Knock Model Evaluation – Gas Engine
Nishchay Sharma
Master of Science Thesis MMK 2018: TRITA-ITM-EX 2018:600
KTH Industrial Engineering and Management
Machine Design
SE-100 44 STOCKHOLM
+
i
Master of Science Thesis MMK 2018:
TRITA-ITM-EX 2018:600
Knock Model Evaluation – Gas Engine
Nishchay Sharma
Approved
2018-August-27
Examiner
Dr. Andreas Cronhjort
Supervisor
Dr. Johan Fjällman
Commissioner
AVL, Södertälje
Contact person
Jonas Modin
Sammanfattning Knack i en förbränningsmotor är en typ av onormal förbränning. Det är ett komplicerat fenomen
som beror på flera fysiska faktorer och resulterar i högfrekventa tryckoscillationer inuti
förbränningskammaren. Dessa oscillationer kan skada motorn och fenomenet hämmar motorns
effektivitet. Knack kan uppstå på två sätt i en Otto-motor och detta examensarbete kommer att
handla om självantändning. Självantändning, i detta fall, är när ändgasen börjar brinna utan att ha
blivit påverkad av flamfronten eller gnistan från tändstiftet. Det finns flera olika matematiska
modeller som i olika grader kan prediktera knackfenomenet. I detta examensarbete studeras några
av de tidigare publicerade prediktionsmodellerna för knack i Otto-förbränning och modelleras för
analys. Huvudsyftet med detta projekt är således att bedöma noggrannheten hos olika typer av
knackmodeller.
Extra fokus har lagts på empiriska korrelationsmodeller, särskilt till de som är baserade på kemisk
kinetik avseende förbränningsprocessen av metan. Dessa modeller förutsäger den tid det tar för
ändgasen att självantända, baserat på dess koncentration av luft och bränsle. Knackmodellerna
bedöms sedan utifrån det beteende som de förutsäger över motorns driftområde och dess
överensstämmelse med kända motorkalibreringsstrategier. Resultatet av knackpredikteringen för
de olika knackmodellerna utvärderas och valideras i en motorsimuleringsmodell i mjukvaran AVL
BOOST. BOOST-modellen kalibreras mot experimentellt uppmätta motortestdata.
Baserat på resultaten från de valda knockmodellerna så blev den modell som bäst korrelerar med
kända motorkalibreringsstrategier analyserad djupare. Den utvalda modellen var en ECM modell
och den utvärderas ytterligare med avseende på variation i predikterad knack-parameter. Detta
görs genom att modifiera två förbränningsparametrar: tändvinkel och förbränningsduration. Det
visade sig att modellerna predikterade en linjär ökning då tändningen tidigareläggs och ett linjärt
minskande vid längre förbränningsduration, vilket är i enlighet med motortestdata. Vidare visade
det sig att variationer i tändvinkel resulterade i en högre gradient i knackpredikteringen vid högre
motorbelastningar och korresponderande minskning vid lägre belastning.
Nyckelord: Knack, Otto-förbränning, Förbränningsmotor, Empiriska korrelationsmodeller,
BOOST, Förbränningsfasning, Tändvinkel, Förbränningsduration, Knackprediktering.
ii
iii
Master of Science Thesis MMK 2018:
TRITA-ITM-EX 2018:600
Knock Model Evaluation – Gas Engine
Nishchay Sharma
Approved
2018-August-27
Examiner
Dr. Andreas Cronhjort
Supervisor
Dr. Johan Fjällman
Commissioner
AVL, Södertälje
Contact person
Jonas Modin
Abstract Knocking is a type of abnormal combustion which depends on several physical factors and results
in high frequency pressure oscillations inside the combustion chamber of a spark-ignited internal
combustion engine (ICE). These oscillations can damage the engine and hamper its efficiency,
which is why it is important for automakers to understand the knocking behavior so that it can be
avoided during engine operation. Due to the catastrophic outcomes of knocking a lot of research
has been done in the past on prediction of its occurrence. There can be several causes of knocking
but when it occurs due to auto-ignition of fuel in the end-gas it’s called spark-knock. There are
various mathematical models that predict the phenomenon of spark-knock. In this thesis, several
of the previously published knock prediction models for heavy-duty natural-gas engine are studied
and analyzed. The main objective of this project is to assess the accuracy of different types of
knock prediction models.
Amongst all the types of knock prediction models emphasize has been given to empirical
correlation models, particularly to the ones which are based on chemical kinetics pertaining to the
combustion process of methane. These are the models that claim to predict ignition delay time
based on concentration of air and fuel in the unburned zone of the cylinder. The models are
assessed based on the knocking behavior they represent across the engine operation range. Results
pertaining to the knock prediction models are evaluated in a 1D engine simulation model using
AVL BOOST. The BOOST performance prediction model is calibrated against experimentally
measured engine test-cell data and the same data is used to assess the knock prediction models.
The knock prediction model whose results correlate with experimental observations is analyzed
further while other models are discarded. Using the validated model, variation in knock occurrence
is evaluated with change in the combustion phasing. Two of the parameter that are used to define
the combustion phasing are spark-advance and combustion duration. It was found that when the
brake mean effective pressure is kept constant the knock prediction parameter increases linearly
with increase in spark advance and decreases linearly with increase in combustion duration. The
variation of knock prediction parameter with spark advance showed increasing gradient with
increase in engine torque.
Keywords: Knocking, Spark-Ignited, Internal Combustion Engine, Spark-Knock, Empirical
Correlation Models, BOOST, Combustion Phasing, Spark-advance, Combustion Duration, Knock
Prediction Parameter.
iv
v
ACKNOWLEDGEMENTS
“Extraordinary results require extraordinary efforts!”, these are the words that my parents have
sown in my character. First, I would like to thank them for always encouraging me to pursue my
passion and at the same time always reminding me to do it with complete devotion.
This thesis work has taught me the importance of guidance and team-work. For those two
invaluable lessons I would like to thank my supervisors Johan Fjällman and Johannes Andersen. I
have always grown from our discussions and their guidance has helped me to stay focused on
course in a complicated project where I could have easily digressed. Throughout the thesis they
have not only been working together with me but have also been acquainting me to the Swedish
culture. I find myself lucky to have them as my colleagues and have the utmost respect for them.
I would like to thank Dr. Andreas Cronhjort, his guidelines on the planning of a research project
have been extremely beneficial to me. It was through his feedback during the earlier project work
that I could learn from my mistakes as a student and then improve on them to put myself ahead as
an organized research professional.
Ludvig Adlercreutz has played a crucial role in realization of this thesis. His contribution with the
experimental data and trails on the test-cell engine proved as big stepping stones which laid the
foundation of this project. Not only did he help me to understand the working of test-cell and the
engine but also demonstrated knocking combustion experimentally. I envy his skills and hope that
I will be bestowed with more learning opportunities from him in the future. Hereby, I thank him
for his immense support.
I would also like to thank the friendly and motivated team members at AVL, especially my
department manager Jonas Modin. It was because of him that I never had to worry about any of
the organizational issues and could focus on my work. Moreover, he and other colleagues also
introduced me to a new sport “Innebandy” which made me feel like a part of the company and
among friends.
Throughout the project work, I met a lot of people at AVL and gained the experience of working
on a challenging research project. Being a part of the AVL team here has developed me
professionally and for that I express my sincere gratitude.
Nishchay Sharma
Stockholm, August 2018
vi
vii
NOMENCLATURE
Notations
Symbol Description
τ Ignition delay time [s]
t Time [s]
Kc Knock prediction parameter [-]
Tu Unburned-zone temperature [K]
p Cylinder pressure [bar]
h Specific enthalpy [J/kg]
V Volume [bar]
cp Specific heat capacity at constant pressure [J/kg.K]
mb Mass of burned zone [kg]
mu Mass of unburned zone [kg]
….. …….
Abbreviations
SI Spark Ignition/Ignited
ICE Internal Combustion Engine
ECM Empirical Correlation Model
CKM Chemical Kinetics Model
CFDM Computational Fluid Dynamics Model
CI Compression Ignition/Ignited
CAD Crank Angle Degree [°]
TDC Top Dead Center
BDC Bottom Dead Center
AIT Auto-ignition Temperature [K]
EGR Exhaust Gas Recirculation
SCR Selective Catalytic Reduction
RON Research Octane Number
HC Hydrocarbon
CNG Compressed Natural Gas
LNG Liquified Natural Gas
RPM Revolutions per Minute [rpm]
viii
BMEP Brake Mean Effective Pressure
IVC Intake Valve Closed
SOC Start of Combustion
EOC End of Combustion
PR Propane Ratio
ER Equivalence Ratio
LHV Lower Heating Value [J/kg]
WHSC World Harmonized Stationary Cycle
CD Combustion Duration [°]
ROHR Rate of Heat Release [J/°]
CR Compression Ratio
… …
ix
TABLE OF CONTENTS
Sammanfattning (Swedish) i
Abstract (English) iii
Acknowledgements v
Nomenclature vii
Table of Contents ix
1 INTRODUCTION 1
1.1 Background 3
1.2 Purpose 7
1.3 Delimitations 7
2 LITERATURE REVIEW 9
2.1 Knock Prediction Models Based on Arrhenius Function 9
2.2 Knock Prediction Models Based on Chemical Kinetics 10
2.3 Knock Prediction Models Based on Thermodynamics 12
3 IMPLEMENTATION 14
3.1 Boost Model Calibration 15
3.2 Calculating Knock Prediction Parameter (Kc) 16
4 RESULTS 18
4.1 Models based on Ignition Delay Time 18
4.2 Calculating Knock Prediction Parameter (Kc) 20
5 ANALYSIS 22
5.1 Boost Model Calibration 22
5.2 Calculating Knock Prediction Parameter (Kc) 24
6 CONCLUSIONS 28
7 FUTURE WORK 29
x
8 REFERENCES 30
APPENDIX A1: Kc vs CAD (KM1-KM10) 32
APPENDIX A2: Kc vs CAD (KM11) 38
APPENDIX B: Contour Plots of Knock Models 44
1
1 INTRODUCTION
In this chapter, a brief introduction to the different types of ICEs and the properties of the fuels
used for them is presented. The working of spark-ignited engines is explained, which under certain
operating conditions undergo knocking. Different types of knocking phenomenon and their causes
are elaborated upon. Lastly, different ways to predict knocking are outlined.
An Engine is defined as a device that converts one form of energy in to another (Ganesan, 2008).
Generally, in an engine it is the chemical energy of a fuel that is converted into useful mechanical
work by combusting the air-fuel mixture in a controlled manner.
