Thermodynamic Simulation of

HCCI Engine Systems

Olof Erlandsson

∂W = p⋅∂V


Doctoral thesis
Lund 2002
To my parents Sigurd and Maritha,
and my brother Henrik.

Department of Heat and Power Engineering

Lund Institute of Technology
P.O. Box 118
SE – 221 00 LUND
Sweden
ISBN 91-628-5427-5
ISSN 0282-1990
ISRN LUTMDN/TMHP--02/1007--SE

© 2002 by Olof Erlandsson. All rights reserved.

Printed in Sweden by Media-Tryck, Lund University
Lund 2002
Abstract

This thesis focuses on engine system simulation using thermodynamics and chemical kinetic
models to investigate the performance and efficiency of Homogeneous Charge Compression
Ignition (HCCI) engines for stationary applications.
It includes the development of software as well as models for engine, turbocharger,
intercooler, inlet and exhaust manifolds, wastegate valve, inlet air humidifier, inlet air heater
and more. The engine model can be classified as a one-zone, zero-dimensional and includes
sub-models for in-cylinder heat transfer, exhaust port heat transfer, heat release and the valve
flow process. The turbocharger model is developed in steps and is based on a polynomial fit
to experimental compressor and turbine performance data. Inlet and exhaust manifolds are
treated as well-mixed volumes. A simplified model for the humidifier is adopted in order to
simulate a humid air motor (HAM) concept. The whole engine system is zero dimensional
and the different system components are linked by means of mean values for mass flow,
temperatures, pressures and gas composition. The NASA-polynomials are used for calculating
thermal and transport properties. The extended Zeldovich mechanism is used as an indicator
of significant NO formation. The models are validated through comparisons with experiments
on mainly diesel and HCCI engines, both cycle and system results.
HCCI engine cycle simulations are made, showing the influence of various engine parameters
e.g. compression ratio, engine speed, air-fuel ratio, exhaust gas recycling, inlet pressure and
valve timing. Mainly natural gas and landfill gas are studied. The self-stabilizing feature of
HCCI ignition timing is investigated using chemical kinetics. HCCI engine system
simulations are made, both with (to find ignition timing) and without chemical kinetics,
investigating turbocharging and the influence of turbine size. The HCCI - HAM concept is
investigated and compared to simpler systems.

i
Preface

The mystery.
The Homogeneous Charge Compression Ignition (HCCI) engine is without a doubt a very
thrilling concept because it uses a completely different form of ignition and combustion
principle from the common engines. It does not use spark plug and propagating turbulent
flame like the spark ignited engine, nor does it use high pressure fuel injection and luminous
diffusion flame like the diesel engine. The combustion in a HCCI engine seems to have no
single origin, it appear in many locations simultaneously with a light-blue fluorescence.

The usefulness.
The reason why HCCI is interesting is the never-ending search for power-sources with higher
efficiency and lower emissions. HCCI engines use high compression ratio and lean
homogeneous mixtures. This is also the key to high efficiency and low NOx emissions in
internal combustion engines.

The invention.
There is very little new under the sun - that is in general very true. However, inventions and
other artifacts can be before its time, and there exist several examples of that. When it comes
to HCCI, the timing is perfect. Today we have all the basic electronics to monitor and control
the HCCI engine, which is fundamental for performance, efficiency and emissions.

The challenge.
HCCI has drawbacks, and in this work, I have tried to tackle mainly one of them: the limited
specific power output that follows from the lean mixtures. The goal is to achieve acceptable
performance without loosing the benefits of HCCI.

The tool.
This thesis covers both use and development of engine system simulation software. The
thesis is written with an increasing degree of simulation complexity, after a short introduction
comes the theory behind the simulation software, then follows various validations, chemical
kinetics, cycle simulations and finally the program is used to study HCCI engine systems.

ii
Acknowledgements
First, and foremost, I would like to express my gratitude to Professor Bengt Johansson for
allowing me to do this work, and then pushing me to finish it. He is a professor with many
ideas and big demands and besides being a guru for me and other Ph. D. students at the
division, he really put Lund on the map when it comes to HCCI engine research.
Another great professor is Gunnar Lundholm who really opened my eyes for internal
combustion engines and allowed me to do my master thesis about variable compression ratio
engines. Together with excellent lectures by Krister Olsson, I got the necessary basics to
continue within this field.
I would also like to thank my colleagues at the division for all help. Either in the lab or in the
office, they are always there for support. I would especially like to mention my office
colleagues whom I had the pleasure of sharing office with these years: Magnus Christensen,
Ola Stenlåås and Petter Strandh. Also thanks to Per Amnéus, a fellow from the Combustion
Physics division, for giving excellent support with the chemical kinetics.
The financial support from the Swedish Energy Agency has been essential, so thanks to all
Swedish taxpayers.

Olof Erlandsson

Lund, October 31, 2002

iii
Previous publications by the author
This thesis is in a way a collection of material presented by the author in papers and other
publications, although the material is updated and much more results are introduced. The idea
is that the reader can read this thesis without having access to those papers. Below is a list of
papers and publications the author somehow has been involved. Some of these, not all, are
reviewed in this thesis.
• “Demonstrating the Performance and Emission Characteristics of a Variable
Compression Ratio, Alvar-Cycle Engine”, Olof Erlandsson, Gunnar Lundholm, Fredrik
Söderberg, Bengt Johansson, Victor W. Wong, SAE 982682
• “Alvar Engine, An Engine with variable compression ratio, Experiments and tests”,
Olof Erlandsson, Master thesis, ISRN LUTMDN/TMVK-5304--SE
• “Hydrocarbon (HC) Reduction of Exhaust Gases from a Homogeneous Charge
Compression Ignition (HCCI) Engine Using Different Catalytic Mesh Coatings”, Olof
Erlandsson, Bengt Johansson, Fredrik Silversand, SAE 2000-01-1847
• “Experiments and Simulation of a Six-Cylinder Homogeneous Charge Compression
Ignition (HCCI) Engine”, Jan-Ola Olsson, Olof Erlandsson, Bengt Johansson, SAE
2000-01-2867
• “Development of an Engine System Simulation Software Package – ESIM”, Olof
Erlandsson, Licentiate thesis, ISRN/LUTMDN/TMVK – 7043
• “Early Swedish Hot-Bulb Engines – Efficiency and Performance Compared to
Contemporary Gasoline and Diesel Engines”, Olof Erlandsson, SAE 2002-01-0115
• “Analysis of a 6 Cylinder Turbocharged HCCI Engine Using A Detailed Kinetic
Mechanism”, Guiseppe Cantore, Luca Montorsi, Fabian Mauss, Per Amnéus, Olof
Erlandsson, Bengt Johansson, Thomas Morel, ASME Internal Combustion Engine
Division, Spring Technical Conference, April 14 – 17, 2002 Rockford, Illinois
• “Simulation of HCCI - addressing compression ratio and turbo charging”, Olof
Erlandsson, Patrik Einewall, Bengt Johansson, Per Amneus, Fabian Mauss, SAE 2002-
01-2862

iv
Contents
1 Introduction .................................................................................................. 1
1.1 Internal Combustion Engines......................................................................................1
1.1.1 The Diesel engine .............................................................................................................. 1
1.1.2 The Spark Ignited (SI) engines .......................................................................................... 2
1.1.3 Modern and novel engine technology................................................................................ 2
1.1.4 Fuels for IC engines........................................................................................................... 4
1.2 Homogeneous Charge Compression Ignition (HCCI)................................................4
1.2.1 The basic principle of the HCCI engine ............................................................................ 4
1.2.2 HCCI combustion characteristics ...................................................................................... 4
1.2.3 HCCI background – 2 stroke engines................................................................................ 5
1.2.4 HCCI background – 4 stroke engines................................................................................ 7
1.2.5 Emissions and exhaust gas after treatment ........................................................................ 7
1.3 “ATAC in Systems” - project description ..................................................................8
1.3.1 The challenge..................................................................................................................... 8
1.3.2 Goals, limits tools and material ......................................................................................... 9
1.4 Engine related definitions .........................................................................................11
1.4.1 Performance..................................................................................................................... 11
1.4.2 Efficiency......................................................................................................................... 13
2 The tool - ESIM .......................................................................................... 15
2.1 Engine simulations....................................................................................................15
2.2 The ESIM package....................................................................................................15
2.2.1 Description....................................................................................................................... 15
2.2.2 Limitations....................................................................................................................... 15
2.3 General assumptions .................................................................................................16
2.3.1 Ideal-gas equation of state ............................................................................................... 16
2.3.2 Thermal and physical gas properties ............................................................................... 17
2.3.3 Energy relations............................................................................................................... 21
2.3.4 Numerical solutions......................................................................................................... 22
2.4 System buildup..........................................................................................................24
2.5 Engine model ............................................................................................................25
2.5.1 Simple heat release model ............................................................................................... 29
2.5.2 In-cylinder heat transfer model........................................................................................ 29
2.5.3 Exhaust port heat transfer model ..................................................................................... 30
2.5.4 Governing equations........................................................................................................ 31
2.5.5 Numerical problems ........................................................................................................ 32
2.5.6 Diesel engine combustion model..................................................................................... 33
2.5.7 Engine friction model ...................................................................................................... 34
2.6 Turbocharger models ................................................................................................35
2.6.1 Performance data ............................................................................................................. 36
2.6.2 Performance data representation...................................................................................... 37
2.6.3 Compressor mass flow model.......................................................................................... 38
2.6.4 Compressor efficiency model .......................................................................................... 39
2.6.5 Turbine mass flow model ................................................................................................ 40
2.6.6 Turbine efficiency model................................................................................................. 41
2.6.7 Model accuracy................................................................................................................ 41
2.6.8 Improved model using interpolation and extrapolation techniques................................. 42
2.6.9 Governing equations........................................................................................................ 45
2.6.10 Comments.................................................................................................................... 46

v
 2.7 Manifold models .......................................................................................................47
2.7.1 Inlet manifold .................................................................................................................. 48
2.7.2 Exhaust manifold............................................................................................................. 48
2.8 Charge air cooler (Intercooler) model.......................................................................49
2.9 Humidifier.................................................................................................................51
2.10 Other components .....................................................................................................53
2.10.1 Heater .......................................................................................................................... 53
2.10.2 Wastegate .................................................................................................................... 53
2.10.3 Engine control ............................................................................................................. 53
3 Simple heat release ..................................................................................... 54
3.1 Validation of HCCI Cycle Simulation......................................................................54
3.2 Validation of System Simulation – Diesel operation................................................56
3.3 Validation of System Simulation – HCCI operation ................................................58
3.4 Turbocharger considerations.....................................................................................62
3.5 Lean-burn spark ignition...........................................................................................65
3.6 The effect of ignition timing on efficiency...............................................................67
4 Single zone chemical kinetic model........................................................... 68
4.1 Mathematical model..................................................................................................68
4.2 Validation of chemical kinetics ................................................................................72
4.2.1 Natural gas....................................................................................................................... 72
4.2.2 Isooctane.......................................................................................................................... 73
4.3 Cycle simulations with chemical kinetics.................................................................74
4.3.1 The effect of various engine operating parameters.......................................................... 75
4.3.2 Stability and late ignition phenomena ............................................................................. 77
4.3.3 The effect of valve timing................................................................................................ 82
4.3.4 The effect of exhaust gas recirculation - EGR................................................................. 85
4.4 The effect of humidification of inlet air – the HAM concept ...................................89
5 More realistic prediction of heat release rate .......................................... 92
5.1 “Slave” cycle and estimation of NO formation ........................................................92
5.1.1 Compression ratio optimization....................................................................................... 93
5.1.2 Inlet temperature control model....................................................................................... 95
5.2 System simulation.....................................................................................................96
5.2.1 The effect of turbine size ................................................................................................. 96
5.2.2 Quest for higher loads...................................................................................................... 98
6 System simulations with chemical kinetics ............................................ 100
6.1 Simple system with natural gas...............................................................................100
6.2 Reformed natural gas fuel.......................................................................................102
6.3 The HAM system in combination with reformed fuel............................................105
7 Summary ................................................................................................... 110
8 References ................................................................................................. 112
Symbols and Abbreviations............................................................................ 119

vi
1 Introduction

1.1 Internal Combustion Engines

The Internal Combustion (IC) Engine is perhaps the most wide-spread apparatus for
transforming liquid and gaseous fuel to useful mechanical work. The reason why it’s so well
accepted can be explained by its overall appearance regarding properties like performance,
economy, durability, controllability but also the lack of other competitive alternatives.
However, the IC engine is certainly not the best apparatus in every aspect but seems to be a
good compromise overall. It is possible to further develop the IC engine to be better in some
property but this is usually at the “cost” of another. For example, it is not difficult to develop
a highly powerful engine for racing, but at the same time keeping manufacturing cost and
durability as for a regular car engine is quite difficult. It is also possible to develop a very
cheep and lightweight engine for handheld equipment like chainsaws, but at the same time
limit exhaust emissions becomes difficult. The delivered power per mass unit has to be high
in order for the machine to be handheld, and the exhaust gases are expelled to the atmosphere,
close to the human working area.
New engine applicable technology emerges continuously that, if or when economically
sensible, can provide improvements in some or several aspects without cost in others. An
example of this is the introduction of electronics in engine control that, combined with
continuous development of engine design, has provided spark ignited engines with minimum
emissions and with both higher power and efficiency. However, the demands are constantly
increasing; less emissions, lower fuel consumption, more power and lower costs. A
continuous development is always required.

1.1.1 The Diesel engine

The diesel engine has through its high compression ratio a very good thermal efficiency but
due to the combustion principle (diffusion flame) it produces lots of nitrogen oxide (NOx)
emissions. In addition, high peak pressures and soot formation limit the diesel engine load.
The ignition principle is self-ignition and a high ignitability is wanted in order to get well-
controlled combustion timing. The ignitability is measured by the cetane number (CN) of the
fuel.
The high compression ratio causes high peak pressures that require a rigid engine design,
which brings up the mass of the design. Nevertheless, this engine has found its major
applications in especially trucks, trains and entrepreneur, agriculture and forestry machinery,
mainly due to its combination of fuel economy, durability and the fact that engine mass plays
a minor role in these applications. Turbo charging has opened a door for the diesel engine in
the lighter car applications. The load control of the diesel engine is very simple, only the
amount of injected fuel is controlled.
The major emissions of concern from the diesel engine are, besides CO2, NOx and soot
particulates. It is possible to use soot traps that collect a major part of the particulates but the
traps have to be cleaned somehow. One way is to burn off the trapped particulates by
temporarily increasing the temperature in the trap by using additional fuel injection during the

1
exhaust stroke or even upstream the trap. To better succeed with this catalytic materials can
be used in the trap. It is more difficult with the NOx emissions because of the idea is to
convert them into N2. This is very difficult to do in diesel exhaust gases because the high
amount of oxygen present. DeNOx catalysts exist but are in general very sensitive to sulfur,
and most diesel engine fuels contain some amounts of sulfur. Technology that uses a
reducing media, like urea or ammonia has also been tested but it then requires “refueling” in
order to work properly.

1.1.2 The Spark Ignited (SI) engines

The SI engine uses a premix of fuel and air and it utilizes a spark plug (forced ignition) that
sets of premixed turbulent flame combustion. This combustion principle makes it sensitive to
self-ignition, “knocking”, which can result in structural damage, mainly to the piston. This
phenomenon limits the compression ratio of the SI engine, which in turn limits the thermal
efficiency. The fuels for the SI engines have a certain degree of knocking resistance,
measured by the octane number (ON) of the fuel.
The SI engine has through the lower compression ratio a possibility to use higher rpm’s in
order to get high specific power because the pressures are limited, thus a more lightweight
design can be used that allows higher rpm’s. Load control is achieved by throttling which
means that the inlet pressure, and hence the air mass flow through the engine, is controlled by
a restriction, usually a butterfly valve. This forces the engine to work like a compressor at part
load, which also reduces efficiency. In car applications, the major operating point of the
engine is low to medium load, which means that the overall efficiency becomes quite low. For
the diesel engine, it is almost the opposite situation. This has contributed to an increase of the
number of diesel-propelled cars.
The SI engine uses (in general) stoichiometric mixtures, which opens a door to exhaust gas
after treatment using a 3-way catalyst. This means that carbon monoxide, CO, and unburned
hydrocarbons, HC, can be oxidized, and nitrogen oxides, NOx, can be reduced,
simultaneously. The spark-ignited engine with 3-way catalyst is therefore the environmentally
friendly alternative. It becomes even better if bio-fuels are used, which do not contribute to a
net CO2 increase in the atmosphere in the long perspective, like petroleum products do.

1.1.3 Modern and novel engine technology

The dream of every engine designer is to find a combination of the diesel engine and the
spark ignited engine that only inherit the good properties of the diesel and SI engine i.e.
efficiency like a diesel engine and exhaust emissions that are as clean as from the SI engine,
or at least possible to after-treat to the same level. This is quite difficult and there have been
very few attempts that have reached any commercial success. There are two such concepts
that are worth to be mentioned: lean burn SI engines with either homogeneous mixtures or
stratified charge.
Lean burn SI engines simply uses a mixture with less fuel per mass air, and hence burns with
lower temperature. By lowering the combustion temperature, NOx formation (heavily
temperature dependant) is limited, but unfortunately this limits the power output of the
engine, so this technology is often used as a part load operation mode of a, in other aspects,
conventional SI engine. One difficulty with lean mixtures are the ignition process itself, it
requires improved ignition technology to get as good early flame propagation with lean
mixtures as with stoichiometric ones. Turbo charging together with lean mixtures becomes
difficult due to lower exhaust gas energy levels, however in large stationary engines with

2
quite low heat transfer losses, this is a successful technology. The problem to be solved is the
ignition process. Wärtsilä NSD uses a prechamber were a stoichiometric mixture is ignited by
a spark plug, the burning gas then flows out of the prechamber, into the main combustion
chamber through several nozzles, resulting in a very good and secure ignition process of an
overall diluted mixture. However, the compression ratio is still limited by knocking
combustion.

Figure 1 Wärtsilä pre combustion chamber (PCC) spark ignited natural gas engine [1].
Stratified charge SI engines uses direct injection of the fuel towards the spark plug and by
using the proper injection timing and duration it is possible (but certainly not trivial) to get a
fairly stoichiometric mixture around the spark plug surrounded by pure air. The idea is that
load control is achieved by varying the amount of air-fuel mixture around the spark plug,
instead of control the total amount of mass inside the cylinder, and this limits the throttle
losses. At maximum load, the engine works more or less like a normal SI engine.
Unfortunately, it becomes difficult to achieve a perfect homogeneous mixture and the
emissions, especially particulates, increase with this technology.
It is the compression ratio of the diesel engine that are the main design factor that give the
diesel engine its high efficiency, and it is the diluted homogenous mixtures that give low NOx
emissions form the lean burn SI engine. The compression ratio of the latter cannot be
increased due to knocking phenomena that destroy the engine. Instead of looking at knocking
as a negative phenomenon, it is actually possible to use it as an ignition process and this is the
basic idea of Homogeneous Charge Compression Ignition - HCCI. This means an engine
without spark plug or direct injection of the fuel, i.e. in its basic configuration, a very simple
design. Unfortunately, the ignition process becomes uncontrolled, or to be more specific;
there is no parameter that directly controls ignition timing like a spark plug or the injection
timing in the diesel engine. Everything is determined when the intake valve closes: Will the
temperature and pressure become high enough in order to get auto-ignition and combustion?
Will ignition occur at a desired crank angle degree? The advantage is that we will get high
efficiency, ultra low NOx emissions and no particulates, because it uses high compression
ratios and well-diluted mixtures (leaner mixtures than used in lean burn SI engines).

3
1.1.4 Fuels for IC engines
The IC engine transforms chemically bonded energy of a combustible fuel into useful
mechanical work. As long as the fuel is relatively safe, efficient and convenient to handle,
transport and store, the IC engine an excellent choice for automotive applications. Since the
IC engine is a compromise between properties the whole engine development becomes very
sensitive to fuel accessibility and technology. This has allowed two major IC engine types to
evolve: the diesel engines and the gasoline engines. If all combustion engines are considered,
the third engine type is the gas turbines; these will not be discussed any further in this thesis
but has its most common application in aircrafts due to its superior power density ratio
(power-to-weight). Oil drilling, cracking and distilling gives a spectrum of different
petroleum products with different combustion characteristics. This fuel refining is obviously
adapted to the major markets, the automotive and aeronautical applications, so proper blends
of distilled products become suitable for a specific application e.g. gasoline for SI engines.
Natural gas is the most economical alternative for stationary applications as long as the access
is sufficient and as long as it can be transported in pipelines.
Fuels that are “renewable” start to appear on the market today, which means fuels that do not
contribute to a net increase of CO2 in the atmosphere, unlike coal and oil products (fossil
fuels). During rationing in wartimes these were quite widely used, a typical example was
gasifiers (gas generators) mounted on European vehicles, during the 2’nd world war, in order
to run the cars on wood-chips. Today it is perhaps ethanol mixed into the gasoline that is the
most frequent example. Other examples are landfill gas or similar biogases.

1.2 Homogeneous Charge Compression Ignition (HCCI)

1.2.1 The basic principle of the HCCI engine

Homogeneous charge compression ignition (HCCI) uses a lean premixed air-fuel mixture that
is compressed with a high compression ratio. During the end of the compression stroke,
ignition occurs through self-ignition in the whole combustion chamber at once.
Since the mixture is lean, the maximum temperature, both locally and overall, becomes low
compared to other engines, which effectively reduces NOx formation. However, at richer
mixtures the combustion becomes too fast and knocking, or ringing, occurs. Therefore, if a
higher load is desired, supercharging, or turbocharging is necessary. The load limit (without
supercharging) is said to be either the engine structure capabilities (knocking limit) or NOx
emissions.
The emissions from the HCCI engine are mainly hydrocarbons (HC) caused by incomplete
combustion and quenching due to wall effects. HC emissions are in general considered more
treatable than NOx emissions. The problem with the HCCI engines is related to the lean
mixtures, the fast combustion, and the high compression ratio (high engine efficiency) that
causes the exhaust temperature to become quite low. This can make it difficult to get both
turbocharging and oxidizing catalysts to work.

1.2.2 HCCI combustion characteristics

The HCCI engine relies on self-ignition of the fuel-air mixture. This means that there is no
mechanical or electrical device controlling the exact time of ignition. The condition for
ignition to take place is determined the moment the intake valve closes and compression
starts. The only factor that could affect the ignition timing somewhat at this point is if the

4
engine speed would be changing tremendously. In this way the volume – time curve, and
hence pressure and temperature trace, prior combustion would be changing and allow
shorter/longer time for the chemistry resulting in earlier or later ignition timing.
As mentioned earlier, the HCCI engine uses a high compression ratio; this is generally true
but depends on the fuel that we would like to run the engine on. A fuel that easily ignites, like
a modern diesel fuel, requires low compression ratio in order to get ignition close to TDC. I
fact diesel is a quite poor fuel to run the HCCI engine on because the required compression
ratio becomes as low as 7:1. This would mean that we actually would get lower efficiency
than in a spark-ignited engine and that is certainly not our goal. The selected fuel should give
us an optimum compression ratio of about 15:1 – 17:1, this is a good compromise, if we use
higher compression ratio, not only do we increase the mechanical load on the engine, we also
increase the area-to-volume ratio and the heat transfer, crevice and blow by losses become
more significant. With this high compression ratio and lean mixtures, high octane number
fuels like natural gas and gasoline are more adequate. Anyway, some sort of parameter for
combustion timing control is required in order to optimize both efficiency and emissions.
The combustion duration at high load operation has shown to be extremely short and it is easy
to get knocking combustion, i.e. the combustion produces an oscillating pressure trace after
the main combustion, and in general a “pinging” noise outside the engine. This phenomenon
is the same as found in spark-ignited engines. The question is what degree of knocking is
acceptable from an engine durability point of view. In general, the combustion duration
depend on several parameters like fuel, air-fuel ratio, burned gas fraction (EGR), engine
speed, compression ratio and several other parameters, but also the degree of homogeneity in
terms of e.g. fuel and thermal distribution. Turbulence smoothes out most of these
distributions (and it is perhaps in this way turbulence plays its most important role). The
turbulence itself is distributed over the combustion chamber.
However, there are better methods to control combustion. It can be done by varying the
compression ratio or by using blends of two or several different fuels or additives that affects
the ignition timing and combustion duration. A variable compression ratio also makes it
possible to run the HCCI-engine on practically any fuel with an acceptable ignition
temperature of the air-fuel mixture. CFR-, Atkinson -, Alvar-cycle [2, 3] and variable valve
timing is examples of engines and methods used to achieve a variable compression ratio in an
engine.

1.2.3 HCCI background – 2 stroke engines

The first published results of using compression ignition of a homogeneous mixture was made
by Onishi et. al. [4] at Nippon Clean Engine Research Institute Co. with the Active Thermo
Atmosphere Combustion, ATAC concept. The basic idea was to run a 2-stroke engine with
lean mixtures during part load operation. At throttled operation, lots of burned gas was
trapped inside the cylinder during the scavenging phase. This increased the mean charge
temperature to a degree where auto ignition (self ignition) occurred. Since lean mixtures were
used, it was possible to use an oxidizing catalyst and hence achieve clean exhaust emissions.
Onishi also provided an ideal model (Figure 2) of the ATAC principle, which became very
famous and discussed in many later HCCI and ATAC papers. It explains in a simple way the
basic difference between an ATAC and a spark-ignited engine. In a spark-ignited engine, the
fuel is burnt in a reaction zone – the flame front, which propagates across the combustion
chamber and hence consumes the fuel. This means that there is a clear separation between a
burned and an un-burned zone. In the ATAC case all the fuel is consumed simultaneously, but
at a lower rate. The fuel is split up into intermediate species during the oxidation process,

5
which terminates when there is no combustible species left. Onishi also showed that the
cycle-to-cycle variations became very small compared to normal spark ignition operation,
which improved the overall efficiency of the engine. Detailed studies of the ATAC
combustion followed shortly after Onishi’s first publications [5-16].
The ATAC technology was adopted in commercial products, first small generators (“NiCE”,
electrical power of 1 kW) but later Honda Motorcycle Company introduced a similar concept
where they adjusted the burned gas fraction with an exhaust port valve. They implemented the
technology in small motorcycles and they gave the concept the name Active Radical
Combustion [9, 16]. A brief but comprehensive summary of ATAC and HCCI publications is
made by A. Hultqvist [17].

Specific Spark ignition Specific ATAC

heat, q heat, q
Q= ∫ q ⋅ dmf Q= ∫ mf ⋅ dq

Flame Heat
propagation Release

Fuel mass, mf Fuel mass, mf

Burned mass Unburned mass
Figure 2 Onischi´s ideal model of spark ignition and ATAC. The area inside the rectangle
symbols total amount of heat released, Q.
Another, commercial 2-stroke product, although in every aspect much smaller scale, are
engines for model airplanes, often denoted as “diesel” engines by model plane enthusiasts.
The more common engine type in these lightweight applications is the glow-plug engine. The
major difference between these two engines is that the “diesel” model engine has an
adjustable compression ratio, which is set for best performance, and does not have an ignition
improvement part i.e. a glow plug or spark plug. These facts imply that these “diesel” engines
are some kind of ATAC engines. This type of engine is reported to be more powerful and
less noisy compared to the glow plug engine. The fuel blend used for these engines is a
mixture of 15% castor oil, 50% paraffin, 33% ether and 2% isopropyl nitrate. The fuel blend
varies with model and degree of run-in. Compared to the ATAC 2-stroke engines mentioned
earlier these engines probably run with much richer air-fuel mixtures (λ well below 1). Heavy
heat transfer and lots of residual gas in these small engines probably gives an acceptable heat
release rate and knocking is not a problem. One interesting feature is that it can be cold-
started without any preheating of inlet air, fuel or cylinder head.

6
 3
4

2
5

Figure 3 The PAW 09 TBR – R/C engine. Weight: 140g Power: 0.24hp at 19500rpm. 1 –
propeller shaft, 2 – carburetor, 3 – compression ratio adjusting screw, 4 – Cylinder, 5 –
exhaust pipe, 6 – crankcase [18].

1.2.4 HCCI background – 4 stroke engines

There have been numerous experiments with port injected four-stroke HCCI engines where
either diesel engines or so-called CFR engines are rebuilt to work with HCCI combustion [19-
93]. Diesel engines are often very robust and they work more or less with the same overall
air-fuel mixtures as the HCCI engines do. The fuel injector can also be removed and an
ordinary port injector from a SI-engine can be placed in the inlet port. In order to obtain auto
ignition temperature and be able to adjust the point of ignition, an inlet air heater is often
used. If the inlet air is heated the temperature over the whole engine cycle is elevated. This
means that auto ignition temperature is reached earlier in the cycle and thereby combustion is
phased earlier. The HCCI engine has proved to have very small cycle-to-cycle variations,
although if the combustion is phased very late, unstable or modal behavior may occur. The
perhaps most comprehensive HCCI experimental work is made by M. Christensen who has,
on one and the same engine, performed a massive parametric study regarding e.g. load,
compression ratio, fuel and inlet conditions [94].
Besides a few of the 2-stroke concepts described earlier, the most successful commercial
concept is made by Nissan Motor Co. who, through their Modulated Kinetics (MK) system
[37, 49], has created something that is quite close to HCCI and with the same advantages.
Basically, it’s a 4 stroke direct injected diesel engine with quite high compression ratio. The
injection timing is retarded and by using EGR, the ignition delay becomes longer. This means
that there is longer time for vaporization and mixing. In combination with high swirl gas
motion, the premixed part of the combustion increases. EGR and late ignition timing reduces
the NOx formation and soot is suppressed by the high fraction of premixed combustion.

1.2.5 Emissions and exhaust gas after treatment

As stated earlier, the NOx emissions are extremely low due to the very lean and homogeneous
mixture. The major emissions are unburned hydrocarbons from the fuel and lubricating oil
(from cylinder walls and piston ring crevices). The ignition process is temperature sensitive
and thus cold spots near the combustion chamber walls can escape combustion, especially at

7
very low load (high λ-values). In addition, fuel can be trapped in top land crevices1, and
evaporate late in the expansion-, or even during the exhaust stroke. The temperature is then
too low for this mixture to ignite or burn completely, which results in UHC emissions. Some
CO emissions are also produced from partial burn in regions where combustion rate is
reduced due to the expansion.
To completely get rid of the UHC and CO emissions some sort of oxidizing catalyst is
required. The temperatures in the exhaust manifold from the HCCI engine is quite low due to
several factors:
• High compression ratio – high efficiency
• Fast combustion
• Lean mixtures
This may require more expensive catalyst technology e.g. coatings that have a low activation
temperature, or complex regenerative catalyst designs. Very few solutions have been
presented to this problem.

1.3 “ATAC in Systems” - project description

1.3.1 The challenge

There are two major problematic concerns to overcome before the HCCI engine is a viable
alternative: engine control and specific power. HC emissions are seen upon as a smaller
problem because the HCCI engine is constantly running on lean mixtures and that there is
oxidizing catalysts that should work if properly sized. Several experimental studies have
proved that NOx emissions are ultra-low, if the air-fuel mixture is lean enough.
Engine control is much more than controlling power output on a HCCI engine. The problem
is to find a control system for the ignition and combustion itself. This system or method has to
be fast in order to cope with transients if the engine is to be used in a vehicle.
Regarding specific power, the traditional way has been supercharging or turbocharging i.e.
increasing the inlet air pressure (or density) of the engine and thereby increasing the mass
flow through the engine. Since the HCCI-engine has quite high efficiency, the energy left in
the exhaust gas is limited and thus it becomes difficult to achieve turbo-charging. However,
one can argue that since the HCCI engine has about the same compression ratio and operates
with lean mixtures like the diesel engine, it should be possible to successfully turbocharge the
HCCI engine as well. The difference is that it is possible to temporarily use richer mixtures in
the diesel engine, before there is any boost pressure, resulting in higher exhaust temperature
and hence, higher boost pressure. This will not work in a HCCI engine due to too fast
combustion resulting in knocking phenomena. In addition, fore some fuels the inlet air
temperature has to be slightly higher in the HCCI engine in order to ensure auto ignition,
which reduces the mass flow through the engine. Therefore, the turbine will probably have to
be smaller on the HCCI engine. Furthermore, the exhaust temperature of the diesel engine
becomes higher at higher loads, not only because of the higher amount of fuel injected, but
also due to the slower combustion that is limited by the mixing between the fuel-spray and
air. At higher loads, the injection duration becomes longer which also delays the combustion,

1
Crevices formed by the small volume between piston crown and cylinder wall, above the top piston ring.

8
causing a lower efficiency and a higher exhaust temperature. From a turbocharging point of
view, the higher exhaust temperature provided, the more energy is available to be transformed
in the turbine, which then gives a higher boost pressure. Supercharging by using a crank
driven compressor is not an option if a high efficiency system is wanted. One exception may
be as a temporary action in automotive applications, when performance is more important
than fuel economy.
The project that includes this study has the goal to find working HCCI engine system
solutions and this study provides a tool, or workbench, for that purpose. The primary task for
the project is to find a HCCI engine solution to stationary engine applications such as small
combined heat and power plants. The purpose of choosing stationary applications as a first
step is that the control problems are believed to be smaller since the control system does not
have to cope with heavy transients, as is the case in a mobile application.

1.3.2 Goals, limits tools and material

The more descriptive name of this project is: “Homogeneous Charge Compression Ignition of
Natural Gas and Biofuel for High Efficiency and Ultralow NOx”. The project is funded by the
Swedish Energy Agency (former Swedish National Energy Administration, former NUTEK,
project number P10025). The goal of this work is to find a cogeneration power plant that
utilizes the HCCI principle in order to limit NOx emissions and increasing the efficiency. The
first stage in the project started November 1997 with a quite wide goal formulation; the goal
was to study:
• Turbocharging/Supercharging
• The influence of pressure gradients on the engine structure
• Exhaust gas composition – FTIR measurements
• Experiments with low expansion, catalyst and turbocharger
• Exhaust gas after treatment (low temperature HC oxidation in catalysts)
Most work was put into the first one, it was decided that the major tool should be simulation
and that in-house software should be produced in order to gain knowledge and competence.
This decision obviously caused long periods of modeling and programming but allowed a lot
of freedom since all the code was accessible. The second goal would probably have required
complex FEM methods that can handle not only load, but also vibration and thermal loads.
There is no effort made in this field in this project, and to the author’s knowledge, very little
is made elsewhere. No structural damage is reported in experimental HCCI engines (often
rebuilt diesel engines) that can be explained by normal HCCI combustion itself. One or two
engines with optical access have experienced cracked or even exploding quartz rings, but this
is quite common in optical engines. One engine was exposed to too high peak pressures
through stoichiometric mixtures, causing a slight compression of the connecting rods. Other
than the connecting rods, there was no structural damage (mechanical adaptive conrod’s!).
A small experiment with FTIR was made in order to evaluate the method, it was concluded
that the major exhaust HC emission was the fuel itself (in this case isooctane). There were
also a wide range other species but it was difficult to get a good quantitative figure if more
species (spectra) with individually low concentrations were added to the analysis, the method
was abandoned but could perhaps be more suitable if simpler fuels like hydrogen, methane,
methanol, or carbon monoxide were studied.

