the needle. The
magnet's motion is detected by the stationary Hall Effect sensor, which outputs a
voltage that is proportional to the distance that the needle has moved. A full bridge
strain gauge circuit is used to measure the hoop strain around the nozzle needle
spring cage. This deformation is a result of fluctuating fuel injection pressures. As the
cage expands and contracts, the resistance of the strain gauge circuit changes and the
bridge becomes unbalanced. This alters the circuit's output voltage, which can be
related directly to the fuel injection pressure. In Figure 1 the injection pressure for a
typical pilot-main injection strategy is compared with the injection pressure for a
single injection event.

Figure 1 Comparison of typical instantaneous fuel injection rates between a pilot ± main and
a single injection strategy, experimental measurements
 In-cylinder reduction of PM and NOx emissions from diesel combustion 87

2.2 Emissions measurements

An AVL Combustion Emissions Bench II (CEB-II) is used to sample, condition and
measure exhaust gas constituents. Analysers quantify the amount of CO, CO2, O2,
total hydrocarbons (THC), NO and NOx in the exhaust gases and the CO2 levels in
the intake manifold. To quantify the amount of soot produced by the engine, an
AVL 415S Variable Sampling Smoke Meter is used. The smoke meter functions by
drawing a known volume of exhaust gas through a filter paper and the paper
blackening is used to determine the Filter Smoke Number (FSN) as defined in ISO
10054. The FSN is closely related to the elemental carbon, or soot, emission as long
as the emitted particulates contain more than 15% carbon. The soot concentration is
calculated from the FSN by using Equation (1), as given by Christian et al. (1993):
ÿ 

C ˆ 12:22 FSN exp…0:38 FSN† mg m3 : …1†
The system has a resolution of 0.001 FSN, which is 0.01% of full scale, and produces
measurements with a standard deviation   0.005 FSN+3% of the measured value.

3 Description of computational models

For the numerical simulations performed in this work a modified version of the Los
Alamos multi-dimensional code KIVA-3V (Amsden, 1993, 1997; Amsden et al.,
1989) has been used. The modifications include the fuel injection, wall impingement,
ignition, combustion and soot formation models, as described in the following
paragraphs. Since a six-hole nozzle is used, only a 60 sector has been modelled,
assuming axial symmetry. The computational grid consists of 27 cells in the radial, 20
cells in the azimuthal direction and approximately 20,000 cells overall. An image of
the grid showing the piston geometry is given in Figure 2.

Figure 2 Mesh used for the CFD analysis

3.1 Fuel injection model

The fuel spray has been modelled by assuming a liquid core emerging from the
nozzle, which disintegrates very fast into droplets, with a diameter equal to the nozzle
diameter. The WAVE breakup model (Liu and Reitz, 1993; Liu et al., 1993; Reitz,
1987; Reitz and Diwakar, 1987) has been adopted in this study for the primary and
secondary atomisation modelling of the resulting droplets. This model is based on a
linearised analysis of a Kelvin±Helmholtz instability of a stationary, round liquid jet
immersed into a quiescent, incompressible gas. A general dispersion equation relates
88 C.A. Chryssakis et al.

the growth rate of an initial surface perturbation to its wavelength. From numerical
solutions it is shown that the maximum growth rate,
and its corresponding
wavelength, ÿ are approximated by:
ÿ
ÿ

 1 ‡ 0:45Z 0:5 1 ‡ 0:4T 0:7
ˆ 9:02 …2†
ro …1 ‡ 0:87We1:67 †0:6
 3 0:5
L ro 0:34 ‡ 0:38We1:5

ˆ ; …3†
 …1 ‡ Z†…1 ‡ 1:4T 0:6 †
where Z ˆ …WeL †1=2 =ReL , T ˆ Z…WeG †1=2 . Under the assumption that the size of the
stripped off product droplets are proportional to the length of the fastest growing
surface wave and that the rate of droplet generation is proportional to the maximal
jet disturbance growth rate,
, one obtains the expression for the radius, r, and the
time constant, , of the stripped off product droplet as:
r ˆ Bo  …4†
ro
 ˆ 3:726B1 ; …5†


where the constants Bo ˆ 0:61 and B1 are subject to further debate. In this work, a
value of B1 ˆ 4:5 was selected.