Based on their construction and working principles engines can be classified as rotary type and
reciprocating type. An example of a modern rotary type engine would be a Mazda RX7 rotary
engine which was originally invented by Felix Wankel (Heywood, 1988) and an example for
reciprocating type engine would be Spark-Ignited (SI) engines which is based on Otto Cycle
(1876). Another classic example of reciprocating type engines is Compression-Ignited (CI)
engines which are based on Diesel Cycle (1892) (Çengel and Boles, 2015). In the automotive
industry, most of the vehicles use reciprocating type engines apart from that there have been a few
vehicles based on electric drivelines. In a reciprocating engine the air-fuel mixture is burned in a
cylinder that has a reciprocating piston which translates along the axis of the cylinder. The piston
transmits power to the driveshaft through the connecting rod and crank mechanism (Heywood,
1988). Since, the fuel is burned internally in a closed space in these engines, they are called as
‘Internal Combustion Engines’. Figure 1 shows the basic geometry and construction of an SI
engine:
Figure 1. Geometry and construction of an SI Engine (Ganesan, 2008)
Conceptually, as far as the combustion of fuel in SI and CI engines is concerned oxygen (from the
air) is necessary but SI and CI engines have different ways of igniting the fuel in presence of air.
Conventionally in an SI engine the air and fuel are mixed outside the cylinder and then inducted
into the cylinder. This pre-mixed charge is ignited by a spark plug later in the engine cycle. While
in a CI engine, only air is inducted into the cylinder which is then compressed to a high pressure
and temperature before the injection of fuel, the high temperature of the surrounding compressed
air auto-ignites the injected fuel. Auto-ignition is the tendency of the fuel to spontaneously ignite
2
in presence of air without any external source of energy at a given temperature (Johansson, 2014).
It is due to this difference in auto-ignition properties of petrol and diesel that qualifies them as
suitable fuels for SI and CI engine respectively.
Most of the modern SI and CI engines operate on the four-stroke cycle. Each stroke is marked by
the motion of the piston from one extreme position to another. Moreover, during the piston
movement, from one extreme to another, the crank shaft rotates by an angle of 180°. Thus, the
completion of an entire four-stroke cycle, amounts for 720° crank angle degrees (CAD). Since this
work is specifically about SI engine, the construction and general working principle of a four-
stroke SI engine is discussed below.
Figure 2 shows the four strokes during an engine cycle (Ganesan, 2008) respectively. As per Figure
2, the four strokes of an SI engine can be described as:
1. Intake stroke: Suction of the air-fuel mixture starts when the piston is at the Top Dead Center
(TDC); top most position of the piston in the cylinder. Here the intake valve opens and piston
moves down, this downward piston motion results in suction of air through the intake port
inside the cylinder until the piston reaches the Bottom Dead Center (BDC).
2. Compression stroke: The intake valve closes after the intake stroke. The charge induced in
the cylinder is then compressed by the upward motion of the piston as the piston moves from
BDC to TDC. During the end of this stroke, when the piston is moving towards TDC, the
spark plug ignites the compressed charge.
3. Power or Expansion stroke: With both intake and exhaust valve still closed, the ignited air-
fuel mixture at high pressure pushes the piston down to BDC hence doing work on the piston.
This is the only stroke amongst all the four strokes that leads to power generation.
4. Exhaust stroke: After the expansion stroke, the exhaust valve opens when the piston is at
BDC and is starting to move towards TDC. Upward motion of piston pushes the exhaust gas
through the exhaust port and out of the cylinder.
Figure 2. Working principle of a four-stroke SI Engine (Ganesan, 2008)
The type of fuel plays a crucial role in determining the suitable ignition strategy to be used for its
efficient combustion. As mentioned earlier, one of crucial fuel properties is its tendency to auto-
ignite, this tendency is physically represented by the fuel’s Auto-Ignition Temperature (AIT); it is
the temperature at which the fuel gets ignited in presence of oxygen in its surrounding. The value
of AIT poses as one of the most important parameters for when the design of an SI engine is
concerned. This is because if the fuel has a low AIT then it limits the compression ratio of the
engine and a lower compression ratio leads to lower pressure and temperature at the end of
3
compression stroke which reduces the power output and hence efficiency of the engine (Ganesan,
2008). The power delivery and efficient fuel combustion of an ICE is highly dependent on its
geometrical construction. For example, thermodynamically, the efficiency of an ICE depends on
the compression ratio; ratio of maximum cylinder volume to minimum cylinder volume during an
engine cycle (Heywood, 1988). But increasing the compression ratio increase the temperature of
the charge at the end of compression stroke which leads to auto-ignition of fuel resulting in
abnormal combustion.
There is another important aspect of an ICE that has been the area of enormous amount of research
work for decades now: emissions. The quest of designing an ICE with less harmful emissions
without affecting its performance has been a challenge for engineers. To mitigate the harm caused
to the environment by emissions after combustion, numerous after-treatment technologies have
been developed like Exhaust Gas Recirculation (EGR) and Selective Catalytic Reduction (SCR).
EGR is the recirculation of a portion of the exhaust gas after cooling into the cylinder along with
the fresh charge, this reduces the temperature in the combustion chamber and hence the formation
of harmful oxides of nitrogen (NOx) emissions. But there is also a downside of using excessive
EGR because it increases the formation of soot particles (Heywood, 1988). While, SCR is a device
that is used to chemically reduce the emitted NOx to nitrogen and water (Johansson, 2014). Apart
from all the methods mentioned above, one of the most basic requirements to deal with emissions
is to have a clean and efficient combustion of fuel, which is easier said than done since there are
possibilities of abnormal combustion amongst other challenges.
Another solution that has been explored by researchers for reduction of harmful emissions from
an ICE is the use of unconventional fuels instead of gasoline and diesel. The present thesis
specifically considers the use of natural gas as a fuel for heavy-duty SI engines. Since the fuel is
burned in an SI-engine that are susceptible to knocking, the focus of this thesis is on the
phenomenon of knocking which is defined as a type of abnormal combustion. This phenomenon
hampers the combustion process and reduces the engine efficiency, in worst cases it may damage
the engine as well. In the next sections, the motivation, background and purpose of this research
work are outlined in detail.
1.1 Background
Most of the heavy-duty automotive applications are powered by diesel engines. The emission
legislations are very tough and for the heavy-duty automotive industry this means complex and
expensive after treatment systems for the diesel engines. An alternative approach to this problem
is to use gas-engines because natural gas as a fuel has numerous advantages. For example; it is
composed mainly of methane (CH4) so it has 75% carbon mass compared to petrol and diesel
which have 86-88%. It produces less CO2 per unit of energy as compared to petrol and diesel, it
has a high-octane number (RON 110-130) and hence can be used in an engine with high
compression ratio (at least when compared to petrol engine). Since, natural gas has a higher auto-
ignition resistance it requires a higher spark energy for ignition as compared to other hydrocarbons
(HCs). Its higher flammability range allows the engine to operate on a lean fuel mixture. However,
being gaseous leads to lower volumetric engine efficiency since the air breathing capacity of the
engine is reduced. Moreover, the lower energy density requires vehicle onboard storage in
compressed (CNG) or liquefied (LNG) form leading to issues with the heavy pressurized storage
cylinders (McLean and Lave, 2002). Still, considering the overall evaluation, natural gas comes
out as a better fuel compared to gasoline and diesel.
Natural gas as a fuel is a potent prospect for the heavy-duty automotive industry, but incorporation
of this fuel in the powertrains based on existing ICEs has its own peculiar challenges. One of these
challenges is knocking, to understand knocking in a better way at first the combustion process is
to be analyzed in detail. Conceptually, for a natural gas engine the combustion process is like that
4
of a conventional gasoline engine: mixing of the air and fuel in the intake system, induction of air-
fuel and EGR mixture into the cylinder, mixing of the inducted charge with the left over residual
gas from the previous engine cycle. Then, as mentioned in the earlier section the four-stroke engine
cycle takes place, where power is generated by combustion of the charge via an electric discharge
from the spark plug that initiates the ignition (Heywood, 1988). Taking a deep dive into the
phenomenon of combustion, the burning of the fuel starting from the time of spark can be divided
sequentially into the following stages:
1. Spark initiation or electrical discharge: A high potential difference across the electrodes of
the spark plug leads to electrical breakdown of air in the gap between them. Thus, leading
to a spark in the combustion chamber. This occurs during the compression stroke, usually
around -30° to -20° CAD for part load, -10° to 0° CAD in case of full load. This spark forms
a localized high temperature zone which initiates the combustion of the compressed charge
in the cylinder. The local high temperature gas near the spark plug forms plasma and within
the plasma surface the chemical reaction starts to form radicals from the reactants.
2. Start of combustion and early flame development: When the number of radicals formed
reach beyond a critical number a chain reaction is started which leads to formation of more
radicals by consuming the reactants. In the next chapter, the chemical reactions forming
radicals are discussed in detail.
3. Flame propagation: The increase in reactant consumption due to chain reactions propagates
the reaction zone further away from the spark plug. The intensity and movement of the
reaction zone is influenced by various factors like local air-fuel composition, inert gas
fraction (EGR and residual gas), turbulent-laminar flow profile, in cylinder motions – squish,
tumble and swirl etc. This propagating reaction is called a flame front. The mixture of air,
fuel, EGR and residual gas in the unburned zone is called end-gas. Relevant to this project,
one of the most important parameters is the flame speed. The flame propagation can be
categorized based on its speed, namely: (1) laminar flame speed; (2) expansion speed; and
(3) turbulent flame speed. At first, the flame front propagates with laminar flame speed
which is determined by the reactivity of the charge due to the fact that the charge velocity
around the spark plug is very low. Areas with high turbulence in the cylinder increase the
flame propagation speed due to relatively faster intake of unburned gas into the reaction
zone. The velocity at which the unburned gas enters the reaction zone is defined as the
turbulent flame speed. The laminar flame speed is around 0.2-0.5 m/s, whereas the effective
flame propagation in turbulent regime ranges from 10-50 m/s. Moreover, the burned zone is
at a higher temperature as compared to the unburned zone, due to which there is thermal
expansion. This thermal expansion leads to even higher flame speed which is also called as
the expansion speed. The expansion speed of the flame front can be twice as high as the
turbulent flame speed (Johansson, 2014).
4. Flame quenching: Under normal combustion circumstances, the advancing flame reaches
the cylinder wall thus consuming all the charge present in the cylinder. The flame reaches
the walls at around 15° CAD. Even when the entire combustion chamber is ablaze by the
flame there is still approximately 25% of the charge that is yet to burn. Thus, the combustion
continues around the combustion chamber till around 25° CAD. Ultimately, all the charge
gets burned which extinguishes the flame (Heywood, 1988).
