Managing Transient Behaviors of a Dual Mode Spark Ignition / Controlled
Auto Ignition Engine with a Variable Valve Timing System
by
Halim G. Santoso
S.M. Mechanical Engineering
Massachusetts Institute of Technology, 2002
B.S. International Development Engineering
Tokyo Institute of Technology, 2000
SUBMITTED TO THE DEPARTMENT OF MECHANICAL ENGINEERING IN PARTIAL
FULLFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
AT THE
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
December 2004
@ Massachusetts Institute of Technology
All Rights Reserved
Signature of Author:__
Department of Mechanical Engineering
December 2004
Certified by:
Wai K. Cheng
Professor of Mechanical Engineering
Thesis Supervisor
Accepted by:
Lallit Annand
Chairman, Departmental Graduate Committee
MASSACHUSETTS INSTiTUTE
OF TECHNOLOGY
BARKERIR
MAY 0 5 2005
LIBRARIES
Managing Transient Behaviors of a Dual Mode Spark Ignition /
Controlled Auto Ignition Engine with a Variable Valve Timing System
by
Halim G. Santoso
Submitted to the Department of Mechanical Engineering on December 15, 2004 in partial
fulfillment of the requirements for the Degree of Doctor of Philosophy
ABSTRACT
Gasoline Homogeneous Charge Compression Ignition (HCCI) engine has the potential of
providing better fuel economy and emissions characteristics than current spark ignition engines.
One implementation of this technology employs a Variable Valve Timing (VVT) system and is
also often referred to as Controlled Auto Ignition (CAI) combustion in the literature. The
objective of the study can be divided into two topics. First, the dynamic nature of load trajectory
and several important phenomena in CAI mode were investigated. Second, the issues
encountered during mode transition between SI and CAI regime were considered.
Despite wide-open-throttle operation, pumping loss in CAI mode was not negligible. A major
source of pumping loss in CAI mode was the heat transfer to cylinder wall during the
recompression process due to the high in-cylinder residual gas temperature. The influence of fuel
air equivalence ratio on combustion stability was analyzed to explain the misfires phenomenon
in fuel rich condition during transient operation. Heat release analysis has been conducted to
characterize the combustion process in CAI mode. Large variations of the 50% burned point due
to fluctuation of residual gas mass and temperature were observed.
Small step changes in valve timings (EVC, EVO, and IVC) and fuelling resulted in a new steady
state within 3-4 engine cycles at 1500 rpm. These small step changes are reversible in nature.
Sudden large step change in load required much longer time to reach steady state due to the time
required for thermal stabilization. Misfires were observed in large low-load-to-high-load step
change but not in high-load-to-low-load step change. A simple open loop controller based on
linear interpolation of fuel injection and valve timing events was implemented to assess the time
scale required to avoid misfires during large load perturbation.
Mode transition from the SI to CAI regime may be categorized as failed, non-robust, and robust.
In a failed transition, the engine would not reach steady state CAI combustion. In a non-robust
transition, one or more intended CAI cycles misfire, although the cycles progress to a
satisfactory CAI operating after a few cycles. In robust transition, every intended CAI cycle is
successful. A mode transition strategy based on late IVC and fuel metering strategy has been
proposed. Smooth and robust modes transitions both from SI to CAI and vice versa have been
experimentally demonstrated by implementing this strategy.
Thesis Supervisor: Wai Cheng
Title: Professor of Mechanical Engineering
3
Acknowledgements
I am sincerely grateful to many individuals that have made significant contributions to this
project. I couldn't have finished the project without their supports.
Prof Wai K.Cheng, who has been my thesis advisor for both my master and doctoral projects. He
made sure that I could focus on my research without having to worry about anything else. He is
also always available to give me his sage guidance whenever I'm in doubt. His conviction that
"Everything can be done, just DO it!" has taught me additional confidence in tackling any new
problem.
Prof William Green, who has given numerous excellent suggestions in the modeling part of the
project. He made sure I had the access to the right modeling tools, which has saved me a lot of
time. It was also a pleasure to collaborate with some of his students.
Prof. John Heywood, who has taught me the fundamentals of internal combustion engine. It is also
through his assistance that Daimler Chrysler donated the original VVT system to us. His remarkable
presentation skill is one of the things that I am still trying to learn.
Prof Bora Mikic, who is a wonderful teacher and an important member of my thesis committee.
Heat and Mass Transfer has been my favorite subject since my undergraduate education and I
have learned a lot more about it from him.
Dr.Jeff Matthews, who has been a wonderful colleague and personal friend. We shared so much
time troubleshooting the experimental apparatus. Working with him, some of the most frustrating
tasks became enjoyable. I also learned so many things about daily life from him.
Peter Morley and the rest of MIT Machine shop, who helped me to make, modify, and repair
some of the experimental apparatus. I don't know how many times I said, " I need it fast " and
got it the next morning. I am grateful for their assistance.
Thane Dewitt, our laboratory supervisor, who always made sure I got the parts I need on time.
Morgan Andreae, Yuetao Zhang (Tony), and Kevin Lang, who have been wonderful friends. I
truly enjoyed our discussions about engines, new inventions, politics, etc. during our afternoon
strolls to the student center. I wish you all the best with your Phds.
Dr. Tom Asmus and Daimler Chrysler Corporation who has donated the original VVT system to
us. Without them, the project would not be completed in such a short period of time.
US Department of Energy, which has funded the project through HCCI University Consortium.
Finally I would like to thank my family. My wife, Catherina Wijaya, who is always there to
encourage me during the difficult times. My daughter, Quinna Halim, whose presence always
brings joy to my heart.
4
Table of Contents
A B S T R A C T ..........................................................
. .....
. .. ..............
......... . 3
A cknow ledgem ent........................................................4
T ab le of C o ntents..............................................................................................5
List of Sym bols and A bbreviation.........................................
............................
8
L ist of F igu res............................................................................................
.. 9
CHAPTER 1 Introduction...................................................13
1.1
B ackground .............................................................................
. . 13
1.1.1
HCCI combustion literature review..........................13
1.1.2
Controlled Auto Ignition .................................................
16
1.2
O bjectives..............................................................................
1.3
Thesis Framework..............................................17
. .. 17
CHAPTER 2 Experimental Apparatus & Modeling Tools......................................
2.1
2.2
Experimental Apparatus and Procedures.............
..............................19
2.1.1
Engine and Dynamometer ................................................
2.1.2
Fuel Injection System...................
2.1.3
Variable Valve Timing System..................................................22
2.1.4
Intake Air Temperature Controller...........................................24
2.1.5
Pressure Sensor..
2.1.6
Lambda Sensor............................................26
2.1.7
Labview Data Acquisition System..............................................26
.......
19
.......
.............................................
19
............. 20
........ 25
Modeling Tools.............................................27
2.2.1
Ricardo wave I-D code......................................................27
2.2.2
Kinetics Model..............................................28
2.2.3
H eat Release C ode................................................................28
5
CHAPTER 3 Consideration and Constraints for a SI/CAI engine.........................29
. 29
3.1
Introduction ..............................................................................