3.2 Wall impingement model

A wall impingement model, developed by Grover and Assanis (2001) has been used
to improve the prediction capability of spray±wall interactions. The model conserves
mass, tangential momentum and energy of an impinging parcel. This model focuses
on spray impact on dry and wet surfaces below the fuel's Leidenfrost temperature, a
scenario encountered under typical engine operating conditions (Han et al., 2000).
Three splashing parcels and one wall film parcel are used to represent the shattering
of a splashing droplet upon impact with the surface. It is assumed that the impulsive
force on an impinging droplet normal to the surface is dominant, thus allowing one
to treat the magnitude of its tangential momentum component constant after impact.
The viscous dissipation of an impinging droplet and kinetic energy of the wall film
are accounted for in the energy conservation equation.

3.3 Ignition and combustion models

The Shell Ignition model (Halstead et al., 1975) is based on a general eight-step
chain-branching reaction scheme, which uses lumped kinetic parameters and other
terms as coefficients in the conservation equations. The species involved are RH, the
hydrocarbon fuel of the form CnH2n, O2, R radical formed from the fuel, B
branching agent, Q intermediate species and P products consisting of CO, CO2 and
H2O in specific proportions. The five conservation equations that describe the above
mechanism consist of a series of coupled differential equations describing the
concentrations of the chemical species that influence heat release in the auto-ignition
process and the system temperatures.
 In-cylinder reduction of PM and NOx emissions from diesel combustion 89

The model used in this work is an enhanced version of the original Shell Ignition
model (Hamosfakidis and Reitz, 2003). Three main deficiencies have been recognised
and eliminated in this improved version. The first one concerns the calculation of the
heat release, which was based on an assumed fixed ratio CO/CO2 in the products.
This assumption has been removed and the heat release is given instead by an energy
balance. The second modification is related to the inert products of the two
termination reactions. In the previous scheme, radicals are removed from the
reactants pool by converting them into N2. Here, it is assumed that the two radical
termination reactions lead to the same species that would result from the combustion
of the same initial mixture of reactants. The third modification is related to the Shell
species R , B and Q. In the previous scheme the contribution of these species to the
energy balance was ignored. In the present scheme enthalpy values for these generic
species have been assigned.
The characteristic-time combustion (CTC) model (Han et al., 1996) is used for
combustion simulations, assuming that seven species (fuel, O2, N2, CO, CO2, H2,
H2O) are necessary to predict the thermodynamic equilibrium temperatures. The rate
of change of the mass fraction of species m, Ym , due to chemical reaction is given by:
dYm Ym ÿ Ym
ˆÿ ; …6†
dt c
where c is the characteristic time of combustion and Ym denotes the local
equilibrium mass fraction. The characteristic time is assumed to be the same for all
seven species and it is the weighted sum of a laminar time scale, l , and a turbulent
time scale, t , given by:
c ˆ l ‡ ft ; …7†
where the weight function f simulates the influence of turbulence on combustion after
ignition has taken place. The turbulent time scale is proportional to the eddy
turnover time
t ˆ CM k=": …8†
The constant CM is an input variable that acts as a scaling parameter between the
different engines and their injection configurations.

3.4 NOx and soot formation models

The extended Zeldovich Mechanism, as given by Heywood (1988) has been used to
predict NO formation from diesel combustion. The principal reactions for NO
formation and destruction are:
O ‡ N2 $ NO ‡ N
N ‡ O2 $ NO ‡ O
N ‡ OH $ NO ‡ H:
The soot formation model developed by Hiroyasu and modified by Han et al. (1996)
has been used throughout this work. The model predicts the production of soot mass,
_ sf , and the
Ms , by a single-step competition between the soot mass formation rate, M
_
soot mass oxidation rate, Mso according to:
90 C.A. Chryssakis et al.

dMs _ sf ÿ M
_ so :
ˆM …9†
dt
The Arrhenius formation rate is proportional to the fuel vapour mass, Mfv , and the
formation coefficient is a function of pressure and temperature, according to:
ÿ 

