Improving the NOx-BSFC Trade Off of a Turbocharged Large

Diesel Engine Using Performance Simulation

Ki-Doo Kim and Dong-Hun Kim
Hyundai Heavy Industries Co., Ltd.

ABSTRACT
The purpose of this study is to determine the
optimum intake valve closing time of a large diesel
engine having lower fuel consumption and lower NOx
emission.
The performance simulation has been conducted for
this purpose, and a phenomenological combustion
model is verified by experimental data of heat release
rate and NOx emission in order to enhance the
prediction quality of the performance simulation. The
results of performance simulation are compared with
measured data to confirm the modeling method and
results. The fuel injection system simulation has been
also performed to get fuel injection rate, and the results
is also verified by experimental data of fuel injection
pump pressure and injected fuel mass.
The performance simulation investigate the
application of Miller cycle to a large diesel engine, and
so, the intake valve closing time is determined at the
condition of reducing NOx emission and fuel
consumption at the same time. As that result, Miller
cycle has a feature that the maximum reduction of NOx
emission is 15.7% while the improvement of specific fuel
oil consumption is 1.7g/kWh.

INTRODUCTION
Large diesel engines are widely used for stationary
power generation and ship propulsion. But it is important
to reduce their emissions to a reasonable level because
diesel engines are one of the major sources for air
pollution. In recent years, the maximum quantities of
oxides of nitrogen (NOx) are limited for marine
applications by the IMO regulation. In case of land
installations for generating electricity, there are many
different regulations specific to the various countries and
regions. It has become clear that the development of
diesel engines must concentrate on reducing exhaust
gas emissions. Many methods for reduction of the
emissions, especially for NOx, however have a
contradictory effect on fuel oil consumption of an engine.
It therefore is desirable to reduce both fuel oil
consumption and NOx emission at the same time for a
competitive engine.

It is important to determine optimum operating

conditions and basic geometries of an engine for an
economical and ecological development. All design
parameters must be optimized together. However the
interaction of all these parameters cannot be overlooked
without simulation tool, namely performance simulation.
In recent years, the latest simulation tool of the engine
performance has a feature combined with threedimensional flow analysis. But it is difficult to verify threedimensional combustion and flow analysis in combustion
chamber because of the requirement of visualization
tool. In addition, it is impossible to catch the real
situation of three-dimensional combustion with moving
piston. The other method to predict combustion is using
a phenomenological multi-zone model in a performance
simulation. This simulation method is useful in respect of
time and cost. But it needs many experimental data of
engine performance, and cannot figure out the effect of
three-dimensional configuration in combustion chamber
on the combustion and flow. The experimental data
needed to verify a phenomenological combustion model
are mainly heat release rates, cylinder pressures, and
NOx emissions including engine performance, and can
be got easily from usual engine performance
measurement. The heat release rate also can be
obtained easily from data of cylinder pressure.
The combustion analysis always needs a fuel
injection rate, which can be obtained from the simulation
of a fuel injection system. It is also difficult to verify this
simulation because of the difficulties in collecting
experimental data of the fuel injection rate. However, it is
easy to measure the pressure of injection pump and
injected fuel mass in an engine.
This paper introduces the practical experience for an
application of the engine performance simulation to get
lower fuel consumption and lower NOx emission on a
diesel
engine.
The
possibility
to
apply
a
phenomenological
combustion
model
to
the
performance simulation has been investigated using
usual experimental data of engine performance for
specifying optimum engine operation parameters. And
the possibility of using fuel injection simulation verified
by injection pump pressure and injected fuel mass has
also been surveyed for a successful performance
simulation.

MODELING OF PERFORMANCE SIMULATION

The performance simulation S/W, WAVE from
RICARDO is used to model a direct injection,
compression ignition, and turbocharged diesel engine.
The engine geometries and its specifications are
summarized in Table 1.

Bore

Table 1 Engine Specifications

210 mm

Stroke

320 mm

Connecting Rod Length

640mm

Rated Power

200kW/cyl.