Depending upon the engine operating conditions, the duration of flame development and
propagation typically lies between 30° and 90° CAD. This is typically the case when the
combustion is normal. In SI engines, the abnormal combustion phenomenon can occur in many
ways but in the present thesis work the phenomena of knocking is explored, which when severe
may lead to catastrophic engine damage. When it is not severe, it limits the engine performance
and generates noise. Knocking got its name due to the type of noise that it transmits through the
engine body and it can occur due to two reasons; localized auto-ignition in the unburned zone and
surface ignition. The auto-ignition in the unburned zone creates another flame front which then
5
collides with the normal combustion flame front (see figure 3). This causes high intensity pressure
oscillations in the combustion chamber, thus, making knocking an acoustic phenomenon. Surface
ignition, on the other hand, is the burning of the charge due to a hot spot on the combustion
chamber walls. The hot spot can be an overheated spark plug or valve, hot cylinder wall or a heated
scrapped oil deposit. Any source of combustion other than the normal spark ignition classifies as
surface ignition, see figure 4. When the surface ignition occurs before the spark discharge, it’s
called as pre-ignition. Contrarily, when it occurs after the spark discharge, it’s called as post-
ignition. It should be noted that surface ignition can lead to knocking, especially pre-ignition.
Moreover, several plausible combinations of knock and surface ignition can occur depending upon
various parameters like, pressure and temperature of end-gas, spark advance, combustion phasing,
in-cylinder turbulence, etc. (Heywood, 1988).
Figure 3. Auto-ignition of end-gas that eventually leads to knocking (Hpcomb.kaust.edu.sa, 2018)
The repeated occurrence of auto-ignition of end-gas instead of normal combustion is called spark-
knock. Although, knocking is a highly sporadic phenomenon and varies greatly from one engine
cycle to another. It is possible to have knocking at the same operating conditions repetitively for a
certain number of cycles and none during other cycles. As far as spark-knock is concerned, it can
be controlled by changing the combustion phasing. This can be done by changing the spark
advance which is defined as the CAD value during the latter part of the compression stroke at
which spark discharge from the spark plug ignites the charge. When the spark advance is retarded
both the knock severity and intensity decrease. Concurrently, advancing the spark event increases
the possibility of knocking (Heywood, 1988).
Figure 4. Various types of surface-ignition (Hpcomb.kaust.edu.sa, 2018)
6
The root cause for occurrence of spark-knock is the auto-ignition of the end-gas ahead of the flame
front. The auto-ignition happens because the end-gas reaches at or beyond the AIT of the fuel. To
understand how and why this happens, it is essential to emphasize on the thermodynamics involved
in the combustion process. Based on the first law of thermodynamics, if the end-gas is considered
as an open system, then its internal energy (which is directly related to the temperature) can only
be increased in two ways: (1) adding heat to the system; (2) doing work on the system (Çengel and
Boles, 2015). There are several ways in which heat and work are added to the end-gas.
Chronologically, they occur as:
• After the induction of the charge in the cylinder during the intake stroke, the piston
compresses the charge during the compression stroke before the spark discharge from the
spark plug. This volumetric work done by the piston increases the temperature and pressure
of the entire charge.
• Spark-ignition occurs late during the compression stroke initiating the combustion of fuel.
The fuel burning in the reaction zone releases heat to the surrounding end-gas. It should be
noted that during this time, the piston moving towards TDC still continues to compress the
charge, hence adding work. Moreover, the combustion of fuel leads to increase in pressure
and temperature of the reaction zone along with expansion of burned gas. The high-pressure
reaction zone pushes the end-gas, thus compressing end-gas further.
• Cylinder wall around the combustion chamber is at high temperature due to constant
exposure to heat release from combustion. Same is the case with piston top and exhaust
valves. The portion of end-gas close to these hot physical boundaries gets heated from them.
On top of that, the same portion of end-gas, since located farthest away from the spark-plug
experiences the maximum compression work and heat addition from the propagating
reaction zone (Heywood, 1988).
As described above, knocking is a complex abnormal combustion phenomenon which affects the
engine in terms of performance and in worst case damages its parts thus stalling the engine. Due
to this knocking has been a topic of research and exploration for a long time, where the area of
interest has been prediction of its occurrence for a given engine operating condition. The prediction
of occurrence of knocking has been done majorly through three different types of models:
1. Empirical Correlation Models (ECMs): these models predict knocking by using empirical
correlations that have been tuned to experimentally acquired data. The advantage of using
these types of models is that they can be tuned to work for any engine if the parameters used
are independent of the engine geometry. However, the accuracy of prediction of occurrence
of knocking from these models is not very high. Interestingly, some of the empirical
correlation models use the fundamentals of chemical kinetics while others are based on
purely mathematical expressions and curve fitting. Apart from these, there are several other
models which take into consideration the variation of thermodynamic parameters during the
engine cycle to predict knocking. The mathematical expressions on which all these models
are based require inputs in terms of physical and thermodynamic parameters. These
parameters are related to the combustion process, composition of the charge in the cylinder
and the concomitant state of end-gas.
2. Chemical Kinetics Models (CKMs): the combustion process is nothing but a chemical
reaction where the fuel is oxidized. These models use the chemical reactions involved in the
oxidation of the fuel. The chemical oxidation of fuel involves numerous intermediate
reactions as well as radicals. CKM models rely entirely on the chemical kinetics
fundamentals to evaluate the concentration of reactants and products to predict the
possibility of auto-ignition in the end-gas. Most of the CKMs involve iterative numerical
methods to calculate the mass fraction of the chemical species.
3. Computational Fluid Dynamics Models (CFDMs): the physical parameters that define the
state of the end-gas can be estimated by solving the Navier-Stokes equations. In a CFDM,
for a given engine geometry, the combustion process can be simulated. The simulations to
7
solve the Navier-Stokes equation are based on various numerical approximations, e.g.
Reynolds-Averaged Navier–Stokes (RANS) equations, Large Eddy Simulation (LES), etc.
Setting up a combustion case and solving the RANS equations for it using different types of
iterative solution methods can result in evaluation of parameters of interest. The only
drawback with CFDMs is that they are highly geometry dependent and for every engine
operating point a separate simulation is to be run.
1.2 Purpose
Numerous researchers have proposed knock prediction models for gas engines, the problem with
these models is that they claim to predict knocking accurately but their predictions do not coincide
with the experimental observations. Therefore, the first and the foremost objective of this thesis is
to evaluate the accuracy of end-gas auto-ignition claimed by a selected few knock prediction
models. To check the accuracy of a given knock prediction model its results are to be compared
with previously acquired experimental data.
Through a thorough literature survey several knock prediction models are were shortlisted for
chosen for evaluation. They were then analyzed and compared with each other in terms of their
accuracy. Out of the assessed models the ones with plausible results are to be used for further
analysis by performing simulations while changing the spark timing and combustion duration. The
engine modelling is to be done in BOOST.
Out of the three mentioned categories of knock prediction models in the previous section ECMs
are the focus of the current work. On that basis, the following research questions have been
formulated. These research questions are aimed to explicitly define the objectives of the thesis:
• What are the different types of ECMs currently available that can be used to predict knocking
in gas engines?
• What are the thermodynamic assumptions and considerations these models are based on?
According to these models, what are the parameters on which knocking depends? Can these
parameters be related to other physical parameters e.g. fuel concentration?
• Which ECMs show plausible knocking behavior?
• How does the knocking behavior vary with change in parameters like spark advance and
combustion duration?
1.3 Delimitations
As mentioned in the previous section, prediction of knocking can be done by using three different
types of models. In this project the target is to assess a zero-dimensional (0D) ECM. 0D models
have no geometric dependency rather they only have time dependency (Holzbecher, 2012). Apart
from the type of models there are several other assumptions as well as limitations that have been
deliberately imposed to get a better understanding of the complex knocking behaviour. These
assumptions and limitations are:
1. Fuel consideration: The fuel of interest is natural gas but we are using pure methane, both
for the ease of modelling but also for the easier chemical estimations. To model the
combustion phenomenon as close as possible to reality, the use of a certain fraction of
propane in the fuel should be considered, but in this work the fuel is assumed as pure
methane. This is because it has been found that the presence of propane changes the kinetics
of methane combustion (Lifshitz et al., 1971).
2. Only spark-knock consideration: Knocking is an abnormal combustion phenomenon which
is highly complex in nature, one of the factors that adds to its complexity is the cause that
leads to knocking. As discussed earlier, there can be several causes that may lead to
8
knocking, e.g. spark-knock and surface ignition (especially pre-ignition). In this work, the
aspect of auto-ignition of end-gas is studied and modelled. No emphasise has been given to
the surface ignition phenomenon, except that the cylinder wall temperature is considered
when the heat transfer in the combustion chamber is analysed in BOOST.
3. No acoustics involved: Another aspect that has not been considered in this work is the
acoustics of knocking. This is because the work is focused on predicting the occurrence of
knocking, that is, for a given engine operating condition will there be knocking or not? While
the acoustics on the other hand come into play when knocking is occurring already because
it is signified by oscillation of cylinder pressure.
4. Emphasis on when knocking starts to occur: The prediction of occurrence of knocking has
two important aspects; (1) predicting the CAD at which knocking starts to occur which is
otherwise called as the ‘knock onset’, (2) predicting the intensity of the knocking
phenomenon which is the amplitude of the pressure oscillation that knocking is going to
result in. In this work the modelling is focused on the prediction of knock onset, but the
possibility of estimating knock intensity has not been considered. To relate this to the
pressure oscillations due to knocking, the focus is to predict the point of time at which the
oscillations begin (knock-onset) and not the amplitude of those oscillations (knock
intensity).
5. Consideration of spark-advance only in terms of CAD: Another essential aspect that is
directly related to knocking is the spark advance. In this work, the influence of spark advance
on knock onset is emphasized upon but not on knock intensity. Rather, the spark advance is
considered as an event that controls the combustion phasing. Spark advance has only been
considered in terms of CAD to assist in marking the beginning of combustion of fuel at any
engine operating point.
6. Air-fuel ratio consideration: In an engine the air-fuel ratio varies with respect to the
operating conditions, this variation in the air-fuel ratio also plays a crucial role when it comes
to knocking. An air-fuel mixture is called lean which has more than the required amount of
air to burn the fuel completely. Contrarily, if the amount of air required to burn the fuel is
lower, then the mixture is called rich. Whereas, a stoichiometric mixture has the exact
amount of air required to burn the fuel completely. It has been observed that, lean mixtures
have a higher tendency to knock as compared to rich mixtures. This is because lean mixtures
attain higher temperature at the end of the compression stroke as compared to stoichiometric
and rich mixtures (Grandin et al., 2002). This leads to more heating of the end-gas and may
also lead to surface ignition (especially pre-ignition). Therefore, for the sake of simplicity,
the air-fuel mixture is considered as stochiometric. The stoichiometric air-fuel ratio (by
mass) for pure methane is 17.23, whereas for natural gas it is 14.5 (Heywood,1988).