Kf ˆ Asf P1=2 exp ÿEsf RT : …10†
The Arrhenius oxidation rate is proportional to the soot mass and the oxidation
coefficient is a function of pressure, temperature and the oxygen mole fraction. The
original model has been modified by replacing the Arrhenius global oxidation
rate equation with the experimentally based oxidation rate of Nagle and
Strickland-Constable (NSC). The NSC oxidation rate mechanism offers improved
accuracy in predicting the oxidation rate because it is based on oxidation experiments
of carbon graphite in an O2 environment over a range of representative partial
pressures (Han et al., 1996).

3.5 Model validation

The prediction capabilities of the CFD models have been verified by comparing the
predicted combustion characteristics and experiments for single and split injection
cases. The calculations and comparisons were carried out for a range of injection
timings and loads. Figure 3 shows the cylinder pressure comparisons for a
representative single injection and a pilot-main injection case. Both cases correspond
to an engine speed of 1500 rpm, 3.25 bar Brake Mean Effective Pressure (BMEP).
The injection timing for the single injection was 7 CA BTDC, and for the split
injection strategy, the injection timings were 28 and 7 CA BTDC for the pilot and
the main injection events, respectively. The agreement between CFD and
experiments is excellent for both cases. Similar agreement was observed for most
of the other operating conditions considered in this investigation. Some discrepancies
were observed only in case of extremely long dwell angles (more than 40 ), when cold
flames were observed before the main event. Further refinement of the models is
needed for resolving the increasingly complex phenomena associated with such
conditions. Comparison of emissions predictions with measured data for a wide
range of conditions is given in the next section.

Figure 3 Comparison of CFD predictions and experimentally measured cylinder pressure for
a single and a pilot-main injection strategy, at 1500 rpm, 3.25 bar BMEP
 In-cylinder reduction of PM and NOx emissions from diesel combustion 91

4 Effect of pilot injections on mixing and emissions

Pilot injections are typically used in diesel engines for controlling combustion noise
during idle. Recent advances in capabilities of injection systems enable extending this
concept to a wider range of conditions in order to enhance the fuel±air mixing and
achieve a less stratified charge, compared to a single injection strategy. In addition,
an increased amount of fuel is burned under relatively low temperatures, thus
limiting the amount of NOx formed during combustion. Optimisation of the pilot
injection timing and amount of fuel injected is crucial for reaching the desired effect
of reducing emissions. If the injection occurs too early, fuel evaporation is poorer due
to lower cylinder temperatures and a significant amount of fuel impinging on the
piston bowl, creating a wall-film and resulting in soot formation and unburned
hydrocarbons. Furthermore, extremely early injections can result in fuel spray
missing the bowl and liquid fuel reaching the liner, mixing with oil, thus leading to
both high soot formation rates and unburned hydrocarbons.
A preliminary study of the influence of pilot injection has been performed
experimentally, showing the effects of timing in reducing NOx emissions and its
relation to PM found in the exhaust. After analysing the measurements and
characterising the effect of the pilot injection timing, a series of complementary CFD
simulations were conducted in order to explain the experimentally observed trends.
Five different dwell angles are investigated at each condition, with the smallest dwell
angle as the limit imposed by engine control software. The discussion focuses on the
1500 rpm, 3.25 and 7 bar BMEP cases, with the main injection timing set to 7 and
3.3 BTDC respectively, and the pilot dwell at 15, 21, 26, 36 and 46 , but a wider
range of engine speeds and loads with similar pilot timings has been investigated.
Because of the injection system hardware and software limitations, the quantity of
fuel in the pilot is not controllable. At any one operating condition, however, the
pilot fuel mass remains about the same as the pilot dwell is lengthened. Total fuel
injection quantity at each point remains the same, as does main injection timing. The
fuel mass for each injection event is given in Table 2. The EGR valve is closed and the
VGT vanes are fixed to remove the influence of these two controllable inputs.
Standard calibration fuel injection pressure is used throughout this study.