3.2
C onstraints w ithin CAI regim e ........................
3.3
Importance of Mode Transition..................................34
30
...................................
CHAPTER 4 Steady State and Transient Operation within CAI regime.............39
. . 39
4 .1
Introduction .............................................................................
4.2
M apping of the CAI regim e............................................................39
4.3
Effects of Fuel Metering on Combustion Stability....................................
4.4
Pumping Loss in CAI mode.....................................57
4 .4 .1
Introduction ........................................................................
4.4.2
Sources of Pumping Loss in CAI Mode
49
57
59
.......................
4.5
Heat release analysis and combustion phasing characteristics....
......... 63
4.6
Spark Assisted Controlled Auto Ignition..................
......... 70
4.7
Oxidation in the Recompression Process.........................................73
4.8
Engine Transient Response to Step Change in CAL..
4.9
75
..................
......... 75
4.8.1
Fuel Metering Effect..................
4.8.2
Step Change in IVC.....................................
4.8.3
Step Change in EVC.......................................77
4.8.4
Step C hange in EV O ..............................................................
78
Open Loop Load Control..........................................
81
76
4.9.1
Large Step Change from High to Low Load...................................81
4.9.2
Large Step Change from Low to High Load...........
4.9.3
Load Modulation by Fuel and Valve Timing Interpolation...............89
......... 86
CHAPTER 5 Mode Transition Management between SI mode and CAI mode................92
....... 92
5.1
Early Work on SI-CAI Mode Transition...........
5.2
Late IVC as a Load Control Mechanism in SI mode..................................96
5.3
CAI to SI Mode Transition.....................................99
5.3.1
CAI to SI mode transition phenomena......................................99
5.3.1
CAI to SI mode transition strategy................................100
6
5.3.2
5.4
CAI to SI mode transition demonstration....................................100
SI to CAI Mode Transition... ..........
5.4.1
............. 103
SI to CAI Mode Transition Phenomena....................................103
5.4.1.1 Failed Transition........................................................103
..... 106
5.4.1.2 N on-robust Transition.............................................
5.4.2
SI to CAI Mode Transition Strategy...............................110
5.4.3
SI to CAI Mode Transition Demonstration.................................113
120
CHAPTER 6 Summary and Conclusion.....................................
6.1
6.2
APPENDIX
W ithin CA I Regim e......................................................
...... 120
120
6.1.1
Steady State CAI Phenomena........................
6.1.2
Transient CAI Phenom ena............................................121
SI-CAI Mode Transition.............................................
. . 122
6.2.1
Transition from SI to CAI M ode....................................122
6.2.2
Transition from CAI to SI Mode.....................
TWo-zone Heat Release Code ..................................
REFEREN CE S .............................................................................................
124
126
127
7
List of Symbols and Abbreviations
ABDC
ATDCi
ATDCf
A/F
BBDC
BMEP
CA
CAI
CA50
EGR
EVC
EVO
F/A
FTP
GIMEP
HCCI
IMEP
IVC
IVO
MAP
MBT
NIMEP
P
PFI
PMEP
RON
RPM
RVP
SI
SOC
TDC
WOT
After Bottom Dead Center
After Top Dead Center Intake
After Top Dead Center Firing
Air Fuel Ratio
Before Bottom Dead Center
Brake Mean Effective Pressure
Crank Angle
Controlled Auto Ignition
Crank Angle corresponding to location of 50% mass fraction burned
Exhaust Gas Recirculation
Exhaust Valve Closing
Exhaust Valve Open
Fuel Air Ratio
Federal Test Procedure
Gross Indicated Mean Effective Pressure
Homogeneous Charge Compression Ignition
Indicated Mean Effective Pressure
Intake Valve Closing
Intake Valve Open
Manifold Absolute Pressure
Maximum Brake Torque
Net Indicated Mean Effective Pressure
Pressure
Port Fuel Injection
Pumping Mean Effective Pressure
Research Octane Number
y
Specific heat ratio
Air fuel equivalence ratio
Ignition delay (ms)
Fuel air equivalence ratio
(1
Revolution Per Minute - engine speed
Reid Vapor Pressure
Spark Ignition
Start of Combustion
Top Dead Center
Wide Open Throttle
8
List of Figures
Chapter 1
Figure 1.1
Illustration of Diesel, Spark Ignited, and HCCI Combustion................ 14
Chapter 2
20
Table 2.1
Summary of engine geometry...................................
Table 2.2
Certificate of Analysis ofUTG-91..................................21
Figure 2.1
Fuel injector calibration curve.........................................................
Figure 2.2
Cross Section Diagram of VVT Pod..........................
...................... 23
Table 3.1
Physical Variables of CAI combustion...............
..................
Table 3.2
Engine Control Variables .......................................
Figure 3.1
EPA Highway Fuel Economy Drive...................
Figure 3.2
Fed'eral Test Procedure for Ford Taurus 3.0 L....................................
Figure 3.3
CAI Operating Regime in FTP test...............................34
Figure 3.4
Details of Test Cycle#5 during FTP test................................................35
Figure 3.5
Residence Time in CAI Regime................................36
22
Chapter 3
30
31
..........................
..... 32
33
Chapter 4
...
.....
...
........
40
Figure 4.1
Map of CAI Operating Regime...
Figure 4.2
IVC Effect on Nimep.........................................41
Figure 4.3
EV C Effect on N imep..................................................
Figure 4.4
Four Consecutive Pressure Traces at High Load Limit......
Figure 4.5
Average of 300 cycles Pressure Trace at High Load Limit......................43
Figure 4.6
Average of 300 cycles Pressure Trace at Low Load Limit.......
Figure 4.7
Schematic of In-cylinder Condition Estimation Procedures.................. ........ 45
Figure 4.8
In-cylinder Temperature, Residual Mass Fraction and Charge
................ 42
............43
......... 44
Mass Estimation................................................46
9
Figure 4.9
Residual Mass Fraction vs Gimep...................
......
....... 48
Figure 4.10
Residual Mass Fraction vs In-cylinder Temperature................
....... 48
Figure 4.11
Fuel M etering Effect on Combustion Stability.........................................49
Figure 4.12
Fuel Equivalence Ratio Effect on COV and Average Nimep.....
Figure 4.13
Wave Simulation and Experimental Pressure Data......
Figure 4.14
In-cylinder Temperature Profile at Different Fuel Air
........
...............
Equivalence Ratio...........................................
50
51
52
Figure 4.15
Equivalence Ratio Effect on Total Charge Mass........................53
Figure 4.16
Equivalence Ratio Effect on Peak Charge Temperature.................. 53
Figure 4.17
Equivalence Ratio Effect on Charge Temperature at 30BTDC......................54
Figure 4.18
Equivalence Ratio Effect on Residual Gas Mass Fraction...........................54
Figure 4.19
Fuel Injection Effect on Initial Charge Temperature...............................56
Figure 4.20
Fuel Injection Effect on Average Gamma during Compression... ...................