Engine RPM

900rpm

The performance simulation is in need of

calculation of combustion process for more accurate
results of NOx emission, peak combustion pressure, fuel
consumption, and exhaust gas temperature according to
the variations of operating conditions and basic
geometries. Three methods have been used in diesel
combustion modeling: zero-dimensional, phenomenological multi-zone and multi-dimensional computational
fluid dynamics models. Zero-dimensional models are
widely used for overall performance simulation.
However, such models cannot give any details of
temperature and composition distribution in the cylinder,
and hence they are unsuitable for investigations of
exhaust gas emissions. Phenomenological multi-zone
models take into account the spatial differences in
temperature and chemical compositions by dividing a
cylinder into two or more different zones. Because the
model can give local temperature and compositions, it
can be used to predict exhaust emissions and cylinder
pressures. However, it needs many experimental data of
an engine for accurate simulation results. Multidimensional simulation has negative effects on time and
cost.
In this study, a phenomenological model, Diesel
Jet Combustion in WAVE S/W is used to predict engine
performance for variable operating conditions and basic
geometries. It has the ability to predict combustion heat
release rate from user-specified fuel injection rate and
injector geometry. It also predicts NOx emissions using
the Zeldovich mechanism, but has many parameters
that affect velocities, turbulence levels, and gas mixing
effects in the combustion chamber. To enhance the
prediction quality of the performance simulation, these
parameters must be suitably determined with
experimental data in the target engine. The parameters
using experimental results of heat release rate and NOx
emissions are determined from the target engine.

Turbocharger modeling needs performance maps

of its compressor and turbine used in a target engine. It
directly affects the simulation results of air flow rate, and
intake and exhaust pressure. These are obtained from a
turbocharger manufacturer, ABB.
It is important to get flow coefficients and swirl
ratios of intake and exhaust head port in order to
calculate more accurate air flow rate and intake
pressure. These also have effects on compression
pressure of combustion chamber because of the
pressure loss in intake port. In this study, these have
been obtained from a steady state flow test rig.
Heat transfer in combustion chamber affecting
thermal efficiency, compression pressure has been
calculated using the Woschni correlation in WAVE S/W.
It also needs suitable coefficients of the correction and
wall temperatures of cylinder head, liner, and piston on
the target engine to enhance the simulation results of
heat loss in combustion chamber. These coefficients
have been determined from performance data, and the
temperature data have been taken from experimental
data of the engine.
Friction mean effective pressure(FMEP) is also
important to get accurate result of fuel consumption rate.
In this study, the Chen-Flynn correlation in WAVE S/W
has been used with appropriate coefficients determined
from experimental data of the target engine.

VERIFICATION OF PERFORMANCE
SIMULATION
The performance simulation results were
compared with experimental data in heat release rate
and pressure of cylinder, intake and exhaust as well as
general engine performance data for the verification.
The most basic and useful factor for studying the
combustion process is the heat release rate that
representatively indicates entire combustion conditions
including start, duration and shape of combustion.
Figure 1 shows the comparison of simulation results and
experiment of heat release rate in combustion chamber
according to the variation of engine load. No significant
effect on the premixed combustion is shown up in the
heat release profile. The combustion durations over all
load conditions are shorter than 60degree due to very
high fuel injection and cylinder compression pressure.
The simulation results are similar to experiment in all
load conditions except 25 % due to the simulation with a
lot of the late combustion. And hence, it is found that the
used combustion model is well modeled in the variation
of load, but some improvements are needed for the
calculation for lower load.
Figure 2 shows experiment and simulation results
of intake, exhaust and cylinder pressure at 100% load. It
indicates that the simulation modeling is constructed well

5.0

Cylinder

Experiment
Simulation

Load 100%

3000

Simulation

Intake

4.0
3.5
3.0

Exhaust

2.5
2.0

75%
50%
25%

1500
1000
500
0
0

10

20

30

40

Crank Angle(deg. ATDC)

50

60 120 180 240 300 360 420 480 540

Crank Angle(deg.ATDC)

Figure 2 Verification of Simulation Results with

Experiment in Intake, Exhaust, and Cylinder Pressure

2500
2000

Experiment

4.5

-180 -120 -60

3500

Heat Release Rate(J/deg.)

complex NOx generation mechanism. And so, it is

concluded that the simulations of NOx emissions show
satisfactory results. The maximum error rate of the
simulation results in the maximum cylinder pressure has
6.43 at 25% load. That has the same reason as specific
fuel oil consumption. The simulation results of the other
parameters ; intake pressure, exhaust temperature and
air flow, are also similar to those of the experimental
data. In conclusion, every results of performance
simulation are verified with experimental data.

Pressure(bar)

in the intake and the exhaust system. The simulation

result of the intake pressure shows lower fluctuation
than that of the experiment, but this average value is
almost equal to that of the experiment. That is because
of the air receiver modeled with one junction that has
large volume. The pressure fluctuation of intake
pressure is however small, therefore it has very small
effects on the overall engine performance. The similarity
between simulation result and experiment in average
value of the intake pressure indicates that the flow
coefficient of intake valve and the turbocharger are
modeled successfully. The exhaust pressure of
experimental data during entire one-cycle shows three
peak pressures, which are the characteristic of threepulse turbocharging system. The exhaust pressure in
the simulation has a shape similar to that of the
experiment. But the minimum value is lower than that of
the experiment, and hence, it needs improving the
friction model of exhaust pipe or turbocharger turbine
model. The cylinder pressure profile in Figure 2 indicates
whether flow coefficients of the intake and the exhaust
valve are correct or not. This simulation result of cylinder
pressure is similar to that of the experiment, and
therefore it can be concluded that the flow coefficients
are modeled very well.