Table 2 Fuel mass injected in each injection event in mg

3.25 bar 7.00 bar

Dwell Pilot (mg/stroke) Main (mg/stroke) Pilot (mg/stroke) Main (mg/stroke)

0 0 17.7 0 31.9
15 8.33 9.36 6.01 25.92
21 8.34 9.3 6.54 25.6
26 8.43 9.19 6.48 25.2
36 6.4 11.35 8.49 23.58
46 5.39 12.51 12.45 19.28
92 C.A. Chryssakis et al.

4.1 NOx emissions ± measurements and CFD analysis of the dominant effects
The effect of pilot dwell on NOx has been investigated experimentally, as shown in
Figure 4, for the 1500 rpm, 3.25 and 7 bar BMEP operating conditions. Figure 4
reveals that as the pilot dwell increases, NOx emissions decrease. However, when a
short pilot dwell is compared to a single injection, there is initially an increase in NOx
concentrations for both cases. This can be linked to the fact that the timing of the
main injection event was not re-optimised, hence leading to relative advancement of
burning when a portion of fuel is injected as pilot. In the 7 bar BMEP case,
lengthening the pilot dwell from 15 to 46 results in a NOx reduction of 60.6%. The
impact is less effective at NOx reduction as the pilot timing advances; changing the
pilot dwell from 15 to 21 accounts for 73.8% of the NOx decrease. Similar
conclusions are reached for the 3.25 bar BMEP conditions. This drastic drop is
attributed to the lower cylinder temperatures reached for longer dwells. NOx
production is highly sensitive to temperatures and as the pilot dwell goes up, the peak
cylinder temperatures tend to decrease and occur at a later crank angle. In Figure 4
the CFD predictions have also been plotted for the 3.25 bar BMEP case,
demonstrating a very good agreement with the experimental measurements.

Figure 4 NOx emissions with pilot dwell sweep at 1500 rpm, 3.25 and 7 bar BMEP

KIVA-3V was used to generate temperature profiles at specific crank angles to

obtain a fuller understanding of NOx production trends. Analysis reveals that as the
pilot timing is brought closer to the main injection, temperatures get higher and
encompass a greater proportion of the cylinder. Figure 5 demonstrates how the
temperatures change at the longest and shortest investigated pilot dwells at the
1500 rpm, 7 bar BMEP operating condition. In the histogram, the temperature
distribution as a function of crank angle is shown. The temperature values were
distributed in 250 K bins and in the vertical axis the proportion of the combustion
chamber volume that corresponds to each temperature bin is shown. Cylinder
temperature predictions from CFD in the 16 dwell case increase sooner than the 46
dwell case due to the effect of a more pronounced premixed spike, as shown by the
heat release rate curves in Figure 6. As combustion progresses, a greater proportion
of the cylinder is consistently above 2000 K when a 16 pilot dwell is used. These
higher-temperature conditions that occur over an extended crank angle interval are
more likely to produce higher NOx levels.
 In-cylinder reduction of PM and NOx emissions from diesel combustion 93

Figure 5 Cylinder temperature predictions with CFD with 16 and 46 pilot dwells at
1500 rpm, 7 bar BMEP

Figure 6 Heat release rate for 1500 rpm, 7 bar BMEP, experimental measurements
94 C.A. Chryssakis et al.

4.2 Soot emissions

Figure 7 shows the influence of pilot dwell on soot emissions at the 1500 rpm, 3.25 and
7 bar BMEP operating conditions. In the 7 bar BMEP case, the soot emissions decrease
when a short pilot dwell is introduced compared to the single injection. This can be
tied to the same explanation made in the previous section about initial increase of
NOx with the introduction of pilot. As the pilot dwell increases at the 7 bar BMEP
operating condition, the soot emissions first increase and then decrease. This trend
tends to follow peak temperature data; that is, when the peak cylinder temperature
goes up, the soot emissions go down. A hotter flame promotes more complete
oxidation of the soot particles. About a 25% difference exists between the smallest and
largest soot concentrations as the pilot timing is swept. In the 3.25 bar BMEP case, the
soot emissions drop compared to the single injection as a short dwell is introduced,
but they increase as the pilot dwell lengthens. The CFD predictions agree reasonably
well with the experimental measurements for the 3.25 bar case, shown in Figure 7.