Figure 4.21
Schematic of P-V diagram in light load SI combustion..............................58
Figure 4.22
Schematic of Pmep definition.......................................
60
Figure 4.23
Illustration of two sources of pumping loss in CAI combustion... ..................
60
Figure 4.24
Experimental Data of Pmep during steady state CAI mapping......
Figure 4.25
Detailed of Pmep from Wave Modeling.............................................
63
Figure 4.26
Fuel Mass Fraction Burned Profile..................................
64
Figure 4.27
Normalized Heat Release Profile................................64
Figure 4.28
Location of CA50 vs Imep.........................................65
Figure 4.29
Location of CA50 vs Imep........................................65
Figure 4.30
Correlation between CA50 and Peak Cylinder Pressure.........66
Figure 4.31
Relative Cum ulative H eat Loss..........................................................67
Figure 4.32
Details of Energy Release Curve.................................68
Figure 4.33
Comparison Between Late and Early Burning Cycle...............
Figure 4.34
Pressure Trace of a Spark Assisted CAI Cycle
56
....... 62
......... 69
during SI to CA I mode transition........................................................71
Figure 4.35
An Example of Spark Assisted CAI Cycle at steady state
high load operation .........................................................................
Figure 4.36
71
Combustion in the Recompression Process..............................................74
10
................ 74
Figure 4.37
Oxidation During Normal Combustion in CAI mode...
Figure 4.38
Step Change in Fuel Injection Amount..............................
Figure 4.39
Engine Response for a Step Change in IVC..........................
76
Figure 4.40
Engine Response for a Step Change in IVC..........................
77
Figure 4.41
Engine Response for a Step Change in EVC...........................................78
Figure 4.42
Effect of EVC timing.........................................
Figure 4.43
Effect of Retarding EVO timing....
Figure 4.44
Effect of Retarding EVO timing..................................80
Figure 4.45
Large Step Change from High Load to Low Load....................................81
Figure 4.46
Details of Pressure Trace before and after Step Change in Load...
Figure 4.47
Wave Simulation Results for High Load to
... .......
.
75
........
79
79
.............................
82
........
Low Load Step Change........................................83
84
.........................
Figure 4.48
Summ ary of W ave Simulation...................
Figure 4.49
Ignition Timing Estimation..........................................
Figure 4.50
Large Step Change from Low Load to High Load........
Figure 4.51
Details of Low Load to High Load Transient.........................................87
Figure 4.52
Wave Modeling Result........... .........
Figure 4.53
M odulation Period of 60 Cycles..........................................................90
Figure 4.54
M odulation Period of 30 Cycles.........................................................90
Figure 4.55
Modulation Period of 14 Cycles.........91
Figure 4.56
Details of Failed Transient at 14 Cycles Modulation Period........................91
86
............
...............
85
..... 88
Chapter 5
............... 92
Figure 5.1
Mode Transition by Fuel Metering.................
Figure 5.2
Non-synchronized Mode Transitio.....................................93
Figure 5.3
Non-robust SI-CAI M ode Transition..................................................95
Figure 5.4
Robust SI-CAI Transition with Throttle Movement...
Figure 5.5
P-V Diagram of Throttleless SI Engine with Late IVC...............................98
Figure 5.6
Spark Sweep for SI Engine with Late IVC....................................98
Figure 5.7
Non-robust CAI to SI M ode Transition.................................................99
Figure 5.8
Successful CA I to SI Transition........................................................101
.................
96
11
Figure 5.9
Details of CAI to SI M ode Transition........................
Figure 5.10
Ricardo Wave Simulation for CAI-SI Mode Transition .................
102
Figure 5.11
Failed SI-CAI Mode Transition... ..............................
103
Figure 5.12
Pressure Trace of Failed SI-CAI transition............................................104
Figure 5.13
Gimep and Nimep of Failed SI-CAI transition.......................
Figure 5.14
D etails of Failed Transition.............................................................106
Figure 5.15
N on-robust SI-CA I Transition..........................................................107
Figure 5.16
Details of non-robust SI-CAI transition................................................107
Figure 5.17
Wave and Kinetic Simulation Results for SI-CAI
................... 102
106
Mode T ransition ...........................................................................
108
Figure 5.18
Sensitivity of Nimep History to IVC Programming.................................112
Figure 5.19
An Example of Implementation of Mode Transition Strategy......
Figure 5.20
Robust SI-CAI Transition.....................................
Figure 5.21
Details of robust transition with slight pressure oscillations....... ..................... 114
Figure 5.22
An example of robust SI-CAI transition.............................
Figure 5.23
Details of robust transition without pressure oscillations..........................116
Figure 5.24
Wave Sinmlation Results for SI-CA mode transition ..................
Figure 5.25
Low Speed Low Load, SI-CAI Transition.............................................117
Figure 5.26
Low Speed High Load, SI-CAI Transition............................................118
Figure 5.27
High Speed Low Load, SI-CAI Transition..................
Figure 5.28
High Speed High Load, SI-CAI Transition.......
...... 113
114
.......................
..................
115
116
118
119
12
Chapter 1: Introduction
1.1
Background
1.1.1
HCCI combustion literature review
Conventional internal combustion engine can be classified into spark ignition (SI) and
diesel engine. In a Port-Fuel-Injection (PFI) spark ignition engine, the combustion process can be
described as follow: fuel is injected into the intake port; fuel evaporates and mixes with fresh air
in the intake port; fuel and air mixture is drawn into the combustion chamber by the downward
piston movement when the intake valve is open; fuel and air continue to mix as the charge is
compressed by the upward piston movement, the charge is ignited by a spark discharge which
creates a flame kernel near the spark plug region, the flame propagates across the mixture of air,
fuel, and residual gas in the cylinder until it extinguishes at the combustion chamber wall. The
load in SI engine is controlled by adjusting the airflow with a throttle in the intake system while
maintaining stoichiometric fuel and air equivalence ratio. On the other hand, in a diesel engine,
only air is induced into the cylinder during the intake stroke. Liquid fuel is injected directly into
the cylinder near piston top dead center (about 30 BTDC) where it atomizes into drops,
evaporates to form fuel vapor, and mixes with air to form a combustible mixture and autoignites. Therefore in diesel engine, the start of combustion is controlled by the fuel injection
event. The combustion takes place in a locally fuel rich (globally lean) environment leading to
the formation of soot. The combustion process in a diesel engine can be divided into two distinct
processes. The first one is the auto-ignition of fuel and air which mixes during the ignition delay
period. This stage is marked by rapid heat release in the pressure trace. The second stage is the
main combustion process where the combustion is controlled by turbulent mixing of fuel and air.
The heat release rate is diffusion controlled and is slower than the first phase of combustion.