9
Some experimental efforts were made with several catalysts. Mesh-type catalysts were
developed in Lund by Catator AB (former Katator AB) and could be doped with various
precious metals. The results were documented in a paper [95] and showed that carbon
monoxide was quite simple to handle regardless of coating; however higher amount of
platinum nitrate (PtNO3) gave much higher oxidation. It was a little bit more difficult with HC
emissions (Figure 4). By adding some ceria or stannic oxide (CeO2 or SnO2) some gain could
be seen, although there were some variations between the different experiments, e.g. the
amount of HC emissions upstream catalyst; the conclusions were not that precise. Anyway, it
seemed like if the catalyst were properly sized it would be possible to reduce the CO and HC
emissions to acceptable levels.
100 100

99.5 90

THC reduction [%]

CO reduction [%]

99 80

98.5 70
2 2
PtNO 10g/m PtNO3 10g/m
3 2
98 PtNO 20g/m2 60 PtNO3 20g/m
3
CeO2 + PtNO3 CeO + PtNO
97.5 50 2 3
SnO2 + PtNO3 SnO + PtNO
2 3
97 40
250 300 350 400 250 300 350 400
Temperature pre catalyst [°C] Temperature pre catalyst [°C]

Figure 4 Reduction in CO (left) and THC (right) emissions for various catalyst coatings
The second stage in the project started April 1999 and this time the goal was more explicit
into three part goals: 1 – further develop the simulation software, 2 – show a full scale HCCI
engine system and 3 – make new experiments with catalysts. The funding was however quite
limited which limited all experimental efforts within the project alone. A new project started
with separate funding and with a new Ph. D. student. A 6-cylinder diesel engine was
converted to HCCI by adding a dual-fuel port injection system. Varying the fuel mixture of n-
heptane and isooctane i.e. the RON number of the fuel controlled the ignition timing.
Cooperation resulted in a joint paper [68] and this fulfilled the second part goal. There were
however very little boost pressure and thus the engine load was quite small. In addition,
valuable validation was made of the software. The engine was fitted with its diesel size
turbocharger. Later, smaller turbines were tested with promising results. The results with the
catalysts described above were also published during this stage. Discussions were made with
catalyst developers about a regenerative catalyst design but it felt like it was a little bit too
early to start the development of full-scale catalysts at this time.
During the third stage, which started November 2000 the simulation software was fitted with
a chemical kinetics part that provided a prediction of the ignition timing. This is quite
demanding calculations that were not that suitable to use in direct combination with full
system simulations. Instead, cycle simulations were used to find proper control strategies (fuel
composition, inlet temperature or compression ratio) depending on engine speed, inlet and
exhaust conditions and air-fuel ratio. Then system simulation was then performed with a fixed
heat release function and the previously determined control model. Another part goal was still
to do experiments with catalysts but this felt again too early i.e. there were no bright ideas on
how to overcome the problem more than making a sufficiently large catalyst. The funds were
also quite limited and therefore catalysts experiments were considered a less prioritized area
at that time. Other part goals were simulations with exhaust gas recirculation (EGR) and
variable compression ratio (VCR). EGR has an effect of dilution and can be used to decrease
the air-fuel ratio for the same amount of fuel (heat) input. This reduces the heat release rate

10
during the combustion and hence suppresses knocking combustion. However, the ignition
temperature increases with EGR and higher inlet temperature reduce the mass flow through
the engine. In spark ignition and diesel engines EGR usually decreases the NOx formation. In
the HCCI engine the required higher inlet temperature balances this effect, however the NOx
levels are overall ultra low in comparison with SI and diesel engines.

1.4 Engine related definitions

In order to make discoveries and results understandable for others it is important to use the
same language and terminology, especially in the field of combustion engine that is very
broad. In order to reduce the risk of being misinterpreted, some important properties and
relations are shown. A Sankey diagram illustrating the definitions and the energy flow of an
arbitrary engine is seen in Figure 5.

− FuelMEP − − Fuel 100% −

− QhrMEP − − ηc −
QemisMEP Emissions

QhtMEP Heat
transfer
IMEP
gross η
i,g

QlossMEP Heat loss

(exhaust) (exhaust)

IMEP
− net − −η −
PMEP i,n Pump work
FMEP ηb Friction

BMEP Work

Figure 5 Sankey graph illustrating the energy flow of an engine, in this case a naturally
aspirated, spark ignited, car-size engine running at 3500 rpm at WOT conditions, selected to
give significant pumping and friction. Mean effective pressure representation to the left and
efficiency representation to the right.

1.4.1 Performance
Starting at the end, the performance of an engine is in general described by the maximum
brake2 output power of an engine. This power, Pb [W] is the product of speed, n [rpm] and
torque, Tb [Nm] on the outgoing shaft and can be expressed as:
2π ⋅ n
Pb = ⋅ Tb Eq.1
60

2
The word “brake” refers to the usual way to determine the performance of an engine; connect it to a
dynamometer and brake the engine to a desired torque (load) and speed.

11
In general, it is sometimes necessary to compare, or to normalize, with the respect to the size
of the engine in order to make relevant comparisons between different engines and this is
usually done by looking at what Brake Mean Effective Pressure, BMEP [Pa], is delivered by
the engine. This is a ratio between the delivered torque and the displaced volume, Vd [m3] of
the engine. In addition, the number of revolutions per power strokes is considered in order to
make it possible to compare 2-stroke (nR = {revolutions per cycle} = 1) and 4-stroke engines
(nR = 2). Another way to look at it is that it is the delivered work per cycle and displaced
volume [96].
2π ⋅ nR ⋅ Tb 60 ⋅ nR ⋅ Pb
BMEP = = Eq.2
Vd n ⋅ Vd
The engine is an apparatus that transforms energy and the transformation can be separated
into different steps, each represented by a mean effective pressure. The difference between the
steps of this transformation may often be considered as losses (or gains) and can be expressed
as differences in mean effective pressures. Into the engine flows the energy carrier – the fuel.
The fuel flow rate introduces a certain amount of fuel per cycle, mf [kg]. Fuel corresponds to a
certain amount of energy through the heating value of the fuel, QLHV [J/kg]. This means that
we can express this fuel flow as a delivered energy per cycle, Fuel Mean Effective Pressure,
FuelMEP defined as:
m f ⋅ QLHV
FuelMEP = Eq.3
Vd
Note that the lower heating value (LHV) of the fuel is used, normally any condensation of the
water content is difficult to implement in combustion engines. The lower heating value
corresponds to the maximum possible heat released per mass fuel during combustion if
combustion is allowed to go to completion. The energy carried by fuel is released through
combustion inside the cylinder, at least some of it, depending on how effective the
combustion process is. This means that the amount of energy that is released, Qhr, may not be
as much as the amount that was carried by the fuel. The amount of energy released may be
expressed as a Heat Release (Qhr) Mean Effective Pressure, QhrMEP defined as:

Q
v∫ dQhr
QhrMEP = hr = cycle
Eq.4
Vd Vd
The difference between FuelMEP and QhrMEP corresponds to energy that is chemically
bonded to unburned components like hydrocarbons and carbon monoxide or any other species
(e.g. NO, NO2, H2, H2O2, ...) that is an undesired product of combustion and that “steals”
energy by escaping out of the engine with the exhaust gases or into the crank case. We can
interpret this as an Emission (Qemis) Mean Effective Pressure, QemisMEP:
QemisMEP = FuelMEP − QhrMEP Eq.5
The released energy, or heat, then increases the temperature and pressure during combustion
and, thanks to the cyclic operating principle of the engine, mechanical work can be extracted
by the moving piston. The energy that is transformed into work per cycle is called indicated
work, Wc , and an Indicated Mean Effective Pressure, IMEP, can be calculated. In 4-stroke
engines, it is sometimes convenient to separate this work into IMEPnet and IMEPgross
depending on the gas exchange, is considered or not.

12
The definitions are:

W
v∫ p ⋅ dV
IMEPnet = c = cycle

Vd Vd
V (540° )

∫
V (180° )
p ⋅ dV
IMEPgross = Eq.6
Vd
360° ⇔ TDC combustion
The released energy that that is not transformed into work, is partly transferred to the cylinder
walls by the temperature difference between the gas and the walls and some escape with the
exhaust gas; exhaust temperatures are significantly higher than the intake temperatures of the
engine. Heat Transfer-, QhtMEP, and the Heat Loss Mean Effective Pressures, QlossMEP,
can be calculated as:

Q
v∫ dQht
QhtMEP = ht =
cycle

Vd Vd Eq.7

QlossMEP = QhrMEP − IMEPgross − QhtMEP

An internal combustion engine is partly an open system, it breathes fuel, air and burned
exhaust gases. The breathing part of the process is difficult to do totally without work input
since the valves and other inlet and exhaust devices are restrictions in some sense. To force a
mass flow through these devices requires pumping work. In engines, this pumping work,
Pump Mean Effective Pressure, PMEP, is defined as the difference between the gross and net
indicated mean effective pressure:
PMEP = IMEPgross − IMEPnet Eq.8
The internal combustion engine consists of several moving parts. They are very well
lubricated but not totally frictionless, so some mechanical work has to be used to keep them
moving. This work can be expressed as a Friction Mean Effective Pressure, FMEP, and is
defined as the difference between indicated and brake mean effective pressures:
FMEP = IMEPnet − BMEP Eq.9
With this final definition, the whole chain from fuel input to shaft output is covered and the
engine can be thermodynamically analyzed by looking at these factors.

1.4.2 Efficiency
Using the previously defined mean effective pressures it is possible to normalize one more
step by using the FuelMEP as reference, and by doing so, several efficiencies are easily
defined. To start with, the combustion efficiency, ηc , that simply state how much of the
energy in the fuel is converted into heat during the combustion:

QhrMEP
v∫ dQhr
Vd
v∫ dQhr
ηc = = cycle
⋅ = cycle
Eq.10
FuelMEP Vd m f ⋅ QLHV m f ⋅ QLHV

13
Two different indicated efficiencies could be defined based on IMEPnet and IMEPgross
respectively. The next step is incorporates the basic cycle and in the same way as there are
two different IMEP, two indicated efficiencies, net – whole cycle, and gross – compression
and expansion, are defined as:

IMEPnet
v∫ p ⋅ dV
ηi ,n = = cycle

FuelMEP m f ⋅ QLHV
V (540° ) Eq.11

IMEPgross
∫
V (180° )
p ⋅ dV
ηi , g = =
FuelMEP m f ⋅ QLHV
And the more interesting brake efficiency that includes the frictional loss is simply calculated
through either the mean effective pressures, or in the more common way from experimental
data; from power output and fuel flow:
BMEP 2π ⋅ nR ⋅ Tb 60 ⋅ nR ⋅ Pb Pb
ηb = = = = Eq.12
FuelMEP m f ⋅ QLHV n ⋅ m f ⋅ QLHV m f ⋅ QLHV
Sometimes it is interesting to only look at the thermal process and not include combustion.
Then the reference is changed from being the heat equivalent of the fuel to the actual released
heat instead, and thermal conversion efficiency can be calculated as:
V (540° )

IMEPGross
∫
V (180° )
p ⋅ dV
ηi , g
ηTh. = = = Eq.13
QhrMEP v∫
cycle
dQhr ηc

The gas exchange process also has some degree of efficiency, the gas exchange efficiency,
and it is simply the ratio between the indicated net and gross mean effective pressures, or
indicated efficiencies:
IMEPnet PMEP η
ηG . e. = =1− = i ,n Eq.14
IMEPgross IMEPgross ηi , g
The mechanics of transforming the piston motion to shaft rotation has mechanical efficiency,
which is the ratio of work leaving the engine shaft and work delivered to the pistons:
BMEP FMEP η b
ηM .e. = = 1− = Eq.15
IMEPnet IMEPnet ηi ,n
The brake efficiency can now also be calculated as a product of several part efficiencies, each
representing the efficiency of an essential step of energy transformation in the IC engine:
ηb = ηc ⋅ ηTh. ⋅ ηG .e. ⋅ η M .e. Eq.16

14
2 The tool - ESIM

2.1 Engine simulations

Engine calculations along with experimental activities have been performed ever since the
beginning of the internal combustion engine era. The first man to actually use a
thermodynamic approach to study heat engines was Rudolf Diesel [97]. There are often
several reasons behind a theoretical approach and some of them can be connected to terms
like dimensioning of engine mechanics, system behavior, performance studies or
understanding of in-cylinder processes.
There exist several different levels of simulation, some includes both turbulence and
chemistry calculations of the in-cylinder gas, whereas others only treat the engine as a black
box with some inputs and outputs.
The program package that is developed here is primarily for thermodynamical system
analyses and performance studies. “System” refers to the engine and surrounding components
like manifolds, turbocharger, intercooler etc.

2.2 The ESIM package

2.2.1 Description
The Engine SIMulation (ESIM) software package is written in MATLAB© and consists of
several functions that correspond to a certain engine system components. These different
components are linked together by state, mass and energy conservation equations. The engine
part is a zero-dimensional, one-zone model that considers heat transfer and mass-flow
calculations over the valves. This section focuses on describing the different models and how
to connect them, building an entire engine system, e.g. a turbocharged HCCI engine.

2.2.2 Limitations
The engine part handle only thermodynamics, i.e. there are no mechanical calculations
regarding e.g. friction losses. That means that result as mean effective pressure and efficiency
is shown as indicated, i.e. based on the in-cylinder gas calculations only. Various friction
models often use the indicated mean effective pressure, IMEP and engine speed to determine
friction mean effective pressure, FMEP. Then a simulated brake mean effective pressure,
BMEPsim can be calculated and hence brake power, Pb, sim.
The program is not fully dynamical i.e. it cannot fully handle transients. However, it can cope
with changes in the operating situation if these are not to swift. The potential and kinetic part
of the total energy is somewhat neglected yet commented on in this report when necessary.
At first the combustion is treated empirically through a simple model i.e. there are no
calculations of the chemical reactions. The heat release is described by a shape function; the
level is determined by the amount of fuel present and a combustion efficiency factor. As an

15
alternative, an external chemical kinetic program can be linked to the engine module that
provides an estimation of ignition timing.
When linking the different system components, mean values are used i.e. mass flow into and
out of the engine is averaged when linked to the manifolds. The process itself, flow over
intake or exhaust valves, works with a varying in-cylinder pressure and a constant manifold
pressure. Flow is allowed ”backwards” over the valves and regarding the inlet valve this
process is assumed to work as a plug flow process i.e. mixing does not occur between the
back-flowing gas and the inlet manifold gas.

2.3 General assumptions

In order to limit some calculations and computational time there has to be some
simplifications. Therefore it is important as user to know what the limitations are and when
they have to be compensated for.

2.3.1 Ideal-gas equation of state

All gases are assumed to be ideal i.e. the ideal-gas law, or the ideal-gas equation of state, are
assumed to be valid:
p ⋅V = n ⋅ R ⋅ T Eq.17
The universal gas constant is set to 8.3145 J/(mol·K) [98]. For a gas mixture the law becomes:
N
p ⋅ V = ∑ (ni ) ⋅ R ⋅ T Eq.18
i =1

This means that there are no mixing effects accounted for.

In order to show the validity of this law for engine applications the following simple analysis
is done. Among different gases there are some that does not behave as an ideal gas. The usual
way to describe this is to modify the right hand side of the ideal gas law, Eq.17 by multiply
with a compressibility factor, Z that depends on both pressure and temperature.
p ⋅ V = Z ( p, T ) ⋅ n ⋅ R ⋅ T Eq.19
However, it is shown that the Z factor is approximately same at the same reduced temperature
and pressure, defined by Eq.20, for different pure species.
p T
p r= T r= Eq.20
pcr Tcr
pcr and Tcr is the critical pressure and temperature, specific for different substances. The
relation between reduced states and Z is e.g. given in the Nelson-Obert generalized
compressibility charts [reviewed in 99] and some examples are shown in Table 1. It has been
shown that very low pressure (pr<<1, regardless of temperature) and at high temperature
(Tr>2) the gases can be treated as ideal gases with good accuracy. The error becomes larger
close to the critical point. Gas mixtures of real gases are a little more complicated due to the
influence of dissimilar molecules. In combustion engines the state are usually close to normal
atmosphere in the inlet manifold and during the compression stroke both pressure and
temperature increases. Higher pressure results in an increasing compressibility factor but this
is balanced to some extent by the increasing temperature. During combustion, expansion and
exhaust stroke, the temperature is high and the gas can be treated as ideal. Liquid fuel, or fuel

16
that undergoes vaporization is definitely not an ideal gas and situations were vapor is present,
e.g. fuel or water injection must be treated a little bit careful. So the conclusion is that in an
engine the gas can be treated as ideal without any larger error. In addition, most of the
working gas is nitrogen, which is a gas that is very close to ideal gas in all states, this also
give a high accuracy. In this study all gases are treated as ideal gases and equation Eq.17 and
Eq.18 are assumed valid. However, this assumption requires that the mixture be well known
which maybe difficult in some cases. At high temperatures, species can dissociate, affecting
number of moles and mixture composition, a phenomena that is not treated here unless
chemical kinetics is involved.
Table 1 Examples of compressibility factors, Z for some gases at engine related pressures and
temperatures. The table is based on generalized compressibility chart readouts.
Gas T[K] p[bar] T[K] p[bar] T[K] p[bar] T[K] p[bar]
300 1 700 50 1000 100 2500 250
Z Z Z Z
N2 1,00 1,02 1,04 1,02
O2 1,00 1,01 1,02 1,01
Ar 1,00 1,02 1,03 1,02
CO2 1,00 0,99 1,01 1,01
H2O n.a. 0,93 0,97 1,01
CH4 1,00 1,02 1,03 1,02
C2H6 1,00 1,00 1,03 1,02
C3H8 0,99 0,95 1,01 1,04

2.3.2 Thermal and physical gas properties

Thermal and physical properties for pure substances have been taken from the NASA-
polynomials [98]3. It provides a compact way to calculate gas properties for different species
at different states. The data are a large set of polynomial coefficients. For some compounds of
substance such as ”gasoline” or ”diesel", there exist no polynomials. In those cases it is
easiest to either assume that they have the same properties as similar pure substance, or set up
a polynomial set that are equal to a constant specific heat i.e. enthalpy becomes a straight line.
The errors will probably be small in engine calculations since the fuel fraction of most fuel-air
mixtures is quite small.
The species that are treated in this study is Nitrogen (N2), Oxygen (O2), Argon (Ar), Carbon
dioxide (CO2), Water vapor (H2O) and Fuel. Fuel is treated here as single specie regardless
of what it is. For example, gasoline is treated as a ”synthetic” fuel C7.76H13.1, diesel as
C10.8H18.7 and natural gas is set up depending on its individual mole fractions, of typically
methane, ethane, propane and so on. Thermal properties for fuels like gasoline and diesel is
taken from data reviewed by Heywood [96], although converted to NASA-polynomials.
Thermo physical properties for gasoline and diesel vapor is missing completely but since
viscosity and conductivity is only needed in the exhaust gas, were fuel is only present as a
very small unburned fraction, viscosity and conductivity properties for butane was used
instead.

3
Another very useful source for this type of data (and more) is the NIST Chemistry WebBook,
http://webbook.nist.gov/chemistry/.

17
The easiest way to display how this data is stored and handled is by an example. For nitrogen
(N2) the coefficients for thermal properties are shown in Figure 6.
N2 J 3/77N 20 00 00 0G 300.000 5000.000 1
0.28532899E+01 0.16022128E-02-0.62936893E-06 0.11441022E-09-0.78057465E-14 2
-0.89008093E+03 0.63964897E+01 0.37044177E+01-0.14218753E-02 0.28670392E-05 3
-0.12028885E-08-0.13954677E-13-0.10640795E+04 0.22336285E+01 4

Figure 6 Example of a set of polynomial coefficients for calculating thermodynamic

properties, in this case for nitrogen, N2.
The first line contains various information e.g. name, reference and valid temperatures
(between 300 and 5000K). Then there are 14 coefficients, first 7 high- (>1000K) and then 7
low temperature coefficients. Seven coefficients c1–c7, either the high temperature or the low
temperature set, is then used to calculate heat capacity (Cp), enthalpy (h0) and entropy (s0) 4:
C p (T ) = R ⋅ ( c1 + c2T + c3T 2 + c4T 3 + c5T 4 )  J ( mol ⋅ K )
 T2 T3 T4 T5 
h 0 (T ) = R ⋅  c1T + c2 + c3 + c4 + c5 + c6  [J mol ] Eq.21
 2 3 4 5 
 T2 T3 T4 
s 0 (T ) = R ⋅  c1 ⋅ ln (T ) + c2T + c3 + c4 + c5 + c7   J ( mol ⋅ K )
 2 3 4 
Now it is very important to point out that enthalpy calculated from the expression above is
calculated at a reference state at 298.15 K and includes enthalpy of formation, for liquids the
vaporization heat is also included. This means that enthalpy above is written as:

( 0 0
)
h 0f ,298 (T ) = h (T ) − h 298 + ∆ h f ,298 + ∆ h vap ,298
0 0
[J mol ] Eq.22

When calculating chemical reactions it is very convenient to have the formation enthalpy
included because heat from combustion will be added directly following the specie
concentration. However, if there are species present like gasoline or diesel, where
polynomials like those above is missing but the heating value is known, then it is easier to
ignore the formation enthalpy, as shown in equation Eq.23, and instead add energy as heat
during combustion. This approach is used mostly in this study.
0
h298 (
(T ) = h 0f ,298 (T ) − h 0f ,298 (298) = h (T ) − h 298
0 0
) [J mol ] Eq.23

The reverse calculation, from enthalpy to temperature becomes also important. Several
numerical methods are possible but here is the well-known Newton-Rhapson method chosen
manly because in this case it becomes simple since the needed derivative of enthalpy, is the
same as specific heat and the enthalpy - temperature relation is monotonically increasing for
most species. The iterative process of the Newton-Rhapson can thereby be written as:
0
h298 (Tk ) − h298
0

Tk +1 = Tk − [K] Eq.24
C p (Tk )

4
Note: Notation of the coefficients and the polynomials is different here than in the original reference, hopefully
this notation is a little bit simpler to grasp.

18
The reference state for entropy is at 0K and 1bar in order to only having to ”table” a function
of temperature only. Entropy change for an arbitrary process is then calculated as:
p 
∆s = ( s2 − s1 ) = ( s20 − s10 ) − R ⋅ ln  2   J ( mol ⋅ K ) Eq.25
 p1 
In analogy with the ideal gas law, it is assumed that there are no mixing effects and specific
heat, enthalpy and entropy for the mixture is calculated with respect of the property of each
individual species and their mole fractions e.g. enthalpy for a mixture is calculated as:
N
hmix (T ) = ∑ hi (T ) ⋅ yi [J mol ] or
i =1
N
Eq.26
H mix (T ) = ∑ hi (T ) ⋅ ni [J ]
i =1

Mole fraction, yi is calculated as:

ni
yi = N
Eq.27
∑n
k =1
k

Viscosity and thermal conductivity for pure substances are taken from the same source as the
thermodynamical data [98]. An example is shown Figure 7.

N2 V2C2 GORDON; NASA TM86885, OCT 1984

V 300.000 1000.000 0.60443938E 00-0.43632704E 02-0.88441949E 03 0.18972150E 01
V 1000.000 5000.000 0.65060585E 00 0.28517449E 02-0.16690236E 05 0.15223271E 01
C 300.000 1000.000 0.94306384E 00 0.12279898E 03-0.11839435E 05-0.10668773E 00
C 1000.000 5000.000 0.65147781E 00-0.15059801E 03-0.13746760E 05 0.21801632E 01

Figure 7 Thermophysical coefficients for nitrogen, N2.

The first line contains the substance name, phase and some reference data. Then there are four
lines that start with a ”V”, for viscosity, or a ”C”, for conductivity. The two first values in
each of these rows are the temperature boundary, which the coefficients are valid. The
following four values are the coefficients, A–D for the specific property and temperature
interval. Viscosity and conductivity is then calculated as:
 B C 
Viscosity: η (T ) = exp  AV ⋅ ln (T ) + V + V2 + DV  [3]
 T T 
Eq.28
 B C 
Conductivity: λ (T ) = exp  AC ⋅ ln (T ) + C + C2 + DC  [: FP ⋅ . ]
 T T 
Note that viscosity given by Eq.28 is the so-called dynamical viscosity5, which is temperature
dependent only. Kinematical viscosity is calculated as:
η (T )
υ (T , p ) = Eq.29
ρ (T , p )

5
Note: In some literature, µ is used as symbol for dynamical viscosity, instead of η.

19
Now, viscosity and conductivity calculated from Eq.28 gives properties for pure substances.
When mixing substances it is not as simple as with thermodynamical properties. Several
approximate methods exist for mixture calculations and the one is chosen for this work is
derived from Brokaw (1961), as reviewed by Ferguson [101]. This method is briefly
reviewed here.
Thermal conductivity in pure species is split into a translational- and internal transport of
energy:
λi = λi′ + λ ′′ Eq.30
The internal contribution, λ´´ is approximately given by:
 2 C p ,i 
λi′′ ≈ 0.88  − 1  λi′ Eq.31
5 R 
Equation Eq.30 and Eq.31 gives:
   2 C p ,i 
′
λi ≈ λi  1 + 0.88  − 1 
  5 R  Eq.32
 ′′
λi ≈ λi − λ′ i

Now, the two parts is calculated for the mixture:

 N
λi′
λmix′ ≈ ∑

N yj
i =1
1 + ∑ Ψi, j
 j =1 yi
 j ≠i
 Eq.33
 ′′ N λi′′
λmix ≈ ∑ N yj
 i =1
1 + ∑ Φi, j
 j =1 yi
 j ≠i

where y denotes mole fraction and:

   η 1 2  M 1 4 
 1 +  i   j  
   η j   M i  
Φ i , j =  

12
 M 
 2 2 1 + i  Eq.34
 M j 
 


Ψ = Φ

1 + 2.41
( M i − M j )( M i − 0.142 M j ) 
 i, j  ( ) 
i, j 2
  M i + M j 
The conductivity for the mixture is then calculated as:
′ + λmix
λmix = λmix ′′ Non reacting media
Cp Eq.35
λmix = (λmix ′′ )
′ + λmix Reacting media
C p , frozen

20
In this study, all calculus that needs conductivity is using the first expression in Eq.35 e.g.,
heat transfer calculations in exhaust port and manifold are assumed to be under non-reacting
conditions. Viscosity for mixtures is calculated trough:
N
ηi
ηmix ≈ ∑ Eq.36
N yj
i =1
1 + ∑ Φ i, j
j =1 yi
j ≠i

with Φi,j given by Eq.34.

2.3.3 Energy relations

One fundamental relation in this study is the first law of thermodynamics for an open system
(Eq.37), which is based on the conservation of energy principle of a control volume, as
illustrated in Figure 8.
Q − W + ∑ E IN + ∑ EOUT = E2 − E1 = ∆ECV Eq.37
Q is heat transferred to the system (or control volume, CV), W is work done by the system,
EIN and EOUT is energy related to mass flowing into and out of the system respectively,
E2 - E1 is the change of energy in the control volume.

CV EIN EOUT
EIN W
∆ECV
∆ECV Q
Q
CV

W
EOUT

Olof E.

Figure 8 Left: Energy flow in and out of a system. Right: Engine interpretation.
The total energy, E in the system can be written as a sum of internal, potential and kinetic
energy:
E = U + PE + KE Eq.38
Assuming potential and kinetic energy very small and insignificant (gas has usually a quite
low density), the total energy difference in the control volume will essentially be the same as
difference in internal energy. Consider small changes and assume that the flow can be
described uniform. Now, this assumption may be a limitation but normally, when solving the
following equations, these are usually made over small time-steps over which the properties
of the fluids may me treated uniform, or averaged, without any large errors. For a short time-
step and with the assumptions mentioned above, we get:
∂Q − ∂W + ∑ ∂nIN ⋅ eIN + ∑ ∂nOUT ⋅ eOUT = U 2 − U1 Eq.39

21
For a flowing fluid, the flow work has to be considered (which can be seen upon as a part of
potential energy) when the flowing media is passing the control volume boundaries:
E = U + p ⋅V (+ KE + PE ) or
Eq.40
e = u + p⋅v on mass-unit basis
It is very convenient to introduce enthalpy into the calculations for three reasons: Enthalpy
depends on temperature only (see section), gas properties are given in enthalpy and the
definition includes the p·v term:
H = U + p ⋅ V or
Eq.41
h = u + p⋅v
The change in heat is governed by heat transfer models e.g. heat transfer between in-cylinder
gas and inner cylinder surface, the combustion is also seen upon as heat transferred to the
system as mentioned earlier. The definition is set so that positive ∂Q means heat transferred to
the control volume. Work is boundary work, and positive values on ∂W means work is done
by the media in the control volume. In addition, mass transfer, e.g. valve flow, is governed by
models also which also depends on various state variables.
∂Q = ∂Q (T , p,... )
Eq.42
∂W = p ⋅ ∂V
Inserted in Eq.39 we get:
∂Q (T , p,... ) − p∂V + ∑ ∂nIN ⋅ hIN + ∑ ∂nOUT ⋅ hOUT =
Eq.43
( H (T ) − p V ) − ( H (T ) − p V )
2 2 2 1 1 1

This is an ordinary differential equation and it is, like most other large, differential equations,
quite often impossible to solve analytically, numerical solutions seem more appropriate.

2.3.4 Numerical solutions

There exist many different methods for solving ordinary differential equations. Although it
would be appropriate, there has not been any room in this study to test and select an optimum
method. Instead, one has been selected that works and is quite simple to handle and
understand. It is called trapezoid method and can simply be described as an integrating
method where states are represented by their mean values over a time-step.
H, p, T

T, p
Iterations

∂Q
f(T, p, …)
∂W
Time, V
i i+1
dt, dV

Figure 9 Principle of trapezoid method

22
Schematically this can be described as shown in Figure 9. From one step to another, work,
heat and mass transfer are calculated as functions of the mean value representation of
temperature and pressure between the two steps. The state in the one point (i) is used as start
values for the next point (i+1). Trough iterative calculations, pressure and temperature are
calculated in the next point. This means that Eq.43 is written as:

( )
∂Q T , p,... − p∂V + ∑ ∂nIN ⋅ h IN + ∑ ∂nOUT ⋅ hOUT ≈

( H (T ) − p
i +1 V
i +1 i +1 ) − ( H (T ) − p V )
i i i

with: Eq.44
Ti + Ti +1 pi + pi +1
T= p=
2 2
Note that the left-hand terms in expression Eq.44 are functions of the mean state variables and
these causes a change in state between two points, which is described on the right-hand side.
The ideal gas law Eq.18 then connects the state variables in the next point (i+1):
pi +1 ⋅ Vi +1 = ni +1 ⋅ R ⋅ Ti +1 Eq.45
The relation between enthalpy and temperature is then given by Eq.23 and Eq.24. It is thus
possible to set up an iterative calculation scheme (simplified):
1. Given: Ti, pi, Vi, ni (including mole fractions) and Vi+1.
2. As start values for the next point, set Ti+1=Ti , pi+1=pi and ni+1=ni (or a better
guess).
3. Calculate mean values T , p

(
4. Use models to calculate heat and mass transfer, functions of T , p,... )
5. Calculate mole fractions in the next point, ni+1 , which changes due to combustion
and mass transfer.
6. Use Eq.44 to calculate enthalpy in the next point, H(Ti+1)
7. Use Eq.24 to calculate temperature, Ti+1 from enthalpy and Eq.45 to calculate
pressure, pi+1
8. Iterate 3 to 7, until stable pressure and temperature in the next point.
The description above is somewhat influenced by in-cylinder calculations i.e. to calculate in-
cylinder temperature and pressure trace. The time step used is set to be equal to the time
corresponding to 1 crank angle degree (CAD) at the specific engine speed. Test were made
were the step was set to 0.2 CAD with almost the same results (about 1% difference in
resulting variables). At 10 CAD per step the still generate acceptable results as shown in
Figure 10, indicating that the numerical method is very robust.
The expressions and method is very similar when calculating the whole engine system. In
some cases there has to be some numerical ”dampers” in order to achieve convergence due to
e.g. singularities in mass transfer functions. These dampers are inside the iteration loop above
and do not affect the results; they ensure convergence but cost calculation time.

23
 2 2
10 0.2° (3.2253bar) 10 ESIM 1°/step (3.2199bar)
1° (3.2199bar) Experiment (3.201bar)
10° (3.1694bar)
Pressure [bar]

Pressure [bar]
1 1
10 10

0 0
10 10
−1 0 −1 0
10 10 10 10
Volume [liter] Volume [liter]

Figure 10 Comparison between different calculation resolution of 0.2, 1, and 10 CAD (left)
and a comparison with experiment (right). IMEPnet is shown in brackets. Volvo TD100
engine running with HCCI combustion at λ = 3, isooctane.

2.4 System buildup

In this section, the different models of engine system components are described in more
detail. The idea is also to point out what effects are considered and which are not.
The different components are in the software represented by different functions and it
becomes the user role to give the necessary inputs and to connect the different components
together.
When linking these functions together, with respect of pressure, temperatures, mixtures and
mass flows, an engine system can be simulated. The engine cycle calculation can be treated a
little bit different from the other components. By leaving the engine calculations out when
calculating the surrounding components, computational time can be saved without any larger
errors in the calculations. The error that this causes is taken out over time, i.e. if the flow into
the engine is under-predicted then the next cycle will experience a higher inlet pressure that
increases the flow. In a way, this procedure is an assumption that the dynamic changes is
small, and that the engine cycles have a damping effect in the system. An example of a
system is schematically described in Figure 11.
If better accuracy of dynamic, multi-cylinder engine simulations is needed, then the cylinders
should be calculated separately, but overlapped in time, and together with the rest of the
system i.e. inside the iteration loop shown in Figure 11, resulting in a tremendous increase in
computational time.

24
 Set time-step

ENGINE CYCLE
x6 (flow)
TURBOCHARGER Linking of:
– Pressure
– Temperature
– Mixture
INTERCOOLER – Mass flow

INLET

EXHAUST

Adjust timestep?
<300

Maximum
300 Convergence?
iterations No
Yes!

Warning!
No convergence ENGINE CONTROL

Figure 11 Schematical illustration of the system calculation, in this case a 6 cylinder engine.