Figure 7 Soot emissions with pilot dwell sweep at 1500 rpm, 3.25 and 7 bar BMEP

Soot production and the impact of pilot dwell is highly dependent on engine
operating conditions. Most conditions considered experienced an increase and a
decrease in soot emissions as the pilot was advanced. These alterations are similar in
magnitude to those displayed in Figure 7, but the shapes of the curves vary widely.
Most operating conditions achieved lower soot emissions with a properly-timed pilot
than they produced with a single injection. The presence of these variances and lack
of any concrete link between soot production and pilot dwell indicates a highly-complex
soot-production process and its dependence on engine operating conditions.
However, relatively smaller changes of soot with dwell and simultaneous large
reductions of NOx with dwell offer a promise of improved overall trade-off.

4.3 Soot-NOx tradeoff

From the soot and NOx trends identified in the preceding two sections, it is clear that
changing the pilot dwell does not function on the traditional soot-NOx trade-off curve.
While NOx emissions tend to decrease markedly with increasing dwell, soot emissions
may either go up or down as the pilot timing is advanced. Figure 8 illustrates the
 In-cylinder reduction of PM and NOx emissions from diesel combustion 95

measured soot±NOx curve at the 1500 rpm, 7 bar BMEP operating condition generated
by sweeping the pilot dwell. Also shown on the graph is the soot±NOx condition
obtained with a single injection. This chart demonstrates that pilot dwell has a
relatively large impact on NOx production and a lesser influence on soot emissions,
thus allowing the optimum point to move significantly SW, i.e. towards the origin of the
graph, compared to the single injection point. Significant improvement over the single
injection case is realised ± a 26.1% reduction in soot and a 44.1% decrease in NOx.

Figure 8 Soot±NOx trade-off with pilot dwell sweep at 1500 rpm, 7 bar BMEP, experimental
measurements

4.4 Hydrocarbon emissions

Unburned hydrocarbon emissions increase as the pilot dwell lengthens. The
hydrocarbon emissions from the 1500 rpm, 3.25 and 7 bar BMEP conditions are
shown in Figure 9. As the pilot dwell increases from 15 to 46 , the hydrocarbon
emissions increase by a factor of three. This indicates that larger quantities of fuel
are either only partially burning or not combusting at all. This is most likely due to
over-mixing of the fuel as the pilot advances, as well as increased wall impingement.

Figure 9 Hydrocarbon emissions measurements with pilot dwell sweep at 1500 rpm, 3.25 and
7 bar BMEP
96 C.A. Chryssakis et al.

4.5 Engine performance

The relationship between pilot dwell and Brake Specific Fuel Consumption (BSFC)
at the 1500 rpm, 3.25 and 7 bar BMEP condition is shown in Figure 10. The total
mass of fuel injected is kept constant at each speed/load combination. As the pilot
dwell increases from 15 , BSFC increases slightly, reaches a plateau and finally
decreases slightly again for the extremely advanced pilot. The variance between the
maximum and minimum BSFC values while using a pilot is 2.4% for the 7 bar BMEP
case. The BSFC with a pilot is somewhat greater than BSFC obtained with a single
injection, no matter what the timing. The observed BSFC trends can be attributed to
changes in combustion phasing due to pilot injection. The heat release resulting from
the pilot injection causes a premature increase of cylinder pressure before TDC (see
Figure 6), hence reducing somewhat the indicated work from the cycle. These
premixed spikes are more advanced in case of the 21, 26 and 36 dwell, while the
extreme case of 46 demonstrates slower burning and a lower peak of the initial rate
of heat release profile. The timing of the main injection event was intentionally held
constant, and given the main objective of the paper there was no attempt to re-optimise
its timing for each case. However, there is no fundamental reason why it could not be
done, hence minimising the BSFC penalty in the practical application of the
proposed strategy.