60

Figure 1 Verification of Simulation Results with

Experiment in Heat Release Rate
Figure 3 indicates the comparison of simulation
results with experiment in main performance parameters
with error rates over the all load range. The simulations
of specific fuel oil consumption have very good results
with lower error rate than 2% except 25% load whose
error rate is only 3.57%. Higher error rate at 25% load is
attributed to lower combustion pressure due to a lot of
late combustion. The simulation results of NOx
emissions have lower error rate than 5.7% over all load
conditions. It is difficult to predict NOx emissions
because it is related to cylinder pressure, air flow and

Experiment

3.57

195

200
160

0.88

3.16

5.69

-3.52
0.77

13
12
11
10

Error Rate(%)
1.58

120
80

0.42

0.76 -0.66 -1.65

0.43

0.33

-4.46

-6.43

3
2

500

450
400
3

350

2
1
0

25

50
75
Load (%)

100

Intake Pre. (bar)

210

NOx(g/kwh)

Simulation

225

Air Flow(kg/h)

Exh. Temp.(oC) Max. Cyl. Pre.(bar) SFOC(g/kWh)

240

Figure 3 Comparisons of Simulation Results with

Experiment in Main Performance Parameters

MILLER CYCLE
In the 1940s, Miller(1947)[1] proposed a new
system, Miller cycle which has early intake valve closing,
and so, shorter effective compression stroke than
expansion. Thus it is conceptually equivalent to an
Atkinson cycle, and can get high cycle efficiency. There
are also many studies to get reducing knocking in
gasoline and gas engines, and reducing NOx emissions
in diesel engines using Miller cycle because it can
reduce the compression pressure and trapped air
temperature in cylinder.
Bolton and Assanis(1994)[2] have used a
comprehensive computer simulation to determine
optimum breathing strategies, and concluded that
variable valve timing cycle can improve power and fuel
economy by up to 14% at high speeds for a given peak
pressure constraints. The results also show that the
success of Miller cycle depends on creating increased
air pressure using the turbocharger.
According to Stebler et al., (1996)[3] Miller cycle had
that NOx emissions were reduced by 15 ~ 20% and fuel
consumption was decreased by 0.5 ~ 2% at the same
time, and either variable inlet valve timing or variable

turbine geometry could overcome the reduced air-fuel

ratios at low part load.
Mavinahally et al., (1997)[4] have shown that the
insulated Miller cycle engine with intake valve closing
time of 80deg.aBDC improved in specific fuel oil
consumption of 4.23% at 2600RPM and 93% load.
Stebler et al., (1998)[5] have shown that NOx
emissions could be reduced to around 20% and fuel
consumption 0.5 ~ 2% in the high load range. They also
suggested the necessity of a variable turbine nozzle
geometry to solve low air-fuel ratio at low part load.

PERFORMANCE ANALYSIS OF MILLER CYCLE

Miller cycle is investigated to apply it to the target
engine using performance simulation. The performance
simulation is carried out to get a optimum intake valve
close timing of the target engine. The simulations is
executed with changing the intake valve closing
time(IVC); -30, -20, -10, 0, 10, 20, 30, and 46deg.aBDC
in the intake stroke. Figure 4 shows the valve lift profiles
used in this simulation.
Figure 5 indicates the simulation results of Cylinder
Pressure-Volume diagram according to the variation of
intake valve closing time. It also shows intake pressure
in order to calculate effective compression ratio, which is
defined as cylinder volume at the crank angle of the
cylinder pressure equal to intake pressure divided by
combustion volume. In the Figure 5, all of the cylinder
pressures are steadily decreased with rotating
crankshaft and decreasing cylinder volume in exhaust
process. The cylinder pressure keeps at a little uniform
level in intake process, and then it is rapidly decreased
due to the expansion of sealed cylinder volume after the
intake valve closing time in the cases of having the valve
closing time of before BDC. After 16 Volume/Clearance
Volume, the cylinder pressures are increased due to the
compression by piston motion. In the case of 20deg.
IVC, the cylinder pressure has the equal value to the
intake pressure at A point. The effective compression
ratio of this case is therefore 13.75 different from
geometrical compression ratio, 16. The cylinder
pressure in this case is lower because of lower
compression pressure. And so, lower NOx generation is
also expected. Fig. 8 also indicates the pumping loss,
namely Pumping Mean Effective Pressure(PMEP),
defined by the enclosed area by Pressure-Volume
diagram during intake and exhaust process. The
pumping loss has a positive value when intake pressure
is higher than exhaust. It is found that the more intake
valve time is delayed, the more pumping loss is high in
the Figure 5.