9
2 LITERATURE REVIEW
In this chapter, the summary of existing knowledge and former performed research about ECMs
used for knock prediction is presented. Several types of knock prediction models in the ECM
category have been identified and mentioned. A comparative analysis of these models has been
done which outlines their similarities and differences. The physical and numerical parameters that
these models consider are analyzed and discussed.
Most of the knock predicting ECMs are based on ‘ignition delay time’ which is defined as, “A key
physicochemical property of combustible fuel-air mixture used for engines running on various
principles. The time or the turn angle of a piston engine crankshaft from the start of fuel injection
into the combustion chamber to the instant of appearance of flame (cold flame glow) or rise of
pressure in the chamber due to heat evolution upon combustion of the fuel is considered as the
ignition (spontaneous ignition) delay time.” (Keshavarz et al.,2013). By some authors like,
Heywood (Internal Combustion Engine Fundamentals, 1988) the ignition delay time is also
referred as induction time. The parent correlation on which most of ECMs are based that uses
ignition delay time to predict the onset of auto-ignition in end gas is (Livengood and Wu, 1955):
𝐾𝑐 = ∫1
𝜏
𝑡0
𝑑𝑡 = 1 (1)
In the above equation, 𝐾𝑐 is the knock prediction parameter and signifies onset of knocking at the
time when its value reaches 1, τ is the ignition time delay and t is the elapsed time from start of
compression (closure of intake valve) to auto-ignition. τ has been estimated in several ways by
several approaches: (1) by fitting the engine’s experimental data to an Arrhenius function
(Reference 99 in chapter 9 by Heywood, 1988); (2) by chemical kinetics that consider the oxidation
of the fuel, which in this case is methane. There are several models that are studied and analysed
in this thesis that use these approaches for estimation of ignition delay time.
2.1 Knock Prediction Models based on Arrhenius Function
The Arrhenius function that is used for matching the experimentally measured data is:
𝜏 = 𝐴𝑝−𝑛exp (𝐵
𝑇𝑢) (2)
Here 𝑝 is the cylinder pressure, 𝑇𝑢 is the unburned zone temperature and 𝐴, 𝑛 and 𝐵 are constants
which depend on the fuel type. In general, the accuracy of knock prediction models based on this
equation is ambiguous and depends upon how the parameters 𝐴, 𝑛 and 𝐵 are calibrated with the
experimentally obtained data (Heywood, 1988). To minimize the error in prediction of knock least
square method is used for optimization of the parameters (Douaud and Ezyat, 1978). The following
sets of values for the constant parameters were proposed by researchers to predict occurrence of
knocking based on equation (2):
Table 1. ECMs based on Arrhenius function
Model No. A n B Reference
KM1 0.021 1.7 3800 (Elmqvist-Möller, 2006)
KM2 0.0071 1.325 3296 (Elmqvist-Möller, 2006)
KM3 32.87 1.7 3800 (Douaud and Ezyat, 1978)
KM4 0.985 0.887 6167 (Soylu, 2005)
10
Amongst the models mentioned in Table 1, the ignition delay time estimation in knock prediction
model KM3 is based on the following equation, it should be noted that this equation gives the
ignition delay time in milliseconds:
𝜏 = 17.68 (𝑂𝑁
100)
3.402
𝑝−1.7𝑒𝑥𝑝 (3800
𝑇𝑢) (3)
In equation (3), ON stands for the octane number of the fuel which is 120 for methane. (Heywood,
1988). From equation (3) it is observable that ignition delay time is dependent on the octane
number which in turn is a fuel property, this supports that accuracy of the knock prediction model
is sensitive to the type of fuel. This high dependency of the accuracy of knock prediction models
on the type of fuel also forces to consider the fact that fuels used for running the engine are not
pure and when two or more different types of hydrocarbons are present in the fuel then the
parameters of the knock prediction models need to be calibrated accordingly. As far as dependency
on the quality of fuel is concerned, the parameter 𝐵 for only KM4 considers the ratio of propane
present in the natural gas. The following equation shows how the propane ratio and equivalence
ratio are considered in the evaluation of 𝐵 (Soylu, 2005):
𝐵 = (−0.575 + (10.058 ∗ 𝑃𝑅 − 54.053 ∗ 𝑃𝑅2)) ∗ 𝐸𝑅 + (1.456 + (−8.703 ∗ 𝑃𝑅 + 43.615 ∗
𝑃𝑅2)) ∗ 7000 𝐾 (4)
In equation (4), PR is propane ratio by mass and ER is the equivalence ratio of the charge inducted
in the cylinder.
It should be noted that KM1 was developed based on equation (3) which is same as for KM3, the
only difference was that KM1 was developed for an engine running on gasoline fuel while the
parameters in KM3 are corresponding to methane. Moreover, the parameters in KM2 were
obtained from the optimization of parameters in KM1, this was done to improve the accuracy of
knock prediction by KM1 (Elmqvist-Möller, 2006). Therefore, the parameters in KM1 and KM2
need to be calibrated and optimized if the models are to be used for knock prediction on a methane
fuelled engine.
Apart from the dependency on the fuel properties like propane ratio and octane number as well as
other curve fitting parameters it is also observable that for models KM1-KM4 the ignition delay
time reduces with increase in cylinder pressure and the temperature of the unburned zone.
Consequently, from equation (1) it is straightforward that decrease in the ignition delay time
increases the value of knock prediction parameter, thus suggesting higher possibility of knocking.
2.2 Knock Prediction Models based on Chemical Kinetics
Chemical kinetics theory for study of knocking is based on the fact that combustion in an engine
is nothing but the chemical oxidation reaction of fuel with oxygen present in air. The chemical
reaction mechanism is complex for the process of combustion, since it not a single or even few-
steps process but consists of numerous concurrent reactions called as chain reactions. There is an
initiation step in the chain reactions which results in formation of highly reactive intermediate
species called as radicals which are produced by dissociation of stable molecules like fuel and
oxygen. Radicals formed in the initiation step react with the reactants to form intermediate
products and new radicals which continue the chain reactions. There are some reactions which
result in generation of two radicals by consuming one, these are called as chain branching
reactions. While, the termination reactions are those which mark the end of chain reactions by
consuming the radicals such that no more radicals are left. The following reactions explain the low
temperature oxidation of methane (Glassman, 1996):
𝐶𝐻4 + 𝑂2 → ��𝐻3 + 𝐻𝑂2 } (chain initiating) (5)
��𝐻3 + 𝑂2 → 𝐶𝐻2𝑂 + ��𝐻 } (chain propagating) (6)
11
��𝐻 + 𝐶𝐻4 → 𝐻2𝑂 + ��𝐻3 } (chain propagating) (7)
��𝐻 + 𝐶𝐻2𝑂 → 𝐻2𝑂 + 𝐻��𝑂 } (chain propagating) (8)
𝐶𝐻2𝑂 + 𝑂2 → 𝐻𝑂2 + 𝐻��𝑂 } (chain branching) (9)
𝐻��𝑂 + 𝑂2 → 𝐶𝑂 + 𝐻𝑂2 } (chain propagating) (10)
𝐻𝑂2 + 𝐶𝐻4 → 𝐻2𝑂2 + 𝐶𝐻3 } (chain propagating) (11)
𝐻𝑂2 + 𝐶𝐻2𝑂 → 𝐻2𝑂2 + 𝐻��𝑂 } (chain propagating) (12)
��𝐻 → 𝑤𝑎𝑙𝑙 } (chain terminating) (13)
𝐶𝐻2 → 𝑤𝑎𝑙𝑙 } (chain terminating) (14)
𝐻𝑂2 → 𝑤𝑎𝑙𝑙 } (chain terminating) (15)
However, there is a change in the above reaction mechanism when high temperature oxidation of
methane occurs. This is because the 𝐻2𝑂2 molecule does not dissociate into ��𝐻 radicals at lower
temperature, the dissociation is significant only above 900K (Glassman, 1996). Thus, the chain
reaction mechanism changes with the increase in temperature and hence it becomes relevant to the
study of combustion in a gas engine since the temperature in the burned zone at certain operating
conditions can reach above 2500 K. Therefore, apart from the low temperature oxidation of
methane the reaction mechanism at high temperature needs to be considered as well. Following
reactions showcase the major reaction path for oxidation of methane at high temperatures
(Glassman, 1996):
𝐶𝐻4 + 𝑀 → 𝐶𝐻3 + 𝐻 + 𝑀 (16)
𝐶𝐻4 + 𝑋 → 𝐶𝐻3 + 𝑋𝐻 (17)
𝐶𝐻3 + 𝑂2 → 𝐶𝐻3𝑂 + 𝑂 (18)
𝐶𝐻3 + 𝑂2 → 𝐻2𝐶𝑂 + 𝑂𝐻 (19)
𝐶𝐻3𝑂 + 𝑀 → 𝐻2𝐶𝑂 + 𝐻 + 𝑀 (20)
𝐻2𝐶𝑂 + 𝑋 → 𝐻𝐶𝑂 + 𝑋𝐻 (21)
𝐻𝐶𝑂 + 𝑀 → 𝐻 + 𝐶𝑂 + 𝑀 (22)
𝐶𝐻3 + 𝐶𝐻3 → 𝐶2𝐻6 (23)
𝐶𝑂 + 𝑂𝐻 → 𝐶𝑂2 + 𝐻 (24)
In the above reactions, 𝑀 represents a chemically stable molecule while 𝑋 on the other hand
represents any of the radicals like 𝐻, 𝑂𝐻 and 𝑂.
As far as occurrence of knocking is concerned, according to Semenov’s theory, the air-fuel mixture
auto-ignites when the energy released by the chemical reactions is more than the amount of heat
that is lost to the surroundings. This results in a situation like self-heating which ultimately leads
to auto-ignition (Glassman, 1996). Several researchers have experimentally performed and
analysed the process of auto-ignition of methane, particularly in shock tubes. From these
experimental studies researchers have claimed to understand the influence of one or more reactions
in driving the process of oxidation of methane. Using the chemical kinetics approach these
researchers have empirically derived relationship between ignition delay time and concentration
of methane in the air-fuel mixture. For example, Seery and Bowman developed a chemical kinetics
model that uses numerical integration method to evaluate the thermodynamic properties of the gas
mixture during the combustion process. There are other similar experimental studies as well which
represent the dependence of ignition delay time on concentration of methane and air present in the
mixture, these studies are based on experiments conducted in a shock-tube to study oxidation of
12
methane. These studies propose that the ignition delay time can be expressed in the following way
as shown in equation (25):
𝜏 = 𝐴. 𝑒(
𝑇𝑎𝑇𝑢
). [𝑂2]𝑥. [𝐶𝐻4]𝑦 (25)
If the evaluated ignition delay time from equation (25) is substituted in equation (5) then the knock
prediction parameter can be estimated. Here 𝑥 and 𝑦 are the exponents for mole density (in
mole/cu.cm) of oxygen and methane in the end-gas respectively and 𝑇𝑎 is the activation
temperature which is equal to the activation energy 𝐸𝑎 divided by the ideal gas constant 𝑅.