Figure 10 Effect of pilot dwell on BSFC, 3.25 and 7 bar BMEP

5 Effect of post-injections on emissions ± a CFD study

Encouraging results obtained with pilot injection motivate further work on a more
sophisticated, three-stage injection schedule. The current hardware and software
used in the test cell does not allow for post injection; therefore post-injection
strategies have been considered only in computational studies. Capabilities of the
CFD simulation tool have been utilised to perform a parametric study investigating
the effect of post-injection timing and amount of fuel injected, in order to
characterise the trends and quantify the trade-offs. The insight obtained through this
systematic study enables exploration of strategies that can lead to further reduction
of both NOx and soot in the cylinder.
 In-cylinder reduction of PM and NOx emissions from diesel combustion 97

The parametric study was based on the 1500 rpm, 3.25 bar BMEP case, because it
showed good agreement with experimental measurements when the pilot injection
strategies were simulated, as shown in Figures 3, 4 and 7. The pilot dwell was set to
21 and the main injection timing was fixed at 7 BTDC. The total fuel mass was held
constant, at 17.7 mg/cycle (ˆ 1.77 g/s), as well as the pilot injection quantity, at 8.4 mg/
injection. The remaining fuel was distributed between the main- and the post-injection.
During this investigation the effects of fuel quantity in the post-injection and the
dwell between the main- and post-injection were simultaneously studied. The post
fuel quantity was varied from 0.5 to 2.0 mg in increments of 0.5 mg, while the main
injection fuel quantity was accordingly adjusted to maintain the total amount
constant. The post dwell values were 6, 11, 16, 21 and 26 . The shortest of these
values (6 and 11 ) may be difficult to realise in a practical engine system but they
demonstrate the potential of post-injection strategies, as well as the advantage of
CFD simulations for understanding physical mechanisms.

5.1 NOx and soot predictions

The CFD calculations for the matrix of 20 points corresponding to the conditions
described in the previous section (4 for fuel quantity, 5 for injection timings) provide
an estimate for the influence of the two parameters under investigation on emissions.
The predictions for NOx emissions, shown in Figure 11, demonstrate a strong trend
towards reducing nitric oxides as dwell between the main and the post increases.
Furthermore, it appears that NOx concentration generally drops as the fuel quantity
of the main decreases (and the quantity of the post increases). Both these trends can
be explained through combustion temperature considerations. As the dwell increases,
the post-injection takes place in a relatively cool environment, resulting in burning a
portion of the fuel in temperatures lower than 2000 K and thus avoiding conditions
that produce higher levels of NOx. In addition, as the fuel quantity of the post
increases, the fuel injected in the main is decreased; therefore combustion
temperatures around TDC are lower since less fuel is being burned. The trend
becomes stronger as the amount of pilot is increased beyond 1.5 mg.

Figure 11 NO concentration predictions with CFD for 1500 rpm, 3.25 bar BMEP
98 C.A. Chryssakis et al.

The soot emissions predictions are shown in Figure 12. Inspection of these results
reveals that soot emissions can be significantly reduced when the main±post dwell is
shortened. Also, for short dwells, there is a tendency to reduce soot emissions as the
post mass increases, but this trend is weaker. This is because the post-injection can
enhance soot oxidation when the timing is optimised. It appears that for early
post-injections, the combustion temperatures are sufficiently high to oxidise the already
existing soot in the combustion chamber. On the other hand, when post-injections occur
later in the expansion stroke, the fuel is burned at somewhat lower temperatures, thus
resulting in further soot formation without the benefits of enhanced soot oxidation.
This is illustrated in Figure 13, where the soot formation and oxidation history in the
cylinder is plotted, for a short- and a long-dwell case. Even though a larger amount
of soot is formed in the short-dwell case, it initiates a faster oxidation process,
thus resulting in low soot levels at the end of the cycle. On the other hand, a late
post-injection produces moderate amounts of soot but it cannot trigger a strong
oxidation reaction, therefore contributing to higher soot concentration in the exhaust.