2.5 Engine model

The engine calculation includes geometry-, heat transfer-, heat release- and valve-flow
calculations. There also exist two different layouts depending on the engine type, port injected
or direct injected engine. The engine is here seen upon as a component with a varying
volume, exposed to heat transfer (including heat release) and connected to two ports with
variable restrictions.
First, before doing any thermodynamic calculations or simulations, some constant data can be
set up representing the geometry of a specific engine. These are typically volume, combustion
chamber area and effective valve flow area. An expression for the piston movement is set up,
which includes an offset in crankshaft-cylinder centerline. TDC is set to be when the piston
reaches its uppermost position, see Figure 12. The following expression for the piston
position can be set up:
 2 
x (θ ) = ( r + l ) ⋅ cos (ς ) −  r ⋅ cos (θ + ς ) + l 2 − (r ⋅ sin (θ + ς ) − δ ) 
 
where: Eq.46
 δ 
ς = arcsin  
r+l

25
This position is then used to calculate in-cylinder volume and combustion chamber area
trough:
π D2
V (0...720° ) = ⋅ x (0...720° ) + Vc
4
A (0...720° ) = Acylinder (0...720° ) + Apiston + Acylinderhead = Eq.47
π D ⋅ ( x (0...720° ) + x0 ) + Apiston + Acylinderhead

Vc is the minimum compressed volume and should include crevice volumes. Area should also
include piston topland area, valve slot area and cylinder head gasket area.
TDC TDC = Top dead center, the piston is in the upper position
x
x = Piston position, x = 0 ⇔ TDC
r+l ς = Angle between the connecting rod (conrod) and the
cylinder when x = 0
Conrod
l = Conrod length
l r = Crank arm length ⇔ Crank radius
ζ
δ = Offset, i.e. shortest distance between the crankshaft and
θ the cylinder centerlines
Crank θ = Crank angle, θ = 0 ⇔ TDC
r radius

δ CL cylinder

Figure 12 Geometry for piston movement

Valve lift is calculated trough the cam profile determined by mechanical analysis (Bensinger,
1968, reviewed by Söderberg [102]). For a given desired maximum valve lift and duration a
cam profile is calculated. This is then used to calculate the lift curve. Further description will
not be made here more than it can be mentioned that the lift curve is determined with e.g.
respect of maximum acceptable valve acceleration and deceleration. One alternative is to use
measured valve lift. This data can be stored in a suitable format and read by the program
during initialization.
The valve lift is used to calculate the minimum geometrical flow area (Amin) but for the valve
flow calculation, an effective valve flow area (AE) is used. There are some flow effects that
depend on valve and port geometry and friction against port and seat surfaces. The ratio
between effective flow area and minimum flow area is usually called “discharge coefficient”
(CD) and can be seen upon as port and valve geometry efficiency (Eq.48). An alternative is
the AE/AC ratio, which is the effective valve flow area divided by the valve curtain area (AC).
The valve curtain area is simply the cylindrical area defined by lift and valve diameter. The
minimum flow area depends on the geometrical layout of the valve and the port and is
calculated through Eq.49, the geometrical parameters given in Figure 13. An example of
calculated lift and flow areas is seen in Figure 14.
AE
CD = [ AE AC AC = π ⋅ Dv ⋅ Lv ] Eq.48
Amin

26
   
 Amin = π ⋅ Lv ⋅ cos ( β ) ⋅  Dv − 2 w + 2 ⋅ sin ( 2 β ) 
Lv
when 0 < Lv ≤ Lv1
  

 Amin = π ⋅ ( Dv − w ) ⋅ ( Lv − w ⋅ tan ( β )) + w
2 2
when Lv1 < Lv ≤ Lv 2

 A = π ⋅ ( D2 − D2 ) when Lv 2 < Lv
 min 4 p s


where valve lifts Lv1 and Lv2 is calculated through: Eq.49
 Lv1 = w
 (sin ( β ) ⋅ cos ( β ))

 2
L =  D p2 − Ds2 
  − w + w ⋅ tan ( β )
2
 v2
  4 ⋅ ( Dv − w ) 
Ds

Dp
β
Lv
w

Dv
Olof E.

Figure 13. Valve geometry

Initially models for the discharge coefficients were adopted, based on measurements reviewed
in Heywood [96]. Although based on rather limited amount of data, it provided at least an
estimate of the effective flow area. The discharge coefficient was dependent on relative valve
lift (lift divided by valve diameter). However, some limitations were found with this
approach. According to the reviewed results the CD factor tended to go down at high lifts but
the review did not say if Amin was constant or not at high lifts. If constant it would mean that
the effective flow area would go down at high lifts and this seems quite unrealistic since it
would also mean that the pressure drop given a certain mass flow then would be higher. If the
effective flow area is determine by the AE/AC ratio it is possible that the effective flow area
becomes higher than the Amin, which also seems quite unrealistic. Therefore, in the absence of
measured data on effective flow area, a constant discharge coefficient, CD of 0.7 is adopted,
regardless of flow direction. If there is some knowledge about the flow process of the real
engine, it is quite simple to add this information into the calculations.

27
 15 Inlet
Exhaust

Lift [mm]
10

0
0 180 TDC 540 720
15 Ainl.,min
A
10 inl.,eff.
Aexh.,min
[cm ]
2

A
5 exh.,eff.

0
0 180 TDC 540 720
CAD
Figure 14 Example of calculated valve lift (top), minimum and effective flow area (bottom).
Note that the CD factor is set constant.
The effective valve flow area and the pressures in the inlet port and in the cylinder are then
used to calculate mass-flow over the valves. The expression below is derived with an ideal,
steady flow and isentropic assumption [96], resulting in:
 (γ −1) γ
2γ   pdown 
1γ
m = 1000 ⋅ A ⋅ p  p  
E up

down
 1−    [g s ]
 1000 ⋅ Tup ⋅ R M  pup  γ − 1   pup 
   

 1000 ⋅ AE ⋅ pup  2 
(γ +1) 2(γ −1)
m choked = γ  [g s ] Eq.50
 1000 ⋅ Tup ⋅ R M  γ +1
γ (γ −1)
p  2 
Choked flow occur when: down ≤  
pup  γ +1
Note that the expressions are written as function of γ, up to now this has been avoided since
gas properties are given by enthalpy. However, in this case it is simpler due to the non-
choked/choked check has to be done, which is easiest described as a function of γ, calculated
through:
C p (T )
γ (T ) = Eq.51
C p (T ) − R
Note that specific heat also depends on what mixture flows trough the valves. The mixture is
set to be a mixture of air-fuel-EGR when flowing over the inlet valves, except when backflow
occurs and when this flow is pulled back into the cylinder again. In this case the mixture is
assumed to be the same as inside the cylinder. This is also is assumed to be the case regarding
the exhaust valve flow. The cylinder is assumed to be a well-mixed volume. In order to fit
into the energy conservation equation, the flow is represented as:

∂n ⋅ h = ∂t ⋅ m
M ( )
⋅h T [J ] Eq.52

28
2.5.1 Simple heat release model
Since this model does not have any built-in chemical kinetic calculations or an empirical
model of the heat release during combustion, a simple shape function is used for the heat
release. The amount of heat released is determined of the amount of fuel (and oxygen) that is
trapped inside the cylinder at IVC. This amount of fuel, its lower heating value and an
assumed combustion efficiency is then used to calculate the heat released. Together with the
shape-function the heat release for each CAD can be calculated and inserted in the left-hand
side of Eq.44. Several different shape functions have been tested and a function first described
by Wiebe, as reviewed by Heywood [96] (see also Vibe [103]) was found appropriate. In a
slightly modified form it can be described as:

(
xb = 1 − exp a ′ (θ − θ 0 )
m′
) Eq.53

where xb is the percentage heat released, or combustion progress. The parameters θ0, a' and m'
is often hard to give values with direct physical meaning, it is easier to think in terms of A0,
A10 and A90, which represent the crank angle of 0,10 and 90%, of total heat release. It is
possible for a set of A0, A10 and A90, find a θ0, a' and m':
θ = A0
 0
  ln (0.9 )   A10 − A0 
m ′ = ln   ln   Eq.54
  ln (0.1)   A90 − A0 

a ′ = ln (0.9 ) ( A10 − A0 )
m′

Heat release fraction is then calculated as:

∂QHR = ηc ⋅ m f ⋅ QLHV ⋅ ( xb,i +1 − xb,i ) [J ] Eq.55

where mf is fuel mass present before combustion takes place and ηc is an assumed combustion
efficiency. An example of the Wiebe heat release shape is showed in Figure 15.
1
A90
0.8

0.6
xb

0.4

0.2
A10
0 A0
−10 −5 0 5 10 15 20
CAD

Figure 15 A typical Wiebe heat release shape representing HCCI combustion.

2.5.2 In-cylinder heat transfer model

Heat transfer is modeled with an expression derived by Woschni [104, see also reviews in 96,
101] The characteristic equation for the this model is:
Nu = 0.035 ⋅ Re0.8 Eq.56

29
It relate the Nusselt number, Nu = α⋅Lref/λ, – the dimensionless representation of the
convective heat transfer coefficient, α, with the Reynolds number, Re = ρ⋅Uref⋅Lref/η , – the
ratio between inertia forces and viscous forces. The Reynolds number is often used as a
characterization of the flow, the degree of turbulence. The average cylinder gas velocity
throughout the cycle is described by Eq.57 were indices “ref” denotes a reference state e.g.
IVO or start of combustion:
Vd ⋅ Tref
w = Cadjust .1 ⋅ 2.28 ⋅ S p + Cadjust .2 ⋅ 3.24 ⋅ 10 −3 ⋅ ⋅ ( p − pmotored ) Eq.576
pref ⋅ Vref
The original expression of Eq.57 used stroke dependant coefficients, the factor 2.28 was 6.18
during the exchange strokes and 3.24 was 0 for the exchange and compression strokes. In
addition compensation factors (Cadjust.1, Cadjust.2) are introduced to give “tuning knobs” for
better agreement with experiments. In this program the second term of the expression is used
from start of combustion (A0) to EVO. Start of combustion is used as reference state and the
motored pressure, pmotored is estimated over the combustion and expansion period by an
isentropic relation:
γ A0

pmotored ( A0! EVO ) = p A0 ⋅ VA0 V ( A0 ! EVO )  Eq.58

 
The correlation between average in-cylinder gas properties and heat transfer coefficient is
according to Woschni (please note the units):

hc W 2  = 3.26 ⋅ B [m ] ⋅ p [kPa ] ⋅ T [K ] [m s]

−0.2 −0.55
⋅w
0.8 0.8
Eq.59
 m K
In cylinder surface is calculated from data about the actual engine’s cylinder head and piston
crown down to the first piston ring. The average wall temperature is assumed and kept
constant, however it would be possible to insert cylinder wall temperature calculations based
on the heat flux over the walls. The total heat transfer fraction is calculated as:

(
∂QHT = ∂t ⋅ A ⋅ hc ⋅ T − T wall ) [J ] Eq.60

2.5.3 Exhaust port heat transfer model

In order to get a good estimation for the turbine conditions, especially the exhaust
temperature, it becomes important to model the heat transfer in the exhaust ports. A model
developed by Hires and Pochmara (1976, reviewed by Ferguson [101]) is used to calculate the
instantaneous heat transfer in the exhaust port. The basic relation for this model is:
Nu = 0.158 ⋅ Re0.8
where Eq.61
 m ⋅ D p 
Re =  

 η ⋅ Ap 

6
According to the original reference [104] the factor 3.24·10-3 has the unit m/(s·°C) which means that Tr should
be in °C. However, this seems strange since this part of the expression is in a way a calculation of the mass
trapped inside the cylinder through the ideal gas equation of state. Therefore, it is believed that there is a misprint
in the original reference and the unit should be m/(s·K).

30
The heat transfer coefficient is calculated as:
Cadjust . ⋅ Nu ⋅ λ
αp = Eq.62
Dp
An adjustment factor Cadjust. in Eq.62, initially set to 1, is can be used to fit the model to
measurements. Note that there is about 4.5 times higher heat transfer in the exhaust port, than
inside the cylinder, according to the basic equations, Eq.56 and Eq.61. One explanation can
be that the flow becomes much more turbulent closer to the walls in the exhaust port, than in
the cylinder, due to varying flow velocities and port geometry. This constantly disturbs and
reduces the laminar boundary layers near the port walls, and hence increases heat transport to
the walls.
The coolant side of the exhaust port wall is assumed to have a constant temperature, 100°C
approximately equal to the boiling temperature of the coolant. The inner side temperature is
calculated from a mean value of the heat flux over the exhaust wall which is calculated in
iterative manner assuming that the wall can be described as a plate as shown in Figure 16. The
thermal dynamics in the wall is not considered. That means that the mean wall temperature
during an exhaust stroke is determined by the heat flux during that, and not previous exhaust
strokes.
Different models for exhaust port heat transfer were compared to measurements by Caton and
Heywood [105], both time-dependence and spatially at the exit of an engine port. The results
indicated that the model selected here gave an ”approximate agreement” on the mean exit port
temperature (so it suits the purpose here fairly well), the time-resolved temperature was less
accurate.

Temperature twall

αp λ
Tport

Exhaust q· Coolant
port side
Tcoolant
Exhaust
port wall

Figure 16 Scematic temperature distribution over exhaust port wall

2.5.4 Governing equations

Besides the energy conservation equation, mass conservation equations and equations
governing species is needed.
Two things affect in-cylinder mixture; mass flow through the inlet valves and combustion. A
mixture of fuel, air and EGR in the inlet manifold gives the gas composition of the mass
flowing into the cylinder. An array of mole (N2, O2, Ar, CO2, H2O and fuel) represents the

31
mixture inside the cylinder. The flow into the cylinder is added to this array with respect of
the mass flow rate and the inlet mixture.
The combustion is, as mentioned earlier, governed by the combustion shape function and
combustion efficiency. With the specie set-up chosen here, it is assumed that if the
combustion is not completed (less than 100% combustion efficiency), unburned fuel is the
only emissions i.e. there are no CO emission or HC emission besides pure fuel, in order to
reduce the specie set-up. The combustion is thereby only affecting the number of moles of
fuel, O2, CO2 and H2O. However, it is possible to expand the set of species, but it increases
the computational time. The overall complete combustion equation for the involved species
becomes:
 b c
[Ca H bOc + dN 2 + eAr + fCO2 + gH 2O ] fuel +  a + −  O2 
→
 4 2
Eq.63
b 
dN 2 + eAr + (a + f ) CO2 +  + g  H 2O
2 
The total change in moles for combustion of one mole of fuel under lean mixtures (λ≥1) can
be written as (please note the array notation):
∆nc ,tot = ∆  nN 2 nO2 n Ar nCO2 nH 2O n fuel  =
  b c b   Eq.64
ηc ⋅  d −a + −  e (a + f )  + g  −1
  4 2 2  
Finally, the total change in cylinder gas moles and composition, due to valve flow and
combustion, can be expressed by:
m inl m exh
ni +1 = ni + ∂t ⋅ ⋅ x inl − ∂t ⋅ ⋅ x cylinder + ( xb,i +1 − xb,i ) ⋅ ∆n c ,tot Eq.65
M inl M cylinder
In summary, Eq.65 combines the mass and atom conservation constraints.

2.5.5 Numerical problems

There is one numerical difficulty regarding the expressions for the valve flow. Consider the
expression for non-choked flow in Eq.50. Lets simplify this function a little by setting γ = 1.3
and removing the dimensional part:

1
0.769
  1 0.231 
m 0 =   8.67  1 −    Eq.66
π  
 π  
When plotting this function close to a pressure ratio (π) 1, e.g. when flow switches direction,
there will be a problem in finding a solution to the differential equation with the method that
are selected here. This is visualized as in Figure 17. The first solution says that it flows into
the cylinder, which means that the pressure increases inside the cylinder. Due to the length of
the time-step, the pressure becomes too high and the new solution says that it flows
backwards, and so on. The explanation of this behavior is that the slope of this curve goes to
infinity close to 1 in pressure ratio. Another similar situation appears in the transition from
non-choked to choked flow.

32
 0.4

0.2

m· 0
-0.2

-0.4
0.95 1 1.05
π

Figure 17 Numerical problem when flow changes direction

Two simple methods are used to solve these problems. The shape of the curve close to
pressure ratio one is changed so the slope does not go to infinity. This is done by inserting a
new function, a 2'nd degree polynomial function when the pressure ratio is about
0.99<π<1.01. The same level and gradient is used at about π=1.01 as Eq.66, and the mass
flow is zero at π=1. This modification has only effect when the pressure ratio is very close to
one, which happens perhaps a few crank angles per cycle at higher engine speeds. Therefore,
it does not have any noticeable effect on the results. The other method used is a damper inside
the iterative loop. The mass flow is only allowed to change to a mean value of the previous
(k) and new (k+1) calculated mass flow inside the iterative loop, as shown in Eq.67.
1
m k +1 = (m k + m new ) Eq.67
2

2.5.6 Diesel engine combustion model

In order to verify calculations and models against measurements on a real engine, diesel
combustion is modeled. Direct diesel injection and combustion is modeled as suggested by
Egnell [106] which simply states that the heat release rate is proportional to the difference
between injected fuel, in terms of heat and accumulated heat release:
Q hr = khr ⋅ (Q f − Qhr ) Eq.68

m f ,inj .

m f ,inj .

SOI τ inj. EOI

10° BTDC
Figure 18 Assumed fuel injection, SOI - Start Of Injection, EOI - End Of Injection
With the assumption that fuel injection process can be described as shown in Figure 18, i.e.
that the fuel flow through the injector nozzles are constant during the injection period, the
differential equation Eq.68 can be solved analytically and the result is shown in Eq.69.

33
 1  
qhr (t ) =  (t − tSOI ) +
τ inj . 
1
(
exp ( − k hr ⋅ (t − t SOI )) − 1  ) tSOI ≤ t ≤ t EOI
khr 
qhr (t ) = 1 + ( qhr (t EOI ) − 1) ⋅ exp ( − khr ⋅ (t − t EOI )) tEOI < t Eq.69
∂Qhr = ηc ⋅ m f ,inj ⋅ QLHV ⋅ ( qhr (ti ) − qhr (ti −1 )) [J ]
It is assumed that a fraction of the fuel that is injected, but has not burned, is in vapor phase.
Vaporization has a cooling effect but is very small in comparison with the overall heat
release. It can then be expressed as in Eq.70 and to illustrate this, heat release and
vaporization are exemplified in Figure 19.
(t − tSOI )
qVAP (t ) = kVAP + (1 − kVAP ) ⋅ qhr (t ) tSOI ≤ t ≤ tEOI
τ inj.
qVAP (t ) = kVAP + (1 − kVAP ) ⋅ qhr (t ) t EOI < t Eq.70
∂QVAP = m f ,inj ⋅ QVAP ⋅ ( qVAP (ti ) − qVAP (ti −1 )) [J ]

150 ∂QHR 4000

100 *∂Q
VAP

3000
100
[J]

[J]

2000
50
1000 QHR
100 *Q
VAP

0 0
340 360 380 400 420 440 460 340 360 380 400 420 440 460
CAD CAD

Figure 19 Example of diesel vaporization and heat release (QHR), note how small the
vaporization part is compare to heat release (a magnification factor of 100 is used here!),
kVAP = 0.5 and step length is 1 CAD. The right plot shows accumulated values.

2.5.7 Engine friction model

There exist numerous of friction models, most of these give an estimate of the friction mean
effective pressure (FMEP) as a function of mainly engine speed. A small comparison is made
in Table 2 between different models. Experimental results from a SCANIA DSC 12 diesel
engine running at 1500 rpm are used for comparison. Typically, IDI and small DI diesel
engines give a higher degree of friction. The model selected for this study is for a direct
injected diesel engine (Eq.71, reviewed by Heywood [96]). Despite the fact that it is stated
that the last term in this model is pumping work it seems like to give a reasonable estimation
of friction for the SCANIA DSC12 engine, when comparing to models for smaller TDI
engines. Most simulation are made at 1500 rpm7, so the speed dependence of the model is
only used in some individual cases, FMEP of 1.71 bar seems to be a fair estimate at this
speed.

7
1500rpm correspond to a synchronous speed of an electrical generator connected to 50Hz AC. Other
synchronous speeds for 50Hz are e.g. 750, 1000 and 3000 rpm, and for 60Hz: 900, 1200 and 1800 rpm.

34
 (
FMEP = 1000 ⋅ 75 + 48 ⋅ n 1000 + 0.4 ⋅ S p
2
) [Pa ] Eq.71

Given this model a mechanical efficiency (ηM.e., Eq.15) can be calculated from engine load
(BMEP) and engine speed. The result is shown in Figure 20 and it clearly shows the
importance of high load operation for efficiency.
100

80

60
ηM.e.

40
1000rpm (1.34bar)
20 1500rpm (1.71bar)
2000rpm (2.13bar)

0
0 5 10 15 20
BMEP [bar]

Figure 20 Speed and load dependence on mechanical efficiency. FMEP is noted in brackets.
Table 2 Friction models. The top one is used throughout this work (1 bar = 100 kPa).
Model FMEP Description
[kPa] [bar]

MEPMotored = 75 + 48 ⋅ n 1000 + 0.4 ⋅ S p

2 1.71 Reviewed by Heywood [96], last term
stated to be pumping work. 8
MEPMotored = 75 + 48 ⋅ n 1000 1.47 Same as above, but without last term.

MEPMotored = 110 + 48 ⋅ n 1000 + 0.4 ⋅ S p

2 2.06 Same as top, but for large swirl
chamber IDI engines.
FMEP = 190 + 2.9 ⋅ (n 60 − N 0 ) 2.18 Model for Audi 1.9 liter TDI engine
from 1990, Wu and Ross [107]
N 0 = 30 ⋅ 3 /(1000 ⋅ Vd ,tot ) [1 s ]
FMEP = 183 + 2.3 ⋅ (n 60 − N 0 ) 2.06 Model for VW 2.46 liter TDI engine
from 1992, Wu and Ross [107]
Experiment 1.5 - 1.9 Truck size TDI diesel engine

2.6 Turbocharger models

The HCCI engine is running on lean mixtures and has limited possibilities to use rich
mixtures in even for short periods. It becomes critical to succeed in supercharge the HCCI
engine in order to achieve acceptable specific power i.e. power output per engine size (weight
or volume). The favored method is turbocharging since the power required for the
compression of the inlet air is taken from the exhaust gas trough expansion. Other methods

8
The friction was investigated on a motored engine. When using the model: FMEP = MEPMotored, or more
correct: (FMEP+PMEP) = MEPMotored

35
are supercharging with a compressor, driven directly by the engine itself, however this
directly costs energy and thereby efficiency. Another, quite advanced method is the Comprex-
charger which uses the blow down - pulse of the exhaust gas to directly compress the inlet air,
however this advanced solution has found very little acceptance on the market.
Turbocharging is selected here since it is the most dominating form of supercharging for
heavy-duty engines. It exist several manufacturers, and dealers, that can provide complete
turbochargers. The method that are used her to model the turbocharger should be applicable
on various superchargers as well.

2.6.1 Performance data

The base for the turbocharger model is the performance map of the compressor and turbine
respectively. Typically, these maps give a relation between:
- pressure ratio
- mass flow
- rotating speed
- isentropic efficiency
The speed and mass flow are often normalized with to certain standard states. For turbines a
third approach is frequent and that is to model the turbine as a restriction with a varying
effective flow area that depends on speed and pressure ratio. Depending on manufacturer and
the standards they follow, different normalizing approaches are used; some typical examples
are shown here.
p2
Compressor: π C =
p1
Eq.72
p
Turbine: πT = 3
p4
The pressure ratio, π (Eq.72), is often given as total to total or total to static, probably due to
what method that were used during the determination of the performance maps. The
difference between the total and static pressure is in general quite small i.e. the dynamic
pressure contribution is small. It is only at the short period of blowdown, the velocities
become so high that there will be a difference and hence it becomes important when studying
pulse-loading turbocharging (see below). In order to calculate from static to total and vice
versa, information about flow velocities and density is needed. From now on all state
variables are considered at stagnation state (total). The index numbers in Eq.72 indicate states
at 1– before compressor, 2– after compressor, 3– before the turbine and 4– after the turbine.
For mass flow, there are two common types of normalization:
T pref .
Compressor: CMF − Corrected Mass Flow = m corr. = m [g s]
Tref . p
Eq.73
T g ⋅ K 
Turbine: MFP − Mass Flow Parameter = m  
p  s ⋅ Pa 
Examples of reference states are 1.013bar/288.15K for compressors and 2bar/773.15K for
turbines and the states are upstream the compressor and turbine respectively. The parameter
representation is simpler since there is no need for a reference state. However, it is easy to

36
convert between corrected- and parameter value representation. A third way to normalize
mass flow is by calculate an effective flow area by rearranging equation Eq.50.
Speed is normalized in the same manner as mass flow but only against temperature:
Tref .
Compressor: CSP − Corrected speed = N corr. = N [1 s ]
T
Eq.74
N  1 
Turbine: SPP − Speed parameter = s⋅ K 
T  
The isentropic efficiency is defined as:
h2,s − h1 {Isentropic}
Compressor: ηC = =
h2 − h1 {Real}
Eq.75
h −h {Real}
Turbine: ηT = 3 4 =
h3 − h4,s {Isentropic}
Index ”s” indicates isentropic states, which is calculated from pressure ratio and temperature
before compressor and turbine respectively.
Typical schematic figures of performance maps are shown in Figure 21.

πC Compressor map MFP Turbine map

ηT
ηC

surge choke
line line πT

CSP
70%
65%
60%
?
?
CMF SPP

Figure 21 Schematic turbocharger performance maps, dotted lines are efficiency. The
questionmarks indicate regions were interpolation is impossible without addition of data.

2.6.2 Performance data representation

A crude way to get the mass flow and efficiency of the compressor or turbine is to simply use
some sort of interpolation method in the table representation of the maps. This has one major
drawback, the performance maps are seldom complete and some operating situations are not
covered, especially the region with low speed and pressure ratio as indicated in Figure 21.
Some extrapolation is needed and this is very difficult, inefficient and often not very
successful, which is a conclusion based on the experience of several attempts by the author.
However, in the region where the data is provided from the maps, interpolation certainly is

37
possible. The problem is that it is not sure that the system reaches this region during startup
for various reasons and it becomes critical that the extrapolation is correct. Models seem to be
a more appropriate solution for handling performance data.
The following models are influenced by the work done by Müller et. al [108], however some
models are modified, others are completely different. Müller has derived models from
physical understanding and correlated this against performance maps and measurements for
small turbochargers for spark ignition engines. However, when trying to directly use these
expressions on turbochargers for a truck-size diesel engine, less accurate correlation was
found here, so this model was modified.

2.6.3 Compressor mass flow model

This part of the model predicts the corrected mass flow (CMF) from the compressor given a
certain pressure ratio (total-total, π) and a corrected speed (CSP). The model consists of first
and second-degree polynomials that are connected. The model is fitted to performance data by
looking at isolines in CSP, which is often the way the data is illustrated in the performance
maps. Each isoline (i) is fitted to a second-degree polynomial with three coefficients:
π = C (i,1) ⋅ CMF 2 + C (i, 2 ) ⋅ CMF + C (i, 3) Eq.76
Given a certain isoline there is a maximum pressure ratio and this maximum is assumed to
correspond to the surge9 point at this speed. The surge point (CMFS, πS) together with the
second-degree coefficient, Aπ is simply another way to describe each isoline:
CMFS (i ) = − C (i, 2 ) (2 ⋅ C (i,1))

π S (i ) = C (i,1) ⋅ CMFS + C (i, 2 ) ⋅ CMFS + C (i, 3)
2
Eq.77

 Aπ (i ) = C (i,1)
The surge point and the second-degree dependence for all isolines are then fitted against CSP.
The pressure ratio and mass flow points of this surge line are assumed to depend on CSP only
and the fit is done so that good extrapolation capabilities is achieved i.e. if the pressure ratio is
1 and the turbo shaft speed is zero, then the mass flow should become nil. The following
equations are used to achieve this.
CMFS CSP = CcmfS (1) ⋅ CSP + CcmfS (2 )

(π S − 1) CMFS = Cπ S (1) ⋅ CMFS + Cπ S (2 ) Eq.78

 Aπ = Cπ A (1) ⋅ CMFS + Cπ A (2 ) ⋅ CMFS + Cπ A (3)
2

This means that there are 7 coefficients that represent the mass flow model for the
compressor. When the model is used in the simulations a certain CSP gives the surge point
and the second-degree dependence from Eq.78. At this moment the isoline is determined at
the given CSP and the polynomial coefficients to this isoline can be determined by solving the
system of equations in Eq.77. The mass flow becomes the roots to Eq.76. The model behavior

9
The surge line is the boundary to instable operation regimes at low mass flows. If the flow is restricted to left of
the surge line there will initially be a reversed flow in some of the compressor flow area. This reduces the
pressure upstream compressor (unless the inlet manifold volume are enormous), the pressure ratio drops
suddenly. But after this drop it is possible that the resulting operation point becomes to the right of the surge line,
in the stable regime, which means that the pressure can be built up again. We get an oscillating, surging,
operation, which in the long run destroys the compressor.

38
in the extrapolation region of the map is exemplified in Table 3. Note that for zero speed, a
pressure ratio higher than one gives a flow in the reverse direction.
Table 3 Extrapolation region in the compressor map. Last point is however covered by the
map and added to show the transition from “extrapolation” to ”interpolation”.
π CSP [1/s] CMF [g/s] π CSP [1/s] CMF [g/s]
model model
1 0 0 0,9 0 51,7
1 100 30,7 1,1 0 -51,7
1 250 66,7 1,05 250 44,2
1 500 146,2 1,2 500 43,9
1 1000 441,8 1,6 1000 285,9*
*) Map value: 275,6

Normally it is the maximum root of Eq.76 that should be selected. However, if the pressure
ratio becomes higher than the surge pressure, we assume that surging occur, in addition the
roots to Eq.76 becomes imaginary. This surging situation is handled here quite simple by
reducing the mass flow quite heavily according to the scheme illustrated in Figure 22 so the
simulation of surging operation becomes quite realistic. The basic principle is that a virtual
pressure ratio, π´ is calculated that is below πS as much as π is above πS. Then, a virtual mass
flow, CMF´ is calculated from π´ and CSP with the model as normal and this time there will
be a real, maximum root to Eq.76 which is CMF´. The estimated mass flow, CMF is then
assumed to be below CMFS as much as CMF´ is above CMFS. Another way to look at it is
that it is the minimum instead of the maximum root that is selected at π´.

π π ´ = 2π S - π
CMF´ = CMF(π´,CSP)
CMF = 2 CMFS - CMF´
π
πS
π´ CSP isoline

CMF
CMF CMFS CMF´

Figure 22 Handling surge operation (exaggerated figure). If the pressure ratio is higher than
the surge pressure ratio, the estimated mass flow will become much lower than the surge
mass flow, it can even become negative with this model. But it is more likely that when the
mass flow drops heavily, so will also the pressure ratio, unless the inlet manifold volume is
enormous.

2.6.4 Compressor efficiency model

The efficiency model works with a very similar principle and again, the isolines in corrected
speed that are used for fitting the model to the data. The isentropic efficiency, η represent the

39
compressor effectiveness. Given an isoline in speed the efficiency is fitted to mass flow
through a second-degree polynomial:
η = D(i,1) ⋅ CMF 2 + D(i, 2) ⋅ CMF + D(i, 3) Eq.79
The representation of this isoline is transformed to maximum efficiency, mass flow for
maximum efficiency and the second-degree coefficient:
CMFη Max (i ) = − D (i, 2 ) (2 ⋅ D (i,1))

ηMax (i ) = D (i,1) ⋅ CMFη Max (i ) + D (i, 2 ) ⋅ CMFη Max (i ) + D (i, 3)
2
Eq.80

 Aη (i ) = D (i,1)
These three coefficients are then fitted to three second-degree polynomials against corrected
speed, this time without any extrapolation considerations since there is no obvious efficiency
when pressure ratio is one and mass flow is zero. There are then 9 coefficients that define the
efficiency model for the compressor. When using the model, the corrected speed gives the
coefficients of the isoline through solving Eq.80 and the efficiency is the calculated directly
by Eq.79, however efficiencies below 1% are not accepted. Note that the compressor can
work like a turbine during speed up if the pressure ratio is below 1.

2.6.5 Turbine mass flow model

The turbine performance maps has also some sort of normalized or corrected parameters like
the compressor. In this case mass flow parameter, MFP and speed parameter; SPP is selected
as base for this model. Note that these parameters do not require any reference state10. This
model is a little bit different than the compressor mass flow model, it requires no isolines and
the fitting is done with the least square method of the following equation:

π = C (1) ⋅ ( MFP ⋅ π ) + C ( 2 ) ⋅ ( MFP ⋅ π ) + C (3) ⋅ SPP 2 + C (4 ) ⋅ SPP + 1

2
Eq.81
The least square fit can be expressed in matrix terms as (n is the number of known points in
the performance map):
y = π (1! n,1) − 1
 ( MFP(1,1) ⋅ π (1,1))2 ( MFP(1,1) ⋅ π (1,1)) SPP (1,1)
2
SPP (1,1) 
 
 ( MFP(2,1) ⋅ π (2,1))2 ( MFP(2,1) ⋅ π (2,1)) SPP (2,1)
2
SPP (2,1)
ù=  Eq.82
 # # # # 
 
( MFP( n,1) ⋅ π (n,1)) ( MFP(n,1) ⋅ π (n,1)) SPP (n,1) SPP (n,1)
2 2

C MFP = ( A T × A ) × AT × y
−1
[4 x 1 matrix ]
CMFP is the model representation; hence only 4 coefficients are required for the turbine mass
flow model. Again, the model has good extrapolation qualities towards low mass flows as
shown in Table 4. When using the model the mass flow parameter is the maximum real root
to Eq.81.

10
It seems strange that a reference state is ever required, even for the compressor maps. A reference state brings
in unnecessary information into the performance data. There may be a historical or practical explanation.

40
Table 4 Extrapolation region in the turbine map. Last point is however covered by the map
and added to show the transition from “extrapolation” to ”interpolation”. The first point with
a pressure ratio below 1 is quite unrealistic for a turbine.
π SPP 103 · MFP model π SPP 103 · MFP model
[1/(s·sqrt(K)] [g/s·sqrt(K)/Pa] [1/(s·sqrt(K)] [g/s·sqrt(K)/Pa]

0,9 0 -23,7 1 10 13,3

1 0 0 1,1 15 28,6
1,1 0 15,2 1,2 20 38,3
1,2 0 25,7 1,3 25 44,6
1,3 0 33,3 1,4 30 48,8*
*) Map value: 48,2

2.6.6 Turbine efficiency model

The efficiency model is similar to the compressor efficiency model but more care is taken to
the extrapolation problem. Isolines for different constant pressure ratios are used to fit the
data. Given a constant pressure ratio, efficiency are fitted against SPP in the following
equation, which can be done through a linear fit against η/SPP:
η = SPP ⋅ ( D (i ,1) ⋅ SPP + D (i , 2 )) Eq.83

This also means that if the speed is zero then the efficiency also becomes zero regardless of
the D-coefficients, which seems to comply with the performance maps as well. Again this
isoline representation is transformed by calculating the maximum efficiency and the
corresponding speed parameter:
 SPPη Max (i ) = − D (i, 2 ) (2 ⋅ D (i,1))
 Eq.84
ηMax (i ) = SPPη Max (i ) ⋅ ( D (i ,1) ⋅ SPPη Max (i ) + D (i , 2 ))
These “coordinates”, SPPηMax and ηMax are then fitted against pressure ratio through second-
degree polynomials, which means that there are 6 coefficients that define the efficiency model
for the turbine.