High NOx level is produced due to the high peak cylinder gas temperature. The load in diesel
engine is controlled by the fuel injection amount, which is limited to fuel air equivalence ratio of
0.75 because of excess smoke. In addition to these two modes of combustion, recently there is a
relatively new type of combustion, which has attracted worldwide attention. This combustion
method is generally called Homogeneous Charge Compression Ignition (HCCI). A simple
illustration of these three different types of combustion mode is attached in Figure 1.1
13
Diesel:
Spark Ignition:
HCCI:
Diffusion Combustion Propagating Flame
Spontaneous Burn
Flam
Fuel Plume and
Diffuion
Bumed G s
Unbumed G s
Aust Keep Flame Temp> 1900K
Multiple Ignition and
Reaction Sites
Figure 1.1 Illustration of Diesel, SI, and HCCI Combustion
An internal combustion engine is said to operate in Homogeneous Charge Compression
Ignition (HCCI) mode if autoignition takes place in a homogeneous, lean or diluted fuel air
mixture, which is ignited by compression. HCCI is receiving increased attention due to its
potential to reduce both the fuel consumption and emission level of an internal combustion (IC)
engine. This type of combustion has been referred to by numerous names within the literature.
Active Thermo
Atmospheric
Combustion (ATAC)
[1],
Toyota-Soken Combustion
[2],
Compression Ignition Homogeneous Charge (CIHC) [3], Premixed Charge Compression Ignition
(PCCI) [4], Homogeneous Charge Compression Ignition (HCCI) [5], Premixed Lean Diesel
Combustion (PREDIC) [6], Controlled Auto Ignition (CAI) [7], and Optimized Kinetic Process
(OKP) [8]. To a certain degree, each of these different engine design names has its own unique
characteristics. While these names are often interchangeable, recent development in the literature
tends to use the term HCCI to generally refer to the combustion method where a relatively
homogeneous mixture is auto-ignited through compression.
The first application of HCCI combustion technology was demonstrated by Onishi et al.
[1] in a two-stroke gasoline engine. They observed an increase in fuel economy, very low NOx
level, and low cyclic variability when the engine was operated in HCCI mode. Building on the
previous work in two-stroke engine, Najt and Foster (1983) extended the work to four-stroke
engines and attempted to gain additional understanding of the underlying physics of HCCI
combustion. They conducted experiments using PRF fuels and intake preheating. By means of
heat release analysis and cycle simulation, they pointed out that HCCI combustion process was
14
governed by low temperature (<950 K) hydrocarbon oxidation kinetics. As such, they concluded
that HCCI combustion is a chemical kinetic combustion process controlled by the temperature,
pressure, and composition of the in-cylinder charge. Thring (1989) further extended the work of
Najt and Foster in four-stroke engines by examining the performance of an HCCI engine
operated with a full-blended gasoline. The operating regime of a single-cylinder engine was
mapped out as a function of air fuel equivalence ratio, EGR rate, and compression ratio. Lavy et
al [7], investigated the effect of both internal and external residual gas on HCCI combustion in
the collaborative 4-SPACE project. They found that the occurrence of CAI is dominated by the
degree of mixing between fresh charge and exhaust gas residuals. Four stroke concepts were
developed to mimic the internal fluid dynamic and mixing effects of the 2-stroke when running
in CAI mode, resulting in the first 4-stroke engine capable of achieving CAI over a limited load
and speed range without the use of external charge heating or high compression ratio. This was
achieved by internally recycling burnt gases using modified camshafts. Law et al [9] presented 2
methods to control the EGR rate by using an electro-hydraulic valve actuation system. The first
was similar to Lavy et.al. [7], and the second involved the storage of exhaust gas in the exhaust
manifold prior to readmission to the combustion chamber during the intake stroke. Koopmans
et.al. postulated that SI to HCCI transition can be achieved in a single cycle if the valve timing
and fuel injection amount can also be adjusted within one engine cycle. However, in their
demonstration the first 4 HCCI cycles resulted in lower Nimep compared to the steady state
value due to advanced ignition timing in those particular cycles.
Currently [2004] there are two commercial engines that partially run on HCCI mode over
their duty cycles. Nissan worked on diesel HCCI combustion technology, which is called
Modulated Kinetics system. At low load, the engine operates with high swirl, high EGR, and
retarded injection timing so that near homogeneous combustion occurs. At high load the engine
is switched back into regular diesel operation. This MK engine is commercially available in
Japan in 2.5-liter unit and 3.0 liter unit. In the gasoline side, Honda has developed Active Radical
motorcycle (2-stroke) engine. It operates as a spark ignition engine for cold start, at very low and
high load. The engine has a compression ratio of 6.1. It is operated in HCCI mode by throttling
the exhaust gas to trap high fraction of residual gas [14]. The mode transition is made in a predetermined transition region where both HCCI and SI combustion modes are feasible.
15
1.1.2
Controlled Auto Ignition (CAI)
In their experiments, Law et al [9] showed that auto-ignition timing could be controlled
by changing the residual gas amount using a VVT system. Since this method relies on trapping a
large amount of residual (up to 70% by volume), there is a question whether the in-cylinder
mixture is truly homogeneous. Oakley et al [10] argued that the degree of mixing between the
residual gases with the fresh charge might be used to control combustion timing and duration.
They also pointed out that the presence of temperature and residual stratification is what
distinguishes stratified CAI combustion from other HCCI engine designs. Therefore, this
combustion mode is simply referred to as CAI combustion throughout this work.
Since CAI combustion relies on controlling the residual gas amount to control the
ignition timing, the practical implementation of this technology requires the use of a Variable
Valve Timing (VVT). Closing the exhaust valve before piston's top dead center (TDC) leads to
residual trapping (internal EGR) method, while closing the exhaust valve after TDC results in
exhaust reinduction method. The latter method produces colder residual gas due to heat transfer
to the exhaust port. Therefore this work focuses on the residual trapping (internal EGR) method
to avoid excessive inlet preheating required to achieve CAI combustion.
There are several challenges that need to be overcome before a practical CAI engine can
be designed. Reported CAI operating regimes are small compared to that of spark ignition
engine. The high load operating point is limited by the excessive pressure oscillations similar to
knocking pressure trace in spark ignition engine. Low load operating point is limited by cycle-tocycle variation caused by the high dilution level, which creates excessive thermal and
concentration stratifications inside the cylinder. One feasible solution to this limitation is a dual
mode SI/CAI engine where the engine is only operated in CAI mode when possible. The
consideration and constraints of such an engine will be discussed in more detail in Chapter 3.
16
1.2
Objectives
While an improvement of in fuel economy and reduction of NOx emissions level has
been widely reported in the literature [8,9,10,11], only few publications addressing the problems
encountered during the practical implementation of CAI combustion technology are available.
This work seeks to develop an improved understanding of the phenomena encountered during
both steady state and transient operation of a dual mode SI/CAI engine.