Figure 12 Soot concentration predictions with CFD for 1500 rpm, 3.25 bar BMEP

Figure 13 Soot formation and oxidation history for two different post-injection timings,
as predicted with CFD
 In-cylinder reduction of PM and NOx emissions from diesel combustion 99

Comparison of Figures 11 and 12 shows that conditions that are favourable for
reducing one of the two pollutants being considered may result in increasing the
other one. The most obvious case is the lower right-hand point of the graphs,
representing the longest dwell with the largest amount of fuel injected in the post.
Even though the NOx emissions are considerably reduced, the additional soot formed
is significantly high. On the other hand, operating with a large amount of fuel in the
post-injection (2 mg) and a moderate dwell allows significant reduction of soot
combined with relatively low NOx.
This is demonstrated in Figure 14, where the comparison between the single,
pilot-main and pilot-main-post strategies is shown. Pilot injections have the potential
to reduce both emissions compared to the single injection if timing is appropriately
selected. Post-injections can further reduce the emissions if an optimisation of timing
and fuel quantity is performed. In this investigation, only the two points shown in
Figure 14 offered improved results, while the remaining points were not advantageous
compared to the pilot injections. The two points represent main-post dwell of 6 and
11 , while the fuel mass in the post is 2 mg in both cases. The two points are marked
in Figure 12 as A and B, respectively. Other cases suffered from undesired effects of
non-optimised post-injection on emissions formation, especially in soot formation.
When the main-post dwell lengthens, the fuel is injected in a relatively cool
environment, favouring soot formation but eliminating its oxidation. The same effect
occurs when the amount of fuel in the post is not sufficient to enhance oxidation of
the soot previously formed.

Figure 14 Comparison of injection strategies for 1500 rpm, 3.25 bar BMEP,
CFD predictions

It is expected that the magnitude of emissions reduction depends on the engine

operating conditions, such as speed and load, and therefore more pronounced effects
could be observed if other conditions are investigated. Realising the full potential of
the pilot±main±post strategy requires optimisation of all injection parameters for any
given set of conditions. The insight obtained with the current investigations provides
a clear guidance for more applied studies.
100 C.A. Chryssakis et al.

6 Conclusions

The effect of multiple injections on reducing pollutants from diesel combustion has
been investigated through a synergistic approach combining experiments and CFD
computations. It has been shown experimentally that pilot injections have the
potential to reduce NOx and PM emissions simultaneously when the timing of the
pilot is selected appropriately. In general, large dwell was favourable, leading to a
significant reduction of NOx with a small soot penalty. Furthermore, CFD analysis
of these cases revealed that these improvements are due to better mixing and
reduction of the size of high-temperature zones during the main combustion phase.
The possibility to further reduce emissions with post-injections was studied using
additional CFD analysis. Parametric studies showed that increased amount of fuel in
the post-injection reduces NOx by lowering the mass of fuel and peak temperatures in
the main event, while simultaneously enhancing soot oxidation when the timing is
optimised. Consequently, the further improvement of the soot-NOx trade-off was
demonstrated for a maximum mass of fuel in the post-injection and a moderate dwell
between the main- and the post-injection event. In general, it was observed that the
influence of advanced injection strategies varies depending on the engine operating
conditions, and optimisation of injection parameters, such as timing and fuel
quantity, has to be performed at each engine speed and load combination for
minimum pollutant formation.

Acknowledgements

The authors would like to acknowledge the technical and financial support of the
Automotive Research Center (ARC) by the National Automotive Center (NAC), and
the US Army Tank-Automotive Research, Development and Engineering Center
(TARDEC). The ARC is a US Army Center of Excellence for Automotive Research
at the University of Michigan, currently in partnership with the University of
Alaska-Fairbanks, Clemson University, University of Iowa, Oakland University,
University of Tennessee and Wayne State University. In addition, the Dual Use
Science and Technology Program `Simulation-Based Design and Demonstration of
Next Generation, Near-zero Emission Diesel Technology' sponsored by TARDEC,
Ford Motor Co. and International Truck and Engine Corporation supported
development of the experimental capabilities important for this study. International
Truck and Engine Corp. is acknowledged for providing the V-8 engine and
information about engine geometry required for setting up the CFD simulation.