2.6.7 Model accuracy

When using the model, the pressure ratios, the upstream gas temperature and rotational speed
are used to calculate the mass flow through the device (turbine and compressor respectively)
and the resulting shaft work, or shaft power. In addition the isentropic enthalpy is required to
calculate the temperature downstream the device. This resulting work then accelerates the
turbo shaft according to the total inertia (assumed) of the shaft. This inertia has a great effect
on the response of the system. So when determine lack of fit it is important to make the
comparison as the model is supposed to be used i.e. from speed and pressure ratio determines
mass flow and efficiency.
There exist numerous of factors that limit the turbocharger model from giving a correct
description of a real turbocharger. First of all, the performance data itself who often are
determined from very few individual compressors and turbines, perhaps not even in engine
like operation. Then we have a lack of fit of the model and its limitation of handling some
phenomena, in this case there has not been taken any special measures to handle choking on
either compressor or turbine, although some is already handled in the model itself. Choking

41
means that the flow reaches sonic velocities and the flow will not depend on the pressure
ratio.
Compressor Compressor
60 15

50
10
40
CMF error [%]

EFF error [%]

30 5

20 0
10
−5
0

−10 −10
100 200 300 400 500 600 700 100 200 300 400 500 600 700
CMF [g/s] CMF [g/s]

Turbine Turbine
2 2

1.5
0
1
MFP error [%]

EFF error [%]

0.5
−2
0

−0.5
−4
−1

−6 −1.5
10 20 30 40 50 60 70 10 20 30 40 50 60 70
SPP SPP

Figure 23 Turbo model fit to performance maps. Mass flow – left, efficiency – right,
compressor – top and turbine – below.
At this stage, the lack of fit is the only possible error to study. Measurements could have been
useful if turbine speed was measured. In Figure 23 is a fit to a compressor and turbine for a
truck-size diesel engine shown with lack of fit plots. The turbine model show excellent
agreement, the compressor model does not show that good agreement. It must be pointed out
that close to surge a small deviation in pressure ratio will lead to big change in mass flow and
that can explain the poor agreement in some points, unfortunately the points with the poorest
agreement are in the low load region.

2.6.8 Improved model using interpolation and extrapolation techniques

Despite the great effort of searching for a proper and general model for the turbocharger an
exact fit of the map data is not possible as seen earlier. There will be better fit in some
operating regions than in others. To overcome some of this error a combination of
interpolation techniques and modeling is tested. Given the corrected speed value, CSP, the
speed-isoline at this speed is first determined by finding the closest isolines from the map at
speeds over, CSP2 and below, CSP1 the actual speed. These are fit by 3´rd degree
polynomials:
πC CSP1
= p3 (CMF ) πC CSP 2
= p3 (CMF )
Eq.85
ηC CSP1
= p3 (CMF ) ηC CSP 2
= p3 (CMF )

42
The coefficients for each isoline are then weighed together against speed according to a
factor:
CSP − CSP1
x= Eq.86
CSP2 − CSP1
Once the new isoline is determined, the problem is to select the correct root of the
polynomial. Here it is assumed that it is the maximum root that is the correct one, which
makes it important that the isoline really decreases at higher mass flows.
In order to handle surge operation it is assumed that the point with lowest mass flow in each
isoline corresponds to the surge point. This point is weight against CSP in the same manner as
the coefficients. (Surging is treated as previous model as seen in Figure 22.) Given the new
mass flow, the efficiency can be calculated directly.
When moving into the low load regions (indicated by question marks in Figure 21) the isoline
with the lowest speed is shifted parabolic in the y direction and linear in the x direction by
moving the surge point according to:
 CSP
CMFS = CMFS ,1 ⋅
 CSP1
 π ,1 − 1
π S = 1 + CS ⋅ CSP CS = S 2
2
Eq.87
 CSP1
 η − ηC ,min
ηC , S = ηC ,min + CSP ⋅ C ,S ,1
 CSP1
The isolines (pressure ratio and efficiency) is moved and interpolated again to a 3’rd degree
polynomial that is used in the same way as the interpolated line. A figure explaining the
interpolation and extrapolation scheme is shown for an arbitrary compressor in Figure 24.
2.2 Map data 0.8
Surge point
2 0.7

Interpolation 0.6
1.8
0.5 Interpolation
1.6
ηC
πC

0.4
1.4 0.3 Extrapolation

1.2 0.2
Extrapolation
0.1 Map data
1
0
0 100 200 300 400 500 0 100 200 300 400 500
CMF [g/s] CMF [g/s]

Figure 24 Improved compressor model. Interpolation and extrapolation of mass flow (left)
and efficiency (right). The isolines with the two lowest speeds of the map data are shown.
The turbine model is improved in a similar way but it is more straightforward to use. Given a
certain pressure ratio the isolines in pressure ratio above and below the actual isoline is
determined and these isolines are fitted to 2’nd degree polynomials (turbine data is usually
less complex than for compressors):
MFP π T ,1 = p2 ( SPP ) MFP π T ,2 = p2 ( SPP )
Eq.88
ηT π T ,1
= p2 ( SPP ) ηT π T ,2
= p2 ( SPP )

43
These isolines are then weighed together according to a factor x:
π − π T ,1
x= Eq.89
π T ,2 − π T ,1
The speed the directly give the mass flow. If extrapolation is required, the two isolines with
the lowest pressure ratio are used to calculate mass flow at zero speed (SPP = 0), MFP01 and
MFP02 respectively. Together with the point were the turbine is in rest, (SPP = 0, MFP = 0, πT
=1), a 2’nd degree polynomial can be fitted for zero speed. Given the actual pressure ratio a
new mass flow at zero speed can be calculated, MFP0 . The isoline with the lowest pressure
ratio is then moved in the y-direction according to this value. The efficiency for this isoline is
shifted in a linear manner simply by scaling against the pressure ratio. It is assumed that the
efficiency is corresponding to a small value (ηT,min = 1%) when the turbine is at rest. The
geometrical interpretation in the performance map is shown in Figure 25.
0.04 0.8
Map data Map data
0.75
MFP [(g*sqrt(K))/(s*Pa)]

Interpolation
0.035 0.7
0.65
T

0.03 0.6
η

0.55
Extrapolation
0.025 0.5
0.45 Interpolation
Extrapolation
0.02 0.4
0 10 20 30 40 50 0 10 20 30 40 50
SPP [1/(s*sqrt(K))] SPP [1/(s*sqrt(K))]

Figure 25 Interpolation and extrapolation of the turbine performance map, mass flow
parameter (left) and efficiency (right).
This improved model show much better agreement to the map data due to the interpolation
technique, as seen in Figure 26 and Figure 27. There is still some points in the compressor
map that has a quite high error, it should be pointed out that it will be difficult to predict the
mass flow close to the surge line; a small deviation in pressure ratio give a big change in mass
flow.
Compressor Turbine
0.04
4.5 Map
Model 0.038
4
0.036
3.5
estMFP

0.034
3
π

0.032
2.5
0.03
2
Map
1.5 0.028
Model
1 0.026
0 100 200 300 400 500 600 10 20 30 40 50 60 70
CMF [g/s] SPP

Figure 26 Improved model, fit to performance data

44
 Compressor Compressor
15 3

10 2
CMF error [%]

EFF error [%]

1
5
0
0
−1
−5 −2

−10 −3
100 200 300 400 500 600 700 100 200 300 400 500 600 700
CMF [g/s] CMF [g/s]

Turbine Turbine
1 1

0.8
0.5
0.6
MFP error [%]

EFF error [%]

0.4
0
0.2

0
−0.5
−0.2

−0.4 −1
10 20 30 40 50 60 70 10 20 30 40 50 60 70
SPP SPP

Figure 27 Improved turbo model, fit to performance maps. Mass flow – left, efficiency – right,
compressor – top and turbine – below.

2.6.9 Governing equations

In the system calculation the mass flow and efficiency is determined from pressure ratio and
speed. The mass flows and efficiencies is then used to calculate the power transformed in the
turbine and compressor:
WC = − nC ⋅ (h2 − h1 )
Eq.90
WT = nT ⋅ (h3 − h4 )
The resulting (accelerating) work on the turbocharger axis is then calculated and it results in a
changed kinetic energy and hence a rotating speed:
∂W = (ηmech . ⋅ WT + WC ) ⋅ ∂t
2 ⋅ ∂W Eq.91
ω 2 = ω12 +
J TC
The calculation proceeds as follows:
1. Given a pressure ratio and a speed the mass flow and efficiency is determined, either from
interpolation in maps or through a model.
2. The pressure ratio and the temperature before the compressor/turbine are used to calculate
isentropic temperature after the compressor/turbine, which also gives the isentropic
enthalpy.

45
3. The efficiency then gives the real enthalpy and together with the mass flow the work
needed/developed in the compressor/turbine is calculated.
4. The difference in work between compressor and turbine then gives an accelerating work,
torque or power.
5. Depending on the inertia of the turbo-shaft, the new rotating speed can be calculated.
With the numerical method described earlier, it is appropriate to use mean state values of
pressures, temperatures, sped and mass flow over one time-step.

2.6.10 Comments
Despite the good behavior of the turbocharger models, the low-mass flow, or low load, region
there has to be some caution when using the models for the turbocharger. Especially since it is
not clear what happens with the efficiency when extrapolating in this region. The more the
maps are covering; the better the model become. The low load region should also be given
special attention because it set the possibilities to reach higher loads. Depending on
turbocharger performance and what is possible to achieve in exhaust temperature and mass
from the engine, the performance in the low load region determine if the speed and thereby
the boost pressure become higher or not. It is like a threshold. In some engines or
applications, the engine operating range is limited, then the mass flow and energy of the
exhaust gas is limited, and perhaps not sufficient to accelerate the turbo so that boost pressure
increases. An example of this can be found in the experimental work by Olsson [68], where
the air-fuel ratio cannot be reduced due to fast combustion and thereby limits the exhaust
temperature. The inlet pressure does not exceed 1.05 bar in this case.
The reason why the models are not tested on a variety of compressors and turbines is that
there is very few of these are distributed freely. The manufactures are not willing to give
everyone free access to these, probably due to several reasons. The major one is probably to
protect the manufactures own consultant activity.
In this work, constant-pressure turbocharging is considered, compared to the pulse-loading
turbocharging. The geometrical difference is shown schematically in Figure 28. In constant
pressure turbo-charging, an exhaust manifold with significant volume is inserted between the
engine and turbine collecting the exhaust gas from all cylinders. It has one exit to the turbine
that then will work with a constant upstream pressure (constant pressure ratio). In the pulse
loading system, this volume is very small, the turbine is mounted very close to the exhaust
port so that the pulse from blow-down is directly transferred to the turbine. It becomes
important to design the manifold properly on a multi-cylinder engine; turbines with two
entries are often used on 6-cylinder engines with pulse loading system.
Watson and Janota [109] and Heywood [96] state that there will be a loss of kinetic energy
when high velocity gas from the blow-down, mixes with low velocity gas in a constant-
pressure system. However, the kinetic energy from the pulses are not lost completely, this
energy must be, at least to some extent, transformed via turbulence to heat, increasing the
overall temperature in the exhaust manifold, providing more energy to be transformed in the
turbine. The part that remains in turbulence through the turbine is the real loss. In pulse-
loading turbocharging, the high kinetic energy of the blowdown pulse is directly transferred to
the turbine. The idea is that the high pressure at EVO is transferred to the turbine through a
small exhaust manifold with very small losses. However, it is not clear that a pulsating but
higher energy level is of benefit; the turbine and turbine entry must be designed properly to
handle this flow.

46
There are two other good reasons why pulse-loading system is better. The response of the
system is faster, because of the fast acting of each pulse into the turbine (compare with
pressure wave supercharging, the Comprex charger, with even faster response). The heat
transfer in the exhaust manifold is probably smaller due to the smaller volume and surface
area of the exhaust. However, these advantages depend on the success of the design of the
turbocharger system, how well the turbine is designed for this pulsating flow from a multi-
cylinder engine and how well the exhaust pipes are designed with respect of the number of
cylinders and firing order.
If a pulse-loading system should be simulated, the dynamics of the exhaust pipes should be
handled as well. It also requires that the performance data for the turbine is determined under
similar pulsating flow. This is often not the case. The constant-pressure model is used here
regardless of what the real system is, due to its simplicity and that the turbine performance
data is determined under steady state flow, but it is important to remember the limitations of
this approach.
Turbocharger related phenomena like surging and choking is quite difficult to correctly
handle in this type of simulations. A simple method of handling surging operation is
employed, however far from validated. Choking occur when the velocities become sonic, then
the mass flow will more or less depend on the pressure ratio only, choking are normally
treated in the performance maps. Compressor choking is not treated at all.

Figure 28 Constant pressure turbocharging (left) and pulse-loading system (right). From
Watson and Janota [109].

2.7 Manifold models

The inlet and exhaust manifolds are volumes where different components are connected.
These volumes provide pressures, temperatures and mixtures for the other components e.g.
the inlet manifold has a certain pressure that affects the mass flows trough the compressor and
engine. The manifolds are assumed well-stirred volumes and the thermodynamics of the
manifolds are assumed to follow the first law and the ideal gas law according to the general
assumptions described earlier. The modeling determines a new pressure and temperature after
a period (time-step), given the mass flow to and from the volumes. Both the mixture and
temperature of the flowing media has to be considered. With the numerical method used here,
the mass flow trough other components are calculated under mean value assumptions and it
becomes the role of the manifold model to provide the points that are used for these mean
values.

47
Eq.44 become:

( )
∂Q T , p,... + ∑ ∂nIN ⋅ h IN + ∑ ∂nOUT ⋅ hOUT ≈ H (Ti +1 ) − H (Ti ) − V ⋅ ( pi +1 − pi ) Eq.92

Again, the new enthalpy is calculated and the ideal gas law Eq.18 is then used to calculate the
pressure in the (i+1) point.
If the numerical method described earlier for solving differential equations is used, then some
numerical damping of the calculations are often required in order to obtain convergence. The
easiest way to this is only allowing, for example, 1/10 of change in manifold pressure and
temperature, compared to what the direct calculation give. This damping does not affect the
results since the system is calculated in an iterative way, however it increases computational
time. This damping should also be set with respect of what time resolution that is chosen. This
affects the resolution in time for the simulations. Higher resolution (smaller steps in time)
requires smaller damping in general. If dynamic processes are of interest, then of course, the
resolution should be set quite high. A good start value to test with is 0.1s steps, which is
sufficient if some dynamics should be treated as well, another more general solution is to vary
the resolution depending on the number of iterations needed.

2.7.1 Inlet manifold

Into the inlet manifold flows in general pure air from e.g. compressor, fuel from fuel injectors
and EGR from exhaust via an EGR valve. In this work, it is assumed that all liquid fuel
vaporization takes place in the inlet manifold if port fuel injection is used. Heat transfer is
neglected but it is quite easy to insert heat transfer if needed (e.g. as shown in Exhaust
manifold below). The cooling effect of vaporizing fuel is simply calculated as:
∂Q = − n fuel ⋅ M w , fuel ⋅ Qvap , fuel ⋅ ∂t Eq.93

2.7.2 Exhaust manifold

The exhaust manifold connects the exhaust ports of the engine (see section 2.5.3) with turbine
and waste gate valve, if one is used. In addition, the EGR gases are taken from the exhaust
manifold. A heat transfer model for the exhaust manifold is used developed by Malchow,
Sorenson, and Buckius (1976, as reviewed by Ferguson [101]) and the basic equation for this
model is:
Nu = 0.0483 Re0.783 Eq.94
This model is actually derived for straight tubes with D/L = 0.3 only but are used here as a
general model anyway in order to limit input data. A constant convective heat transfer
coefficient is applied on the outside of the manifold and is set to 15W/(m2K). Mean values of
mass flow are used when calculating Reynolds number. The surface areas of exhaust ports
and manifolds are set to a value that seems to represent the real engine. The temperature
distribution over the exhaust manifold wall is illustrated in Figure 29. The temperatures both
inside and outside the manifold walls and the heat flux are calculated under the assumption
that the manifold wall is described as a plate. The heat transfer relation's ends up with a
system of equations with three unknowns, which is quite easy to solve.

48
 Temperature twall

α1 λ
Texh.
Surrounding
.
Exhaust q
manifold Manifold
α2
wall
Tsurr.

Figure 29 Schematic temperature distribution over exhaust manifold wall

2.8 Charge air cooler (Intercooler) model

A model for charge air coolers (CAC) is developed, however verified against very limited
data. The model is based on weighing two different models, first the simplest possible relation
for an optimum, ideal heat exchanger:
T2,CAC ,ideal = T0,CAC Eq.95
T2,CAC is the exit temperature of the air that is to be cooled and T0,CAC is the coolant
temperature (surrounding air temperature) before entering the coolant side of the charge air
cooler.
The other model is based on a residence-time view, the longer the gas stays in the CAC, the
closer it comes to the CAC coolant temperature. Consider a portion of air, a virtual control
volume or its mole representation, n [mol], equal to the flow times the residence time, τ [s], as
illustrated in Figure 30.

T(t) (α⋅A)
.
n=n⋅τ

.
Q

T0,CAC

Figure 30 Virtual control volume for residence time dependant CAC model
Assume also that this volume has a certain heat transfer coefficient (αA) [W/K], which is
determined for different charge air coolers, by fitting the model to data. Heat balance for the
control volume then gives a differential equation for the temperature development during the
residence time:
−T ⋅ C p ⋅ n ⋅ τ = (α ⋅ A) ⋅ (T (t ) − T0,CAC ) Eq.96

49
This can easily be solved with the boundary conditions of inlet and exit temperatures,
corresponding to t = 0 and t = τ respectively, and with steady state assumptions, resulting the
following expression for exit temperature:
 (α ⋅ A) 
T2,CAC ,τ = T0,CAC + (T1,CAC − T0,CAC ) ⋅ exp  −
 C ⋅ n 
Eq.97
 p 
The two expressions Eq.95 and Eq.97 are weighed together with a constant, K, and the
resulting CAC model become:
T2,CAC = K ⋅ T2,CAC ,ideal + (1 − K ) ⋅ T2,CAC ,τ
 (α ⋅ A)  Eq.98
T2,CAC = T0,CAC + (1 − K ) ⋅ (T1,CAC − T0,CAC ) ⋅ exp  − JJG 
 C ⋅ n 
 p 
When the mass flow is very large, then Eq.98 becomes:
T2,CAC ≈ T1,CAC − K ⋅ (T1,CAC − T0,CAC ) Eq.99

This relation indicates that K can be seen upon as minimum heat exchanger efficiency at high
flows. When the flow is very small, then Eq.98 become close to the ideal heat exchanger:
T2,CAC ,ideal ≈ T0,CAC Eq.100

An example where K and (αA) is fitted to measured CAC data are shown in Figure 31 where
also the pressure drop are fitted against corrected mass flow, representing a very simple
model for the pressure drop over the charge air cooler.

7
340
Data
Data 6 Model 19.3 [kPa/(kg/s)]
335 Model

330 5

325 4
∆p [kPa]
T2CAC [K]

320
3
315
2
310

1
305

300 0
4 6 8 10 12 14 16 18 20 0 0.05 0.1 0.15 0.2 0.25
Flow [mol/s] CMF [kg/s] @ 298K and 1.013bar

Figure 31 Example of CAC model in comparison with measured data, note that temperature
varies before the charge air cooler (T1,CAC.). Measurements are made on a charge air cooler
for a SCANIA DSC12 diesel engine.

50
2.9 Humidifier
In order to simulate the Humid Air Motor concept a humidifier has to be modeled. The
humidifier adds water to the inlet air after the compressor. This air has an elevated pressure
and temperature so it has a capability to can carry substantive amount of water vapor (steam).
Liquid water is pumped to the top of a humidification tower as illustrated in Figure 32, before
it enters the humidifier it is heated (if necessary) through a heat exchanger, heat can be taken
from either the coolant system of the engine or the engine exhaust. The purpose of this
preheating is to get above the saturation temperature in the humidifier entry. Depending on
this heating, the temperature and water content will change in the humidifier exit. The
pumping work for the liquid water become more or less insignificant compared to engine
delivered power. In the humidifier, the air is cooled down and the water content increases due
to the vaporization. Some of the liquid water passes down through the humidifier (against the
direction of the airflow) is not vaporized and is collected in the bottom of the humidifier. This
water is then recirculated. The air after the humidifier will have a high relative humidity, close
to 100%.
The purpose of the humidifier model is to predict the water content and the temperature of the
air leaving the humidifier. The required input are the airflow, temperature and composition in
the entry, the liquid water temperature and the pressure (assumed uniform in the humidifier).
Further more, an estimation of the relative humidity on the humidifier exit is required, i.e. an
estimation of how “good” the humidifier is.
The humidifier layout is simplified a little bit, as shown in Figure 32, by removing the heat
exchanger and use only a single entry for the water i.e. the water flow correspond to the
vaporization rate. This gives maximum cooling of the air. If circulation and higher water flow
is used, like in the original design, the temperature on the exit will increase corresponding to a
heating of the media, and more water will be vaporized due to this temperature increase.

Warm and Cooled and T2 T2,HAM

dry air humidified air

Humidifier

Heat TH2O
m H 2O ,vap

H2O

Figure 32 Schematic layout of the humidifier (left) and simplified representation (right)
It is quite clear that an iterative approach is required since both the water content and
temperature is to be predicted, and they are dependent on each other. The thermodynamic
properties are again calculated with the NASA-polynomials. The data for liquid water is only
valid up to 500K, but this limit is above what we expect from the compressor (about 450K at

51
2 bar boost pressure). In addition, the liquid – vapor saturation curve is required. The Antoine
equation used to calculate the vapor pressure given the temperature. The parameters,
coefficients A, B and C, are taken from data reviewed by NIST Chemistry Webbook [110]
(selection: Bridgeman and Aldrich, 1964).
 B 
log10 ( psat . ) = A −   Eq.101
T +C 
The water molar fraction is calculated trough:
psat .
x H 2O = ⋅ϕ Eq.102
p
The calculation procedure for the humidifier follows:
1. Determine the enthalpies for the air in the humidifier entry and the liquid water. Note
that the enthalpy representation must include heat of vaporization.
2. Guess an exit temperature, T2,HAM (previously value)
3. From the assumed relative humidity, ϕ, pressure, p, and (guessed) temperature, in the
exit of the humidifier, calculate the water content through Eq.101 and Eq.102.
4. The mass flow of air and calculated water content on the exit gives the water
vaporization rate (mass balance) and the exit mixture.
5. Given the enthalpy and mass flow of air and the enthalpy and vaporization rate of the
liquid water, the enthalpy of the humidified air can be calculated (energy balance).
6. The mixture and enthalpy of the humidified air give a new exit temperature through
Eq.24.
7. The steps 3 to 6 are iterated until stable exit temperature. Numerical damping is used,
the newly calculated temperature is not allowed to change too much after each
iteration (does not affect the results).
After this iterative process the temperatures, mass flows and gas mixtures are known for the
humidifier.

52
2.10 Other components

2.10.1 Heater
When simulating HCCI engine system it is sometimes necessary to include an inlet air heater
in the system. This can be done by inserting a heat-source and a pressure drop. The pressure
drop can be modeled as in the charge air cooler described earlier. Input to this model would
then be the mass flow and the heating power. The exit temperature, T2,HTR is calculated from
the exit enthalpy as shown in Eq.103.
Q HTR
h (T2,HTR ) = h (T1, HTR ) + [J mol ] Eq.103
n
Q HTR [W] is the power of the heater (electrical power) and n is the flow [mol/s].

2.10.2 Wastegate
A waste gate valve is a component used for controlling boost pressure by affecting the
pressure ratio over the turbine. The waste gate is simply modeled as a restriction with variable
throat area and a constant discharge coefficient. The poppet-valve geometry (Figure 13) can
be used and the mass flow is calculated trough equations Eq.50. Input to the model is valve
opening (valve lift) and pressure ratio.

2.10.3 Engine control

This part is the least developed part in this work and is now only used to calculate fuel and
EGR flow from desired values. I addition various heat release models are adjusted here e.g.
when direct injected diesel engine are simulated, then the fuel injection duration has to be set.

53
3 Simple heat release

In this first step a simple heat release function (see sections 2.5.1 and 2.5.6) was adopted and
fitted to measurements made on a single cylinder and on a 6-cylinder HCCI engine. The
purpose was to validate the simulation software.

3.1 Validation of HCCI Cycle Simulation

This validation is a first step that is used to see if the cycle simulation can fit measured cycles.
In this case, measurements on a single cylinder TD100 engine were used as reference.
Comparison is made on pressure trace, and resulting parameters like exhaust temperature,
indicated mean effective pressure and efficiency. Regarding measured pressure traces it must
be pointed out that the measurements gave a relative pressure, which means that the pressure
trace has to be calibrated so that if the measurement were made at atmospheric pressure the
reading should be 1.023 bar. However, the pressure transducer has very good linearity, which
means that indicated work could be calculated directly without knowing the absolute pressure
since:
WC = L∫ pabs. ⋅ dV = L∫ ( p rel . + pref . ) ⋅ dV =
cycle cycle
Eq.104
L∫ prel . ⋅ dV + L∫ pref . ⋅ dV = L∫ prel . ⋅ dV + 0
cycle cycle cycle

For comparison it is much more desirable if the absolute pressure is known but this is a little
bit tricky to do, at least if a good agreement with reality is desired. There are several ways to
do this:
• Assume a pressure at a certain CAD
• Assume an average intake stroke pressure
• From “simulation” of a part of compression stroke with a assumed γ = Cp/Cv value
Here an average intake pressure, determined from transducer measurements (absolute
pressure) in the intake manifold is assumed between 20 and 160 CAD ATDC during intake
stroke. Mathematical expressed:
pabs . = prel . − prel . 20−160CAD + pinlet Eq.105
The inlet pressure in the simulations are shifted a little bit to get a good agreement at the
beginning of the compression stroke.
Below is a comparison of pressure trace expressed as pressure – CAD diagrams. As input to
the simulation the heat release determined from measured traces are used through angles for
0, 10 and 90% burned fraction. The 0% burned fraction is estimated through angle for 1%,
minus 1 CAD (the angle for 0% burned fraction is very difficult to determine
experimentally). In addition, the combustion efficiency determined from exhaust gas

54
emission analysis is also used as input to the simulation. A question mark is the heat transfer
and compression ratio. From analyzing pressure traces without firing the engine, the
compression ratio was determined to be about 17.7:1 which is used in the simulations.
Engine geometry is shown in Table 5. In this case, the heat transfer correlation of Woschni
[104] is used straight off without any compensation. There seems to be only a small
difference indicating that the compression ratio and heat transfer is properly estimated.
2
10 1.4
Experiment Experiment
Simulation Simulation
1.3

1.2
Pressure [bar]

Pressure [bar]
1
10 1.1

0.9
0
10
−1 0
0.8
10 10 0 0.5 1 1.5
Volume [dm3] Volume [dm3]

Figure 33 Comparison between simulation and measurement. Left - In-cylinder pressure

against volume in log-log scale. Right – Gas exchange linear scale, note especially the
difference during the intake stroke. Volvo TD-100 engine with HCCI combustion at approx.
1000rpm, λ = 3 with isooctane.

80
Experiment
70
Simulation
60
Pressure [bar]

50

40

30
20

10
0
−90 −45 0 45 90
CAD

Figure 34 Comparison between measured and simulated in-cylinder pressure during late
compression, combustion and early expansion.
Very good fit can be seen during compression and expansion stroke but the agreement is less
during intake and exhaust stroke which is not that surprising since there is nothing in the
simulated model that handles flow dynamics inside and outside the cylinder. Adding
dynamics to the model would probably give much better agreement, but also longer
computational time. It is always a compromise between levels of detail in the model and
execution time; if more detail is to be considered it usually also requires more computational
time.

55
It is also interesting to look at resulting parameters like indicated efficiency and indicated
mean effective pressure (Table 6). Both indicated pressure, and especially efficiency, is
overestimated. The difference in efficiency can be explained by the lower pump work
required in the simulations despite the lower inlet pressure. The pressure inside the cylinder
during the intake stroke is lower in the real engine compared to the simulated one. During the
exhaust stroke there is no clear difference; the flow produces pressure oscillations in the
experiments as shown in Figure 33. In summary, the difference is quite small and mostly
dependant on the poor intake and exhaust stroke prediction.
Table 5 Engine geometry for the Volvo TD-100 engine used for HCCI experiments in Lund.
The reference point (0°) for valve timings (zero lift) refers to combustion-TDC.
Bore 120.65mm Camshaft NG-SI
Stroke 140mm EVC -350
Conrod length 260mm IVC -140
Compression varied EVO 120
Cylinders 1 IVO 340

Table 6 Parameters and results

Input: Measurement Simulation Results: Measurement Simulation
rC 17.7:1 17.4:1 pinlet [bar] 0.989 0.970
Speed [rpm] 994 994 IMEPnet [bar] 3.201 3.226
λ 3.04 3.04 IMEPgross [bar] 3.395 3.355
Tinlet [K] 369 369 PMEP [bar] 0.194 0.130
pexhaust [bar] 1.015 1.015 ηi,net 37.5% 38.9%
ηc 91% 91% ηi,gross 39.8% 40.5%

3.2 Validation of System Simulation – Diesel operation

A direct injected, six-cylinder SCANIA DSC12 turbocharged diesel engine with a displaced
volume of 11.7 liter is simulated under varying speed and load (see Figure 35). In the
simulations, speed and fuel injection time are adjusted linear during 4 seconds, then kept
constant during 8 seconds and so on. The results from the simulations are then compared to
stationary measurements on a real engine. Engine speed is set to be equal to the real engine
and the fuel injection time is set to result in a fuel flow equal to the fuel flow measurements.
The simulation is made without any knowledge about the specifications of the real engine
regarding injection strategy (injection timing is set to 10°BTDC in the simulations) or fuel
injector characteristics (fuel rate is set to 30g/s per injector). Another difficulty is absence of
heat release knowledge (the heat release factor kHR is set to 300). However, some simulations
were made first to set these parameters to realistic values, before presenting the results here.
The real engine uses pulse-loading turbocharging, which has not been possible to simulate
correctly, due to model and data limitations, as discussed earlier. Instead, a constant pressure
system is simulated.
In Figure 36 IMEPnet and maximum pressure are shown. IMEPnet shows quite good
agreement, whereas the maximum pressure has become too high, which probably can be
explained by somewhat early or fast heat release.

56
In the last point (at 72s), the boost pressure becomes too low, which can be a result of the
difference in turbocharging technique, constant-pressure turbocharging in the simulations and
pulse-loading turbocharging in the real engine. In this low speed, high load regime the pulse-
loading technique probably results in higher turbine efficiency and thus higher boost pressure.

Engine speed Engine λ

2500 3
SIM
2.5 MEAS
2000
2
[rpm]

1500 1.5

1
1000
0.5

500 0
0 20 40 60 80 0 20 40 60 80 100
Time [s] Time [s]

Figure 35 Engine speed (left) and calculated air-fuel equivalence ratio, λ (right). Note the
good agreement in λ, indicating correct air and total mass flow to the engine. Notation: MEA
(crosses)- experiments, and SIM (lines)- simulations.
IMEPNET
25 200
SIM
20 MEA

15 150
[bar]

max

10 SIM
p

MEAS
5 100

−5 50
0 20 40 60 80 0 20 40 60 80
Time [s] Time [s]

Figure 36 IMEPnet (left) and maximum cylinder pressure (right).

TC Temperatures TC Pressures
1000 3.5
T1
800 3
T
2
T p
3 2.5 1
600 T p
4 2
[bar]
[°C]

T MEA 2 p
2 3
400 T MEA p
3 4
T MEA 1.5 p MEA
4 2
p MEA
200 1
3

0 0.5
0 20 40 60 80 0 20 40 60 80
Time [s] Time [s]

Figure 37 Temperatures (left) and pressures (right) around the turbocharger.

Notation: 1- before compressor, 2- after compressor, 3- before turbine and 4- after turbine.
In figure Figure 37, the temperatures and pressures before and after the turbocharger are
shown. The temperatures show quite good agreement in the beginning of the simulation (high
engine speeds). Later in the simulations, the temperature after the turbine becomes much

57
lower, and in the last point, the boost pressure becomes too low resulting in lower airflow into
the engine and higher exhaust temperature. Please note that the measured pressures are static.
Total pressure is a little higher depending on where the static pressures are measured. In the
simulations, all pressures are total pressures. The difference in pressures cannot fully be
explained by the difference between static and total pressure, especially on the inlet side (p2),
which is quite surprising since temperatures, air-fuel ratio and IMEP shows better agreement.

3.3 Validation of System Simulation – HCCI operation

The experimental work and the validation below were published at the SAE Fuel and
Lubricants Conference in Baltimore 2000 [68]. The primary purpose was to demonstrate a
working full-size multi-cylinder HCCI engine and to verify the calculations in the simulation
software (ESIM). The experimental efforts were made by Jan-Ola Olsson and were sponsored
through other channels.
A 6-cylinder turbocharged diesel engine (SCANIA DSC12, normally used in trucks) was
modified to be operated as a HCCI engine by removing the diesel injectors and inserting low-
pressure fuel injectors in the inlet manifold. The engine had 4 valves per cylinder so each
cylinder had 2 inlet ports in which a fuel injector was mounted. Since each cylinder was fed
with 2 fuel injectors it was possible to use two different fuels. The engine control system then
handled the fuel injection individually for each cylinder and each fuel. All fuel injectors were
mounted in a spacer between the cylinder heads and the original inlet manifold. In addition,
an electrical inlet air heater (maximum power of 37kW) was used in some points to ensure
ignition. The system is schematically shown in Figure 38. Iso-octane and n-heptane were used
as fuel which makes it possible to run the engine on different research octane numbers, RON
between 0 (n-heptane only) and 100 (iso-octane only), the mass percentage of iso-octane
gives the RON number.
HEATERS

TURBO-
CHARGER

ENGINE
n-C7H16
i-C8H18

Figure 38 Engine system

Pressure transducers replaced the original fuel injectors in the cylinder heads, which gave an
opportunity to monitor cylinder pressure during operation. These in-cylinder pressure traces
were then evaluated by heat release analysis in order to find ignition timing and combustion
duration. No other changes were made to the engine i.e. pistons, cylinder head, engine block
were left unchanged. The properties of the engine are summarized in Table 7.

58
Validation was first made to find out how well cycle and system simulations fitted with
experiments. The simulation software was fitted with a simple heat release shape function,
Eq.106 originally developed by Wiebe (see section 2.5.1). This function gives an accumulated
heat release expressed as burned fraction. The total amount of energy released (100%)
depends on the amount of fuel present, the lower heating value of the fuel and the combustion
efficiency.