Activation energy is the amount of energy that needs to be added to the reactants to initiate a
chemical reaction amongst them (Glassman, 1996). The activation energy to initiate the fuel
oxidation reaction is provided by the spark event in an SI engine.
The following set of values for the parameters 𝑥, 𝑦 and 𝑇𝑎 are proposed by the corresponding
authors to determine the ignition delay time for methane based on chemical kinetics:
Table 2. ECMs based on chemical kinetics function
Model No. A x y Ta Reference
KM5 7.65E-18 -1.6 0.4 25900 (Seery and Bowman, 1970)
KM6 3.62E-14 -1.03 0.33 23400 (Lifshitz et al.,1971)
KM7 2.50E-15 -1.02 0.32 26700 (Tsuboi and Wagner, 1974)
KM8 1.19E-18 -1.94 0.48 23333 (Cheng and Oppenheim, 1984)
KM9 4.40E-15 -1.03 0.33 26360 (Grillo and Slack, 1976)
KM10 4.99E-14 -1.31 -0.38 9575 (Petersen et al.,1999)
Based on the values of parameters in Table 2, it can be commented that across different models a
parameter’s value changes significantly. E.g., the value of 𝑦 is positive for all the models except
KM10. Extended analysis on this based on equation (25) reveals that if 𝑦 is negative, then the
concentration of methane in end-gas is inversely proportional to ignition delay time. A further
analysis on the same reveals that the ignition delay time is dependent on the engine load, because
more number of moles of fuel (higher molar density!) are burned in the same cylinder volume at
high engine loads if the engine speed is kept constant. Thus, it would be an interesting exploration
to know which of the models here are correct and what is their corresponding value of 𝑦. Similarly,
there are different values of 𝑇𝑎 for different models, this can be accounted to the fact that
researchers have used different values of activation energy for oxidation reaction of methane
(Lifhitz et al., 1971). The value of parameter 𝑇𝑎 for all the models are approximately 2.5 times to
that in KM10. This is expected to cause huge differences in the evaluation of ignition delay time
for KM10 as compared to other models in this category because 𝑇𝑎 has an exponential influence
on the value of ignition delay time.
2.3 Knock Prediction Models based on Thermodynamics
The model based on thermodynamics considers that the pre-reactions in end-gas must become
sufficiently intense and release enough energy to cause auto-ignition of a portion of the mixture
that is yet to be consumed by the propagating flame. These types of models consider two-zones
combustion for analysis and evaluation of the thermodynamic parameters; burned zone and
unburned zone. Based on the total energy released due to pre-reactions activity in the unburned
zone per unit volume the following dimensionless knock criteria to predict occurrence of knocking
was formulated (Karim and Gao, 1992):
13
𝐾𝑐 =(
𝐸𝑛𝑒𝑟𝑔𝑦 𝑟𝑒𝑙𝑒𝑎𝑠𝑒𝑑 𝑏𝑦 𝑒𝑛𝑑𝑔𝑎𝑠 𝑟𝑒𝑎𝑐𝑡𝑖𝑜𝑛𝑠
𝑉𝑜𝑙𝑢𝑚𝑒)
𝑡
(𝐸𝑛𝑒𝑟𝑔𝑦 𝑟𝑒𝑙𝑒𝑎𝑠𝑒𝑑 𝑏𝑦 𝑐𝑜𝑚𝑏𝑢𝑠𝑡𝑖𝑜𝑛
𝑉𝑜𝑙𝑢𝑚𝑒)
𝑡0
(26)
In the above equation (18), when the energy released by end gas reactions is evaluated using
change in enthalpy of the unburned zone from start of combustion to any time 𝑡 the following
expression is obtained which evaluates the knock prediction parameter (Shrestha, 1999):
𝐾𝑐 =(
(ℎ𝑠𝑡−ℎ𝑡)∗𝑚𝑢𝑉𝑡
)
(𝑚𝑜ℎ𝑜
𝑉𝑜)
(27)
In equation (27), ℎ𝑠𝑡 is the specific enthalpy of the charge at start of combustion, 𝑚𝑢 is the
unburned zone mass, ℎ𝑡 is the specific enthalpy of the unburned zone at time 𝑡 after start of
combustion, ℎ𝑜 is the effective heating value of the fuel and 𝑚𝑜 is the initial or total mass, 𝑉𝑜 is
the cylinder volume at the start of combustion and 𝑉𝑡 is the volume of the unburned zone after
time 𝑡 has elapsed after the start of combustion. This model has been referred as KM11 from here
onwards in this text. It should be noted that according to KM11, the time 𝑡 at which value of the
𝐾𝑐 becomes equal to or more than 1.5 signifies the onset and further occurrence of knocking.
Following the approach proposed by Karim and Gao, their two-zone thermodynamic model has
been implemented and tested by several researchers to determine the safe operating conditions for
gas engines that avoid knocking (Saikaly et al., 2008). Although it has been found that different
chemical kinetics schemes are used for the same two-zone thermodynamic model, like by Saikaly
et al. and Attar. Due to the use of different chemical kinetics schemes different numerical solutions
are obtained for energy released in the end-gas reactions. Hence, the thermodynamic parameters
in equation (27) would have different values with different chemical kinetics schemes. This is
similar to what was observed in case of models based on chemical kinetics in the previous section
where different models proposed different values of parameters for the same fundamental
equation. This once again highlights the underlying problem of not having accurate knock
prediction models because the chemical reactions in the end-gas are not fully understood. E.g., to
evaluate the parameters in equation (27), Attar uses the chemical kinetics scheme that predicts the
energy released by end-gas reactions using a 155 reaction steps and 39 species. Whereas, the
chemical kinetics scheme implemented by Saikaly et al. uses 325 equilibrium reactions and 53
species, this scheme was proposed by (Smith et al., GRI-Mech). It should be noted that, these
different chemical kinetics schemes are not evaluated in this thesis.
There is another model which follows the approach laid out by Karim and Gao, the interesting
highlight of this model is that the following differential equation has been proposed to evaluate
the value of 𝐾𝑐 (Sierra Parra et al., 2017):
𝑑𝐾𝑐 =1
𝑉[𝑉𝑐 ∗ 𝐶𝑅
((ℎ𝑠𝑡−ℎ𝑢)𝑑𝑚𝑏+𝑚𝑢∗𝑐𝑝,𝑢∗𝑑𝑇𝑢)
𝐿𝐻𝑉− 𝐾𝑐 ∗ 𝑑𝑉] (28)
In the above equation (28), 𝑑𝑚𝑏 is the change in mass of burned zone, 𝑐𝑝,𝑢 is the constant pressure
specific heat capacity of the unburned zone, ℎ𝑠𝑡 − ℎ𝑢 is the change in specific enthalpy of the
unburned zone since after spark, 𝑉 is the cylinder volume with respect to which the equation is to
be integrated to obtain 𝐾𝑐. However, it was observed that the proposed differential equation (28)
is dimensionally incorrect and that’s why this model is discarded from study in this thesis.
14
3 IMPLEMENTATION
The approach followed to calibrate an existing BOOST engine performance prediction model
against experimentally obtained data has been explained. How the engine simulation boost model
has been used to evaluate the value of knock prediction parameter (Kc) proposed by 11 selected
models (KM1-KM11) has been described briefly.
BOOST is a fully integrated IC engine 1D simulation software. It can simulate a wide variety of
engines, 4-stroke or 2-stroke, spark or auto-ignited. A virtual engine model created in BOOST can
be used for prediction of engine performance, analysis of gas exchange, study of combustion and
emissions. For this thesis, the assessment of knock prediction models was done using a 1D virtual
engine model developed in BOOST. In that 1D engine model, the 0D (only time dependent) knock
prediction models were incorporated in relation to the combustion process. Then the knock
prediction parameter was evaluated for all the knock prediction models separately by running the
engine cycle simulations in BOOST.
The BOOST engine model was calibrated using the steady state data collected from a single-
cylinder heavy-duty natural gas test engine. The data was captured at the load points which are a
part of the World Harmonized Stationary Cycle (WHSC) test cycle. WHSC test is a well-known
standard steady-state engine dynamometer schedule (DieselNet, 2018). There were in total 13
different engine operating conditions which are represented by the points marked in the following
plot in figure 5:
Figure 5. WHSC engine test steady-state operating points
The single cylinder engine cycle simulations were run in BOOST to replicate the engine
performance which was experimentally obtained by running the engine on the 13 operating points
represented in Figure 5. The BOOST model was calibrated using experimentally obtained data
P2
P3
P4
P5
P6
P7
P8
P9
P10
P11
P12
Max_PowLow Speed
BM
EP
[bar
]
Engine Speed (RPM)
LOAD
15
from the test cell engine. It should be noted that running the engine in the test cell to acquire
experimental data was not in the scope of this thesis, rather the experimental data available from
previously conducted tests was used as a reference for modelling. The technical specifications of
the single cylinder test engine that was used to design the model in BOOST are (see table 3):
Table 3. Technical specifications of the heavy-duty natural-gas engine
No. of Cylinders 1
Compression Ratio (CR) 12
Swept Volume 2 L
Using the engine test cell data corresponding to operating points mentioned in Figure 5 and engine
specifications from Table 3 the following 1D engine model was designed in BOOST (see figure
6):
Figure 6. Engine cycle simulations model in BOOST
3.1 BOOST Model Calibration
The most important test cell data that was used to calibrate the engine cycle simulations model in
BOOST was the in-cylinder pressure trace. The in-cylinder pressure was measured from the test
cell engine and recorded in crank angle resolution for all 13 operating points mentioned in figure
5. These measured in-cylinder pressure traces were averaged for at least 10 engine cycles and then
the averaged pressure traces were used for calculating parameters related to combustion such as:
(1) start of combustion (SOC), (2) combustion duration (CD), (3) Wiebe function shape parameters
(Wiebe,1956), (4) rate of heat release (ROHR) and (5) CAD corresponding to 5%, 10%, 50% and
90% mass fraction burned.
16
The BOOST simulations were performed using the ‘Multiple Vibe 2-Zone’ combustion model.