This study can be divided into two topics with each topics consisting of experimental and
simulation work. First, the phenomena encountered during steady state and transient operation
within CAI regime are reported. Within this section, some steady state phenomena are analyzed
more thoroughly than previously reported in the literature. The engine transient response to a
step change in valve timings and fuel metering was documented. These transient experiments are
important to identify the main physical constraints of engine behaviors during load change. A
very simple load controller based on valve timing interpolation and fuel metering was tested to
characterize the engine transient behaviors. Simulations to estimate the in-cylinder conditions
were conducted using a heat release model and a Wave simulation program. In the second part of
the thesis, experiments and simulations were conducted to illustrate the phenomena encountered
during mode transitions both between SI to CAI mode and vice versa. After the constraints have
been experimentally identified, a strategy is devised and demonstrated to achieve fast and
smooth transition between the two engine modes.
1.3
Thesis Framework
This thesis is presented with the following framework. Chapter 1: INTRODUCTION, gives a
historical overview of the development of CAI combustion and explains the motivation of
conducting this study. Chapter 2: EXPERIMENTAL APPARATUS & MODELING TOOLS,
describes the experimental hardware and simulation programs used in this study. Chapter 3:
CONSIDERATIONS AND CONSTRAINTS FOR A SI/CAI ENGINE assesses some practical
issues in the implementation of a dual mode SI/CAI engine. In this chapter the issues for
practical implementation of a dual mode SI/CAI engine is discussed carefully and the objectives
of this study is defined more systematically. Chapter 4: STEADY STATE and TRANSIENT
17
OPERATION WITHIN CAI REGIME, illustrates a few selected steady state CAI combustion
phenomena encountered for this particular engine set-up. The results of some transient
experiments and simulations are also presented to characterize the engine behavior during
transient operations. Chapter 5: MODE TRANSITION MANAGEMENT BETWEEN SI AND
CAI MODE, describes the issues encountered during modes transition. A strategy to control the
mode transition based on late intake valve closing and fuel metering is proposed. An engine
controller based on this strategy was then designed and implemented to demonstrate fast and
smooth SI/CAI transition. Chapter 6: SUMMARY AND CONCLUSION, summarizes the
important findings from this study.
18
Chapter 2: Experimental Apparatus and Modeling Tools
2.1
Experimental Apparatus and Procedures
2.1.1
Engine and Dynamometer
The engine set-up used in this study consists of a Ricardo Hydra crankcase, a Ross
Racing piston, custom made engine block and cylinder liner, and a modified Volkswagen TDI
1.9 L engine head. The Ricardo crankcase was installed before the start of this study. The
estimated time required to replace the crankcase and driveline would have imposed an
unacceptable down time and hence, it was decided to keep the original crankcase. The original
Ricardo Hydra MK IV engine head exhibited a lot of carbon deposits and was deemed unsuitable
to be used for this study. From adaptability to the crankcase point of view, an engine head with a
bore size of 80 mm was preferable. The nature of CAI combustion, which sometimes indicates
violent pressure oscillations, makes a diesel engine head more attractive compared to an SI
engine head due to its structural strength. The mass production VW TDI 1.9 L engine head
satisfies these two requirements. Some modifications were then made to accommodate a single
cylinder dual mode SI/CAI engine. The fuel injector hole was modified to fit a small NGK
R847-1 1 spark plug. Port fuel injection system was selected to ensure better mixing between fuel
and air in the intake port. Labview RT PXI system was used to control the spark and fuel
injection timing. The glow plug hole was modified to fit a Kistler 6125A pressure sensor. A
flame arrestor based on design by General Motors was used to minimize the thermal shock and
noise in the pressure signal. Aluminum engine block and cast iron cylinder liner were made to
adapt the new engine head to the Ricardo crankcase. The cylinder liner is surrounded by water
jacket to prevent overheating of the combustion chamber. By design, the geometric compression
ratio of this engine can easily be changed by inserting/removing some spacers (precision
washers) between the engine block and the crankcase. A custom order aluminum flattop Ross
Racing piston was used to match the cylinder liner. The piston is cooled with a single oil jet
sprayed by a piston cooling nozzle attached to the crankcase. A summary of the engine geometry
is attached in Table 2.1.
19
Bore
80.26 mm
Stroke
88.90 mm
Connecting Rod Length
158 mm
Displacement Volume
449.77 cc
Fuel Injection
Intake Port
Number of valve
1 intake, 1 exhaust
Geometric Compression Ratio
12.28 (manually adjustable, maximum 16)
Number of cylinder
I
Throttle Position
Wide Open
Table 2.1 Summary of Engine Geometry
A water-cooled Dynamatic Dual Mode Dynamometer with a Digalog controller was used
to control the engine speed. The dynamometer could be operated in motoring, absorbing, or duo
mode. Normally the dynamometer was operated in motoring mode until the desired speed was
obtained. The operating mode was then changed into the duo mode before combustion was
initiated. The duo mode is preferable to the absorbing mode in the transient experiments to keep
the engine speed roughly constant even when there are some missed engine cycles.
2.1.2
Fuel Injection System
The fuel injection system consists of fuel tank, fuel pump, accumulator, fuel pressure
regulator, and fuel injector. This study used a fully blend gasoline UTG-91 from Chevron Philips
to represent commercial low-grade gasoline fuel. The average octane rating of the fuel is 87.4.
Details of the fuel properties can be found in Table 2.2. The differential fuel pressure regulator
maintained a constant pressure difference (38 psi) across the fuel line and the intake manifold
during experiments. This constant pressure difference was necessary to track cycle-to-cycle fuel
injection amount. The calibration curve for the injector is attached in Figure 2.3. The Bosch fuel
injector was aligned to spray the fuel toward the back of the intake valve. The solenoid injector is
actuated by a TTL pulse signal generated by Labview Real Time engine controller. The fuel is
injected to the intake port at the expansion stroke (closed valve injection).
20
Specific Gravity, 60/60
0.7407
API Gravity
59.54
Phosphorous, g/gl
<0.0011
Sulfur, ppm
40.5
Corrosion 3 hrs@50 C
IA
Manganese, g/l
<0.0011
Hydrogen, wt%
13.41
Carbon, wt%
86.36
Net Heat of Combustion, btu/lb
19179
Reid Vapor Pressure
8.98
Lead (ml/gal)
< 0.0008
Benzene Content, lv%
0.93
Hydrocarbon Type, vol%
Aromatics
27.8
Olefins
7.5
Saturates
64.7
Research Octane Number.
91.3
Motor Octane Number
83.5
Antiknock Index
87.4
Existent Gums (mg/100 ml) (washed)
0.2
Table 2.2 Certificate of Analysis for UTG-91
21
50
45
40
35
30
25
y=2.2783x - 0.5732
oo'
R 2 = 0.9999
20
LL.