(
xb = 1 − exp a ′ (θ − θ 0 )
m′
) Eq.106

Table 7 Basic geometry of the SCANIA DSC 12 engine. Valve timings refer to zero lift.
Cylinders 6
Swept Volume, total 11 705 cm3
Compression Ratio 18:1
Bore 127 mm
Stroke 154 mm
Connection Rod 255 mm
Exhaust Valve Open 96° BBDC
Exhaust Valve Close 52° ATDC
Inlet Valve Open 54° BTDC
Inlet Valve Close 78° ABDC

This shape function was fitted to the experimental data from the heat release analysis,
however not individually but as an average for each cylinder. It is possible to do simulation
for each cylinder individually but that required longer computational time and also demands
better models of inlet and exhaust manifold in order to become valuable. The crank angles for
10, 50 and 90% burned fraction (xb) are used to fit the function and an example of such fit are
shown in Figure 39.
1 5
Model
A10 4.5
0.8 A50
A90 4

0.6 3.5
p2/p1
xb

3
0.4 2.5

2
0.2
1.5
?
0 1
−15 −10 −5 0 5 0 100 200 300 400 500 600 700
CAD CMF [g/s]

Figure 39 Left - Example of heat release fit at one operating point. Right - Compressor
performance map that illustrates the problem at low mass-flow and pressure ratios. The “x”
signs indicate isolines in turbine speed.
The engine was fitted with the original turbocharger (Holset HX50), which can be considered
as well matched for diesel engine operation. In HCCI operation, the turbine was found too
big. The compressor (map shown in Figure 39) was acceptable since the airflow should be
about the same as in a diesel operation.

59
The turbocharger is an important engine component that has a great influence of the
maximum power output of an engine. What it does is simply increasing the mass flow trough
the engine and thereby increases the power output. The performance maps of a specific
turbocharger relate pressure ratio, turbine speed, mass flow and isentropic efficiency, and
cubic interpolation from tabled maps are used. Some extrapolation is needed at low mass
flows and low-pressure ratios since the maps lack data in this region. This extrapolation
becomes quite difficult and has to be done very precautious since this region directly sets the
possibility to reach higher engine loads. It becomes even more important in the HCCI case
since rich mixtures cannot be used in order to increase exhaust temperature and hence energy
levels upstream turbine. It would be possible to gain information from the experiments in
order to fill this blank area in the maps but that requires measurements of the turbine speed,
which at the time of the experiments was unavailable. This turbocharger map-problem was
later dealt with, see following section. Here is the extrapolation done manually by adding
points in the unknown region resulting in a smooth interpolation “surface”. A similar
approach was used for the turbine.
Besides engine and turbocharger data and heat release shape some measured values was used
as input in order to compare simulations with measurements as seen in Table 8. The
surrounding temperature was set to 35°C, and the surrounding pressure (upstream compressor
and downstream turbine) to 1.023bar.
Table 8 Input data for the simulation, PH – heating power, ηc – combustion efficiency.
Case Speed λ RON PH A10 A50 A90 ηc
[rpm] [kW] [CAD] [%]
1 1510 5,82 0 0,0 -20,8 -9,3 -0,6 77,7
1510 5,16 19 6,0 -20,9 -10,8 -3,9 94,1
1502 3,78 69 0,0 -7,0 -3,5 -0,6 95,2
1495 3,26 89 0,0 1,8 4,6 7,1 94,8
1477 2,87 93 0,0 5,9 9,1 12,8 94,2
1501 2,34 94 0,0 4,1 5,9 15,1 96,7
2 1991 3,51 86 0,0 1,6 5,4 8,9 90,2
3 1202 2,20 98 0,0 8,1 10,3 16,3 96,2

Simulated net indicated mean effective pressure, IMEP and exhaust manifold temperature
shows very good agreement with measurements as shown in Figure 40. IMEP depends
heavily on the heat release, both timing and the amount (accumulated) heat released, and in-
cylinder heat transfer. The timing and shape is set to correspond to measurements and hence
the good agreement should indicate a proper heat transfer and fuel flow, the latter gives the
accumulated heat release. The exhaust temperature depend mostly on air-fuel ratio, heat
transfer in-cylinder and heat transfer in the exhaust port (the thermocouples were mounted in
the exhaust manifold close to the cylinder head). Again, air fuel ratio was set to the same as
measured (from exhaust emissions) value so the good agreement should indicate that the
valve-flow and heat transfer models (at least together) seem to simulate the reality
(represented here by measurements) fairly well.
There is a much bigger difference if we look at the inlet and exhaust pressures as seen in
Figure 41, and this difference is believed to be connected to the poor turbocharger model.
Another factor is that the simulated systems do not have any model for an air-cleaner, which
normally reduces the inlet pressure somewhat. The intake air heater has also some unknown

60
resistance that would cause a pressure drop on the intake side. Altogether, it is not surprising
that there is a difference in intake pressure. The same goes for the exhaust side, although in
this case it’s probably the model alone that has to be blamed. A possible contribution can be
that the heavy oscillating pressure due to the blow down pulses may cause inaccurate pressure
registrations. The exhaust pressure transducers were mounted on approx 0.5 m long tubes
with 4mm bore directly on the two exhaust channels (from cylinder 1-3 and 4-6 respectively)
entering the turbine, which should give a reading of total pressure. The average pressure
reading of the two channels is shown here.
7 400

6 350
Net IMEP [bar]

5 300

Simulated
4 250

3 Case1 200
Case2
Case3
2 Measured 150

1 100
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 100 150 200 250 300 350 400
φ = 1/λ Texhaust [°C] − Measured

Figure 40 Left – Net indicated mean effective pressure. The measured points (dots) are
surrounded by bars that indicate maximum and minimum of the 6 cylinders. Comparison
between measurements and simulations is possible by looking at a specific φ (=1/λ) value for
which two points exist, one measured and one simulated. Right – Exhaust temperature
comparison between experiment and simulation
1.06 1.2
Case1 Case1
1.05 Case2 Case2
Case3 Case3
Exhaust pressure [bar]

Measured 1.15 Measured

Inlet pressure [bar]

1.04

1.03
1.1
1.02

1.01
1.05
1

0.99 1
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
φ = 1/λ φ = 1/λ

Figure 41 Inlet pressure (left) and exhaust pressure (right) over fuel-air equivalence ratio.
The inlet temperatures (Figure 42) are over estimated about 5 – 10 degrees, which can be
explained by the absence of a heat transfer model for the inlet manifold, the cylinder heads
transfer heat to the intake manifold and in its turn to the intake air. Deviations in inlet pressure
should render in a similar difference in mass flow as seen in Figure 43, but the measured mass
flow is higher than the simulated values, which is quite contradictable. A further analysis
reveals that the mass flow measurements probably somewhat erroneous since the volumetric
efficiency (Eq.107) become higher than 100% in several cases.
60 ⋅ m ⋅ R ⋅ Tinlet ⋅ nR
ηV = Eq.107
M A+ F ⋅ pinlet ⋅ Vd ⋅ n

61
 90
Case1
80 Case2
Case3

Inlet temperature [°C]

Measured
70

60

50

40

30
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
φ = 1/λ

Figure 42 Inlet temperature over fuel-air equivalence ratio

220 110
Case1
Case2 Case1
200
Case3 105 Case2
Measured Case3
Mass flow [g/s]

180 Measured
100
ηV [%]
160
95
140

90
120

100 85
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
φ = 1/λ φ = 1/λ

Figure 43 Total mass flow over fuel-air equivalence ratio (left) and volumetric efficiency over
fuel-air ratio (right).
Before this simulation tool is used to study different engine systems or to match
turbochargers, some improvements have to be made, e.g. regarding the heat release
calculations and extrapolation of the turbo maps. The heat release calculations are to be made
based on empirical knowledge or by looking into the chemical kinetics. The extrapolation of
the turbo maps has to be improved in order to handle e.g. start up situations where the original
maps are not sufficient. Other than that, the calculations seem to represent the measurements
quite well.

3.4 Turbocharger considerations

As mentioned in the previous section the turbocharger performance heavily affects the
performance of the engine system. By forcing higher mass flows through the system higher
engine loads is possible i.e. it becomes more economical since a smaller engine can produce
the same amount of power as a bigger one. In addition, frictional losses are not that load
dependant; higher brake output of an engine usually makes frictional losses less significant
and brake efficiency increase.
The requirements for increased load is that the mass flow and exhaust gas energy upstream
the turbine of the turbocharger is high enough to first accelerate the turbo shaft into regions
were the compressor becomes useful and then provide boost pressure. In spark-ignited (SI)
engines, this is usually not a problem since they can be operated with rich or stoichiometric
mixtures that give high exhaust temperature. In addition, the compression ratio of

62
turbocharged SI engines is around 9 to 10 in order to avoid knocking combustion and too high
peak pressures but this limits the efficiency of the engine so the exhaust temperature and
hence the energy levels always becomes high enough. Lean-burn SI stationary engines can
use a temporary richer mixture and late ignition in order to gain boost pressure and then lean
out to a desired operational level when boost is enough. The same goes for diesel engines
where the slow combustion process and low heat transfer also contribute to high exhaust
temperatures. In a HCCI engine, too rich mixtures cannot be used due to too fast knocking-
like combustion that gives heavy sound emissions and is believed too cause engine damage.
Because of this operational mixture limit of HCCI engines, it becomes important to simulate
the boost build-up period in order to find out if the engine can reach desired operating load.
Since the turbo performance maps has limited information in the low mass flow and low
pressure ratio regions a turbocharger model with good extrapolation qualities in this region
are desired. As shown earlier (section 2.6), the turbocharger model used in this work has been
subject to stepwise improvement, in this case the models described in sections 2.6.3 - 2.6.6
are used which is basically a system of polynomials fitted to performance map data.
When matching the turbocharger to the engine the first step is to select a proper compressor
that will provide the desired boost pressure and mass flow to the engine in the desired engine
operating speed. It also is important to minimize the necessary compressor work, which is
essentially the same as maximizing the efficiency of the compressor. For a truck size engine
radial compressor this efficiency is typically between 75 and 80% in the best point.
The size of the turbine determine the pressure ratio given a certain speed and mass flow. A
small turbine brings up the pressure ratio at lower speeds. The size of the turbine is usually
classified by A/R-ratio, which is the inlet area of the turbine divided by radius of turbine
housing. A basic turbine geometry or type can often be delivered with various A/R-ratios.
Without becoming an expert on turbo-charging, it is possible to model and simulate different
sizes of compressors and turbines and an example is used for the discussion here.
Three different turbine sizes were compared through simulation of a load sweep at 2000 rpm.
TC-1 corresponds to the diesel engine’s original turbine size (geometrical flow area: 25 cm2)
while the two smaller ones, TC-2 and TC-3 produce higher pressure ratios since they have
smaller flow area (16 and 13 cm2 respectively). A photo of two of the turbines is shown in
Figure 44. The simple heat release function was used with fixed parameters (A0 = -5, A10 = 0
and A90 = 10 CAD ATDC) and the results were compared against indicated mean effective
pressure and indicated efficiency i.e. engine friction was not considered. It was also assumed
that HCCI combustion was the case in each operating point, which perhaps is not the case in
the reality, at least not with a single fuel. Isooctane was used as a single substitute of a
mixture of n-heptane and isooctane.
The simulation was made so that the engine was started from idle speed with fuel injection
off, then during the first 24 seconds a 2-step ramp was used to bring up the speed to 2000 rpm
and a fuel injection corresponding to 10g/s into the inlet manifold. Thereafter followed a
reduction to 4g/s and then a stepwise increase in fuel rate ending at 16g/s as illustrated in
Figure 45. It is this last stepwise increase that is studied here. The first peak is to ensure boost
pressure buildup since there is a possibility that there can be two operating modes given a
certain fuel flow in some cases, especially low load. The varying fuel flow results in varying
load, boost pressure and hence air-fuel ratio and resulting efficiency. Air-fuel equivalence
ratio (Figure 45) depend heavily on how much boost pressure the turbocharger can provide
since this directly affects the airflow to the engine.

63
Figure 44 Photo (by J.-O. Olsson) showing the exit of the smallest (TC-3, left) and the largest
(TC-1, right) of the three turbines.
16 8
TC−1
14 7 TC−2
TC−3
12
6
Fuel flow [g/s]

10
5
8
λ

4
6
3
4
2 2

0 1
0 20 40 60 80 100 0 20 40 60 80 100
Time [s] Time [s]

Figure 45 Fuel rate for simulation (left) and resulting air-fuel ratio (right)
A smaller turbine forces the backpressure (exhaust manifold pressure) to increase due to
smaller flow area put it also bring up the pressure level (energy level) on the exhaust side
which means that more work for the compressor work can be delivered, resulting in higher
air-flow and higher air-fuel ratio. However, the increased backpressure also gives high
pumping losses on the engine itself, which reduces the overall efficiency of the engine. A too
large turbine on the other hand has difficulty to extract the exhaust energy but gives lower
pumping losses. A difficulty that is connected to the HCCI combustion is the λ-dependence
on heat release rate. If λ becomes to rich, a too fast combustion occur that bring fourth
oscillating pressure waves inside the cylinder. In spark-ignited engines, this is usually the
result of knocking i.e. self-ignition of the air-fuel mixture outside the flame-front. In a SI
engine, this can be devastating if the phenomenon prevails for a long time with structural
damage on the piston as result. The normal mixture in SI engines are stoichiometric which
means that the energy released per mass mixture is obviously higher than in a lean burn HCCI
engine so structural damage due to too fast combustion seems a less unlikely in a HCCI
engine. However, the noise-emission increases heavily. From experiments it is reported to be
“painful” to run the engine with richer mixtures than about λ = 2.1 (isooctane) and lets
assume that this is the limit for HCCI in this case then the resulting operating range with the
different chargers can be seen in Figure 46. Turbocharger 2 and 3 (medium and small size)
gives the highest mean effective pressures without getting lower than λ = 2.1.

64
 6 1.5
TC−1
TC−2
5 TC−3

PMEP [bar]
4
λ

3
0.5
TC−1
2 HCCI limit TC−2
TC−3

1 0
0 5 10 15 20 0 5 10 15 20
IMEP net [bar] IMEP net [bar]

Figure 46 Operating range at 2000 rpm, in simulation (left) and simulated pump mean
effective pressure (right)
As explained above a smaller turbine also brings up pump losses that reduces efficiency,
which is clearly seen in Figure 46. Now, indicated efficiency is not quite what is interesting in
a car, truck or power plant. The brake efficiency is more interesting and with the friction
model described earlier (section 2.5.7), we get lower efficiency (Figure 46) at with low loads.
Friction has a crucial effect at low loads as shown earlier in Figure 20.
50 45
Net Indicated Efficiency [%]

40
48
Brake Efficiency [%]

35
46
30
44
25
TC−1 TC−1
42 TC−2 TC−2
20 TC−3
TC−3
40 15
0 5 10 15 20 0 5 10 15
IMEP net [bar] BMEP [bar]

Figure 47 Net indicated - (left) and brake - (right) simulated efficiencies over mean effective
pressures at 2000rpm.
It is quite clear that it is the medium size turbine, TC-2 that should be selected in this case, it
seems like a good compromise between efficiency and performance. If more advanced turbo
technology were accepted, like the variable vane turbine, this may be a way to keep the larger
turbine without loosing to much pumping work at high loads. A variable vane turbine is a way
to extend the optimum operating range of a certain turbine by adjusting the flow field in the
turbine and hence optimize it for different mass-flows.

3.5 Lean-burn spark ignition

A small validation was made by simulating a whole engine system working with lean-burn
spark ignition and compare the results with experiments. The purpose was to illustrate what
we can expect from a lean burn natural gas engine, in terms of NOx emissions and efficiency,
and to check the validity of the system simulation software.

65
The experiment was made on a 10 liter Volvo TG103 natural gas engine which has basically
the same design as the TD100 engine mentioned earlier, however, it has a lower compression
ratio of 11.2. In the simulation the compression ratio of the DSC12 engine was reduced to
10:1, the medium size turbine (TC-2) was selected and a simple wastegate model was inserted
to control inlet pressure. A much longer combustion duration (25 CAD) was used to simulate
the SI combustion. Although the engine simulated was not the same as the experimental one
(due to lack of information about the turbocharger), the efficiency estimates were accurate as
shown in Figure 48. One explanation for the difference between simulated and experimental
results can be the friction model used in the simulations. At 1200 rpm the friction mean
effective pressure (FMEP) became 1.48 bar compared to about 1.2 bar for the experiments
(the mean value of net IMEP for the 6 cylinders was used to calculate FMEP). Note the high
NOx emissions, around 20g/kWh.

Spark Ignition, 1200rpm, Natural gas λ = 1.5

50

40

30

20
Sim. η [%]
b
10 Exp. η [%]
b
Exp. NOx [g/kWh]
0
10 11 12 13 14 15 16 17
BMEP [bar]

Figure 48 SI operation. Experimental results with the TG103 engine, compared with
simulations of a spark ignition (SI) converted DSC12 engine.
Lean burn spark ignited engines, especially those with 3-way catalysts, is the main
competitive solution to HCCI regarding both efficiency and NOx. Regardless of technology,
current and future engines must comply with the legislations. Table 9 shows some examples
of legislations for stationary natural gas engines in Europe (reviewed by Nellen and
Boulouchos [111]).

Table 9 Legislations is Germany and Switzerland for stationary natural gas engines.
Emissions in mg/Nm3 at 5% O2 (except for NMHC). 250 mg/Nm3 correspond to about 0.76
g/kWh for a lean burn SI engine [111].
Legislation NOx CO NMHC
TA-Luft 92, Germany 500 650 150
Switzerland 250 650 -
Zurich - city 50 650 -

66
3.6 The effect of ignition timing on efficiency
As in SI engines, the ignition timing in HCCI engines is possible to set to a desired value if a
suitable control system is used. The ignition timing affects not only the emissions but also the
efficiency of the engine. In the following simulation, the ignition timing is varied keeping
constant combustion duration (D0-10 = 8 CAD and D10-90 = 5 CAD) and combustion efficiency
(92%). The case correspond to the simulation in section 3.1 and the resulting indicated
efficiency is shown in Figure 49. Note that it is the angle of 0% burned mass fraction, A0 on
the X-axis. The angle of 50% burned mass fraction, A50, is about 8 + (8-5)/2 = 9.5 CAD later
than A0.
44 80
Indicated efficiency, ηi,n and ηi,g [%]

75
42

Maximum pressure [bar]

70

40 65
60
38 55
50
36 Gross
Net 45
34 40
−10 −5 0 5 10 −10 −5 0 5 10
A0 [CAD ATDC] A0 [CAD ATDC]

Figure 49 Ignition timing and its effect on indicated efficiency (left) and maximum pressure
(right). At A0 > 7 CAD ATDC, the maximum pressure is limited by the pressure after
compression.
With fixed combustion duration, a later ignition timing (from optimum efficiency) reduce the
maximum pressure and hence the indicated work from the cycle and the efficiency goes
down. An earlier timing brings up the maximum pressure and thereby the maximum
temperature. This increases the time at high temperatures and thus the heat transfer. A lot of
heat transfer in the beginning of the expansion stroke reduces the pressure during expansion
stroke and thereby the indicated work. When the indicated work goes down, the efficiency
goes down if the fuel flow is constant.
It appears that the indicated efficiency does not change that much with ignition timing, but it
is important to remember that in reality, there is an influence of combustion duration and heat
transfer, both factors change when changing ignition timing.

67
4 Single zone chemical kinetic
model

Since ignition in HCCI engines is affected by numerous parameters, a proper prediction of

ignition timing is required in order to simulate if, and when, ignition occur. Using chemical
kinetics calculations seems like a reasonable solution. Since the HCCI process is similar to
the knocking phenomena in SI engines, software developed for studying these phenomena
seems appropriate to use. Development of that type of software is made at several combustion
research groups around the world and among them is the group lead by professor Fabian
Mauss at the Combustion Physics Division, department of Physics at Lund Institute of
Technology.
Their software has been linked into the cycle calculation developed here. This link provides
an estimate of air-fuel-residual mixture, temperature and pressure at a fixed CAD during the
compression stroke. Then the chemical kinetics calculations take over, including the same
heat transfer model as for the cycle simulation described in section 2.5. The chemical kinetics
program calculates pressure and temperature in a desired CAD-interval, as long as the intake
and exhaust valves are closed.
However, the program has some limitations. In order to save computational time only one
zone is considered which means that the ignition process does not consider any distribution of
fuel or in-cylinder temperature. It has possibilities to do so but that is not used in this work, if
used, it brings up the computational time even more depending on the number of species and
reactions are used. Fuels like isooctane require longer time than fuels like methane since more
species and reactions are needed to describe the combustion process. Each reaction has a rate
that depends on pressure, temperature and specie concentration. All these reactions and the
thermo dynamical calculations can be expressed in a system of equations that has to be solved
for each time step and this is what the program does. Since the reactions depend exponential
against temperature (see the Ahrenius expression) and some reactions are very fast the system
becomes very stiff which puts large demands on the solver of the system. It all becomes even
more complex because negative concentrations cannot be allowed. Deeper description of this
program will not be made here since the author would like to identify himself as end-user of
this software.

4.1 Mathematical model

The mathematical model used for calculating the chemical reactions in the combustion
process is based upon a set of zero-dimensional time dependent differential equations: One
balance equation for the conservation of energy, and one mass conservation equation for each
chemical species in the reaction mechanism. The differential equations are solved using
Newton’s method and the time is resolved with higher order backward differential functions.
The main characteristics of the chemical kinetics program have been discussed in previous
publications [65], more detailed information is given by Amneus [112].

68
The conservation equation for the chemical species is given by Eq.108. Yj is the mass fraction
of the species, ρ is the density, Mj the molar mass, νj,k the stoichiometric coefficient and ωk
the reaction rate.
N
dY j M j r
= ∑ν j ,kω k Eq.108
dt ρ k =1
The energy conservation equation is given by Eq.109. The first term is the contribution from
chemical reactions, the second is the work, and the third term is the change in internal energy
of all mass within the engine cylinder. The fourth term is the convective heat transfer from
walls. m is the total mass, uj the specific internal energy of species j, p pressure, V volume, t
time, Cv the heat capacity and T the temperature. α is the heat transfer coefficient attained
from the heat transfer model, Aw is the wall surface area and Tw is the average wall
temperature.
The use of internal energy and constant volume heat capacity in the energy equation is
justified by the fact that for HCCI ignition and combustion, the peak values of normalized
pressure derivative are higher than the normalized values of maximum temperature derivative
- Eq.110. Using the constant volume specific heat minimizes the error in each time step and
convergence is reached faster. Using constant pressure specific heat for gas property
representation give the same results, but require more iteration. In each time step, the volume
is set constant.
Ns
 M j Nr  dV
m ∑ uj ∑ ν j ,k ω k  + p +
j =1
 ρ k =1  dt
Eq.109
dT
mCv + α Aw (T w − T ) = 0
dt
dp dV
max max
dt dt
>> Eq.110
p V
The volume decrease and consequently the pressure increase due to the piston movement can
easily be calculated. Thus, the total pressure is calculated as the sum of inlet pressure, the
pressure increase caused by piston movement, and the pressure increase due to chemical
reactions. The instantaneous volume is calculated from Eq.111 (note the absence of piston pin
– cylinder – crankshaft offset, compare with Eq.46) and the pressure is calculated from the
ideal-gas equation of state (Eq.17).

(
V = Vc + π ⋅ B 2 4 ⋅ l + a − a ⋅ cos ( Θ ) + l 2 − a 2 sin 2 ( Θ ) ) Eq.111

The ESIM package uses 6 species to describe the gas inside the cylinder (and in the rest of the
system as well): N2, O2 Ar, CO2, H2O and “Fuel”. The chemical kinetics program handles
about 60 to 100 (or more) different species depending on mechanism, which means that the
linkage has to handle this difference in a logical sense. The 6 species are split into air, fuel
and residuals simply by looking at the fuel and CO2 fraction. The CO2 fraction gives the
amount of burned gases since the amount of CO2 in the air is known and set constant. If water
injection is used the amount of steam also has to be determined. The fraction of fuel is given
directly and the rest is then air, and perhaps steam. Then the specie setup for the kinetics can
be set up, air and fuel is simple, residual gas is assumed by using the mixture from the
previous cycle, when the exhaust valve opens. This is sometimes not a correct assumption if

69
there are many residuals left at this point. Radicals may play an important role and could
affect the combustion timing. Since there is no chemical kinetic calculation during the gas
exchange period, several of the radicals in the residual gas become over estimated. It is quite
simple to study at the other extreme by setting all radical mole fractions to zero in the residual
gas.
The linking process is done quite simple, at 60 CAD BTDC a series of input files (text files)
is created and then the chemical kinetics program is called. When the execution of this
program is completed, the output files are read and interpreted by ESIM. The results are
interpolated to CAD resolution so the representation of the cycle in the ESIM program
becomes correct. Using only a little interval with kinetics, give less interpolation time when
doing the connection.

@ 60°BTDC
MATLAB
System (cycle) Input files

Chemical kinetics
calculation
FORTRAN EXEC.
Output files

@ EVO

Figure 50 Linking between ESIM and the chemical kinetics program

When interpreting the output files the identification of N2, O2, Ar, CO2, H2O is simple and
straightforward. The rest of the species is transformed to “Fuel” by looking at the mass
equivalent. This brings in a small error since some of these species is carrying some O-atoms
that do not originate from the fuel, which means that the mass of fuel is somewhat
overestimated. The effect of this error is very small since there is very little left of these
species and the amount of residual gas is in general very small, usually 5 – 10% of the total
mass when inlet valve closes.
It is obvious that the lack of chemical reactions during the gas exchange period may bring
some degree of uncertainty. So to illustrate the effect simulations are made with two different
starting conditions of kinetic calculation, and also termination of some important radicals; H,
O, OH, HO2, CH and CH3 in order to simulate the effect gas exchange has on these radicals.
The results are shown in Figure 51. Starting the kinetics earlier give only a minor change in
combustion timing, about 0.2 degrees later timing, termination of remaining radicals does not
give any noticeable change in the timing. The conclusion is also that any remaining radicals
seem not to affect ignition timing in any major way, at least not in low speed (1000 rpm) 4-
stroke HCCI engines.
Another limitation is the use of a zero dimensional model with chemical kinetics, since heat,
transfer is affecting the overall (average) in-cylinder temperature. Therefore, if the software is
used to find i.e. inlet temperatures, this temperature will probably be overestimated. In a real

70
engine, the temperature of the gas in the bulk should become somewhat higher than the mean
temperature, Tmean, since it is not exposed to any heat transfer (the temperature in boundary
layer is then lower than Tmean). To illustrate this difference an “adiabatic bulk temperature”,
Tbulk is calculated given a simulated pressure trace. From a initial state at IVC where the
temperature is assumed to be uniform all over the combustion chamber the gas is compressed.
Heat transfer reduces the pressure at the end of the compression stroke compared to adiabatic
conditions. Given the initial state and the pressure trace it is possible to calculate an adiabatic
temperature trace using an adiabatic assumption. First law (Eq.43, Eq.44) become:

( ) (
0 − p ⋅ ∂V = H (Ta ,i +1 ) − pi +1Vi +1 − H (Ta ,i ) − piVi ) ⇒
Eq.112
( )
H (Ta ,i +1 ) = pi +1Vi +1 + H (Ta ,i ) − piVi − p ⋅ ∂V

(Correspond to an isentropic relation, ∆s = 0 in Eq.25) . Given the enthalpy H(Ta,i+1) the

adiabatic temperature Ta,i+1 can be calculated through Eq.24 and this temperature is the
estimation of the bulk temperature, seen in Figure 51. The difference is about 35 degrees at
the end of the compression stroke. In reality, we would have a distribution of temperature
ranging from below the calculated Tmean and above the Tbulk estimation. If the inlet
temperature is trimmed to give the same temperature at the end of the compression stroke, the
inlet temperature should be about 15 degrees lower (if the compression ratio is about 18:1).
This means that the inlet temperature is overestimated by about 15 degrees. However, in this
work different kinetic mechanisms are studied and compared with experiments and a proper
mechanism that agrees with the experiments are selected for the further studies. It would
perhaps be an idea to select a mechanism that would have required 15 degrees higher inlet
temperatures; in order to get a proper prediction of ignition timing. So the simulations over-
estimate the inlet temperature, however, the absence of heat transfer models for inlet manifold
and inlet port compensates this “error” to some degree.
2200 Reference Bulk temperature estimation
Kin. start @ 101° BTDC 50
2000 Termination of radicals
40
Temperature [K]

1800
Tbulk − Tmean [K]

30
1600
20

1400 10

1200 0

1000 −10
−1 0 1 2 3 −150 −100 −50 0
CAD CAD

Figure 51 Left –The effect of different start angles and the effect of termination of some
radicals. “Reference” – calculation starts at 60° BTDC (plot resolution: 0.2 CAD). Right -
Estimation of the difference between a mean temperature and an adiabatic bulk temperature.
As with other numerical methods of solving differential equations, iteration limits may affect
the results. One mechanism is used to study the effect of changing tolerances. The heat
release rate is plotted and compared with measurement, note that the experimental result is an
average of 200 cycles. Medium and fine (more accurate) tolerances provide approximately the
same results. Coarse give a fair prediction of ignition timing but start to oscillate during the
main combustion. ”Medium” tolerances are used as basic setting and correspond to absolute
tolerance of 1E-10 and relative tolerance of 1E-7. “Coarse” correspond to 5E-10 absolute, 5E-
7 relative and “Fine” to 0.5E-10 absolute, 0.5E-7 relative tolerances.

71
 1200 5 − Coarse
1 − Medium
1000 0.5 − Fine
Experiment

Heat release [J/CAD]

800

600

400

200

0
−10 −5 0 5 10
CAD

Figure 52 The results of different tolerances on heat release prediction. Simulation peaks at
about 12000J/CAD.

4.2 Validation of chemical kinetics

To find out how good the zero dimensional chemical kinetics are to predict ignition timing
some comparisons are made with experiments. First some comparisons are made with several
kinetic mechanisms (set of species and reactions) developed for natural gas and then also
mechanisms for iso-octane, however most of these are still under development.

4.2.1 Natural gas

Several natural gas mechanisms have been tested and in Figure 53 a comparison is made
between simulations with two different mechanisms and with experimental results. The fuel
(natural gas from Denmark) composition is shown in Table 10. The experiments are made by
Christensen [65] on the Volvo TD-100 engine under supercharged conditions. Engine data is
shown earlier in Table 5 although the compression ratio were about 17.4:1 in this case. The
measured pressure trace is an average of 100 cycles.
Table 10 Fuel composition and energy content, average of 3´rd quarter 1998.
Specie Mole fraction (A/F)S 16.58
CH4 88.11% QLHV [MJ/kg] 47.8
C2H6 6.35% Qstoich. mix [MJ/kg] 2.72
C3H8 2.89%
N-C4H10 0.57%
I-C4H10 0.43%
C5H12 0.18%
C6 + 520ppm
N2 0.34%
CO2 1.07%

72
 300

250

200

Pressure [bar]
150

100
Experiment (335K)
50 Kin−1 (403K)
Kin−15 (369K)
0
−30 −20 −10 0 10 20 30
CAD

Figure 53 Natural gas cycle simulations with different chemical kinetic mechanisms
compared to experiments. Inlet temperature is shown within brackets.
The simulations are made with the same compression ratio as the experiment in which
compression ratio is determined from heat-release analysis on motored (non-firing) cycle. The
same average inlet pressure and exhaust pressure; engine speed and air-fuel ratio is use as
measured during the experiments. The simulation requires much higher inlet temperature than
measured on the real engine. The difference is about 34 degrees between the experiment and
“Kin-15”. Some of this difference, perhaps 10 – 15 degrees or so, can be explained by the
difference discussed earlier about zero dimensional modeling. The rest has to be explained by
the absence of a proper inlet port heat transfer model and perhaps limitation in the chemical
kinetics. “Kin-1” uses 58 species and 668 reaction while “Kin-15” uses 62 species and 711
reactions. Further description will be left out here.

4.2.2 Isooctane
It is possible to create mechanisms for heavier hydrocarbons as well, however these are still
limited to well defined fuels like n-heptane (N-C7H16) and isooctane and not like commercial
fuels, i.e. gasoline and diesel that is composed of a broad, almost infinite, spectrum of
different hydrocarbons. Again, several different chemical kinetic mechanisms have been
studied and some are illustrated in Figure 54. I this case the same compression ratio
(evaluated from heat-release analysis), inlet and exhaust pressure and the same inlet
temperature was used in the simulations as measured during the experiment. In this case, the
engine was operated at 1000 rpm and under naturally aspirated conditions. Pure isooctane (I-
C8H18) was used as fuel. (The experiments are described further in [67].) Based on the
knowledge earlier about the zero dimensional modeling the conclusion is that this mechanism
ignites slightly to early, however it is still quite impressive how close the ignition timing is to
the experiment.
When looking at the heat release rate, Figure 55, a first peak occurs at about 25 degrees
BTDC. It is possible that this peak occur in the real engine as well, if so, it may have drowned
in the smoothing effect of in-homogeneities or by the averaging process during heat release
analysis. However, Christensen [65] reported such 2-stage combustion when operating with
isooctane with inlet pressure of 3 bar and low inlet temperature, and not during naturally
aspirated conditions. What happens is that a “cool flame”, or two-stage combustion appears.
During this process there is a small temperature increase as seen in Figure 55 and this first
peak causes some backward reactions to temporarily dominate and the ignition is halted
[113].

73
 100 Kin−4
Kin−12
Kin−16
80 Kin−17
Exp.
Pressure [bar]
60

40

20

0
−30 −20 −10 0 10 20 30
CAD

Figure 54 Cycle simulations with different chemical kinetic mechanisms for iso-octane
compared to experiments (“Exp.”). The experimental pressure trace is an average of 201
cycles. Naturally aspirated conditions on the Volvo TD100 engine, inlet temperature of 370K,
engine speed 1000rpm and compression ratio is about 18:1 (17.7:1 was used in the
simulations).
10 20
Simulation Fired
Experiment Motored
8
15
Heat release [J/CAD]

*
dT/dCAD [K]

6
10
4
*
5
2

0 0
−40 −30 −20 −10 0 −40 −30 −20 −10 0
CAD CAD

Figure 55 Left – Heat release rate with isooctane. There is a small first peak at 25 degrees
BTDC, marked with “*”. Maximum heat release rate: Simulation: 2937J/CAD, Experiment:
315J/CAD. Right – Simulated temperature rate (K/CAD) trace during first peak, compared
with motored cycle.

4.3 Cycle simulations with chemical kinetics

The chemical kinetics provides an estimation of ignition timing and is therefore useful to
study some interesting phenomena and new control techniques for HCCI. It must be pointed
out that that the zero dimensional models are a limitation because some concepts may have a
completely different effect if the degree of homogeneity is considered. The chemical effect
will on the other hand be predicted fairly correct; however some wall effects may produce
local variations in both concentrations and reaction pathways due to the temperature
difference.