Choosing a two-zone combustion model allowed calculation of thermodynamic as well as physical
parameters in both burned and unburned zone. Using a two-zone combustion model was important
because it enabled the detection of auto-ignition in the unburned zone, thus, allowing the
possibility to analyse spark-knock if it occurred in the engine simulations.
The above-mentioned data was used to calibrate the BOOST engine model. Following plot shows
the comparison between an experimentally measured cylinder pressure trace and the cylinder
pressure trace obtained by BOOST (see figure 7):
Figure 7. Experimental and simulation cylinder pressure trace comparison against CAD
As shown in Figure 7, a good correlation was observed between the experimentally measured
cylinder pressure trace and a pressure trace obtained from simulations. The plots of cylinder
pressure trace comparison for the remaining 12 operating points showed similar coherence
between experimental and BOOST simulation values.
3.2 Calculating Knock Prediction Parameter (Kc)
The physical and thermodynamic parameters 𝑇𝑢, 𝑝 and ℎ𝑢 that are required to calculate the knock
prediction parameter (𝐾𝑐) were taken from the BOOST simulations, this was done for all 11 knock
prediction models separately. For models based on the Arrhenius function and chemical kinetics
the ignition delay time was calculated at each operating point. Moreover, enthalpy was the
parameter of interest for the thermodynamic model (KM11). Since it was also essential to
understand if the knock prediction models can determine the CAD at which knock occurs 𝐾𝑐 for
all the models was calculated in crank angle resolution.
-360 -300 -240 -180 -120 -60 0 60 120 180 240 300 360
Cyli
nder
Pre
ssure
[bar
]
CAD [°]
Experimental
Boost
Difference
17
The value of 𝐾𝑐 determines whether the engine is going to knock or not. Therefore, it is important
to understand its variation to predict knocking successfully. For understanding this variation,
firstly the critical value of 𝐾𝑐 is to be established as a reference and secondly the variation is to be
studied with respect to that reference. From the literature review we know that the critical value of
𝐾𝑐 which signifies occurrence of knocking is 1 for KM1-KM10 and is 1.5 for KM11.
There are two ways of understanding the behaviour of 𝐾𝑐:
1. Variation of 𝐾𝑐 with respect to crank angle during an engine cycle especially from closure
of intake valve till end of combustion. This will be helpful in determining if the knock
prediction model correctly predicts knocking or not. Moreover, it will also show how does
𝐾𝑐 grow with respect to crank angle during the engine cycle.
2. Variation of 𝐾𝑐 with respect to engine operating conditions like load and engine speed. This
will be helpful in determining which model shows plausible occurrence of knocking for an
operating point and at which CAD, moreover how does the knocking characteristic vary with
change in engine operating condition.
Graphs pertaining to the two analyses mentioned above are shown later in this text.
18
4 RESULTS
In this chapter the evaluated results of all the selected knock prediction models (KM1-KM11) are
presented. Firstly, the variation in important parameters like ignition delay time against crank
angle is presented. Secondly, variation in 𝐾𝑐 with crank angle for all the models is presented for
comparison and analysis.
Out of the 11 models selected for assessment 10 are based on ignition delay time theory proposed
by Livengood and Wu, these models are KM1-KM10. For these models, 𝐾𝑐 is evaluated from
closure of intake valve till end of combustion (see equation 1). A similar evaluation is done for 𝐾𝑐
in case of model KM11 which is based on thermodynamic parameters. It should be noted that
according to model KM11, the knock prediction parameter directly depends on the change in
specific enthalpy of the end-gas.
Plots of 𝐾𝑐 for all the knock prediction models KM1-KM11 are shown for an operating point so
that the models can be compared to each other and their accuracy can be assessed at the same time.
Models which showed plausible results are further analyzed in the next chapter using contour plots
of their 𝐾𝑐 on the engine operating range.
4.1 Models based on Ignition Delay Time
For models KM1-KM10, the knock prediction parameter (𝐾𝑐) depends on ignition delay time, thus
it is essential to understand how the ignition delay time varies during the engine cycle. Figure 8
shows the ignition delay time with respect to crank angle for the ‘Low Speed’ operating point:
Figure 8. Ignition delay time with respect to Crank Angle as per models KM1-KM10 for Low Speed
19
Since 𝐾𝑐 is evaluated from the beginning of compression till the end of combustion the variation
in ignition delay time with respect to crank angle in this time frame is of interest and is shown in
figure 9. From figure 9, it is evident that models from the same category such as KM1-KM4 have
the same trend of 𝜏 because of the similar mathematical formulation (see equation 3). While on
the other hand, 𝜏 for models belonging to chemical kinetics category KM5-KM9 show different
trend since they are based on a different mathematical formulation (see equation 25). However, it
is also observable from figure 9 that the plot for ignition delay time given by model KM10 has
more similarities to the graphs of KM1-KM4 than with KM5-KM9. Another observation from
figure 9 is that there is huge variation (peak values range approximately from 102-1022) in the
estimation of ignition delay time proposed by all these models with respect to crank angle. Models
based on the Arrhenius function (KM1-KM4) tend to estimate relatively lower values of ignition
delay time whereas models based on chemical kinetics (KM5-KM9) tend to estimate high values.
The only outlier amongst the chemical kinetics models is KM10, which shows ignition delay time
values like that of models based on the Arrhenius function.
Figure 9. Ignition delay time with respect to crank angle (from intake valve closure to end of combustion)
Figure 9 shows variation in 𝜏 with respect to crank angle for models KM1-KM10 and this is the
information that will be used in determining if there is occurrence of knocking at the considered
operating point (in this case – Low Speed) or not. This will be done by evaluating 𝐾𝑐 using 𝜏 as
per equation 1. Then the evaluation of 𝐾𝑐 for all models needs to be compared to check which
Chemical Kinetics
Arrhenius Function
KM10
Ign
itio
n D
ela
y T
ime
[s]
20
models show reasonable results. It should be noted that during experiments the engine did not
show any characteristics of knocking while it was run using the WHSC load points. No knock was
therefore observed on any of the 13 operating points which are used for assessment of the knock
prediction models. From this information it would be easier to observe that which knock prediction
models emulate the experimental observations and which do not. This leaves only the plausibly
correct knock prediction models to be considered as reliable for further combustion analysis. In
figure 10 𝐾𝑐 is evaluated for models KM1-KM10 and represented along with the heat release curve
to indicate the combustion process. The plots of 𝐾𝑐 for models KM1-KM10 for the remaining 12
operating points are presented in Appendix-A1.
Figure 10. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (Low Speed)
Out of all the models based on ignition delay time, KM10 gave the best plausible results since its
outcomes strongly agreed with the experimental observations. This is because the value of 𝐾𝑐 for
KM10 was less than 1 for all the operating points thus showing that the there is no knocking,
whereas, the other models did not show such results.
4.2 Model based on Thermodynamics
Apart from the evaluated 10 knock prediction models based on ignition delay time calculation,
there is another model represented by KM11 which hypotheses prediction of knocking based on
21
thermodynamic parameters (see equations 26 and 27). According to the formulation of 𝐾𝑐 as per
model KM11, 𝐾𝑐 is directly proportional to change in specific enthalpy of the end-gas during
combustion and inversely proportional to the total heat release by combustion. Moreover, in case
of KM11 the critical value for 𝐾𝑐 is 1.5 while its 1 for models KM1-KM10.
The variation of 𝐾𝑐 as per KM11 is plotted in figure 11 for the ‘Low Speed’ operating point. As
far as the results pertaining to KM11 go, the model has shown that none of the operating points
are knocking because for all the operating points the value of 𝐾𝑐 has been found to be less than the
defined critical value. Thus, from initial observations of the results the model KM11 shows a good
compatibility with the experimental results.
If a profound look is taken at the formulation of 𝐾𝑐 as per KM11 it is understood that spark timing
plays a very crucial role in predicting occurrence of knocking (see equation 27). This is because
the change in enthalpy (ℎ𝑠𝑡 − ℎ𝑡) is zero until the spark event and if no spark is given during an
engine cycle then there will be no combustion, thus no knocking! This has been observed
experimentally as well, when the charge in the engine is not ignited by a spark there is no knocking.
In addition to that, (ℎ𝑠𝑡 − ℎ𝑡) represents how the specific enthalpy of the end-gas changes after
the spark event, hence taking into consideration the variation in amount of energy that is contained
by the end-gas per unit mass. If this energy contained by the end-gas is high enough to cause auto-
ignition then there will be knocking otherwise the end-gas will be consumed by the propagating
flame front.
Figure 11. 𝐾𝑐 and heat release with respect to CAD for KM11 (Low Speed)
Based on KM11, the 𝐾𝑐 is evaluated for all operating points and their graphs are presented in
Appendix-A2.
22
5 ANALYSIS
The evaluated results are analyzed to assess all the models and subsequently show which one of
them is best suitable for knock prediction in a natural-gas engine. Finally, for the best available
model the variation in 𝐾𝑐 is studied with respect to spark advance and combustion phasing.
Using the evaluated results of all the models and comparing them with experimentally obtained
data the knocking behavior was understood, this was done with the help of contour plots and in-
cylinder pressure traces which were measured from test cell engine. E.g., at certain operating
points some of the pressure traces which were obtained from the test cell engine showed very small
pressure oscillations (<10 bar from peak to peak) during few engine cycles. But these oscillations
were eliminated when an average of all the pressure trace was taken for calibration of the BOOST
model. Therefore, when the simulation results for the same operating points (with little pressure
oscillations) showed knocking tendency it was established that the knock prediction model
correctly predicts the occurrence of knocking at these operating points. Moreover, to get a better
insight into the knocking behavior contours were plotted for 𝐾𝑐 against the engine operating
conditions defined by brake mean effective pressure (BMEP) and engine speed. Contours
corresponding to models KM4 and KM10 were found to show realistic knocking behavior with
variation in engine operating conditions.
5.1 Knocking Behaviour – Contour Plots
Figure 12 shows the variation of 𝐾𝑐 with engine operating conditions for model KM4. From this
figure, the value of 𝐾𝑐 increases with BMEP when engine speed is constant.
Figure 12. Variation in 𝐾𝑐with engine operating conditions as per KM4. Red circle highlights the region which is
most susceptible to knocking and corresponds to high load and low engine speed operating condition
4,5
55,5
6
6,5
6,5
7
7,5
8,5
5 5
5,5
6
7
7,5
8
8,5
7,8794
5,4563
6,9368
8,3099
4,2663
7,8568
4,8931
6,3922
7,9331
6,4758
4,351
6,90628,987
BM
EP
[b
ar]
→
Engine Speed [RPM] →
23
This is because when the engine operates at higher loads parameters like pressure, temperature and
heat released due to combustion are increasing. This high temperature and pressure in the cylinder
also increase the chances of knocking in the end-gas because at higher pressure the auto-ignition
temperature of air-methane mixture decreases (Caron et al., 1999). It is to be noted that 𝐾𝑐 is
highest in the region where the engine operates at high load and low engine speed (marked by the
red circle in figure 12). This is because when the engine operates at low speed the turbulence
intensity in the cylinder is low and results in lower flame propagation speed during combustion.