15
10
5
0
0
5
10
15
20
25
Fuel Pulse Width (ms)
Figure 2.1 Fuel Injector Calibration
2.1.3
Variable Valve Timing System
Most of the hardware for the electromechanical VVT system used in this study had been
donated by Chrysler Motor Corporation. The hardware consists of a valve actuator, Sorensen
DCR 110-90T power supply, and Advanced Motion Controls Brush Type PWM Servo
Amplifiers (50A20T-AUI). The Sorenson power supply requires a 240 VAC input and generates
110 VDC for the servo amplifiers. The servo amplifiers then sends current with a maximum
magnitude of 50A to the valve actuator. The valve actuator houses 4 electromagnets and 2
armatures. Two electromagnets were required for each armature; one to catch it at the bottom
position (valve open position) and one to catch it at the top position (valve close position). The
current from the servo amplifiers energizes the electromagnet and creates a magnetic field, which
attracts the armature up or down. Figure 2.2 illustrates the valve actuator (VVT pod).
22
Sp
Spring Force
Adjusting Screw
Spring
Titanium Spring
- --
Retainer
Top Electromagnet
(Valve Closing)
4-
Armature
Bottom Electromagnet
(Valve Closing)
Aluminum Adapting Plate
(VVT Pod to Cylinder Head)
Armature Extension Foot
Original VW TDI Tappet
Figure 2.2 Cross Section of VVT Pod
The servo amplifiers generate currents depending on the magnitude and duration of an
analog voltage input signal. Labview Real Time 7.0 systems were used to produce this input
voltage signal. Since magnetic force, and hence the current, that is required to hold the armature
in place is much smaller than the current required to move and catch the armature, the input
signal to the servo amplifiers needs to have two distinct amplitudes. Unfortunately, early
development with the controller suggested that the controller was not fast enough to generate an
analog output consistently at the required frequency. However it was fast enough to generate
digital signals. The digital signals were then combined together using summing operation
amplifier to provide four analog signals for the four electromagnets. The magnitude of the input
signal to the servo amplifier can be adjusted using variable resistors installed in the voltage
summing board.
23
General VVT systems operation can be described as follows. When the engine is at rest
the two valves (intake and exhaust) are partially open. In this condition, the position of the
armature is governed by force balance between the valve open spring and the valve return spring.
With the engine still at rest, the top magnets are energized to move the armature to the upper
most position. In this position, the electromagnetic force that is exerted by the top magnet is
larger than the force exerted by the valve open spring. The engine is then motored with the
dynamometer with both valves still in the closed position until the desired engine speed is
achieved. Labview then commands the servo amplifier to de-energize the top magnet at the
desired crank angle and energize the bottom magnet to catch the armature at the bottom (valve
open position). The process is then repeated to close the valve at the desired timing.
Two water-cooled plates were installed to keep the VVT pod from overheating. Since there is no
provision for soft landing in this system, the armature arrives at the electromagnet with high
velocity. The high kinetic energy is then converted into thermal energy and heats up the VVT
pod. This fact could create problems with the VVT performance. The cooling plates eliminate
this problem by maintaining the VVT pod temperature below 65 'C.
2.1.4
Intake Air Temperature Controller
A 2 kW Omega AHF-14240 air heater was installed downstream the intake port. The
heater was used to maintain the air temperature to 120 'C during this study. The heating helped
to widen the operating range of the CAI engine without excessive external preheating. In real
practice, this amount of heating can easily be supplied from the exhaust gas thermal energy. A
type K thermocouple is installed about 7 cm from the intake valve. The output of the
thermocouple is fed to an omega I-series temperature controller. The controller outputs a 5 VDC
signals when the thermocouple reading is less than the set point. This signal is then sent to
actuate a normally open solid-state relay, which power the air heater until the desired
temperature is reached.
24
2.1.5
Pressure sensors
Initial testing was conducted using Optrand glow plug pressure sensors. This first
selection was made due to its economy and adaptability to the existing glow plug hole. However,
after successive failures due to the rapid pressure and temperature increase, Kistler piezoelectric
6125A pressure sensor was installed in the modified glow plug hole. A flame arrester, designed
by engineers in General Motors Company was used to alleviate the harsh thermal environment
during rapid pressure increase. The signal from the pressure sensor was amplified using a Kistler
Type 5010 charge amplifier. The output of the charge amplifier is then fed to the data acquisition
system.
The piezoelectric pressure sensor is sensitive to long term drift. Therefore an Omega
PX176-025A5V absolute pressure transducer was installed in the intake manifold to measure the
absolute pressure during the intake valve open period. This transducer has a measurement range
of 0-25 psia. Since the piston movement rate and therefore the bulk gas movement is minimum
at piston bottom dead center (BDC), the in-cylinder pressure at BDC was assumed to be equal to
the corresponding manifold pressure at that particular time. With this assumption the actual incylinder pressure can be estimated.
There are some noises that are introduced into the pressure signal. Because no
optimization was done to ensure soft landing of the armature for this specific VVT system, the
armature hits the bottom and top magnet with high velocity. This event leads to vibration of the
engine structure and is picked by the pressure transducer as spikes in the pressure signal. We
used this phenomenon to define the intake and exhaust valve closing time. This definition of
valve timing slightly differs from the conventional valve timing definition, which defines a valve
opening/closing time as the timing when the valve lift is at 0.15 mm. The valve open timing is
more difficult to define due to its dependence on in-cylinder pressure condition which changes
from one operating condition to another. While the pressure sensor did not capture any noise at
valve opening event, it did capture the event when the armature hit the bottom magnet (valve at
fully open position). Bench testing with a Kaman position sensor indicated that the duration
required to fully open the valve is about 5.2 ms. Therefore the valve opening time for this study
is defined as the time the armature hit the bottom magnet minus 5.2 ms.
25
2.1.6
Lambda Sensor
A Bosch LSU broadband lambda sensor was installed about 23 cm from the exhaust
valve to measure the air fuel equivalence ratio. It is important to install the lambda sensor close
to the exhaust valve to track the fuel air ratio during engine transient. While the actual air fuel
ratio during transient operation cannot be precisely determined with this system, the system is
good enough to distinguish between rich or lean operation even during transient operation. An
ETAS AWS2 signal-conditioning unit was used to evaluate the lambda sensor signal. The signalconditioning unit has 2 output terminals, which linearly correspond to the pump current and the
internal resistance of the lambda sensor. Lambda sensor internal resistance can be used to
calculate the current sensor temperature but was not used in this study. Lambda sensor pump
current can be used to determine the fuel air equivalence ratio value of the exhaust gas with the
help of a suitable calibration curve. The system takes about 3 minutes to heat up the sensor to its
operating temperature. Once it is fully warmed, the system can detect a change in lambda within
250 ms. Transient experiments indicated that the sensor can capture the change in air fuel ratio
within 2-3 engine cycles at 1500 rpm.
2.1.7
Data Acquisition System
The data acquisition hardware consists of a National Instrument BNC2090 adapter,
National Instrument PCI 6025E multifunction board, and a Dell Pentium III computer. This data
acquisition system can record up to 8 analog input channels when operated in differential mode.