74
4.3.1 The effect of various engine operating parameters
Since the ignition process of HCCI rely totally on the auto-ignition chemistry, any factor that
affect the chemical reactions during this process will affect ignition timing and hence the
combustion process. To illustrate some of these parameters a cycle is simulated repeatedly
varying one parameter at the time from a reference case summarized in Table 11. Again the
Volvo TD100 geometry is used (Table 5) with compression ratio of 17.4:1 running with
natural gas (Table 10) at naturally aspirated conditions. The result is illustrated in Figure 56
by the change of combustion timing, A50 (angle for 50% of accumulated heat release), from
the reference case.
• Changing the compression ratio (rC) has a great influence on combustion timing and
this makes engines with variable compression ratio very suitable for HCCI operation.
Higher compression ratio increases both temperature and concentrations at the end of
the compression stroke. Generally, this promotes chemical reactions.
• Engine speed determines directly how long time the chemical reaction has to start,
higher speed mean less time and hence later timing.
• The lambda value (λ) determines the fuel/oxidizer concentration ratio and lower
lambda means more fuel per amount of oxidizer, this also increases the overall
reaction rate as long as the mixtures are lean, λ well over 1.
• Increasing the inlet temperature (Tinl.), increases the overall cycle temperature and
higher temperature promote reactions in general.
• Higher inlet pressure (pinl.) simply increases both fuel and oxidizer concentrations and
hence the reaction rates.
• Increasing the exhaust pressure (pexh.) force more residual gas to be trapped inside the
cylinder when the exhaust valve closes. In this case the residual gas fraction goes from
5.2% to 7.4%, but it causes a more important thermal effect; the temperature at IVC
rise about 2 degrees. Increasing the residual gas fraction even further give a chemical
effect, e.g. exhaust gas recycling - EGR.
• Adding 20% EGR reduces the oxidizer concentration (the mass of fuel per cycle is
maintained by reducing λ to 2.4), which reduces the overall reaction rate, combustion
is delayed.
• Natural gas composition changes over time and a more up-to-date blend, as shown in
Table 12, give a slightly earlier timing. The explanation is the higher amount of
heavier hydrocarbons like ethane and propane that ignites more easily than methane
(compare with Table 10). Note that iso-butane and heavier hydrocarbons ends up as n-
butane in the kinetic mechanism used here. Fiveland et al has investigated a variety of
natural gas compositions [114], showing the importance of higher order hydrocarbons;
ethane, propane and butane.

75
 COMBUSTION SHIFT
← Earlier Later →
rC + 0.6
speed + 200rpm
λ − 0.2
Tinl.+ 2°
pinl.+ 0.1bar
pexh.+ 0.5bar
EGR 20%
NG June 2002
−4 −2 0 2 4
∆ A50 [CAD]

Figure 56 Shift of combustion timing when some operating parameters are changed.

Table 11 Reference case

Input: Results:
rC 17.4:1 IMEPnet [bar] 3.22
Speed [rpm] 1000 IMEPgross [bar] 3.30
λ 3 PMEP [bar] 0.08
pinlet [bar] 1 ηi,net [%] 41.84
pexhaust [bar] 1 A50 [CAD ATDC] 7.4
Tinlet [K] 405

Table 12 Fuel composition and energy content of natural gas from Denmark (North See),
average during June 2002 [115].
Specie Mole fraction (A/F)S 16.53
CH4 87.15% QLHV [MJ/kg] 47.7
C2H6 6.93%
C3H8 3.11%
N-C4H10 0.63%
I-C4H10 0.44%
C5H12 800ppm
C6 + 500ppm
N2 0.32%
CO2 1.19%

76
4.3.2 Stability and late ignition phenomena
In general, the HCCI combustion is very stable in its nature; the cycle-to-cycle variations are
in general very low and this fact was one of the early discovers by Onishi [4]. One interesting
question is what makes HCCI so stable. From HCCI experiments, we know that there are
operating regimes where the combustion become instable and typically this occur at very late
ignition timing. To study the stability of HCCI cycle simulation of several consecutive cycles
is made during a startup procedure. This approach reveals much of the stabilizing feature
about HCCI.
Consider a warm, non-firing HCCI engine running at 1000rpm with the fuel injection and
inlet air preheating turned off. Suddenly fuel injection and preheating is turned on and we
start to monitor the in-cylinder pressure during the first 10 cycles as seen in Figure 57. The
first cycle (C1) will not fire since the mixture is pure and cold air as seen in Figure 58. Fuel is
present during the second cycle (C2), however, the residual gas from the first cycle is pure air
and the low temperatures of this gas limit the fuel fraction. So there is not enough fuel and the
temperature is too low for ignition. During the third cycle (C3), the residual gas is fuel rich
since there was no combustion in the previous cycle. The gas temperature at IVC is also
higher due to the preheating of the inlet air. Ignition occurs early and combustion timing11 is
about 6 degrees ATDC. In the 4’th cycle, the residual gas is hot and the temperature at IVC is
even higher but there is no fuel present in the residual gas, combustion timing is later than the
third cycle. In the following cycles, C5 – C10, combustion timing stables to about 9 degrees
ATDC. The amount of fuel present corresponds to the injected fuel. In this simulation, the
combustion efficiency is about 100% so there are no HC or CO emissions in the residual gas.
The residual gas fraction is 5.4% for C10.
C1 − C10
80
Odd
Even
70 C3
Pressure [bar]

60
C4
50

40
C1, C2
30
0 5 10 15 20 25
CAD

Figure 57 Inlet temperature of 130°C result in stable combustion with late timing after less
than 10 cycles. Odd cycle numbers indicated by dashed lines and even cycles with solid lines.
Cycle C5 – C10 is fairly the similar and cannot be separated here. (Naturally aspirated
conditions, natural gas at λ = 3, compression ratio of 17.4:1 .

11
Crank angle for 50% heat release, A50 is used here as a measure of combustion timing.

77
If a lower inlet temperature is selected, in this case 124°C, oscillating combustion timing may
occur as seen in Figure 59. In this case, the timing does not get stable as previously, modal
behaviors occur. A cycle with slightly earlier combustion timing produces residual gas that is
slightly colder due to increased heat transfer and efficiency. So, the temperature at IVC for the
following cycle becomes slightly lower as seen in Figure 60. The difference in temperature is
very small but since the combustion is very late, even this small change will delay combustion
timing. This later timing will bring up the residual gas temperature and the following cycle
gets earlier combustion timing. Note that the amount of fuel does not change. Therefore, this
tiny oscillation depends on small variations in initial temperature and the late combustion
timing. If we continue the simulation to 50 cycles, we see that the oscillation is still there,
however phase shifted 1 cycle, indicating that stability will not be reached. Modal behavior
similar to this is also known from the 2-stroke self-ignited (ATAC/AR) engines, see Ishibashi
and Asai [5].
Reducing the inlet temperature even further, to 122°C, the oscillation becomes bigger and the
cycles alternate, as seen in Figure 61, between 100% combustion efficiency and about 10%
combustion efficiency. In this case, it is not the temperature that is of importance but the
amount of fuel that is present: If combustion is poor, there will be lots of fuel and other
hydrocarbons in the residual gas (hydrocarbons and CO ends up also ends up here in “Fuel”
fraction) and this will act as ignition improvement. Ignitions occur early despite the fact that
the initial temperature is lower, as seen in Figure 62.
It would be possible that some of the effects shown previously may be to the radicals and
some intermediates that are allowed to survive from the previous cycle. So in order to
investigate this, termination of some species is made. Besides the species in the fuel and air,
only H2O2, CH2O, CH3OH, C2H2, CO and H2 are allowed to survive to the next cycle. The
result is basically the same as shown in Figure 63, both regarding oscillating and alternating
combustion. However, in the oscillating case the magnitude of the oscillation is bigger and
eventually occur a misfire, a cycle with poor efficiency (C8). The following cycle (C9) ignites
earlier. Note that the inlet temperatures are the same as earlier.

430
T [K]

420

410
3.5
Fuel [%]

2.5
10
A50 [CAD]

8
6
C1 C2 C3 C4 C5 C6 C7 C8 C9 C10

Figure 58 Stable combustion. Temperature (“T”) - top, and fuel mole fraction (“Fuel”) -
middle, at IVC. Combustion timing expressed as crank angle for 50% heat released (“A50”) -
bottom.

78
 C1 − C10
50
Odd
Even C3
45
Pressure [bar]

40

35

C1, C2
30
0 5 10 15 20 25
CAD

Figure 59 Lower inlet temperature, 124°C, result in a oscillating combustion timing after
three cycles. C5, C7, C9 has about 0.2 degrees earlier timing than C4, C6, C8 and C10.

424.5
T [K]

424.4

424.3
3.5
Fuel [%]

2.5
19.5
A50 [CAD]

19

18.5
C4 C5 C6 C7 C8 C9 C10 − C49 C50

Figure 60 Oscillating combustion. The combustion oscillates around 19 degrees ATDC after
10 cycles. The cycles with earlier ignition (odd numbers) has higher temperature at IVC.
After 50 cycles the oscillation is still there, however the order is shifted, odd cycle number has
a late ignition due to a lower temperature. (Note that the resolution of A50 is 0.2 degrees).

79
 C1 − C10
80
Odd
Even
70

Pressure [bar]
60
C4
50

40
C3
C1, C2
30
0 5 10 15 20 25
CAD

Figure 61 Even lower inlet temperature, 122°C, result in an alternating combustion with
either a high combustion efficiency, ~100%, and early timing (C4, C6, C8, C10), or poor
combustion efficiency, ~10%, and late timing (C5, C7, C9).

430
T [K]

420

410
3.5
Fuel [%]

2.5
10
A50 [CAD]

0
C4 C5 C6 C7 C8 C9 C10 − C49 C50

Figure 62 Alternating combustion with either high or low combustion efficiency. Due to the
poor combustion efficiency of the cycles with odd numbers, there will be “fuel” in the
residual gas that brings up the fuel fraction for the following, even number, cycles.

80
 430
425
T [K]

T [K]
420

420 410
3.5 3.5
Fuel [%]

Fuel [%]
3 3

2.5 2.5
30 10
A50 [CAD]

A50 [CAD]
20
5
10
0 0
C4 C5 C6 C7 C8 C9 C10 C4 C5 C6 C7 C8 C9 C10

Figure 63 With termination of radicals. With inlet temperature that provided oscillating
combustion timing earlier, 124°C, the oscillation become heavier and C8 become even a
misfire with poor combustion efficiency (left). With lower inlet temperature, 124°C,
alternating combustion occur, as in Figure 62 (right).
Given the stable operation, (as seen in Figure 57 and Figure 58) it is possible to suddenly
insert a disturbance in terms of a cycle with slightly later ignition timing (C11*, without
chemical kinetics) and see what happens in the following cycles. The result is shown in
Figure 64 and it is clear that the initial temperature can explain the response again. To further
investigate this affects the temperature at EVO, IVO and 60 degrees BTDC (i.e. when the
chemical kinetic calculation start) is shown in Figure 64. It clearly shows that later
combustion timing first brings up the EVO temperature (due to less heat transfer and lower
efficiency), this also increases the residual gas temperature, here illustrated by the temperature
when the intake valve opens (IVO). Finally, the temperature at the end of the compression
stroke in the following cycle (C12) is also increased and if this temperature is increased it will
give the same effect as if the inlet temperature would be higher, ignition occur earlier.
60°BTDC [K]

429.1 612.6
T [K]

429 612.4

428.9 612.2
3.5 750
Fuel [%]

EVO [K]

3 740

2.5 730
10 610
A50 [CAD]

IVO [K]

8 600
6
590
C10 C11* C12 C13 C14 C15 C10 C11* C12 C13 C14 C15

Figure 64 The response after a “disturbance cycle” with later ignition timing, C11*. The
following cycle will experience a higher initial temperature at IVC (T) and hence also at 60
degrees BTDC. Note that each cycle starts and stops at IVC so that the EVO and IVO
temperatures will refer to the result of each cycle respectively.
From the results above it is clear that there are one stabilizing effect due to the temperature of
the residual gas. Another stabling, or “rescuing”, effect activates if combustion is poor, the
residual gas become fuel rich (or rich of other hydrocarbons, CO or any other combustible
specie) the following cycle may occur earlier and give better combustion. Any radicals
surviving have little or no effect on the stability.

81
There is other stabilizing or destabilizing effects that also have importance e.g. heat transfer,
which is thoroughly studied by Olsson et al [116]. This has an important role at high load
operation where earlier ignition timing may cause the wall temperature to increase, causing
less heat transfer and higher temperatures at the end of the compression stroke. The ignition
timing gets earlier and earlier and it becomes critical for the control system to detect and stop
such development. At high load supercharged or turbocharged operation, this may cause too
high peak pressures and eventually engine damage.

4.3.3 The effect of valve timing

As seen earlier, the residual gas temperature plays an important role when it comes to HCCI
timing control. Therefore, it is obviously interesting to use methods of controlling the amount
of residual gas inside the cylinder. Koopmans and Denbratt [117] has recently reviewed and
investigated this experimentally in a camless gasoline engine using mostly negative valve-
overlap (increasing positive overlaps too much may cause the valves to interfere with the
piston), meaning that EVC is moved earlier and IVO later than usual. He also showed that this
concept could be used during a transition from spark ignition to HCCI operation. This
concept force significant residual gas to stay in the cylinder during exhaust stroke and the gas
will actually be compressed during the late end of the exhaust stroke. They also showed and
explained cycle-to-cycle variations in this engine by variations in the initial temperature
[118].
A simple simulation is used here to study this concept simply by using the reference case
described in Table 11 as a start point, and then shift EVC earlier and IVO later up to 90
degrees. Valvelifts for 60 degrees shift is shown in Figure 65. An iterative approach is used so
that the amount of injected fuel per cycle (into the engine inlet port) is the same, about 26
mg/cycle, hence the engine load may change only due to the effect of variation in combustion
timing and pumping work and not how much air-fuel mixture is entrapped in the cylinder.
15 Reference
EVC/IVO shifted 60°
Valvelift [mm]

10

0
0 180 TDC 540 720
CAD

Figure 65 Valve lift curves without (Reference) and with 60 degrees shift of EVC and IVO
This concept allow a lot of residual gas to stay inside the cylinder, over 50% is possible as
seen in Figure 66, and with a constant amount of fuel per cycle this means that the λ-value
goes down towards 1, i.e. stoichiometric mixture. Despite this heavy amount of residual gas,
the combustion timing becomes earlier and earlier as also seen in Figure 66. Lot of hot
residual gas increases the temperature prior ignition here illustrated by the temperature 60
CAD BTDC and this explains the earlier timing. The increasing amount of residual gas and
temperature explain this concepts capability of handling a transition from spark ignition to
HCCI.

82
 Earlier EVC, Later IVO → Earlier EVC, Later IVO →
60 0

50 −2
Residual fraction [%]

∆ A50 [CAD ATDC]

← Earlier Later →
−4
40
−6
30
−8
20
−10
10 −12

0 −14
0 20 40 60 80 100 0 20 40 60 80 100
Shift [CAD] Shift [CAD]

Figure 66 Earlier EVC, later IVO shift and its influence on residual gas fraction and
combustion timing.
Earlier EVC, Later IVO → 2.5
700
Temperature @ 60° BTDC [K]

680 2
Pressure [bar]

660
1.5

640

1
620 EVC/IVO shift
Reference
600 0.5
0 20 40 60 80 100 0 500 1000 1500
Shift [CAD] Volume [cm3]

Figure 67 Earlier EVC, later IVO shift and its influence on temperature at 60 CAD BTDC
(left) and pressure over volume during intake and exhaust stroke (right). … indicate EVO,
∇ - EVC, { - IVO and Ä indicate IVC.
Another way to control residual gas is to use inlet cam phasing, and move both IVO and IVC
earlier (increasing overlap) and thereby affect the valve flow process e.g. the backflow over
the intake valve. This backflow of burned gas will then increase the residual gas fraction. To
study this effect the reference case described in Table 11 is used again and the cam timing is
phased +/- 35 degrees. In addition, moving only the IVO timing is investigated as well. The
Volvo TD100 geometry is used however calculated valve lift (from timing and maximum lift)
is used instead of measured lift in order to modify valve timing more freely. It should be
pointed out that the model does not include any heat transfer model for the inlet manifold; the
heat loss of this hot residual gas is one factor of uncertainty.
If the inlet valve timing is moved earlier by using either IVO/IVC shift (cam phasing) or IVO
shift, the residual gas fraction is increased but the combustion timing changes depending on
what method is used as seen in Figure 68. Using cam phasing give first an earlier timing, then
at about 20 degrees shift, the trend alter an we get later and later timing. This behavior can be
explained by the temperature and pressure during the compression, here showed at 60 degrees
BTDC in Figure 69. With early cam-phase shift down to –20 CAD, the amount of residual gas
increase and cause an increase in temperature that move combustion timing earlier. With
more cam-shift, the effective compression is reduced since IVC has become too early and this
limits both temperature and pressure. The temperature ends up at approximately at the same
level as with zero shift but the pressure is much lower and combustion timing becomes later.

83
Shifting only IVO earlier does not affect the pressure during the compression stroke; the
temperature increase with more residual gas and combustion timing become earlier.
Slightly later cam phasing reduces the amount of residual gas and the temperature at 60 CAD
BTDC is reduced, combustion is delayed. With more shift, the amount of residual gas does
not change, instead is the temperature of the residual gas affected. Later cam phasing
effectively reduces the trapped amount of air in the cylinder; the λ-value is reduced (Figure
70), from 3.0 down to about 2.6, and this brings up the temperature after combustion and
hence the residual gas temperature. Despite the lower pressure, reduced by late intake valve
closing, again the temperature has the greatest influence and combustion timing become
earlier.
← Earlier Later → ← Earlier Later →
9 6
IVO/IVC shift IVO/IVC shift
IVO shift 4 IVO shift
8

∆ A50 [CAD ATDC]

Residual fraction [%]

← Earlier Later →
2
7
0
6
−2

5 −4

4 −6
−40 −20 0 20 40 −40 −20 0 20 40
Shift [CAD] Shift [CAD]

Figure 68 The effect of inlet valve timing on residual gas fraction (left) and combustion
timing (right).
← Earlier Later → ← Earlier Later →
624 4.4
IVO/IVC shift
Temperature @ 60° BTDC [K]

IVO/IVC shift IVO shift

Pressure @ 60° BTDC [bar]

622 IVO shift

4.2
620

618 4

616
3.8
614

612 3.6
−40 −20 0 20 40 −40 −20 0 20 40
Shift [CAD] Shift [CAD]

Figure 69 The effect of inlet valve timing on temperature (left) and pressure (right) at 60
degrees BTDC.
In summary, using inlet cam phasing has very little influence on combustion timing; it is not a
suitable as a single method to control HCCI combustion. Shifting EVC/IVO or using late inlet
cam phasing also introduce pumping work illustrated by the p-V diagrams in Figure 67 and
Figure 70. The resulting pump mean effective pressure for the different solutions is showed
in Figure 71. Shifting EVC/IVO causes an “extra” compression phase. Heat transfer during
this phase causes a loss of energy that results in increased pumping work. It should be pointed
out that this extra heat transfer may be overestimated due to model limitations and the wall
temperature may become higher due to the extra heat transfer. Using late inlet cam-phasing
give a huge pumping loop since enough valve flow area is not allowed to be high enough in
time. Maximum valve lift occurs too late in the intake stroke.

84
 ← Earlier Later →
1.2 3.2
1

1.4 3
1.2
Pressure [bar]

λ
1
2.8
0.8
Early IVO/IVC
0.6 Reference IVO/IVC shift
Late IVO/IVC 2.6 IVO shift
0.4
0 500 1000 1500 −40 −20 0 20 40
3 Shift [CAD]
Volume [cm ]

Figure 70 Pressure over volume (left) for 35 degrees earlier IVO (top) and +/- 35 degrees
cam phasing (bottom). { indicate IVO andÄ indicate IVC. The resulting λ-value (right).
Earlier EVC, Later IVO → ← Earlier Later →
0.4 0.3
IVO/IVC shift
0.35 IVO shift
0.25
0.3
PMEP [bar]

PMEP [bar]

0.25 0.2

0.2 0.15
0.15
0.1
0.1

0.05 0.05
0 20 40 60 80 100 −40 −20 0 20 40
Shift [CAD] Shift [CAD]

Figure 71 Pump mean effective pressure, PMEP for EVC/IVO shift (left), IVO/IVC (cam-shift)
and IVO shift (right).

4.3.4 The effect of exhaust gas recirculation - EGR

Exhaust gas recirculation, EGR, has been used successfully in spark-ignited engines and
lately also in diesel engines to reduce NOx formation. It affects the specific NOx emissions
(g/kWh) in two ways; chemically by reducing the oxygen concentration, and hence NOx
formation rate at high temperatures, and by substitute air with EGR reduces the mass flow
through the system and hence the mass of NOx leaving the system.
In HCCI engines the same result is shown for most fuels, however EGR also affect the
combustion timing through the ignition temperature. Christensen [38] showed that for natural
gas and EGR higher inlet temperature were required and this actually increased the volume
fraction of NOx, the specific NOx emissions were quite un-changed. The specific emissions
of HC, and to some degree CO emissions also, went down due the lower mass flow leaving
the system. The volume fraction of HC and CO was quite constant. Using EGR with a fuel
like natural gas may not give any significant benefits but it is perhaps necessary if the
combustion become to fast, or if knocking appear at high load operation. The zero
dimensional models used here cannot handle the oscillating pressure from knocking
combustion and is not very suitable for studying combustion duration. Instead are the work
focused on understanding the influence of EGR on inlet conditions and the gas properties, and
what consequence it has on performance and efficiency.

85
In this simulation, the amount of fuel is kept constant at 25.79±0.01 mg/cycle corresponding
to the reference case (Table 11, λ = 3 without EGR). The amount of EGR is increased is steps
of 10% up to 50% and the inlet temperature is adjusted so that 50% burned fraction (A50)
occur at 8 degrees ATDC. The required inlet temperatures and the resulting indicated
efficiencies are shown in Figure 72.
435 43
Gross
430 Net
42.5

Indicated efficiency [%]

Inlet temperature [K]

425

420 42

415 41.5
410
41
405

400 40.5
0 10 20 30 40 50 0 10 20 30 40 50
EGR [%] EGR [%]

Figure 72 Required inlet temperature (left) and resulting indicate efficiencies (right) with
EGR. ”Gross” = compression and expansion stroke and “Net” = whole cycle.
Higher inlet temperatures are required with EGR due to altered thermal properties of the
working gas and higher ignition temperatures. With EGR, the specific heat of the inlet
mixture is higher (Figure 73) due to higher specific heat of the burned gas compared to the
fresh charge. During the compression stroke, only a limited amount of energy (work) can be
added to the gas trapped in the cylinder due to a fix compression ratio, which means that both
the pressure and temperature is limited at the end of the compression stroke and after the
combustion, due to the higher specific heat. Therefore, a higher inlet temperature is required
to reach the ignition temperature. The ignition temperature is quite constant at low amounts of
EGR, but increases significantly with EGR levels above 20%, as seen in Figure 74. At high
levels of EGR the O2 concentration degreases since air is replaced with burned gases. In
addition, all concentrations are slightly reduced due to lower pressure and hence density.
Higher specific heat limits the maximum temperature after combustion, which limits the heat
transfer to some degree. Therefore, if the simulation is made without heat transfer, the
efficiency drops even more, as seen in Figure 75. In total, the lower pressure levels give a
slight reduction in efficiency. At higher amount of EGR there is a significant increase in
ignition temperature and even higher inlet temperatures are required. The overall temperature
trace is elevated and the heat transfer increases instead (Figure 74), the efficiency drops even
more. The explanation is quite complex; a schematic figure explaining the relations is shown
in Figure 76, which may support the discussion above.
In a real engine however these very small differences in efficiency is probably quite difficult
to see, other effects may have higher significance e.g. combustion efficiency. Without EGR
the combustion efficiency may be, let say 95%, but with EGR some of the exhaust gas is
recycled and the recycled hydrocarbons gets a second chance to burn, combustion efficiency
increases because less mass of unburned fractions leave the system12. When the combustion
efficiency increases, we can expect the indicated efficiency to increase. Combustion duration
will probably increase at high amount of EGR; the maximum temperature and thereby the

12
Note that the fuel trapped inside the cylinder is used here as reference when calculating efficiencies. This
amount of fuel includes residual and recirculated unburned fuel from previous cycles.

86
heat transfer will decrease somewhat. In summary, a less decrease in efficiency could be
expected in a real engine. The higher inlet temperatures lifts the temperature level over the
whole cycle and since a constant wall temperature is assumed here the heat transfer increases.
If the wall temperature would have been allowed to change, it is quite clear that it would be
higher with EGR than without due to the overall higher cycle temperatures. Nevertheless, the
wall temperature can only increase if the overall heat transfer or if the coolant temperature
increases. A thermostat usually controls the coolant temperature and it should be considered
as fairly constant. Therefore, the change in heat transfer on a real engine may not be as
significant as seen in the simulations here.
31.5 1.4
@ IVC & 300K
@ IVC & 300K @ IVC
@ IVC 1.395
31
1.39
Cp [J/(mol*K)]

30.5 1.385

γ
30 1.38

1.375
29.5
1.37

29 1.365
0 10 20 30 40 50 0 10 20 30 40 50
EGR [%] EGR [%]

Figure 73 Specific heat at constant pressure (left) and ratio of specific heats, γ, (right) for the
in-cylinder mixture at IVC. At different amount of EGR and 300K and at actual temperature
at IVC.
30 435
Inlet Gross
25 @ 1% HR Net
Maximum 430
Temperature change [K]

20
Heat transfer [J]

15 425

10 420
5
415
0

−5 410
0 10 20 30 40 50 0 10 20 30 40 50
EGR [%] EGR [%]

Figure 74 Change in inlet, ignition (expressed as temperature at 1% heat release, 3.6 CAD
BTDC in all cases) and maximum temperatures with increasing EGR (left). Accumulated heat
transfer (right).

87
 0
w. HT
w.o. HT
−0.5

Relative change in ηi,g [%]

−1

−1.5

−2

−2.5

−3
0 10 20 30 40 50
EGR [%]

Figure 75 The relative change in efficiency due to the effect of heat transfer and EGR

Up to 20% EGR Over 20% EGR

EGR (+) Tign(~0) EGR (+) Tign(+)

Cp (+) Cp (+)
Cv (+) Cv (+)

∂U/∂T (+) ∂U/∂T (+)

Tmax(-) Tmax(+)
∂T/∂U (-) ∂T/∂U (-)
T comp. (-) QHT (-) T comp. (-) QHT (+)
p (-) p (-)

ηi,g (-) Tinlet(+) ηi,g (+) ηi,g (-) Tinlet(+) ηi,g (-)

Figure 76 The effect of EGR on efficiency

The influence of the gas property is even known from ideal cycles, e.g. ideal adiabatic cycle
with constant volume at TDC combustion and constant gas properties result in an ideal
thermal efficiency that can be expressed like shown in Eq.113, where. As shown earlier,
specific heats depend on temperature and gas composition. With EGR and/or at higher
temperatures, the specific heat is increasing as seen in Figure 73. This mean that the ratio of
specific heats, γ = Cp/Cv = Cp/(Cp - R) = 1/(1 - R/ Cp) decreases and therefore will the
efficiency also decrease according to Eq.113.
1
ηTh. = 1 − γ −1
Eq.113
r C

88
4.4 The effect of humidification of inlet air – the HAM
concept
Instead of using EGR, it is possible to dilute with other non-NOx generating gases e.g.
vaporized water (or steam). Hugon used water injection into the cylinder in a non-
compression flame-ignited engine to cool the double acting piston back in 1861. Later, when
non-compression engines were abandoned, Prof. Bánki of the Budapest Technical University
increased the compression ratio in a SI engine (to 6.5, a high value at that time) and used
water injection to avoid knocking combustion. The high compression ratio gave a better
efficiency but the water consumption was reported to be high, about 5 times the fuel
consumption (reviewed by Cummins [97]). Later “hot-bulb” engines also employed water
injection to control the temperature of the bulb and hence preventing pre-ignition. The hot-
bulb engine was a very interesting engine that was frequent in fishing boats, tractors and as
stationary engines in Europe from about 1900 to 1940. It was often a 2-stroke engine with
direct injection of the fuel into a hot, un-cooled part of the combustion chamber. The injection
timing was quite early and the engine speed was fairly low. It was vital to keep the
temperature of the bulb correct and this could be a problem if the engine were running at high
load for a long time. Water injection, either mixed with the air, the fuel or injected in a
separate injector, was developed. Later came adjustable fuel injectors (spray angle) solving
this problem without the need of any water [119].
The Humid Air Motor (HAM) concept shown in Figure 77. The patents of this method are
held by Rosén and Olsson [120] and it has been developed by MUNTERS AB as a method of
reducing NOx emissions in diesel engines. They have implemented this technology in large
diesel engines for ferries with good results and they claim up to 70% NOx reduction without
increase in fuel consumption [122]. Other positive side effect is that cylinder and valve
temperatures become lower and the engine become “cleaner” i.e. there is less carbon deposit
on the combustion chamber walls. The water to air mass ratio, W/A is about 0.11 at 2.5
bar/75°C and 98% relative humidity (RH).
Humidification Turbo -
system charger
w.o. HAM
T [K]

w. HAM

Heat

Engine
H2O
CAD
TDC

Figure 77 The HAM concept in a diesel engine (left), Revised from [121]. Schematic
temperature diagram illustrating the effect on combustion (right). Water reduces heat release
rate which limits the maximum temperature and hence NOx formation. Revised from [122]

89
Water injection in HCCI engines has been studied experimentally by Christensen [41]. The
results showed that water reduces heat release rate but the combustion efficiency goes down,
both CO and HC emissions increases. It should be pointed out that the heat release rate is
dependent on combustion timing that was allowed to change during these experiments.
If liquid water is injected downstream the compressor, it is possible to make use of the
vaporizing in terms of cooling of the inlet charge. In HCCI, this may be suitable at very high
engine loads, where cooling and/or water injection could be used as a control parameter. In
the HAM concept, heat from the coolant water of the engine or the exhaust gas downstream
the turbine is used as a vaporization improver in the humidifier. It is possible to increase the
mass flow through the turbine and hence deliver more power to the compressor. This is
making this concept very interesting in HCCI engines were it may solve some of the
turbocharging problem. Vaporized water, i.e. steam has a quite high specific heat; a slight
decrease in efficiency of the engine should be expected but this means that the energy level
upstream the turbine would also increase.
To study this further engine cycle simulation is made at turbocharged conditions and a
slightly richer mixture of λ = 2.5 (without water) is selected for high load operation. The inlet
and exhaust pressures are increased from the reference case 1 bar to 2.5 bar and the engine
speed is increased to 1500rpm. The combustion timing is set constant as previously by
adjusting the inlet temperature. The resulting inlet temperature and relative humidity for
different amount of added water steam is seen in Figure 78. Without additional water, the
inlet temperature is about 365K and this corresponds to an isentropic compressor efficiency of
about 75% (compressor inlet temperature: 293K). Higher inlet temperature is required with
more steam as expected, about 11 degrees per 0.1 in W/A-ratio. The reason why the
temperature has to be higher is not due to altered ignition temperature. The ignition
temperature is quite constant as seen in Figure 79, it even decreases but what happens is that
the specific heat increases which reduces the temperature and pressure during compression
and to compensate for this effect the inlet temperature has to be increased to maintain timing.
The maximum temperature decreases as the specific heat increases.

400 40

35
380
Relative Humidity [%]
Inlet temperature [K]

30
360 25

340 20

15
320 DB
WB 10

300 5
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
W/A W/A

Figure 78 Required inlet (dry bulb - DB and wet bulb - WB) temperature (left) and relative
humidity (right), over water to air mass ratio, W/A.

90
 40 Inlet 32
@ 1% HR @ IVC & 300K
Maximum @ IVC
20 31.5
Temperature change [K]

C [J/(mol*K)]
0 31

−20 30.5

p
−40 30

−60 29.5
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
W/A W/A

Figure 79 The change in inlet, ignition and maximum temperature (left) and specific heat at
IVC (right).
As the maximum temperature decreases the heat transfer decreases, however the higher
specific heat reduces the pressure, which has a greater affect on the indicated efficiency of the
engine as seen in Figure 80. In the same figure, the heat balance of the engine is shown scaled
to a 6-cylinder engine. The fuel energy expressed as heating power calculated from fuel flow
and the lower heating value of the fuel (QLHV) is about 298kW (kept constant). The indicated
work is about 140kW but decreases since the efficiency decreases. The heat transfer to the
combustion chamber walls is about 70kW and decreases because the overall temperature
decreases. The heat that leaves the engine with the exhaust gas is about 90kW and increases
due to lower efficiency and less heat transfer. The vaporization heat at this inlet conditions
(heat required to elevate the water temperature to inlet conditions is not considered) is
increasing with increasing water flow and there is a brake-even with the engine heat transfer
at 0.15 in water to air ratio.
49 140
Gross
Net 120
48
Indicated efficiency [%]

100
Power [kW]

47 80

46 60
Work
40 Exhaust
45 Engine HT
20 Water vap.

44 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
W/A W/A [%]

Figure 80 Indicated efficiency (left) and power levels (right). Fuel heat power (LHV) is about
298kW, “Work” – net indicated work, “Engine HT” – heat transfer over combustion chamber
walls and “Exhaust” fuel heat minus work and engine heat transfer. “Water vap.” – Water
vaporization heat, is calculated from water flow and specific vaporization heat at 2.5 bar and
the present inlet temperature.

91
5 More realistic prediction of
heat release rate

5.1 “Slave” cycle and estimation of NO formation

The chemical kinetic calculation renders a pressure trace with a very high, close to infinite,
pressure rate during combustion, which is not realistic, although HCCI has a very fast
combustion at high load operation. The high combustion rate causes a very high maximum
pressure and hence temperature. A very simple way to get around this problem is utilized
here, however other methods are probably better and the author would like to advise the
reader, if he faces the decision, to select either multi-zone models or so-called PDF models
that give a better prediction of the heat release rate. At the writing moment, it may appear
time-consuming in terms of computational time, but computer performance is constantly
under development.
In this case, the chemical kinetics is used to find ignition timing and a “slave cycle” is then
used with more reasonable estimated combustion duration. When simulating high load
operation the combustion duration, from 10 to 90% burnt, is set to be 4 degrees. The Wiebe
function Eq.53 is used for the heat release shape (see section 2.5.1). This means that the
compression and expansion part of the cycle is calculated twice, first with kinetics and then
with the Wiebe function.

250
Pressure [bar]

200

150 Experiment (335K)

Kin. #1 (403K)
Kin. #2 (369K)
Slave to #2
100

−10 −5 0 5 10
CAD

Figure 81 Slave cycle compared to kinetics and experimental pressure traces. Kin#2 is used
trough out this work. Inlet temperatures are shown within brackets.
A large number of simulations with chemical kinetic based prediction of ignition timing has
been done in order to find models for the full system simulation. The purpose is to reduce the

92
computational time for the system simulation. Depending on what resolution is used in the
cycle simulation part (intake, exhaust stroke) the interaction with the chemical kinetics
program takes quite some time due to interpolation.
Using the temperature and pressure trace from the slave cycle it was possible to estimate the
specific NOx emissions with the extended Zeldovich mechanism as described by Heywood
[96]. It gave at least the correct order of magnitude, as shown in Table 13, considering the
differences in maximum pressure and hence temperatures. This mechanism could however be
used as an indicator of when significant NO appear.
Table 13 Comparison between experiment and cycle simulation regarding IMEP and NOx
Experiment [36] Simulation
IMEPnet [bar] 14.0 13.3
NOx [g/(kWh)i] * 0.0075 0.0082
*) NO counted as NO2 (ISO 8178-1)

5.1.1 Compression ratio optimization

In this case, cycle simulations with chemical kinetics were used to optimize the compression
ratio for maximum load. The following assumptions and limitations were used:
• A maximum in-cylinder pressure of 200 bar.
• A maximum NO fraction of 10 ppm in the exhaust, calculated from the extended
Zeldovich mechanism.
• Combustion timing corresponding to maximum pressure at 8 CAD ATDC
• The engine speed was set to 1500 rpm.