The lower flame propagation speed allows more time for the end-gas to auto-ignite, hence resulting
in knock (Johansson, 2014). Moreover, the value of 𝐾𝑐 is above the critical limit of 1 across the
entire engine operation range, this indicates that model KM4 needs to be calibrated.
In figure 13 which corresponds to KM10, a similar trend as KM4 was observed for 𝐾𝑐.
Highlighting the fact that the engine is more susceptible to knocking at higher loads and lower
engine speeds (marked by the red circle in figure 13). The contour plot of model KM10 also shows
a region at medium-high speed where 𝐾𝑐 is higher (marked by the black circle in figure 13). This
means that when the engine is operated to deliver higher power at high loads the chances of
knocking increase. This is because, to deliver higher power more fuel is burned which leads to
higher heat release rate and subsequently higher temperatures in the cylinder resulting in an
increased chance of auto-ignition in the end-gas. Furthermore, the value of 𝐾𝑐 is below the critical
limit of 1 across the entire engine operation range indicating that there is no knocking. This rightly
matches with the experimental observations because none of 13 operating points showed knocking
in the test-cell engine.
Figure 13. Variation in 𝐾𝑐with engine operating conditions as per KM10. Red and black circles highlight the high
load engine operating conditions which are highly susceptible to knocking
Amongst all models based on ignition delay time only KM4 and KM10 showed reasonable
knocking behaviour. Contour plots for all the remaining models are presented in Appendix-B.
Apart from the models based on ignition delay time there was a model namely KM11 which was
based on thermodynamic parameters. In case of KM11, it was found that none of the 13 operating
0,2
0,2
0,25
0,3
0,35
0,4
0,45
0,5
0,55
0,6
0,7
0,2 0,2
0,25
0,3
0,35
0,4
0,45
0,5
0,55
0,6
0,6
5
0,7
0,63328
0,24198
0,50542
0,67749
0,14149
0,6628
0,21058
0,44678
0,72957
0,43507
0,17082
0,570030,77399
BM
EP
[b
ar]
→
Engine Speed [RPM] →
24
points were knocking which coincided with experimental observations. However, from the contour
plot for KM11 it was established that the knocking behaviour shown by the model is incorrect (see
figure 14). This is because the contour plot indicates that the engine is most susceptible to knocking
in the low-mid load region and low-mid engine speed range (marked by black circle in figure 14).
Since this knocking behaviour shown by KM11 does not agree with the experimental observations,
this model is discarded for further analysis.
Only model KM10 was chosen for further analysis of the knocking behaviour with respect to
combustion phasing. This is because none of the models other than KM4 and KM10 showed
correct knocking behaviour across the engine operating range. Furthermore, to achieve better
accuracy with KM4 the model had to be calibrated. Thus, amongst all models KM10 comes out
on top and was used for further analysis of the knocking behaviour with respect to combustion
phasing.
Figure 14. Variation in 𝐾𝑐with engine operating conditions as per KM10. According to this model the engine
operating range that is most susceptible to knocking is the black circle which is around low-mid engine load
5.2 Variation in Kc with Combustion Phasing
According to Wildhaber combustion phasing can be defined as, “The time in the engine cycle,
specifically the compression and expansion strokes, where combustion occurs. A change in
combustion phasing causes a change in combustion duration” (Wildhaber, 2011). Combustion
phasing plays a crucial role not only on the efficiency of an engine but also in determining if the
engine is going to knock or not. The effect of combustion phasing on knocking can be found by
changing the combustion phasing and evaluating its influence on 𝐾𝑐. There are two ways in which
the combustion phasing can be changed:
1. By changing the spark advance: If the spark advance is increased the chances of knocking
increases and if the spark is retarded then the risk of knocking reduces. The spark advance
then plays a crucial role in determining the efficiency of the engine and it is needed to
0,7
0,75
0,8
0,8
0,85
0,85
0,85
0,9
0,95
0,95
10,
70,75
0,8
0,8
0,85
0,85
0,85
0,9
0,9
0,9
0,95
1
0,9467
0,8923
0,8209
0,7311
0,8092
0,8209
0,9795
1,0359
0,9313
1,0341
0,8882
0,76670,9674
Engine Speed [RPM]→
BM
EP
[b
ar]
→
25
extract higher power but then the engine should be calibrated such that it does not reach
the knocking limits. It has also been found that by using cooled EGR the maximum BMEP
of the engine can be increased with advance in spark timing (Grandin et al., 1998).
2. By changing the combustion duration: When the combustion duration is long then it means
that it takes a longer time to burn the inducted charge in the cylinder, meaning that the
flame propagation speed is low. Slow combustion allows more time for the end-gas to auto-
ignite and to avoid knocking the flame propagation should be fast. Controlling the
combustion duration is extremely difficult. Therefore, generally EGR is added in the
charge to increase the time of end-gas auto-ignition, addition of EGR dilutes the unburned
gas mixture and reduces the peak temperatures thus requiring longer time for the end-gas
to heat up to auto-ignition temperatures (Heywood, 1988). This allows the propagating
flame front to consume the end-gas before it auto-ignites (Wang et al.,2017).
The chosen knock prediction model KM10 is used for analysing the influence of spark advance
and combustion duration on 𝐾𝑐. To maintain consistency in the simulations and to make a direct
comparison of the results BMEP was kept constant for every operating point and then the spark
advance and combustion duration were separately changed from their original value. The variation
in 𝐾𝑐 is plotted for a change of ±5° of CAD (step size 1°) in spark advance and combustion duration
respectively. In the following figure 15, the variation in 𝐾𝑐 with change in spark advance for all
the operating points is shown. It was found that 𝐾𝑐 increases linearly with spark advance and the
gradient of 𝐾𝑐 versus spark advance is approximately 6 times steeper at higher engine loads as
compared to lower loads.
From figure 16, it can be seen that 𝐾𝑐 decreases linearly with increase in combustion duration.
This is because the simulations were run by keeping the BMEP constant and in that case when the
combustion duration is increased then the peak pressure inside the cylinder would reduce hence
decreasing the possibility of knocking.
The BMEP was kept constant while analysing the impact of combustion phasing (both spark
advance and combustion duration) on 𝐾𝑐 so that it could be understood how the combustion
phasing influences knocking when the same power is extracted from the engine. The throttle angle
was varied to keep the BMEP constant while the combustion phasing was modified to carry out
this analysis. Throttle angle was chosen as a parameter to vary in the simulations because it is
practically possible to control, if in case there is any experimentation to be performed in future for
validation of this analysis.
26
Figure 15. Variation in 𝐾𝑐 with respect to spark advance for all operating points. Operating points are colour coded
based on load.
y = -0.0088x + 0.0852
y = -0.0589x + 0.3349
0,00
0,20
0,40
0,60
0,80
1,00
1,20
-22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0
Kc
[-]
SOC [ ]
P2
P3
P4
P5
P6
P7
P8
P9
P10
P11
P12
Max_PowLow Speed
BM
EP
[bar
]
Engine Speed (RPM)
LOAD
27
Figure 16. Variation in 𝐾𝑐 with respect to combustion duration for all operating points. Operating points are colour
coded based on load.
y = -0.0034x + 0.2508
y = -0.0177x + 1.4383
0,00
0,20
0,40
0,60
0,80
1,00
1,20
25 30 35 40 45 50 55
Kc
[-]
Combustion Duration [°]
P2
P3
P4
P5
P6
P7
P8
P9
P10
P11
P12
Max_PowLow Speed
BM
EP
[bar
]
Engine Speed (RPM)
LOAD
28
6 CONCLUSIONS
Outcomes and findings of the thesis are laid out in this section. Arguments are proposed in lieu of
the outcomes. The findings are compared with previously conducted researches.
The phenomenon of knocking is sporadic and is difficult to predict. There are a several types of
ECMs that can be used to predict the occurrence of knocking in an SI engine. Out of all the types
of models available, three types of knock prediction models are assessed in this thesis: (1) models
based on the Arrhenius function, (2) models based on chemical kinetics and (3) model based on
thermodynamics. The assessment of the models was done using 1D engine simulations in BOOST.
The simulated results are compared to experimental observations from a test-cell engine.
Among all the 11 different knock prediction models, it was found that two models showed
plausible knocking behaviour and have results that agree with the experimental observations.
These two models are:
1. KM4 – this is a model based on the Arrhenius function (Soylu,2005)
2. KM10 – this is a model based on chemical kinetics (Petersen et al., 1999)
Out of these models KM10 is the best because its results agree with the experiments. Since KM10
is based on the ignition delay time estimation provided by Petersen it can be concluded that his
formulation to evaluate the ignition delay time is correct (see equation (25) and table 2). It was
also observed that of all the models based on chemical kinetics it was only model KM10 which
claimed that the ignition delay time is inversely proportional to the fuel molar density, hence,
indicating that 𝐾𝑐 depends upon load. This is because the fuel molar density increases with increase
in load at constant engine speed and it has been observed experimentally that at higher loads the
engine is more susceptible to knocking.
Based on the assessment of all the models, KM10 was chosen to analyse the variation in knocking
parameter with a change in combustion phasing and at the same time ensuring that the BMEP
delivered by the engine is kept the same. From this analysis it was found that the knock prediction
parameter increases linearly with spark advance and decreases linearly with an increase in
combustion duration. Another observation was that the change in knock prediction parameter with
spark advance is dependent on engine load, at higher loads a high gradient between 𝐾𝑐 and spark
advance is observed while at lower loads the curve has less slope.
29
7 FUTURE WORK
There are potential areas that can be explored based on the thesis, what can be done further in
the field of knock prediction models is presented in this chapter.
There are always a few constraints and limitations that make technical exploration difficult in
research projects. Based on this work, there are few areas which are interesting to tap into but were
not explored due to several limitations. The author recommends that the following can be a good
extension to the current work:
1. Natural gas a fuel occurs in nature as a mixture of several gases, but in this thesis the fuel
was assumed as pure methane. There are claims that presence of gases like propane change
the combustion properties and knocking behaviour of the fuel. Thus, for future work it is
recommended to use the fuel as a mixture of methane and propane.
2. Model KM4 showed plausible knocking behaviour but was not correct in terms of
evaluation of the knock prediction parameter. This model can be calibrated and then used
to predict knocking.