Due to this limitation of number of channels, .some operation amplifier based summing
circuitries were built to superimpose two or three signals into a single channel.
A data acquisition program was created using National Instrument Labview 7.0 software.
The program allows for simultaneous monitoring, processing, and saving various data from the
engine. A TDC signal was used to trigger the data acquisition system. BEI encoder with I degree
resolution was installed to the crankshaft. The program took the crank angle signal to operate in
external clock mode. This configuration made it easy to monitor and process the data, since the
first data point was always at TDC and each data point was separated by one crank angle degree
interval. Several Labview subroutines (virtual instrument) were built to calculate and display the
26
IMVEP value and valve timing information on the fly. The actual valve timing information was
important to compensate for day-to-day variation of the VVT performance so that experiments
could be repeated easily. The real time IMEP display reduced the number of experiments
required for finding MBT timing in SI mode and constant load mode transition experiments. The
program was also designed to save data without interrupting the update of cycle-to-cycle IMEP
value. This helped to judge whether the saved data file was useful or not. The following inputs
were normally recorded by the data acquisition system: cylinder pressure, manifold pressure,
commanded voltages for valve timings, fuel injection pulse, spark charging/discharging pulse,
engine speed, torque output from the dynamometer, and exhaust lambda sensor. Several Matlab
subroutines were then created to load, process, and plot the desired data.
2.2
Modeling Tools
Simulations programs have been created and used to interpret the engine behaviors
during the experimental study. As such the models are not designed to predict the engine
behaviors but to assist the understanding of the phenomena encountered during the experimental
study.
2.2.1
Ricardo wave 1-D code
Ricardo wave is a 1-D gas dynamics simulation program that is capable of predicting the
mass and energy fluxes during the engine cycle. A simple Ricardo engine model was built to
represent the experimental set-up in this study. The user inputs the engine geometry, valve
timings, fuel injection amount/exhaust fuel air equivalence ratio, and the combustion
characteristics to the model. The valve timing and fuel injection amount were recorded by the
data acquisition system and could be directly used in Wave simulations. Since Wave is
essentially a fluid dynamics program, the user needs to define an appropriate combustion model
to be used in the simulation. A wiebe function based combustion model was used to represent the
combustion process in the cylinder due to its simplicity and accuracy in simulating the engine
behaviors during transient operations. The parameters for the wiebe functions were gained
through conducting the heat release analysis.
27
2.2.2
Kinetics Model
A singlezone model was used to calculate the ignition timing of the charge based on the
initial conditions from Wave. The model assumes a single reaction zone with a uniform
temperature, concentration, and pressure within boundary. There is no mass flux across the
boundary. The energy interactions between the zone and the environment is modeled through
work and heat transfer to the combustion chamber wall. The heat transfer model used is based on
modified Woshni correlations.
Chemical reactions mechanism used is based on the work of Lawrence Livermore
National Laboratory. The kinetics mechanism includes 1034 species and 4238 elementary
reactions. The gasoline fuel is modeled as a mixture of iso-octane and normal heptane with an
octane rating of 90. Standard solvers of differential algebraic equation (DASSL, VODE, etc ) can
take up to 2 hours for calculating a single engine cycle (from intake valve closing to exhaust
valve opening) and therefore is not feasible for this study. A much faster solver, DAEPACK, was
implemented through collaboration with MIT Chemical Engineering department.
2.2.3
Heat Release Code
An equivalent of two-zone heat release model was built to estimate the in-cylinder
residual gas conditions. The model assumes there are burned and unburned mixture. The
unburned mixture is compressed and expanded adiabatically by the piston and the burned gas.
The burned mixture is always at chemical equilibrium. The heat loss to the cylinder walls is
based on the modified Woshni correlations.
The inputs to the model are the in-cylinder pressure data, species mol fractions and
trapped charge mass at intake valve closing. The program outputs the burned gas mass fraction,
the sensible energy, and the cumulative heat loss of the charge. The details of the heat release
program can be found in the appendix section.
28
Chapter 3: Considerations and Constraints
For a SI/CAI engine
3.1
Introduction
There are several challenges that need to be overcome before a practical CAI engine can
be designed. Ignition timing and heat release rate controls, cold start management, and limited
range of operation are among the most important issues that need to be addressed. Unlike SI and
diesel engine where spark and fuel injection trigger the combustion event, CAI combustion has
no triggering device. Once the intake valve is closed, the combustion is governed by mixing and
chemical reactions of the trapped charge. Therefore, efforts to control the ignition timing and
heat release rate must be done prior to intake valve closing by controlling the amount of dilution,
temperature, and fuel air ratio for that particular cycle. It is very difficult to control the CAI
combustion during engine start-up due to uncertainty in fuel delivery and heat transfer during
engine warm-up period. Recent publications have indicated that a ten-degree change in the
coolant temperature has the same order of magnitude effect on combustion timing of a CAI
engine as a ten-degree change in inlet air temperature. A lot of efforts have been done to increase
the operating regime of CAI engine. Yang, et.al [8] presented one of the widest operating regrme
of HCCI engine by heating and boosting the intake air. However, the high compression ratio (15)
that they used for their experiments makes it very difficult to operate in SI mode when high
engine load is demanded. All the published results on CAI operating regime indicate that the
operating range of CAI regime is not large enough to meet the requirement of a commercial
vehicle.
A dual mode SI/CAI eliminates the cold start management and limited operating range
problems by only operating in CAI mode when feasible. Such an engine will start in SI mode
until the coolant temperature reaches steady state and the engine enters the CAI operating
regime. It should also be noted that SI operating regime encompasses the CAI operating regime
and therefore the engine can always be operated in SI mode. There are many control parameters
that have been studied to control the ignition timing and heat release rate in CAI combustion.
This study uses a variable valve timing system to control the start of combustion and heat release
rate.
29
3.2
Constraints within CAI regime
With a variable valve timing system, CAI combustion can be achieved by closing the
exhaust valve early and therefore trapping hot residual gas inside the cylinder. Another method
of initiating CAI combustion is by the exhaust gas reinduction. However preliminary simulation
results indicated that the later method leads to colder residual gas due to heat loss to the exhaust
port. Therefore, the residual trapping method was selected in this study. The physical variables
that affect combustion in CAI mode are listed in Table 3.1.
No
Physical Variables
I
Air Temperature
2
Wall Temperature
3
Residual Temperature
4
Air Mass
5
Fuel Mass
6
Residual Mass
7
Type of Fuel
8
Effective Compression Ratio
9
Combustion Phasing
10
Engine Speed
Table 3.1
Physical Variables of CAI Combustion
Note that not all of the physical variables are independent. For example, residual gas temperature
is affected by the air mass, fuel mass, and residual mass. The air temperature is maintained at
120 'C throughout the experiments by an air heater. The wall temperature was kept
approximately constant by flowing coolant water through the engine head and cylinder liner. The
coolant temperature was kept at 95 'C by a coolant heater. There is no direct control over the
residual gas temperature in this study. The residual gas temperature depends on the combustion
process of the previous cycle. The mass of air is mainly controlled by cylinder volume at intake
30
valve closing. Late intake valve closing reduces the cylinder volume and therefore reduce the
mass of fresh air. Late intake valve closing also results in lower effective compression ratio.