The λ-value basically controls the maximum temperature and hence NO formation. Inlet
pressure determines the maximum pressure, and inlet temperature determines combustion
timing. The selected timing is held constant, based on a compromise between efficiency and
NO formation. (A very detailed study of the relation between combustion timing and
stability, NOx emissions and compression ratio was made by Olsson et. al [116].) An iterative
approach is used to find the proper λ, inlet temperature and pressure for each fuel and
compression ratio. Two different fuels were investigated; natural gas and land fill gas. Gas
composition and energy content are shown in Table 14.
The indicated efficiency became higher with higher compression ratios as shown in Figure 82,
a relation well known from ideal cycles. However, the load (Figure 83) was limited with
higher compression ratios due to the limited maximum pressure restricting the maximum inlet
pressure. When the friction was considered, the simulated brake efficiency peaked at
compression ratios between 15:1 and 17:1 (Figure 82). The friction model has no load or
compression ratio dependence; the results may have been somewhat different if it did. The λ-
values (Figure 83) was fairly constant indicating that the ignition temperature and the NO
formation do not vary that much with different compression ratios.
The efficiency and load became higher with natural gas than with landfill gas, which can be
explained by the higher ignition temperature of landfill gas that raises the required inlet
temperature (Figure 84). There is also a small difference in inlet pressure that probably also
can be explained by the higher inlet temperature. Higher inlet temperature limits the mass
flow, and thereby the amount of fuel burned. If higher combustion temperatures are allowed
due to dilution effects, (the high CO2 content of the land-fill gas) more fuel can be burned by
increasing the inlet pressure.

93
Table 14 Natural gas and landfill gas composition
Mole fractions
Specie
Natural gas Landfill Gas Natural gas Landfill Gas
CH4 88.11% 49.2% (A/F)S 16.58 4.93
C2H6 6.35% 10ppm QLHV [MJ/kg] 47.8 14.3
C3H8 2.89% - Qstoich. mix [MJ/kg] 2.72 2.41
N-C4H10 0.57% -
I-C4H10 0.43% -
C5H12 0.18% -
C6 + 520ppm -
N2 1.07% 15.7%
CO2 - 34.4%
O2 - 0.2%

Simulation Simulation
47 40
Net indicated efficiency [%]

46 39.5
Brake efficiency [%]

45 39

44 38.5

43 Natural gas 38 Natural gas

Landfill gas Landfill gas
42 37.5
12 13 14 15 16 17 18 19 12 13 14 15 16 17 18 19
Compression ratio Compression ratio

Figure 82 Left - Simulated net indicated efficiency against compression ratio. Right -
Simulated brake efficiency at 1500 rpm
Simulation Simulation
15 2.45
Natural gas Natural gas
14 Landfill gas Landfill gas
2.4
13
BMEP [bar]

2.35
12
λ

2.3
11
2.25
10

9 2.2
12 13 14 15 16 17 18 19 12 13 14 15 16 17 18 19
Compression ratio Compression ratio

Figure 83 Left - Simulated brake mean effective pressure at 1500 rpm. Net indicated mean
effective pressure is 1.71 bar higher at this speed (FMEP is 1.71 bar). Right - Simulated λ-
values.

94
 Simulation Simulation
180 4.5
Natural gas Natural gas
Landfill gas Landfill gas
160 4
Inlet temperature [°C]

Inlet pressure [bar]

140 3.5

120 3

100 2.5

80 2
12 13 14 15 16 17 18 19 12 13 14 15 16 17 18 19
Compression ratio Compression ratio

Figure 84 Required inlet temperatures (left) for proper ignition timing and required inlet
pressures (right) for reaching maximum load (maximum pressure).

5.1.2 Inlet temperature control model

The previous results show that the best compression ratio is about 15:1. The efficiency then
becomes almost maximum but it is still a good compromise when it comes to brake power. To
achieve this efficiency the engine has to be turbocharged or supercharged to 3.5 bar (abs)
which perhaps is not possible without losses. To investigate this, a full system simulation with
various exhaust turbines is studied.
The current HCCI-engine setup uses the inlet temperature to control combustion timing. This
is a slow control parameter. Nevertheless, it might be the only solution if no advanced engine
design with e.g. variable compression ratio or dual fuel is economically feasible.
A model for desired inlet temperature is made at a compression ratio of 15:1. The required
inlet temperature for combustion timing is investigated for different air-fuel ratios and inlet-
and exhaust pressures. The speed is kept constant at 1500 rpm. The results are summarized in
Figure 85 and Figure 86, where the dependence of fuel, inlet pressure (x-axis), exhaust
pressure and λ are shown.
Natural gas Landfill gas
460 500
pexh = pinl pexh = pinl
p = p +1bar 480 p = p +1bar
440 exh inl exh inl
Inlet temperature [K]

Inlet temperature [K]

p = p +2bar p = p +2bar
exh inl exh inl
460
420
440
400
420
380
400

360 380
1 1.5 2 2.5 3 3.5 4 1 1.5 2 2.5 3 3.5 4
Inlet pressure [bar] Inlet pressure [bar]

Figure 85 Required inlet temperatures depending on inlet (pinl) and exhaust pressure (pexh).
Left – Natural gas at λ = 2.4 . Right – Land fill gas at λ = 2.3 .

95
 Natural gas Landfill gas
480 500
λ=4 λ=4
λ=3 λ=3
460 480
λ = 2.4 λ = 2.3
Inlet temperature [K]

Inlet temperature [K]

λ=2 λ=2
440 460

420 440

400 420

380 400
1 1.5 2 2.5 3 3.5 4 1 1.5 2 2.5 3 3.5 4
Inlet pressure [bar] Inlet pressure [bar]

Figure 86 Required inlet temperatures depending on λ-values and engine inlet pressure.
The reason why the exhaust pressure is set higher than the inlet pressure is that we know that
the turbine has to be small, compared to diesel engine configuration. This increases the
exhaust pressure, but also allow more energy to be extracted.
The modeling is made through multidimensional non-linear regression of these results. This
model is used in the system simulation to control proper inlet air conditioning (level of
intercooling/heating). It is quite surprising that the required inlet temperature for landfill gas
shows little dependence of λ-value.

5.2 System simulation

A whole engine system is simulated in order to study the effect of varying the turbine size of
the turbocharger. The system is illustrated in Figure 87. Three different turbine sizes are
explored, the largest one, TC-1 corresponds to the diesel engine’s original turbine size
(geometrical flow area: 25 cm2) while the two smaller ones, TC-2 and TC-3 produce higher
pressure ratios since they have smaller flow area (16 and 13 cm2 respectively). Again, the
simulations are made using the two different fuels mentioned earlier. The fuel is added
downstream the compressor (into the inlet manifold), the efficiency of the system is not
compensated for the work required to compress the fuel.

5.2.1 The effect of turbine size

Again, the turbine size is investigated using the same turbine sets from the previous
investigation with isooctane/n-heptane (see section 3.3), however in this case with other fuels.
A more proper estimation of required inlet conditions and the turbo model described in
sections 2.6.3 - 2.6.6 is used instead of plain interpolation.
The simulations are performed with a time trace i.e. under dynamic conditions. However the
reader should be aware that a number of simplifications were made to speed up the
calculations. First, there is no chemical kinetics calculations made in this part, it is assumed
that the timing is achieved with proper inlet air conditioning according to the model described
earlier (section 5.1.2). The engine cycle is treated as a quite large damper in the system and is
withdrawn from the iterative solution of the other system components i.e. for a certain time
step only single-cycle calculations are made based on the inlet and exhaust states provided
from the previous time step. The same goes for the simple “engine control” module used, it
sets the fuel mass flow based on the air mass flow and desired λ (the fuel is injected into the
inlet manifold). It also controls the inlet air “conditioner” based on air mass flow and desired

96
inlet temperature. The conditioner can be interpreted as a controllable charge air cooler
(intercooler) or a heater of some sort. It can either cool the gas or heat the gas by removing or
adding heat to the air stream. The calculation scheme for the system is shown in Figure 87.
Each time step is 0.1 seconds but is allowed to be shorter if there is a problem with
convergence. Then it is reduced by a factor 0.5 after each 100 iteration cycle.

Inlet air
conditioner Cycle calculation

Turbochager

Turbo- Conditioner
charger
Inlet/Exhaust manifold

Engine Engine control

Figure 87 System layout and simplified flow chart of the system calculation, note that the
engine cycle and engine control module are moved outside the iterative loop.
Simulation, Natural gas
3

2.5

2
λ

1.5 TC−1
TC−2
TC−3
1
0 5 10 15 20 25 30
Time [s]

Figure 88 λ over time, 1500 rpm for natural gas. λ=2.4 at 29s. With land-fill gas the λ value
is 2.3 at 29s.
During the simulation, the λ-value is varied over time. At first, the λ-value is set to a fairly
low value of 2 (low for HCCI operation), simultaneously the engine is speeded up to 1500
rpm and then there is stabilization for 10 seconds where the turbine can speed up further.
After that, the λ-value is lowered to the value given by the optimization earlier: 2.4 for natural
gas and 2.3 for landfill gas. Everything is then allowed to reach steady state for 20 seconds.
This procedure is illustrated in Figure 88. The idea is to simulate a start-up procedure under
which the engine goes from idle to full load. The purpose of the middle step with a low λ is to
make sure that the turbocharger speeds up, if possible. This procedure is used with three
different turbine sizes and two different fuels. The resulting BMEP is shown in Figure 89 and

97
it is clear that the turbine does not provide enough work. The mass flow and the energy levels
in the exhaust are too low.

Simulation, Natural gas Simulation, Landfill gas

10 10
TC−1 TC−1
TC−2 TC−2
8 TC−3 8 TC−3
BMEP [bar]

BMEP [bar]
6 6

4 4

2 2

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time [s] Time [s]

Figure 89 Simulated BMEP for three different turbochargers. Neither of them provides
enough boost.

5.2.2 Quest for higher loads

As a final step, adjustments were made in order to reach higher loads. Much later ignition
timing was selected; the maximum pressure was set to occur 14 CAD ATDC and the
combustion duration was set to 5 CAD. The same control strategy for the inlet temperature
was used, although compensated with -10 degrees per step in compression ratio, complying
fairly well with the results shown in Figure 84. In addition, the λ-value was reduced to 2.2.
The resulting loads (at time = 29 s) are shown in Figure 90, and the results seem quite
surprising: A higher compression ratio gave higher load. The explanation to this is the
absence of proper boost pressure with a low compression ratio, which raises the required inlet
temperature. This reduces mass flow through the system and consequently, the turbine cannot
provide enough work. Regardless of the compression ratio the inlet pressure became 2.22
±0.02bar. The required inlet air cooling or heating is also shown in Figure 90 and the
breakeven point is somewhere between compression ratio 15 and 16.
Natural gas, TC−3, pInl 2.22+/−0.02bar Natural gas, TC−3, pInl 2.22+/−0.02bar
10 8

6
Conditioner power [kW]

9.5
4
BMEP [bar]

9 2

0
8.5
−2

8 −4
13 14 15 16 17 13 14 15 16 17
Compression ratio Compression ratio

Figure 90 Left - BMEP with the smallest turbine (TC-3) and for different compression ratios.
Inlet pressure (pInl) is 2.22 ±0.02 bar. Right - Required inlet air conditioning. Positive values
mean heating and negative mean cooling of the air.
Without proper boost, it is the compression ratio that influences brake efficiency (Figure 91)
more than friction. Since the turbocharger cannot provide enough boost, the required inlet

98
temperature become quite high. This effectively limits the mass flow through the system,
making it even more difficult to produce boost. A higher compression ratio can actually allow
higher mass flows thanks to lower inlet temperatures. In the following simulations, five
different compression ratios are studied. Consequently, the efficiency became higher with
higher compression ratio and the specific NO (Figure 91) went down due to the higher load.
At 17:1, the efficiency became 40.3% and the NO emissions were about 22 ppm, 0.162
g/kWh, a figure far below the 20 g/kWh of the lean burn SI engine and also well below the
legislation limits in Germany and Switzerland (see section 3.5) due to the much leaner
mixtures. The specific load of the engine is however limited, only 9.7 bar BMEP at 17:1.
Natural gas, TC−3, pInl 2.22+/−0.02bar Natural gas, TC−3, pInl 2.22+/−0.02bar
40.5 0.23

Specific NO (Zeldovich) [g/kWh]

40 0.22
Brake efficiency [%]

0.21
39.5
0.2
39
0.19
38.5
0.18
38 0.17

37.5 0.16
13 14 15 16 17 13 14 15 16 17
Compression ratio Compression ratio

Figure 91 Left - Simulated brake efficiency. Right - Simulated specific NO emissions at λ =

2.2. At 17:1 the NO fraction is about 22 ppm.

99
6 System simulations with
chemical kinetics

As a final step the chemical kinetics is used to predict ignition timing during system
simulations, however a slave cycle is used to give a more realistic heat release prediction as
shown earlier (section 5.1). The combustion efficiency is set to 95%. A combustion timing
control has to be designed to get a suitable timing. The advantage of using chemical kinetics
when running full system simulation is that accuracy increases, more parameters are allowed
to change and more interesting inlet mixtures can be studied, on the other hand, computational
time increases. The geometry of the SCANIA DSC 12 engine (Table 7) is used in all
modeling although with compression ratio of 17.4:1 (except in one case were 16:1 is used)
and it is fitted with the smallest turbine (TC-3: 13 cm2 geometrical flow area). In addition, the
improved turbocharger model described in section 2.6.8 is used here in order to get better
accuracy in the interpolating regime in the maps.

6.1 Simple system with natural gas

A simple system with engine, turbocharger and a combined intercooler – heater
(“conditioner”, see section 5.2) is simulated using pure natural gas as fuel. The system setup
is shown earlier in Figure 87 and the inlet temperature is used for combustion timing control
as previously. The purpose is to have this simulation as a reference case for further
simulations.
The simulation input, a selected λ-value over time, is shown in Figure 92, in addition the
actual λ value is also shown. The compressor model gives a slight oscillation in mass flow
and this is why the calculated λ-value seems “noisy”, almost like experimental recordings.
The chemical kinetics gives some hystheresis effect in combustion timing, which also give a
noise effect on IMEP, as seen in Figure 92. The boost pressure is very limited and at λ = 2.1
only 3.3 bar BMEP is reached. The inlet temperature is 395K at this point. Previously, the
same system reached 9.7 bar BMEP, 40.3% brake efficiency and 0.162 gNO/kWh at 1500
rpm and λ = 2.1, but in that case a simple model for the inlet temperature were adopted (see
section 5.1.2 and 5.2.2) and richer mixtures were allowed during the first period of the
simulation (λ = 2.0). It is possible that a richer mixture than λ = 2.1 may provide higher boost
pressure during the startup procedure and hence bring up the turbo-speed to a level were
leaner mixtures are allowed. Once the boost pressure goes up it is possible to reduce the inlet
temperature, which brings up the mass flow through the engine. The increased mass flow then
may provide enough energy to the turbine in order to maintain a high boost pressure. On the
other hand, λ = 2.1 is a very rich mixture for HCCI combustion (without the use of EGR or
any other dilute gas) and knocking combustion is very likely to occur.
A simple way of increasing the mass flow is to increase the engine speed, however at higher
speed the pumping and friction losses increases and that will be a penalty on brake efficiency.
In Table 15 are the results of various simulations with different air/fuel ratios shown, engine

100
speed and compression ratio. With higher speed, a slightly earlier timing is used. With the
higher mass flow, boost pressure eventually builds up properly and engine load increases.
With higher speeds, the pressure upstream the turbine also increases and the pumping losses
increase rapidly, at 2000rpm, the PMEP is about twice that of 1800rpm. Friction increases as
well, which reduces the overall efficiency at 2000rpm. This results comply with those in
section 3.4, when increasing the engine speed a bigger turbine as the TC-2 is more suitable.
The maximum pressure is higher than the selected “design limit” used earlier (200bar) and
this why leaner mixtures or a lower compression ratio should be selected. A compression ratio
of 16:1 is tested at 1800rpm and 220kW brake power is reached with 39.3% brake efficiency
and 0.04 gNO/kWh.
2.25 6

5
2.2

MEP [bar]
2.15
3
λ

2.1
2
IMEPnet
2.05 Set 1 BMEP
Actual
2 0
0 20 40 60 80 0 20 40 60 80
Time [s] Time [s]

Figure 92 Left - The λ value, both set-point and actual value. Right – Load expressed as
indicated – and brake mean effective pressure.

Table 15 Results with simple system using natural gas (T1 = 298K, p1 = p4 = 1.013bar) At
1500 rpm 13kW of inlet air heating is required (Tinlet>T2).
Speed pMax NOZeldovich dm/dttot TInlet
rC λ Run
[rpm] [bar] @CAD [ppm] [g/kWh] [g/s] [K]
1500 17.4 2.18 71 13 7.1 0.08 127 397 Reference
1800 17.4 2.33 212 11 3.8 0.03 456 360 Speed +
1800 16.0 2.30 199 11 5.4 0.04 456 374 Speed +, rC -
2000 17.4 2.45 223 11 1.4 0.01 541 357 Speed + +

Speed Pb
IMEP IMEP PMEP FMEP BMEP ηi,n ηb p2 p3 T2 T3 T4
gross net
[rpm] [kW] [bar] [bar] [bar] [bar] [bar] [%] [%] [bar] [bar] [K] [K] [K]
1500 43.5 5.00 4.68 0.313 1.71 2.98 41.9 26.6 1.14 1.35 326 660 636
1800 220 15.2 14.5 0.71 1.96 12.5 46.0 39.8 3.05 3.31 445 676 543
1800 220 15.1 14.5 0.65 1.96 12.5 45.2 39.3 3.15 3.36 452 698 559
2000 237 15.6 14.3 1.34 2.13 12.2 44.6 37.9 3.29 4.01 465 679 527

101
6.2 Reformed natural gas fuel
A way to overcome the problem with the high inlet temperature may be to change the ignition
characteristics of the fuel. A way to do that is to add other fuels that ignite more easily.
However, in most cases it is convenient and more economical feasible to use only one fuel.
Regarding natural gas and many other fuels it is possible to reform them into CO and H2. This
can be done through fuel rich pre-combustion in a so-called reformer. The fuel composition
leaving the reformer may vary due to load, operating pressures and temperatures. In order to
simplify the simulations here, it is assumed that the reformed fuel consists of 40% (volume)
CO and 60% H2 at all conditions. This “synthetic” fuel is then mixed with the natural gas in
order to study the possibility to use this fuel as an ignition improver.
Here is the SCANIA engine used again with the setup shown in Figure 87. A mixture of
natural gas and this fuel is used. Up to 95% reformed fuel fraction is used, if that is not
enough to achieve proper combustion timing, inlet air preheating is also used. As input to the
simulation the natural gas equivalent λ is used, λNG,set. This air-fuel ratio sets a desired
amount of heat added to the system. The fuel flow is then calculated based on this value and
compensated for the actual fuel heating value; the resulting λ become higher if reformed fuel
is used. The heating value for the reformed fuel is only 20.81 MJ/kg (natural gas: 47.8 MJ/kg)
but the stoichiometric air-fuel ratio is 5.63 (natural gas: 16.58) which means that both the fuel
mass flow and λ differs, given the same amount of fuel-bounded heat. The heating value for a
stoichiometric mixture also differs; it is 2.08 MJ/kg for reformed fuel and 2.72 MJ/kg for
natural gas. If all natural gas is replaced with reformed fuel, the mass flow and λ with
reformed, fuel (“ref.”) can be calculated through (assumed the same airflow):
QLHV , NG
m ref . = ⋅ m NG ≈ 2.30 ⋅ m NG
QLHV ,ref .
Eq.114
m air m ref . QLHV ,ref . ( A F )s ,NG
λref . = = ⋅ ⋅λ ≈ 1.28 ⋅ λNG
( A F )s,ref . QLHV , NG ( A F )s,ref . NG
The simulations are performed as previously with a richer mixture during startup, and then
leaning out to the desired set point after 34 seconds as shown in Figure 93. The actual
(resulting) λ is then a result of fuel heating value difference and the control dynamics.
The resulting load and efficiency are shown in Figure 94 and Figure 95. There is no
noticeable difference in indicated efficiency between the two shown cases. The brake
efficiency is higher with richer mixtures as expected. At λNG,set = 2.2, the load is 12.1 bar
BMEP with 39.5% brake efficiency, However, at this point there is 22 ppm NO resulting in
0.17 g NO/kWh. The maximum pressure is about 181 bar, if higher maximum pressures are
allowed it is possible to increase the amount of fuel. At λNG,set = 2.1 about 14.1 bar BMEP is
reached and the brake efficiency at this point is about 40.4%. However, the NO formation
also increases to 0.30 g/kWh. These results and more detailed data can be found in Table 16.
In these tables comparison are also made with natural gas only. One interesting thing is that
when inlet air preheating is not required, intercooling is used to some degree when running
with reformatted fuel. Still, the turbocharging is not sufficient. Smaller turbine is required but
unfortunately, there is not available data on smaller turbines.

102
 2.6 2.6
λNG,set = 2.2
2.5 λ = 2.3 2.5
NG,set

2.4 2.4
λNG,set

2.3 2.3

λ
2.2 2.2
λNG,set = 2.2
2.1 2.1 λ = 2.3
NG,set

2 2
0 20 40 60 80 0 20 40 60 80
Time [s] Time [s]

Figure 93 λ set-point (left) and actual value (right).

15
16
14
12
10
IMEPnet [bar]
BMEP [bar]

10
8

5 6
λNG,set = 2.2 4 λ = 2.2
NG,set
λ = 2.3 λ = 2.3
NG,set 2 NG,set

0 0
0 20 40 60 80 0 20 40 60 80
Time [s] Time [s]

Figure 94 Resulting BMEP (left) and IMEPnet (right).

0.45 0.5

0.4

0.35 0.45
i,n
ηb

0.3
η

0.25 0.4
λNG,set = 2.2 λNG,set = 2.2
0.2 λ = 2.3 λ = 2.3
NG,set NG,set

0.15 0.35
0 20 40 60 80 0 20 40 60 80
Time [s] Time [s]

Figure 95 Resulting brake - (left) and net indicated efficiency (right).

103
Table 16 Results using 95% reformed fuel and 5% natural gas at 1500 rpm, except first case
where only natural gas is used and showed for reference (T1 = 298K, p1 = p4 = 1.013bar).
pMax NOZeldovich dm/dttot TInlet
λNG,set λ Run
[bar] [ppm] [g/kWh] [g/s] [K]
2.3 2.30 199 5.4 0.04 456 374 Natural gas only, 1800rpm, 16:1
2.1 2.36 201 44 0.30 362 333 95% reformed fuel
2.2 2.47 181 24 0.17 330 336 95% reformed fuel
2.3 2.59 154 11 0.08 291 340 95% reformed fuel

Pb Speed
IMEP IMEP PMEP BMEP ηi,n ηb p2 p3 T2 T3 T4
λNG,set gross net
[kW] [rpm] [bar] [bar] [bar] [bar] [%] [%] [bar] [bar] [K] [K] [K]
2.3 220 1800 15.1 14.5 0.65 12.5 45.2 39.3 3.15 3.36 452 698 559
2.1 207 1500 16.0 15.8 0.199 14.1 45.2 40.4 2.78 2.71 435 714 597
2.2 177 1500 14.0 13.8 0.177 12.1 45.1 39.5 2.55 2.47 421 697 592
2.3 145 1500 11.8 11.6 0.164 9.90 45.1 38.4 2.27 2.21 403 680 589

100% 100%
η ηC
C
5% Emissions 5% Emissions

16.2% Heat transfer 15.1% Heat transfer

ηi,g ηi,g

31.3% Heat loss 34.4% Heat loss

ηi,n ηi,n
2.03% Pump work 0.355% Pump work
ηb ηb
6.13% Friction 4.89% Friction

39.3% 40.3%
Work Work

Figure 96 Sankey diagrams showing energy flow in the HCCI engine running at 1800 rpm
and with natural gas (left) and at 1500rpm using reformed fuel (right).

104
6.3 The HAM system in combination with reformed fuel
The HAM system may be a feasible way to increase the mass flow through the turbine and
simultaneously dilute the charge allowing higher engine loads without knocking combustion
or NOx formation. The HAM concept is described earlier and here it is implemented in a
whole turbocharged HCCI engine system simulation using the smallest turbine available (TC-
3), the setup is illustrated in Figure 97. The air after the compressor is humidified to 98%
relative humidity and then is the charged heated to give the necessary ignition criteria. The
temperature of the heated liquid water into the humidifier is assumed to have a temperature of
70°C. There is obviously other possible system buildup but this is perhaps the simplest one.
An inlet air heater for coarse control of combustion timing and reformed fuel is used for the
fine adjustment. It would be better to make use of as much cooling of the humidifier as
possible but that requires another ignition control parameter such as compression ratio, dual-
fuel, fuel additive or any other suitable parameter.
The basic idea is to get additional mass flow from the humidification without any extra effort
from the compressor. The extra water is used to dilute this mixture in a similar way as with
EGR and will probably also reduce the heat release rate; knocking combustion can be
avoided. The extra flow will also give extra work from the turbine and more power will be
delivered to the compressor. As shown earlier (Figure 79), the specific heat of the inlet air
mixture increases with more water, which results in lower indicated efficiency. Higher engine
output will however reduce the friction detrimental effect on brake efficiency (FMEP
assumed constant if constant engine speed).

Humidification Turbo -
system charger

Heat

Heat

Engine
H2O

Figure 97 HAM system for HCCI operation

Reformed fuel allow much lower inlet temperatures, together with the HAM concept effective
turbocharging is feasible through increased turbine mass flow. Water also suppresses NO
formation by replacing oxygen in the inlet mixture which allow lower λ’as. In this simulation
the effect on combustion duration cannot be seen but it would be reasonable to believe that
high amount of steam will slow down combustion. Three different λ - values are tested
expressed on natural gas base. This lambda value (λNG,set) is then compensated because of the

105
water content and the heating value of the fuel, hydrogen brings up the heating value for a
stoichiometric mixture while carbon monoxide has about the same heating value for a
stoichiometric mixture as natural gas. The fuel flow is then calculated and depending on the
response of the system an “actual” lambda system can be calculated. The simulation is, as
previously simulated as a startup procedure at 1500rpm. The λ - set point and the resulting λ
is shown in Figure 98. For the two leaner cases, a slightly richer mixture is used during the
first 34 seconds, then follows a 10 s ramp where the final desired value is set13. If the
maximum in-cylinder pressure reaches 200 bar the λ set-point is increased. The primary
control factor is the reformed fuel fraction but at all these cases some preheating was required;
the reformed fuel fraction was set to it maximum value of 95%. 5% natural gas is left to
simulate a control margin for cycle-to-cycle and cylinder-to-cylinder variation control. In
order to investigate high load operation, combustion is phased quite late, as control parameter
the crank angle for maximum pressure is used and the set-point is 13 CAD ATDC.
The resulting loads, both IMEPnet and BMEP are shown in Figure 99. With the richer mixture,
λ is reduced in order to limit maximum pressure to 200 bar so maximum load became about
14.7 bar BMEP at 1.71 bar FMEP. The efficiency also became higher with richer mixtures,
about 45.2% net indicated and 40.3% brake efficiency as seen in Figure 100. The reformed
fuel allowed reasonable inlet temperatures, which limited the thermal throttling. The
compressor operating point became close to optimal as seen in Figure 101. In Table 17 the
results of each simulation are shown. The pressure difference between engine inlet and
exhaust also became positive despite the small turbine (TC-3) indicating that an even smaller
turbine would be possible to use without higher pumping losses. This could allow higher
boost pressures, although maximum in-cylinder pressures are already reached, so this action
would require a lower compression ratio.
2.5 2.8 λNG,set = 2.2
λNG,set = 2.3
2.7 λ = 2.4
2.4 NG,set

2.6
2.3
λNG,set

2.5
λ

2.2
λNG,set = 2.2 2.4
λ = 2.3
2.1 NG,set
λ = 2.4 2.3
NG,set

2 2.2
0 20 40 60 80 100 0 20 40 60 80 100
Time [s] Time [s]

Figure 98 Left – Input to the simulation, λ for natural gas. Right - λ outcome, differs from set
value (λNG,set) due to vaporized water, fuel composition and maximum pressure limitation.

13
Required computational time (case hours:minutes): λ = 2.2 8:50, λ = 2.3 10:00, λ = 2.4 11:16, on a AMD
Athlon Thunderbird 1200MHz, 256MB RAM, Windows 98 and Matlab 5.3. Functions MEX-compiled when
possible. The computational time was estimated by looking at when the result files were created but the times are
underestimated; it was discovered that the computer clock were delayed about 2 hours (!) after running the three
simulations.

106
 15
16
14
12
10
IMEPnet [bar]

BMEP [bar]
10
8
6 λNG,set = 2.2
5 λNG,set = 2.2
λ = 2.3
4 NG,set λNG,set = 2.3
λ = 2.4
NG,set λNG,set = 2.4
2
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Time [s] Time [s]

Figure 99 Left – Indicated mean effective pressure. Right – Brake mean effective pressure,
FMEP is 1.71 bar
0.5 0.45

0.4

0.45 0.35
ηi,n

0.3
η

0.4 λNG,set = 2.2 0.25 λNG,set = 2.2

λ = 2.3 λ = 2.3
NG,set NG,set
λ = 2.4 0.2 λ = 2.4
NG,set NG,set

0.35 0.15
0 20 40 60 80 100 0 20 40 60 80 100
Time [s] Time [s]

Figure 100 Left – Indicated efficiency. Right – Brake efficiency.

Compressor Turbine
5 0.04
4.5
MFP [s*sqrt(K)/(s*Pa)]

4 0.03
3.5
3
0.02
π

2.5
2
1.5 0.01
Map Map
1 Simulation Simulation
0.5 0
0 200 400 600 800 0 20 40 60 80
CMF [g/s] SPP [1/(s*sqrt(K))]

Figure 101 Compressor map (left) and turbine map (right) with the simulation (λNG,set = 2.2)
shown.
As much as 6.1 vol.-% of the inlet mixture is water, the relative humidity is assumed to be 98
% after the humidifier. The temperature drops in the humidifier due to the vaporization (T2 →
T2,HAM). When the air passes the heater, the relative humidity goes down and the temperature
goes up depending on the heating power. The temperature in the inlet will besides the
preheating also be dependant on the temperature of the fuel, and the assumption here is that it

107
has a temperature of 20°C before entering the inlet manifold which means that the fuel
actually cools the inlet charge, this is especially clear at λNG,set = 2.2 .
The NO formation according to Zeldovich is quite low despite the low λ. Some of this can be
explained by the lower inlet temperatures than with natural gas only. Another factor may be
the dilute effect of the water vapor. At λNG,set = 2.2 the NO fraction in the exhaust gas were 12
ppm corresponding to a specific NO emission of 0.08g/kWh.
Table 17 Results with HAM using 95% reformed fuel and 5% natural gas. (T1 = 298K, p1 =
p4 = 1.013bar)
pMax NOZeldovich dm/dttot dm/dtfuel dm/dtH20,vap xH2O TInlet T2,HAM
λNG,set λ
[bar] [ppm] [g/kWh] [g/s] [g/s] [g/s] [%] [K] [K]
2.2 2.32 200 12 0.08 370 15.6 22.3 6.1 337 338
2.3 2.43 179 7 0.05 337 12.6 19.5 5.5 340 334
2.4 2.57 151 4 0.03 289 9.08 16.0 4.6 344 328

Pb PHeater
IMEP IMEP PMEP BMEP ηi,n ηb p2 p3 T2 T3 T4
λNG,set gross net
[kW] [kW] [bar] [bar] [bar] [bar] [%] [%] [bar] [bar] [K] [K] [K]
2.2 206 1.54 15.9 15.8 0.12 14.1 45.2 40.3 2.91 2.75 445 704 587
2.3 177 3.81 13.9 13.8 0.11 12.1 45.2 39.6 2.66 2.51 429 687 582
2.4 139 6.35 11.4 11.2 0.11 9.5 45.0 38.2 2.30 2.19 406 670 581

Figure 102 shows a Sankey plot illustrating the energy flow through the HCCI engine
compared to a simulation of the diesel engine, running at peak load at 1500 rpm. The heat
release parameters in the diesel engine simulation are adjusted to fit experimental data on
maximum pressure, IMEP, and combustion efficiency. In the HCCI case the emissions steals
a great deal of the energy and since the brake, output is less than in the diesel case the friction
part is higher. The diesel engine loose quite much of the energy through the exhaust due to a
slow combustion process, but gains some of this loss back due to a negative pumping work,
the net indicated efficiency is higher than the gross indicated efficiency. This is possible
because the higher exhaust temperatures, which makes it possible to use a much bigger
turbine.
In summary, HAM in the combination with reformatted fuel seems to be a promising solution
for small-scale power plant engines. With larger engines, it is a little bit questionable if HAM
is required or not because of less heat transfer. In the simulation here it is not stated
wherefrom the heater power for inlet air preheating is taken from. It is of cause suitable to use
a heat exchanger and take heat from the exhaust gas. It should be pointed out that this system
also requires a water supply of 22 g/s, about 80 liter/hour. However, during cold startup this
may be a problem because of the thermal inertia of the whole system. The best operating
point shown here provide 206 kW brake power, at 40.3 % brake efficiency and NO emissions
of 0.08g/kWh. A big question mark is what combustion duration we would expect at this load
and with this water content.

108
 100% 100%
ηC η
C 0.148% Emissions
5% Emissions
14.3% Heat transfer
15.1% Heat transfer
ηi,g
ηi,g

34.4% Heat loss 42% Heat loss

η η
i,n 0.355% Pump work i,n
η −1.17% Pump work
b
ηb 4.89% Friction 3.49% Friction

40.3% 41.2%
Work Work

Figure 102 Sankey diagram showing the energy flow through the HCCI engine (left) using
HAM and reformatted fuel (λNG,set = 2.2) at BMEP 14.5bar, compared to simulation of a
diesel engine at the same speed but at 20 bar BMEP (note the negative pump work),
emissions estimated from experiments .