20

Exhaust

IVC
(deg.ABDC)

Intake

30

15

Valve Lift(mm)

0
-10
-20

10

-30

Intake Valve
Closing(IVC)

0
120

180

240

300
360
420
480
Crank Angle(deg.ATDC)

540

600

660

Figure 4 The Valve Lift of Intake and Exhaust

5.0
IVC(deg.ABDC)

Cylinder Pressure

-30

4.5

Pressure(bar)

-20
0

4.0

Intake Pressure

MODELING OF FUEL INJECTION SIMULATION

30
46

3.5

3.0

2.5
0

4
6
8
10
12
Volume/Clearance Volume

The minimum value of NOx emissions is at 20deg.aBDC, and then the more intake valve closing
time is retarded, the more the NOx emissions are
increased. This is because the more the intake valve
closing time is retarded, the more the maximum cylinder
pressure is high. We can find that the Miller cycle is
useful to reduce specific fuel oil consumption and NOx
emissions at the same time, and the improvement of fuel
consumption is attributed to decreasing pumping loss.
The more intake valve closing time is retarded, the more
the air fuel ratio is decreased because of the reduced
opening area of intake valve.
All simulation conditions are allowable in practical
design because the temperature of turbine inlet is lower
than limit temperature, 650 .
As results of the performance simulations, the
optimum intake valve closing time at -20deg.aBDC is
found out in respect of NOx emissions. When comparing
of -20deg.aBDC with 30deg.aBDC, the intake valve
closing time of -20deg.aBDC is improved in NOx
emissions of 15.58% while the improvement of specific
fuel oil consumption is 1.7g/kWh.

14

16

Figure 5 Variation of P-V Diagram with Changing

Inlet Valve Closing Time(IVC)
Figure 6 shows the simulation results of the main
performance parameters over the variation of intake
valve closing time. The effective compression ratio is
increased with delaying intake valve closing time until
30deg.aBDC, but slightly reduced after 30deg.aBDC.
The maximum value of compression ratio is generated
at 30deg.aBDC delayed from BDC, and this attributes to
the inertia of intake air. The trend of maximum cylinder
pressure is similar to the effective compression ratio
which directly has a effect on the compression pressure,
and then the maximum pressure of the cylinder. The
specific fuel oil consumption has the minimum value at 10deg.aBDC, and then the more intake valve closing
time is retarded, the more the specific fuel oil
consumption is increased. This is because the more the
intake valve closing time is advanced, the more the
PMEP is low as shown in Figure 6, but the combustion
efficiency becomes worse and the combustion duration
is increased because of lower compression pressure of
the cylinder with advancing intake valve close timing.

Fuel injection rate is prerequisite data to calculate

combustion in a performance simulation because it
influences significantly the combustion. There are two
methods to have the fuel injection rate. One is
experimental method using the test rig of fuel injection
system, but needs a lot of cost and time. The other is
computational simulation, but needs verifying simulation
results. The latter is used in this study because the
pressure measured in fuel injection pump and the fuel
mass injected to the cylinder are easily obtainable, and
so, these data were used to verify the simulation results.

14
12

200
180
160

140

-0.2
-0.4
-0.6

194
193

10.5

191

Figure 7 The Fuel Injection System

10.0
9.5
9.0

33
27

525

24

500
475

VERIFICATION OF FUEL INJECTION

SIMULATION
Figure 8 shows the comparison of the simulation
results of the injection pump pressure with experimental
data over all load conditions. The simulation results have
the shape similar to experiment over all load conditions,
and therefore it is concluded that the models of fuel
properties(Bulk Modules, Wave Speed), spill port,
delivery valve and needle valve are constructed well.