3. The experimental results are available for only 13 operating points which are used to assess
and analyse different types of knock prediction models. More test cell data can be acquired
to further analyse and then quantify the knocking behaviour. The quantification should be
done in such a way that the values of 𝐾𝑐 are able to distinguish between no knock,
borderline knock and sever knock.
4. The focus of the thesis was on spark-knock and surface ignition was not considered for
simulations. As an extension, surface ignition can also be taken into consideration.
30
8 REFERENCES
Ganesan, V. (2008). Internal Combustion Engines. 4th ed. New Delhi: Tata McGraw-Hill, pp.1-7.
Cengel, Y. and Boles, M. (2015). Thermodynamics - An Engineering Approach. 8th ed. New
York, NY: McGraw-Hill Education, pp.70-77, 490-496, 506.
Heywood, J. (1988). Internal Combustion Engine Fundamentals. New York: McGraw-Hill, pp.4,
9-12, 102-103, 371-412, 450-457, 837-839, 915.
Johansson, B. (2014). Combustion engines. Lund: Department of Energy Sciences, Lund
University, pp.227-268, 299.
MacLean, H. and Lave, L. (2003). Evaluating automobile fuel/propulsion system technologies.
Progress in Energy and Combustion Science, 29(1), p.28.
Hpcomb.kaust.edu.sa. (2018). Home - Research. [online] Available at:
https://hpcomb.kaust.edu.sa/Pages/Advanced%20Spark%20Ignition%20Engines.aspx [Accessed
12 Mar. 2018].
Holzbecher, E. (2012). Environmental modeling. Berlin: Springer., p.35.
Lifshitz, A., Scheller, K., Burcat, A. and Skinner, G. (1971). Shock-tube investigation of ignition
in methane-oxygen-argon mixtures. Combustion and Flame, 16(3), pp.311-321.
Grandin, B., Denbratt, I., Bood, J., Brackmann, C., Bengtsson, P., Gogan, A., Mauss, F. and
Sundén, B. (2002). Heat Release in the End-Gas Prior to Knock in Lean, Rich and Stoichiometric
Mixtures with and Without EGR. SAE Technical Paper Series.
Livengood, J. and Wu, P. (1955). Correlation of autoignition phenomena in internal combustion
engines and rapid compression machines. Symposium (International) on Combustion, 5(1),
pp.347-356.
Douaud, A. M., Eyzat, P.; Four-Octane-Number Method for Predicting the Anti-Knock Behavior
of Fuels and Engines; SAE Technical Paper 780080.
Elmqvist-Möller, C. (2006). 1-D simulation of turbocharged SI engines: focusing on a new gas
exchange system and knock prediction (Licentiate dissertation). KTH, Stockholm. Retrieved from
http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-4218
Soylu, S. (2005). Prediction of knock limited operating conditions of a natural gas engine. Energy
Conversion and Management, 46(1), p.130.
Glassman, I. (1996). Combustion. 3rd ed. San Diego: Academic Press, pp.90-94, p.38, p.327.
Karim, G. and Gao, J. (1992). A Predictive Model for Knock in Spark Ignition Engines. SAE
Technical Paper Series, 922366.
Attar, A. (1997). Optimization and knock modeling of a gas fueled spark ignition engine. Ph.D.
Thesis. The University of Calgary. pp.66-79.
Sierra Parra, A. F. and Díaz Torres, A. G. (2017). Improvement of a knock model for natural gas
SI engines through heat transfer evaluation. International Journal on Interactive Design and
Manufacturing. Springer Paris, pp. 1–11. doi: 10.1007/s12008-017-0452-6.
Seery, D.J., Bowman, C.T. (1970). An experimental and analytical study of methane oxidation
behind shock wave. Combustion and Flame. Vol 14. pp.37-47.
Lifshitz, A., Scheller, L., Burcat, A., Skinner, G.B. (1971). Shock-tube investigation of ignition in
methane-oxygen-argon mixtures. Combustion and Flame, 16. p.331.
31
Tsuboi, T., Wagner, H.G. (1974). Homogeneous thermal oxidation of methane in reflected shock
waves. Proceedings of the Combustion Institute, 15. p.883.
Cheng, A.K., Oppenheim, R.K. (1984). Autoignition in methane-hydrogen mixtures. Combustion
and Flame, 58. pp.125-139.
Grillo, A., Slack, M.W. (1976). Shock tube study of ignition delay times in methane-oxygen-
nitrogenargon mixtures. Combustion and Flame, 27. pp.377-381.
Petersen, E.L., Davidson, D.F., Hanson, R.K. (1999). Kinetics modeling of shock-induced ignition
in low-dilution CH4/O2 mixtures at high pressures and intermediate temperatures. Combustion
and Flame, 117. pp.272-290.
Emission Test Cycles: World Harmonized Stationary Cycle (WHSC). Revision: 2008.08a.
Available at: https://www.dieselnet.com/standards/cycles/whsc.php (Accessed: 29 June 2018).
Caron, M. et al. (1999). Pressure dependence of the auto-ignition temperature of methane-air
mixtures. Journal of Hazardous Materials, A65. pp.1–11.
Wildhaber, S. (2011). Impact of combustion phasing on energy and availability distributions of
an internal combustion engine. Masters Theses. Available at:
http://scholarsmine.mst.edu/masters_theses/4930.
Grandin, B. et al. (1998). Knock suppression in a turbocharged SI engine by using cooled EGR.
SAE Technical Papers, (724). doi: 982476.
Wang, Z., Liu, H. and Reitz, R. D. (2017). Knocking combustion in spark-ignition engines.
Progress in Energy and Combustion Science. Elsevier Ltd, 61. pp.101–105. doi:
10.1016/j.pecs.2017.03.004.
32
APPENDIX A1: Kc v/s CAD (KM1-KM10)
Figure A1-1. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P2)
Figure A1-2. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P3)
33
Figure A1-3. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P4)
Figure A1-4. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P5)
34
Figure A1-5. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P6)
Figure A1-6. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P7)
35
Figure A1-7. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P8)
Figure A1-8. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P9)
36
Figure A1-9. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P10)
Figure A1-10. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P11)
37
Figure A1-11. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (P12)
Figure A1-12. 𝐾𝑐and heat release with respect to CAD for models KM1-KM10 (Max_Pow)
38
APPENDIX A2: Kc v/s CAD (KM11)
Figure A2-1. 𝐾𝑐and heat release with respect to CAD for model KM11 (P2)
Figure A2-2. 𝐾𝑐and heat release with respect to CAD for model KM11 (P3)
39
Figure A2-3. 𝐾𝑐and heat release with respect to CAD for model KM11 (P4)
Figure A2-4. 𝐾𝑐and heat release with respect to CAD for model KM11 (P5)
40
Figure A2-5. 𝐾𝑐and heat release with respect to CAD for model KM11 (P6)
Figure A2-6. 𝐾𝑐and heat release with respect to CAD for model KM11 (P7)
41
Figure A2-7. 𝐾𝑐and heat release with respect to CAD for model KM11 (P8)
Figure A2-8. 𝐾𝑐and heat release with respect to CAD for model KM11 (P9)
42
Figure A2-9. 𝐾𝑐and heat release with respect to CAD for model KM11 (P10)
Figure A2-10. 𝐾𝑐and heat release with respect to CAD for model KM11 (P11)
43
Figure A2-11. 𝐾𝑐and heat release with respect to CAD for model KM11 (P12)
Figure A2-12. 𝐾𝑐and heat release with respect to CAD for model KM11 (Max_Pow)
44
APPENDIX B: Contour Plots of Knock Models
Figure B1. Variation in 𝐾𝑐with engine operating conditions as per KM1
Figure B2. Variation in 𝐾𝑐with engine operating conditions as per KM2
1
1
1,5
2
2,5
3
3,5
4
1
1,5
2
2,5
3
3,5
4
3,8536
1,0607
2,6407
3,969
0,7884
3,0846
0,9759
1,9975
3,8406
1,992
0,8429
2,99223,8048
BM
EP
[b
ar]
→
Engine Speed [RPM] →
1,5
2,25
3
3,75
4,5
5,25
1,5
2,25
3
3,75
4,5
5,25
6
6,0326
1,6604
4,1339
6,2132
1,2342
4,8287
1,5277
3,127
6,0122
3,1184
1,3196
4,68415,9562
BM
EP
[b
ar]
→
Engine Speed [RPM] →
45
Figure B3. Variation in 𝐾𝑐with engine operating conditions as per KM3
Figure B4. Variation in 𝐾𝑐with engine operating conditions as per KM5
2,25
3
3,75
4,5
5,25
6
2,25
3
3,75
4,5
5,25
6
6,3424
2,305
4,5952
6,5769
1,9536
5,0918
2,1617
3,5832
5,9496
3,6779
1,9592
4,83316,2995
BM
EP
[b
ar]
→
Engine Speed [RPM] →
0,02
0,02
0,03
0,03
0,04
0,04
0,04
0,05
0,05
0,02
0,02
0,03
0,03
0,04 0,0
5
0,01768
0,04170
0,03035
0,02359
0,01214
0,06358
0,03200
0,05681
0,03764
0,05022
0,02094
0,034400,05541
BM
EP
[b
ar]
→
Engine Speed [RPM] →
46
Figure B5. Variation in 𝐾𝑐with engine operating conditions as per KM6
Figure B6. Variation in 𝐾𝑐with engine operating conditions as per KM7
0,004
0,0060,0080,01
0,01
0,004
0,006
0,006
0,008
0,008
0,01
0,00366
0,01172
0,00635
0,00464
0,00409
0,01167
0,00897
0,01168
0,00696
0,01058
0,00622
0,006700,01011
BM
EP
[b
ar]
→
Engine Speed [RPM] →
22
3
3
4
4
2
2
3
3
4
1,0299
4,494
2,1252
1,4131
1,2395
4,2405
3,2833
4,4381
2,2135
3,8668
2,1499
2,17263,4862
BM
EP
[b
ar]
→
Engine Speed [RPM] →
47
Figure B7. Variation in 𝐾𝑐with engine operating conditions as per KM8
Figure B8. Variation in 𝐾𝑐with engine operating conditions as per KM9
0,3
0,4
0,4
0,4
0,5
0,10,2
0,3
0,3
0,4
0,5
0,20769
0,30901
0,30288
0,26419
0,09789
0,60285
0,24293
0,49228
0,41271
0,43593
0,16113
0,363520,54824
BM
EP
[b
ar]
→
Engine Speed [RPM] →
0,0020,002
0,003
0,003
0,004
0,004
0,004
0,002
0,002
0,003
0,003
0,004
0,00125
0,00505
0,00245
0,00167
0,00154
0,00493
0,00380
0,00506
0,00263
0,00452
0,00254
0,002550,00413
BM
EP
[b
ar]
→
Engine Speed [RPM] →