109
7 Summary

This thesis presents the predicted performance and efficiency of turbocharged natural gas
driven Homogeneous Charge Compression Ignition (HCCI) engine systems. From these
simulated results, we can expect that the engine load in terms of brake mean effective
pressure will be lower than with conventional diesel engines, mainly due to the maximum in-
cylinder pressure. The heat release rate during HCCI combustion is very high compared to
diesel combustion and the peak pressures become higher. Nevertheless, for this rather small
stationary engine it is possible to reach efficiency comparable to diesel engines, and with an
acceptable specific power. The real benefit of HCCI is the low NOx emissions. The Zeldovich
model used here give only an estimation of the NO formation, however this can still be used
as an indicator of when NOx starts to form.
The most interesting cases are summarized in Table 18 and Table 19. In Table 18,
comparisons are made at constant air-fuel equivalence ratio (normalized to natural gas) and in
Table 19 at constant maximum pressure. The first case is the simple system with inlet air
conditioner. In this case, the engine speed had to be increased in order to get proper
turbocharging. The compression ratio was decreased in order to stay below a fictive design
limit of 200bar peak in-cylinder pressure. Increasing the speed to 1800rpm gave both higher
friction and pumping losses and the efficiency became lower.
Keeping the speed at 1500rpm and using a higher compression ratio of 17.4:1 and instead
affecting the ignition characteristics of the fuel gave other results. This was done by simulate
reformed natural gas, a CO (40 vol.%) and H2 (60 vol.%) mixture. The ignition characteristics
of the reformed fuel allowed lower inlet temperatures to be used and the mass flow through
the turbine could be increased trough less thermal throttling of the engine. The boost pressure
was lower compared to the high-speed case; the load also became lower. Using richer
mixtures obviously produced more power and gave better efficiency, but the NO formation
increased from 0.08g/kWh to 0.30g/kWh. The brake efficiency reached above 40% in the best
operating conditions.
By also using the Humid Air Motor (HAM) concept, the mass flow through the engine and
turbine could be increased at leaner mixtures. Engine load increased and hence also
efficiency. Using a richer mixture, the load became higher as expected, and performance and
efficiency became comparable with the case without HAM. On the other hand, the NO
formation became much lower. Despite the richer mixture, the NO stayed at 0.08g/kWh. This
final case would be even more interesting if the combustion duration were affected (longer)
by the increased humidity. This could not be predicted with the zero dimensional models
used.

110
Table 18 Summary of full system simulation, keeping λNG,set at 2.3
IMEP
Speed pMax NOZeldovich Pb PHeater BMEP ηb
Fuel/Case rC λ net
[rpm] [bar] [g/kWh] [kW] [kW] [bar] [bar] [%]
NG 16.0 1800 2.33 199 0.04 220 - 14.5 12.5 39.3
Reformed 17.4 1500 2.59 154 0.08 145 - 11.6 9.9 38.4
Ref. + HAM 17.4 1500 2.43 179 0.05 177 3.81 13.8 12.1 39.6

Table 19 Summary of full system simulation, keeping pMax about 200bar

IMEP
Speed pMax NOZeldovich Pb PHeater BMEP ηb
Fuel/Case rC λ net
[rpm] [bar] [g/kWh] [kW] [kW] [bar] [bar] [%]
NG 16.0 1800 2.33 199 0.04 220 - 14.5 12.5 39.3
Reformed 17.4 1500 2.36 201 0.30 207 - 15.8 14.1 40.4
Ref. + HAM 17.4 1500 2.32 200 0.08 206 1.54 15.8 14.1 40.3

Based on these results a suggestion would be to try these concepts on a real engine. If
reformatted fuel is not possible, a temporary action could be to speed up the engine, if
possible, in order to build up boost and then use the HAM concept to go down in engine
speed and maintain boost pressure.
Some facts points towards better results in reality e.g. the limitation of the model about the
overestimation of the inlet temperature indicate that the mass flow through the engine will
become higher. Using HAM may affect combustion duration allowing slightly richer mixtures
to be used without reaching maximum pressure, significant NOx formation or knocking
combustion. This would lead to higher output and higher efficiency.
If bigger cylinders (larger displaced volume per cylinder) were used scale effects would
probably appear. Less heat transfer would lead to higher efficiency, lower inlet temperatures
and higher exhaust temperatures leading to less thermal throttling and less pumping losses if
the compressor and turbine is matched properly.

111
8 References

1. “Pure Energy, Gas Cogeneration Plants”, pamphlet from Wärtsilä NSD Corporation,
November 1997 edition
2. O. Erlandsson, ”Alvar engine An engine with variable compression ratio Experiments
and test”, ISRN LUTMDN/TMVK-5304--SE
3. O. Erlandsson, Gunnar Lundholm, Fredrik Söderberg, Bengt Johansson,Victor W.
Wong ”Demonstarting the Performance and Emission Characteristics of a Variable
Compression Ratio, Alvar-Cycle Engine”, SAE 982682
4. S. Onishi, S. Hong Jo, K. Shoda, P Do Jo, S. Kato: ”Active Thermo-Atmosphere
Combustion (ATAC) – A New Combustion Process for Internal Combustion Engines”,
SAE790501
5. M. Noguchi, Y. Tanaka, T. Tanaka, Y. Takeuchi, "A Study on Gasoline Engine
Combustion by Observation of Intermediate Reactive Products during Combustion",
SAE 790840, 1979
6. N. Iida, "Combustion Analysis of Methanol-Fueled Active Thermo-Athmosphere
Combustion (ATAC) Engine Using a Spectroscopic Observation", SAE 940684, 1994
7. P.Duret, S. Venturi, "Automotive Calibration of the IAPAC Fluid Dynamically
Controlled Two-Stroke Combustion Process", SAE 960363, 1996
8. Y. Ishibashi, M. Asai, "Improving the Exhaust Emissions of Two-Stroke Engines by
Applying the Activated Radical Combustion", SAE 960742, 1996
9. J. Yamaguchi, "Honda readies Activated Radical Combustion two-stroke engine for
production motorcycle", Automotive Engineering, January, 1997
10. N. Iida, T. Ichikura, K. Kase, Y. Enomoto, "Self-ignition and combustion stability in a
methanol fueled low heat rejection ceramic ATAC engine- analysis of cyclic variation
at high wall temperatures and lean burn operation.", JSAE 9733684, 1997
11. R. Gentili, S. Frigo, L. Tognotti, P. Habert, J. Lavy, "Experimental Study on ATAC in a
Two-Stroke Gasoline Engine", SAE 970363, 1997
12. N. Iida, "Alternative Fuels and Homogeneous Charge Compression Ignition
Combustion Technology", SAE 972071, 1997
13. Y. Ishibashi, M. Asai, K. Nishida, "An Experimental Study of Stratified Scavenging
Activated Radical Combustion Engine", SAE 972077, 1997
14. H. Oguma, T. Ichikura, N. Iida, "A Study on Adaptability of Alternative Fuels for Lean
Burn Two-Stroke ATAC Engine", SAE 972097, 1997
15. E. Esterlingot, "Thermodynamical and Optical Analysis of Controlled Auto-Ignition
Combustion in Two-Stroke Engines", SAE 972098, 1997
16. Y. Ishibashi, M. Asai, "A Low Pressure Pneumatic Direct Injection Two-Stroke Engine
by Activated Radical Combustion Concept", SAE 980757, 1998

112
17. A. Hultqvist, “Characterization of the Homogeneous Charge Compression Ignition
Combustion Process”, Doctoral Thesis, ISRN LUTMDN/TMVK-1030-SE
18. Homepage of Progress Aero Works, www.paw.ac, 2002-03-20
19. P.M. Najt, D. E. Foster, "Compression-Ignited Homogeneous Charge Combustion",
SAE 830264, 1983
20. R.H. Thring, "Homogeneous Charge Compression Ignition (HCCI) Engines", SAE
892068, 1989
21. V. M. Stockinger, H. Schäpertöns, P. Kuhlmann, "Versuche an einem
gemischansaugenden Verbrennungsmotor mit Selbstzündung", MTZ 53, 1992
22. H. Shoji, Y. Tosaka, K. Yoshida, A. Saima, "Radical Behavior in Preflame Reactions
Under Knocking Operation in a Spark Ignition Engine", SAE 942061, 1994
23. T. Aoyama, Y. Hattori, J. Mizuta, Y. Sato, "An Experimental Study on Premixed-
Charge Compression Ignition Gasoline Engine", SAE 960081, 1996
24. J. R. Smith, S. M. Aceves, C. Westbrook, W. Pitz, "Modeling of Homogeneous Charge
Compression Ignition (HCCI) of Methane", ASME International Engine 1997 Fall
Conference, Madison, Wisconsin, 1997
25. H. Yanagihara, Y. Sato, J. Mizuta, "A study of DI diesel combustion unders uniform
higher dispersed mixture formation", JSAE 9725853, JSAE 9733675 1997
26. H. Suzuki, N. Koike, H. Ishii, M. Odaka, "Exhaust Purification of Diesel Engines by
Homogeneous Charge with Compression Ignition Part 1", SAE 970313, 1997
27. H. Ishii, N. Koike, H. Suzuki, M. Odaka, "Exhaust Purification of Diesel Engines by
Homogeneous Charge with Compression Ignition Part 2", SAE 970315, 1997
28. K. Nakagome, N. Shimazaki, K. Niimura, S. Kobayashi, "Combustion and Emission
Characteristics of Premixed Lean Diesel Combustion Engine", SAE 970898, 1997
29. A.W. Grey III, T. W. Ryan III , "Homogeneous Charge Compression Ignition of Diesel
Fuel", SAE 971676, 1997
30. M. Christensen, B. Johansson, P. Einewall, "Homogeneous Charge Compression
Ignition (HCCI) Using Isooctane, Ethanol and Natural Gas - A Comparison with Spark
Ignition Operation", SAE 972874, 1997
31. P.Amnéus, D. Nilsson, F. Mauss, M. Christensen, B. Johansson, "Homogeneous Charge
Compression Ignition Engine: Experiments and Detailed Kinetic Calculations",
COMODIA, 1998
32. T. Hashizume, T. Miyamoto, H. Akagawa, K. Tsujimura, "Combustion and Emission
Characteristics of Multiple Stage Diesel Combustion", SAE 980505, 1998
33. H. Suzuki, N. Koike, M. Odaka, "Combustion Control Method of Homogeneous Charge
Diesel Engine", SAE 980509, 1998
34. T. Seko, E. Kuroda, Y. Hamano, "Methanol Lean Burn in an Auto-Ignition DI Engine",
SAE 980531, 1998
35. A. Harada, N. Shimazaki, S. Sasaki, T. Miyamoto, H. Akagawa, K. Tsujimura, "The
Effects of Mixture Formation on Premixed Lean Diesel Combustion", SAE 980533,
1998
36. M. Christensen, B. Johansson, P. Amneus, F. Mauss, "Supercharged Homogeneous
Charge Compression Ignition", SAE 980787, 1998
37. Y. Mase, J. Kawashima, T. Sato, M. Eguchi, "Nissan's New Multivalve DI Diesel
Engine Series", SAE 981039, 1998

113
38. M. Christensen, B. Johansson, "Influence Of Mixture Quality on Homogeneous Charge
Compression Ignition", SAE 982454, 1998
39. J. Willand, R-G. Nieberding, G. Vent, C. Enderle, "The Knock Syndrome - Its Cure and
Its Potential", SAE 982483, 1998
40. N. Shimazaki, H. Akagawa, K. Tsujimura, "An Experimental Study of Premixed Lean
Diesel Combustion", SAE 1999-01-0181, 1999
41. M. Christensen, B. Johansson, "Homogeneous Charge Compression Ignition with Water
Injection", SAE 1999-01-0182, 1999
42. H. Akagawa, T. Miyamoto, A. Harada, S. Sasaki, N. Shimazaki, T. Hashizume, K.
Tsujimura, "Approaches to Solve Problems of the Premixed Lean Diesel Combustion",
SAE 1999-01-0183, 1999
43. M. Odaka, H. Suzuki, N. Koike, H. Ishii, "Search for Optimizing Control Method of
Homogeneous Charge Diesel Combustion", SAE 1999-01-0184, 1999
44. Y. Iwabuchi, K. Kawai, T. Shoji, Y. Takeda, "Trial of New Concept Diesel
COmbustion System - Premixed Compression -Ignited Combustion", SAE 1999-01-
0185, 1999
45. T. Miyamoto, A. K. Hayashi, A. Harada, S. Susaki, K. Tsujimura, "A Computational
Investigation of Premixed Lean Diesel Combustion - characteristics of fuel-air mixture
formation, combustion and emissions", SAE 1999-01-0229, 1999
46. M. Richter, A. Franke, M. Aldén, A. Hultqvis, B. Johansson, "Optical Diagnostics
Applied to a Natuarlly Aspirated Homogeneous Charge Compression Ignition Engine",
SAE 1999-01-3649, 1999
47. M. Christensen, A. Hultqvist, B. Johansson, "Demonstrating the multi fuel capability of
a homogeneous charge compression ignition engine with variable compression ratio",
SAE 1999-01-3679, 1999
48. A. Hultquist, M. Christensen, B. Johansson, A. Franke, M. Richter, M. Aldén, "A Study
of the Homogeneous Charge Compression Ignition Combustion Process by
Chemiluminescence Imaging", SAE 1999-01-3680, 1999
49. S. Kimura, O. Aoki, H. Ogawa, S. Muranaka, Y. Enomoto, "New Combustion Concept
for Ultra-Clean and High-Efficiency Small DI Diesel Engines", SAE 1999-01-3681,
1999
50. R. H. Stanglmaier, C. E. Roberts, "Homogeneous Charge Compression Ignition
(HCCI): Benefits, Compromises and future Engine Applications", SAE 1999-01-3682,
1999
51. T. Løvås, D. Nilsson, P. Amnéus, F. Mauss, "Reduction of Complex Fuel Chemistry for
Simulation of Combustion in an HCCI Engine", VDI Berichte Nr. 1492, p. 511, 1999
52. P. Maigaard, F. Mauss, M. Kraft, "Homogeneous Chrge Compression Ignition Engine:
A Simulation Study on the Effects of Inhomogenities", ASME 2000 Spring Engine
Technology Conference of the Internal Combustion Engine Division: April 9-12, 2000
53. J. Martinez-Frias, S. M. Aceves, D. Flowers, J. R. Smith, R. Dibble, "Exhaust Energy
Recovery for Control of a Homogeneous Charge Compression Ignition Engine", ASME
International Mechanical Engineering Congress and Exposition: Orlando, Florida,
November 5-10, 2000
54. A. Hultqvist, M. Christensen, M. Richter, A. Franke, B. Johansson, "Near-Wall
Combustion in a Homogeneous Charge Compression Ignition (HCCI) Engine", AVL 4.
Internationales Symposium Für Verbrennungsdiagnostik, Kurhaus, Baden-Baden, 18-19
May, 2000

114
55. S. M. Aceves, D. L. Flowers, C. K. Westbrook, J. R. Smith, W. Pitz, R. Dibble, M.
Christensen, B. Johansson, "A multi-zone model for prediction of HCCI combustion
and Emissions", SAE 2000-01-0327, 2000
56. D. Flowers, S. Aceves, R. Smith, J. Torres, J. Girard, R. Dibble, "HCCI in a CFR
Engine: Experiments and Detailed Kinetic Modelling", SAE 2000-01-0328, 2000
57. Z.Chen, M. Konno, M. Oguma, T. Yanai, "Experimental Study of CI Natural Gas/DME
Homogeneous Charge Engine", SAE 2000-01-0329, 2000
58. A. Higashino, H. Sasaki, K. Kishiishita, S. Sekiyama, H. Kawamura, H. Iwamoto,
"Compression Ignition Combustion in a Prechambered and Heat Insulated Engine Using
a Homogeneous Natural Gas Mixture", SAE 2000-01-0330, 2000
59. T. Tsurushima, T. Miyamoto. H. Akagawa, Y. Aoyagi, J-H Lee, S. Goto, T. Wakisaka,
T. Ishiyama, H. Kawanabe, "Effects of Initial In-Cylinder Flow Field om Mixture
Formation i a Premixed Compression Ignition Engine", SAE 2000-01-0331, 2000
60. S. B.Fiveland, D. N. Assanis, "A Four-Stroke Homogeneous Charge Compression
Ignition Engine Simulation for Combustion and Performance Studies", SAE 2000-01-
0332, 2000
61. J-H Lee, S. Goto, T. Tsurushima, T. Miyamoto, T. Wakisaka, "Effects of Injection
Conditions on Mixture Formation Process in a Premixed Compression Ignition Engine",
SAE 2000-01-1831, 2000
62. N. Iida, T. Igarashi, "Auto-Ignition of n-Butane and DME/Air Mixtures in a
Homogeneous Charge Compression Ignition Engine", SAE 2000-01-1832, 2000
63. A.Hultquist, M. Christensen, B. Johansson, "The Application of Ceramic and Catalytic
Coatings to Reduce the Unburned Hydrocarbon Emissions from a Homogeneous
Charge Compression ignition Engine", SAE 2000-01-1833, 2000
64. Y. Yamasaki, N. Iida, "Numerical Simulation of Auto-Ignition and combustion of n-
Butane and Air Mixtures in a 4 Stroke HCCI Engine by Using Elementary Reactions",
SAE 2000-01-1834, 2000
65. M. Christensen, B. Johansson, P. Amneus, F. Mauss, "Supercharged Homogeneous
Charge Compression Ignition (HCCI) with Exhaust Gas Recirculation and Pilot Fuel",
SAE 2000-01-1835, 2000
66. J. Lavy, J-C. Dabadie, C. Angelberger, P. Duret, J. Willand, A. Juretzka, J. Schäflein, T.
Ma, Y. Lendresse, A. Satre, C. Schulz, H. Krämer, H. Zhao, L. Damiano, "Innovative
Ultra-low Nox Auto-Ignition Combustion Process for gasoline Engines: the 4-SPACE
Project", SAE 2000-01-1837, 2000
67. O. Erlandsson, B. Johansson, F. A. Silversand, "Hydrocarbon (HC) Reduction of
Exhaust Gases from a Homogeneous Charge Compression Ignition (HCCI) Engine
Using Different Catalytic Mesh-Coatings", SAE 2000-01-1847, 2000
68. J. - O. Olsson, O. Erlandsson, B. Johansson, "Experiments and Simulation of a Six-
Cylinder Homogeneous Charge Compression Ignition (HCCI) Engine", SAE 2000-01-
2867, 2000
69. M. Richter, J. Engström, A. Franke, M. Aldén, A. Hultqvist, B. Johansson, "The
influence of charge inhomogenity on the HCCI combustion process", SAE 2000-01-
2868, 2000
70. J. Martinez-Frias, S. Aceves, D. Flowers, J. R. Smith, R. Dibble, "HCCI Engine Control
by Thermal Management", SAE 2000-01-2869, 2000
71. G. Kontarakis, N. Collins, T. Ma, "Demonstration of HCCI Using a Single Cylinder
Four-stroke Engine", SAE 2000-01-2870, 2000

115
72. S. Kimura, O. Aoki, Y. Kitahara, E. Aiyoshizawa, "Ultra-Clean Combustion
Technology Combining a Low-Temperature and Premixed Combustion Concept for
Meeting Future Emissions Standards", SAE 2001-01-0200, 2001
73. T. Noda, D. E. Foster, "A Numerical Study to Control Combustion Duration of
Hydrogen-fueled HCCI by Using Multi-zone Chemical Kinetics Simulation", SAE
2001-01-0250, 2001
74. D. Law, D. Kemp, J. Allen, G. Kirkpatrick, T. Copland, "Controlled Combustion in an
IC-Engine with a Fully Variable Valve Train", SAE 2001-01-0251, 2001
75. T. Amano, S. Morimoto, Y. Kawabata, "Modeling of the Effect of Air/Fuel Ratio and
Temperature Distribution on HCCI Engines", SAE 2001-01-1024, 2001
76. J. Zheng, W. Yang, D. L. Miller, N. P. Cernansky, "Prediction of Pre-ignition
Reactivity and Ignition Delay for HCCI Using a Reduced Chemical Kinetic model",
SAE 2001-01-1025, 2001
77. S.- K. Kong, C. D. Marriott, R. D. Reitz, M. Christensen , "Modelling and Experiments
oa HCCI Engine Combustion Using Detailed Chemical Kinetics with Multidimensional
CFD", SAE 2001-01-1026, 2001
78. S. M. Aceves, D. L. Flowers, J. Martinez-Frias, J. R. Smith, C. K. Westbrook, W. J.
Pitz, R. Dibble, J. F. Wright, W. C. Akinyemi, R. P. Hessel, "A Sequential Fluid-
Mechanic Chemical-Kinetic Model of Propane HCCI Combustion", SAE 2001-01-
1027, 2001
79. S. B. Fiveland, D. N. Assanis, "Development of a Two-Zone HCCI Combustion Model
Accounting for Boundary Layer Effects", SAE 2001-01-1028, 2001
80. W. L. Easley, A. Agarwal, G. A. Lavoie, "Modeling of HCCI combustion and
Emissions Using Detailed Chemistry", SAE 2001-01-1029, 2001
81. A. Oakley, H. Zhao, N. Ladommatos, T. Ma, "Experimental Studies on Controlled
Auto-ignition (CAI) Combustion of Gasoline in a 4-Stroke Engine", SAE 2001-01-
1030, 2001
82. J. - O. Olsson, P. Tunestål, B. Johansson, "Closed-Loop Contol of an HCCI Engine",
SAE 2001-01-1031, 2001
83. A. Hultqvist, U. Engdar, B. Johansson, J. Klingmann, "Reacting Boundary Layers in a
Homogeneous Charge Compression Ignition (HCCI) Engine", SAE 2001-01-1032,
2001
84. K. Tanaka, H. Endo, A. Imamichi, Y. Oda, Y. Takeda, T. Shimada, "Study of
Homogeneous Charge Compression Ignition Using a Rapid Compression Machine",
SAE 2001-01-1033, 2001
85. S. S. Morimoto, Y. Kawabata, T. Sakurai, T. Amano, "Operating Characteristics of a
Natural Gas-Fired Homogeneous Charge Compression Ignition Engine (Performance
Improvment Using EGR)", SAE 2001-01-1034, 2001
86. N. Milovanovic, R. Chen, "A Review of Experimental and Simulation Studies on
Controlled Auto-Ignition Combustion", SAE 2001-01-1890, 2001
87. T. Tsurushima, A. Harada, Y. Iwashiro, Y. Enomoto, Y. Asaaumi, Y. Aoyagi,
"Thermodynamic Characteristics of Premixed Compression Ignition Combustions",
SAE 2001-01-1891, 2001
88. Y. Nashijima, Y. Asaumi, Y. Aoyagi, "Premixed Lean Diesel Combustion (PREDIC)
using Impingement Spray System", SAE 2001-01-1892, 2001

116
89. M .Christensen, B. Johansson, A. Hultqvist, "The Effect of Topland Geometry on
Emissions of Unburned Hydrocarbons from a Homogeneous Charge Compression
(HCCI) Engine", SAE 2001-01-1893, 2001
90. M. Y. Au, J. W. Girard, R. Dibble, D. Flowers, S. M. Aceves, J. Martinez-Frias, R.
Smith, C. Seibel, U. Maas, "1.9-Liter Four-Cylinder HCCI Engine Operation with
Exhaust Gas Recirculation", SAE 2001-01-1894, 2001
91. D. Flowers, S. M. Aceves, J. Martinez-Frias, J. R. Smith, M. Au, J. W. Girard, R.
Dibble, "Operation of a Four-Cylinder 1.9L Propane Fueled Homogeneous Charge
Compression Ignition Engine: Basic Operating Characteristics and Cylinder-to-Cylinder
Effects", SAE 2001-01-1895, 2001
92. J. - O. Olsson, P. Tunestål, G. Haraldsson, B. Johansson, "A Turbocharged Dual Fuel
HCCI Engine", SAE 2001-01-1896, 2001
93. R. H. Stranglmaier, T. W. Ryan III, J. S. Souder, "HCCI Operation of a Dual-Fuel
Natural Gas Engine for Improved Fuel Efficency and Ultra-Low NOx Emissions at Low
to Moderate Engine Loads", SAE 2001-01-1897, 2001
94. M. Christensen, ”HCCI Combustion – Engine Operation and Emission Characteristics”,
Doctoral Thesis, 2002, Lund Institute of Technology, Lund, Sweden, ISBN 91-628-
5424-0
95. O. Erlandsson, B. Johansson, F. Silversand, ”Hydrocarbon (HC) Reduction of Exhaust
Gases from a Homogeneous Charge Compression Ignition (HCCI) Engine Using
Different Catalytic Mesh-Coatings”, SAE 2000-01-1847
96. J. B. Heywood; ”Internal Combustion Engine Fundamentals”, ISBN 0-07-100499-8
97. C. L. Cummins Jr.; "Internal fire", ISBN 0-89883-765-0
98. B. J. McBride, S. Gordon, M. A. Reno; ”Coefficients for Calculating Thermodynamic
and Transport Properties of Individual Species”, NASA Technical Memorandum 4513
99. Yunus A. Çengel, Michael A. Boles; ”Thermodynamics: An Engineering Approach”,
ISBN 0-07-113249-X
100. O. Erlandsson, “Development of an Engine System Simulation Software Package –
ESIM”, Licentiate thesis, Lund Institute of Technology, ISRN/LUTMDN/TMVK -
7043
101. C. R. Ferguson; ”Internal Combustion Engines - Applied Thermosciences”, ISBN 0-
471-83705-9
102. Fredrik Söderberg; ”A Study of Fluid Flow and Combustion with Variable Valve
Timing”, ISRN LUTMDN/TMHT-98/7029--SE
103. I. I. Vibe; ”Brennverlauf und Kreisprozeß von Verbrennungsmotoren”, VEB Verlag
Technik Berlin
104. G. Woschni; ”A Universally Applicable Equation for the Instantaneous Heat Transfer
Coefficient in the Internal Combustion Engine”, SAE 670931
105. J. A. Caton, J. B. Heywood; ”An Experimental and Analytical Study of Heat Transfer in
an Engine Exhaust Port”, Int. J. Heat Mass Transfer Vol 24, No. 4, pp. 581-595, 1981
106. R. Egnell; ”A Simple Approach to Studying the Relation between Fuel Rate Heat
Release and NO Formation in Diesel Engines”, SAE 1999-01-3548
107. W. W and M. Ross; “Modeling of Direct Injection Diesel Engine Fuel Consumption”,
SAE 971142
108. M. Müller, E. Hendricks, S. C. Sorenson; ”Mean Value Modelling of Turbocharged
Spark Ignition Engines”, SAE 980784

117
109. N. Watson, M. S. Janota, “Turbocharging the Internal Combustion Engine”, 1982,
ISBN 0-333-24290-4
110. The NIST Reference on Constant, Units and Uncertainty; http://physics.nist.gov
111. C. Nellen, K. Boulouchos, “Natural Gas Engines for Cogeneration: Highest Efficiency
and Near-Zero-Emissions through Turbocharging, EGR and 3-way Catalytic
Converter”, SAE 2000-01-2825
112. P. Amneus, "Homogeneous Ignition, Chemical Kinetic Studies for IC-Engine
Applications", Doctoral Thesis, ISRN LUTFD2/TFCP-81-SE, Lund Institute of
Technology, Lund, Sweden
113. J. Warnatz, U. Maas, R. W. Dibble, “Combustion, - Physical and Chemical
Fundamentals, Modelling and Simulation, Experiments, Pollutant Formation”, 1995,
ISBN 3-540-60730-7
114. S. B. Fiveland, R. Agama, M. Christensen, B. Johansson, J. Hiltner, F. Mauss, D. N.
Assanis, “Experimental and Simulated Results Detailing the Sensitivity of Natural Gas
HCCI Engines to Fuel Composition”, SAE 2001-01-3609
115. DONG Energi-Service, Gas quality readings June 2002; http://www.dong.dk,
116. J. - O. Olsson, P. Tunestål, B. Johansson, S. Fiveland, R. Agama, M. Willi, D. Assanis,
“Compression Ratio Influence on Maximum Load of a Natural Gas Fueled HCCI
Engine”, SAE 2002-01-0111
117. L. Koopmans, I. Denbratt, “A Four Stroke Camless Engine, Operated in Homogeneous
Charge Compression Ignition Mode with Commercial Gasoline”, SAE 2001-01-3610
118. L. Koopmans, O. Backlund, I. Denbratt, “Cycle to Cycle Variations: Their Influence on
Cycle Resolved Gas Temperature and Unburned Hydrocarbons from a Camless
Gasoline Compression Ignition Engine”, SAE 2002-01-0110.
119. O. Erlandsson, “Early Swedish Hot-Bulb Engines – Efficiency and Performance
Compared to Contemporary Gasoline and Diesel Engines”, SAE 2002-01-0115
120. P. Rosén, L.-O. Olsson, “Method for supplying vapor to the intake of an internal
combustion engine, and device therefore”, Patent WO 95/23286, 31 August 1995
121. T. Lindquist, P. Rosén, L. - O. Olsson, “Humid Cycles, A Novel Way To Increase
Performance Of Marine Propulsion Systems”, CIMAC World Congress on Combustion
Engine Technology, May 2001, Germany
122. Munters AB, Pre-cooling of diesel engines - H.A.M.; http://www.munters.com/

118
Symbols and Abbreviations

MEP – Mean Effective Pressures

BMEP Brake MEP
FMEP Friction MEP (mechanical friction)
FuelMEP Fuel MEP
IMEPnet Net Indicated MEP
IMEPgross Gross Indicated MEP (without the gas exchange strokes)
QhrMEP Heat Release MEP
QhtMEP Heat Transfer MEP
QlossMEP Heat Loss MEP
QemisMEP Emission MEP (based on heat chemically bounded in emissions)
PMEP Pump MEP
[Pa] or [bar], 1bar = 105Pa ≈ 1atm

η – Efficiencies
ηb Brake efficiency
ηc Combustion efficiency
ηC , ηT Isentropic efficiency for compressor and turbine
ηG.e. Gas exchange efficiency
ηi,n Net indicated efficiency
ηi,g Gross indicated efficiency
ηM,e Mechanical efficiency
ηTh. Thermal efficiency
ηV Volumetric efficiency

Other symbols
α Convective heat transfer coefficient [W/(m2⋅K)]
φ Fuel-air equivalence ratio, (mf/ma)/(mf/ma)s
γ Ratio of specific heats
η Dynamical viscosity [µP] or [Pa⋅s], 1µP = 10-7 Pa⋅s
λ Thermal conductivity [W/(m⋅K)]
λ Air-fuel equivalence ratio, (ma/mf)/(ma/mf)s
ν Kinematical viscosity
πC, πT Pressure ratio for compressor and turbine
ρ Density [g/m3] or [kg/m3]
τ Time period [s]
ω Rotational speed [1/s]

119
A Area [m2]
Amin Minimum geometrical valve flow area [m2]
AE Effective valve flow area [m2]
Ap Port area [m2]
A0, A10 Crank angle for 0 and 10 % burned mass fraction [CAD]
A50, A90 Crank angle for 50 and 90 % burned mass fraction [CAD]
BDC Bottom dead center
CD Discharge Coefficient
Cp Specific heat at constant pressure [J/(mol⋅K)]
CMF Corrected mass flow [g/s]
D Cylinder bore [m]
D0-10, D10-90 Duration 0 – 10% and 10-90% burned mass fraction [CAD]
Dp Port diameter [m]
E Total energy [J]
EVO, EVC Exhaust valve opening and closing (in crank angle degrees)
e Total energy [J/mol]
h0 Enthalpy [J/mol]
h1 , h2 Enthalpy upstream and downstream compressor [J/mol]
h3 , h4 Enthalpy upstream and downstream turbine [J/mol]
h2,s , h4,2 Isentropic enthalpy downstream compressor and turbine [J/mol]
hc Convective heat transfer coefficient [W/(m2⋅K)]
H Enthalpy [J]
IVO, IVC Intake valve opening and closing (in crank angle degrees)
JTC Inertia of the turbo shaft [kg⋅m2]
l Stroke [m]
Lref Characteristic length of the vessel [m]
M Mole weight [g/mol]
m Mass, [g] or [kg]
ma Air mass, [g] or [kg]
mf Fuel mass, [g] or [kg]
MFP Mass flow parameter [g⋅sqrt(K)/(s⋅Pa)]
n Moles [mol]
n Engine speed [rpm]
nR Revolutions per cycle
N Number of species
Nu Nusselt number, Nu = α⋅Lref/λ
p Pressure [Pa]
p1 , p2 Pressure upstream and downstream compressor [Pa]
p3 , p4 Pressure upstream and downstream turbine [Pa]
pcr , pr Critical and reduced pressure [Pa]
Pb Brake power, [W] or [kW]
Q Heat [J]
QLHV Lower heating value of the fuel, [J/g ] or [J/kg]
Qhr Heat released during combustion [J]
R Universal gas constant [J/(mol⋅K)]
rC Compression ratio
Re Reynolds number, Re = ρ⋅Uref⋅Lref/η
s0 Specific entropy [J/(mol⋅K)]
Sp Mean piston speed [m/s], S p = 2 ⋅ l ⋅ n 60

120
SPP Speed parameter [1/(s⋅sqrt(K))]
t Time [s]
T Temperature [K]
T1 , T2 Temperature upstream and downstream compressor [K]
T3 , T4 Temperature upstream and downstream turbine [K]
Tb Brake torque, [Nm]
Tcr, Tr Critical and reduced temperature [K]
u Internal energy [J/mol]
U Internal energy [J]
Uref Characteristic flow velocity [m/s]
V Volume [m3]
Vd, Vd,tot Displaced volume of the whole engine, or a single cylinder [m3]
W Work [m3]
w Characteristic gas velocity [m/s]
x Piston position
xb Burned mass fraction
yi Mole fraction
Z Compressibility factor
See also figures for definitions not covered in the list, e.g. Figure 13 and Figure 14.

Abbrevations
ATAC Active Thermo Atmosphere Combustion (Engine)
ATDC After top dead center
BTDC Before top dead center
ARC Active Radical Combustion
CAC Charge air cooler, Intercooler
CAD Crank Angle Degrees
CFR Cooperative Fuel Research (Engine, used for octane number determination)
EGR Exhaust Gas Recycling (or Recirculation)
ESIM “Engine SIMulation”, software made in-house for engine system simulation
FTIR Fourier Transform Infra Red (analysis of the infrared absorption spectra)
HC Hydrocarbons (emissions)
HCCI Homogeneous Charge Compression Ignition (Engine)
HTR Heater
IC Internal Combustion (Engine)
KE Kinetic energy
MK Modulated Kinetics (diesel engine concept by Nissan Motor Co.)
NOx, NOx Nitrogen oxides (NO + NO2 + N2O)
PCC Pre Combustion Chamber (Engine)
PE Potential energy
RON Research Octane Number (A measure of “auto-ignition resistance”)
SI Spark Ignition (Engine)
TDC Top Dead Center (usually the combustion TDC)
THC, UHC Total-/Unburned hydrocarbons (including methane, CH4. See NMHC)
NMHC Non-methane total hydrocarbons (methane excluded), sometimes NMTHC
VCR Variable Compression Ratio (Engine)
WOT Wide Open Throttle (SI engines)

121
 … and that was the beginning of the end of my Ph. D. studies.

122