References

Amsden, A.A. (1993) `KIVA-3: a KIVA program with block-structured mesh for complex
geometries', Los Alamos National Laboratory LA-12503-MS.
Amsden A.A. (1997) `KIVA-3V: a block-structured KIVA program for engines with vertical
or canted valves', Los Alamos National Laboratory LA-13313-MS, July.
Amsden, A.A., O'Rourke, P.J. Butler, T.D. (1989) `KIVA II ± a computer program for
chemically reactive flows with sprays', Los Alamos National Laboratory LA-11560-MS.
 In-cylinder reduction of PM and NOx emissions from diesel combustion 101

Benajes, J., Molina, S. and Garcia, J.M. (2001) `Influence of pre- and post-injection on the
performance and pollutant emissions in a HD diesel engine', SAE Technical Paper Series
2001-01-0526.
Christian, V.R., Knopf, F., Jaschek, A. and Schindler, W. (1993) `Eine neue messmethodik der
Bosch-Zahl mit erhoÈhter empfindlichkeit', MTZ, Vol. 54, pp.16±22.
Grover, R.O. and Assanis, D.N. (2001) `A spray wall impingement model based upon
conservation principles', Fifth International Symposium on Diagnostics and Modeling of
Combustion in Internal Combustion Engines, pp.551±559.
Halstead, M.P., Kirsch, L.J., Prothero, A. and Quinn, C.P. (1975) Proc. R. Soc. Lond., p.346.
Hamosfakidis, V. and Reitz, R.D. (2003) `Optimisation of a hydrocarbon fuel ignition model
for two single component surrogates of diesel fuel', Combustion and Flame, Vol. 132,
pp.433±450.
Han, Z., Uludogan, A., Hampson, G.J. and Reitz, R.D. (1996) `Mechanism of soot and NOx
emission reduction using multiple-injection in a diesel engine', SAE Technical Paper
Series, 960633.
Han, Z., Xu, Z. and Trigui, N. (2000) `Spray/wall interaction models for multidimensional
engine simulation', International Journal of Engine Research, Vol. 1, pp.127±146.
Heywood, J.B. (1988) Internal Combustion Engine Fundamentals, McGraw-Hill.
Ishikawa, N., Uekusa, T., Nakada, T. and Hariyoshi, R. (2004) `DI diesel emission control by
optimized fuel injection', SAE Technical Paper Series 2004-01-0117.
Liu, A.B. and Reitz, R.D. (1993) `Mechanisms of air-assisted liquid atomization', Atomization
and Sprays, Vol. 3, pp.55±75.
Liu, A.B., Mather, D. and Reitz, R.D. (1993) `Modeling the effects of drop drag and breakup
on fuel sprays', SAE Technical Paper Series 930072.
Park, C., Kook, S. and Bae, C. (2004) `Effects of multiple injections in a HSDI diesel engine
equipped with common rail injection system', SAE Technical Paper Series 2004-01-0127.
Payri, F., Benajes, J., Pastor, J.V. and Molina, S. (2002) `Influence of the post-injection pattern
on performance, soot and NOx emissions in a HD diesel Engine', SAE Technical Paper
Series 2002-01-0502.
Reitz, R.D. (1987) `Modeling atomization processes in high-pressure vaporizing sprays',
Atomisation and Spray Technology, Vol. 3, pp.309±337.
Reitz, R.D. and Diwakar, R. (1987) `Structure of high-pressure fuel sprays', SAE Technical
Paper Series 870598.

Nomenclature

Asf Empirical constant for the soot formation model

Bo ; B1 Empirical constants for the spray break-up model
C Soot concentration
CM Scaling constant for the characteristic-time combustion model
Esf Activation energy
k Turbulent kinetic energy
Kf Arrhenius reaction coefficient
MS Soot mass
P Pressure
R Ideal gas constant
102 C.A. Chryssakis et al.

ro Initial drop diameter

T Temperature
We Weber number
Ym Mass fraction of species

Greek symbols
" Turbulent eddy dissipation
 Wavelength
L Liquid density
 Surface tension
 Time constant

Maximum growth rate