450

-40
-20
0
20
40
60
Intake Valve Closing Time(deg.ABDC)
Figure 6 Variation of Engine Performance with Inlet
Valve Closing Time at 100% Load

The simulation of fuel injection system has been

carried out using FLOWMASTER S/W to have the fuel
injection rate of the target engine over all load
conditions.
The fuel injection system consists of a fuel pump,
a high-pressure pipe, and a fuel pump as shown in
Figure 7. The geometries of all fuel flow path are
modeled from spill port to cylinder in the system, and
operating conditions of fuel cam profile and fuel rack
index, which is measured at the engine test over all
loads are inputted in the model. Needle valve is modeled
to be open at the opening pressure of 450bar, and to
change its opening area according to move-up of the
needle. Nozzle is modeled by one hole equivalent to the

2000

Fuel Injection Pump Pressure (bar)

Turbine Inlet Temp.(oC)

30

Air/Fuel Ratio

NOx(g/kWh)

192

Max. Cylinder Pressure(bar)

16

BSFC(g/kWh)

Effective Compression Ratio

PMEP(bar)

18

sum of total nozzle hole area and a proper discharge

coefficient. The calculated nozzle hole diameter is
0.32mm and the number of nozzle is 10. The simulation
conditions of the fuel injection system are 900rpm in
crank angle and 25, 50, 75, 90 and 100% loads. In the
same operating conditions, the pressure of injection
pump and fuel injection mass are measured at the target
engine to verify the simulation results.

Experiment
Simulation

1600

1200

100% Load
800

75%
50%
400

25%

0
-20

-10

10

20

30

40

50

Crank Angle (deg.ATDC)

Figure 8 The Verification of Simulation Results on

the Fuel Injection Pump Pressure

In the Table 2., the calculation of injected fuel

mass is compared with experimental data. The
maximum error is 5.42% at 25% load, and hence it can
be found that the discharge coefficients of all the valves
are modeled relatively well.
Figure 9 shows the calculated fuel injection rate
which is inputted in the performance simulation.
Table 2. Comparison of Simulation Results of
Experiment in Injected Fuel Mass
Injected Fuel Mass(mg)
Error Rate
Load
Measureme
(%)
Simulation
nt
100
1444
1438
-0.42
90

1271.6

1268.6

-0.24

75

1066.7

1076.6

0.93

50

744.4

754.4

1.34

25

414.8

437.3

5.42

REFERENCES

400

Fuel Injection Rate (mm3/msec)

Nozzle 0.32*10, 900rpm

100% Load

300

75% Load
50% Load
25% Load

200

100

0
-20

-10

10
20
30
Crank Angle (deg.ATDC)

40

specifying optimum engine operating conditions for a

engine design.
(3) The fuel injection simulation verified by
experimental data in fuel injection pump pressure and
injected fuel mass is satisfied as input data of a
phenomenological combustion model.
(4) Miller cycle can reduce not only NOx emission
but also fuel consumption, and the maximum reduction
of NOx emission is 15.7% while the improvement of
specific fuel oil consumption is 1.7g/kWh in this study.
The reduction of NOx emission is attributed to lower
compression pressure in cylinder. And the reduction of
fuel consumption is attributed to lower pumping loss in
intake and exhaust process.

50

Figure 9 The Calculated Injection Rate

COLCLUSIONS
The performance simulation has been carried out in
order to specify the optimum intake valve closing time of
the large diesel engine with lower fuel consumption and
lower NOx emission. Conclusions are drawn and
summarized as follows:
(1) The results of the performance simulations are
verified with experimental data in the maximum error
rate of 6.43% over all load conditions.
(2)
The
performance
simulation
with
a
phenomenological combustion model verified by
experimental data of heat release rate, NOx emission
and main engine performance parameters is useful in

[1] Miller, R.H., 1947, Supercharging and Internal

Cooling Cycle for High Output, ASME Transactions,
Vol. 69, pp. 453-457
[2] Bolton, B.K. and Assanis, D.N., 1994, Optimum
Breathing Strategies for Turbocharged Diesel
Engines Based on the Miller Cycle Concept, ASME
ESDA Conference, Vol. 64-8.2, pp. 253-262.
[3] Stebler, H., et al., 1996, Reduction of NOx emissions
of D. I. Diesel Engines by Application of the MillerSystem : An Experimental and Numerical
Investigation, SAE paper 960844.
[4] Mavinahally, N.S., Kamo, R., and Kamo, L., 1997,
Performance Analysis of A Turbocharged MultiCylinder Insulated Miller Cycle Diesel Engine, ASME
97-ICE-37, pp. 47-54.
[5] Stebler, H., Boulouchos, Eberle, M.K., Geist, M., and
Vlaskos, I., 1998, Reduction of NOx-Emissions of a
Medium Speed D.I. Diesel Engine Using Miller
System, Exhaust Gas Recirculation, Variable Nozzle
Turbocharger and Common Rail Injection, CIMAC
Congress, pp. 857-866.