Originally an electronic throttle plate was installed in the intake manifold to control the mass of
air. However, using a throttle increases the pumping loss and creates additional complication
during SI to CAI mode transition. Therefore the main experiments in this study were conducted
without a throttle plate. The fuel mass is controlled by the fuel pulse width sent to the fuel
injector. During steady state experiments, the air fuel equivalence ratio could be controlled by
changing the fuel pulse width. With the residual trapping method, the amount of residual gas is
mainly controlled by exhaust valve closing time. Because of the short valve opening duration, in
reality the residual gas mass is also affected by the exhaust opening time through the exhaust
pressure dynamics. The fuel used in this experiment is fully blended UTG-91 gasoline from
Chevron Philips. The effective compression ratio is controlled by the intake valve closing and to
some degree the intake valve opening through the pressure dynamics in the intake manifold.
No
Control Variables
Influence
(Independent)
I
IVO
Pumping loss and residual gas temperature through backflow to the
intake port. Set to be symmetric w'ith EVC
2
IVC
Total trapped charge mass and effective compression ratio. Very
late in SI mode to avoid knock, variable in CAI mode
3
EVO
Gimep and trapped residual mass through exhaust pressure
dynamics
4
EVC
Major control parameter for residual gas mass, secondary effect on
residual gas temperature.
5
Injection Duration
Directly control fuel injection amount to the intake port. Not all of
the fuel injected is delivered into the cylinder in a single cycle.
6
Air heater
Intake air temperature. Set at 120 C
7
Geometric
Set the maximum effective compression ratio. Set to 12.28. A
Compression Ratio
compromise between performance in SI mode and CAI mode.
Spark Timing
Control the combustion in spark ignition mode
8
Table 3.2 Engine Control Variables
31
The engine speed in these experiments is controlled with a dynamometer. Most experiments in
Chapter 4 are conducted at 1500 rpm while the speeds were varied between 1250 and 1750 in
Chapter 5. A summary of the engine control variables and their effects on the physical variables
are shown in Table 3.2.
Figure 3.1 and Figure 3.2 illustrate the feasibility of CAI combustion during an FTP test
and an EPA highway test. The CAI operating regime was taken directly from SAE Technical
Paper# 2002-01-0420 [11]. It is clear from both figures that it is not possible to solely drive in
CAI mode during the two tests. However, there is a significant difference in the utilization of
CAI technology between the highway test and urban test. During the highway testing, about 75%
of the time, the engine can be operated in CAI mode while in the FTP test, which simulates an
urban driving pattern in California, only 40% of the time, can the engine be operated in CAI
mode.
30
-
-HCCI
25E
e
15-
10
>
0
0
200
400
600
800
1000
Time (s)
Figure 3.1 EPA Highway Fuel Economy Drive
32
30 --
.0 25
E
d
20
S15
- HCCI Mode
:SI Mode
0
25
I
0
200
600
400
Time (s)
!
%1
I
.
800
1000
Figure 3.2 FTP test for a Ford Taurus 3.0 L Engine
In order to maximize the benefit of operating in CAI mode, the transition between SI and
CAI mode and the transient between one CAI point to another CAI point should be made as fast
as possible. Controlling the engine torque in CAI combustion is very difficult due to cycle-tocycle coupling in CAI engine through heat transfer and residual gas. The coupling affects
combustion, and thus next cycle boundary condition through charge thermal environment and
charge composition. The torque output is governed by charge composition, combustion phasing,
and pumping loss.
Varying the air fuel ratio is one method to control the engine load while operating in CAI
mode. Running the engine in lean conditions leads to extension of the operating regime,
especially in the low load regime. However, running the engine lean will use up the oxygen
storage capacity in a conventional three way catalyst, which leads to NOx breakthrough when
the engine is switched back into SI mode.
33
Chapter 4 is focused on understanding and identifying the phenomena encountered
during load transient operation within CAI regime. Small and large step changes in load were
conducted to characterize the response of a CAI engine and evaluate the underlying physical
phenomena that govern the system response. A simple open loop controller was designed to
investigate the order of magnitude of the time scale required to make a large load change. No
control optimization was conducted in this study. It is not the intention of this study to enlarge
the operating range of CAI combustion or to find the optimized combustion phasing since those
two topics have been explored many times in the literature. However, several interesting
phenomena in steady state CAI combustion were analyzed and presented in the study to provide
a better understanding of CAI combustion.
Importance of Mode Transition Management
3.3
Figure 3.3 illustrates, the load and speed trajectory of a Ford Taurus 3.0 L engine during
a standard FTP test. The data points were taken at 1 s interval. At 1500 rpm, 1 s corresponds to
12.5 cycles per cylinder. The CAI operating range is taken directly from SAE Technical Paper#
9
8
7
6
-#
5
I:
SA-E 2002-01-0420
4
3
2
1
0
-1
S
560
1C0
,*
1500
2000
RPM
2 00
3000
3 E 00
-2
Figure 3.3 CAI operating regime in FTP test
34
2002-01-0420. Out of 1098 data points, 423 of them are within CAI mode operating regime. From
these 423 data points, approximately 220 data points involve a mode transition either from an SI
operating in the previous data point or to an SI operating point in the next data point. Thus mode
transitions are often, occurring every few to every tens of seconds. This result highlights the
importance of mode transition management for a dual mode SI/CAI engine. It can also be inferred
from Figure 3.3 that the number of transition depends critically on the CAI boundary location. An
engine speed of 1250 and 1750 were then selected to represent a low and high engine speed condition
where mode transition occurs.
The details of the operating trajectory of test cycle number 5 (last cycle of Bag 1) of the FTP
drive test are shown in fig. 3.4. The engine starts at idle from point (a) where the engine speed is at
approximately 700 rpm and vehicle speed at 0 mph. The vehicle was then accelerated
60
50
S40
30
20
10
0 0
600
400
200
100
1400
1200
1000
Time (s)
40
24
k
h
h
I
'nn
- - 35
0
19
a.
14
-Gear
-*-- bm e p (b a r)
E
9
-
M
c
-$____
Cu
U)
30
q
C)
4
2CL
-25
a-
-20
-0
s
>
-15
RPM/1 00
t
Av _Velocity
a)
0D
10
0
-1
u
4
440
a
450
460
470
480
490
500
V
- 0
510
Time (s)
Figure 3.4 Details of FTP Test
35
until it reaches 35 mph in about 15 seconds. During this acceleration period the gear was also changed
from the first to the third. Once the vehicle reaches its cruising speed, the required torque output from
the engine drops significantly compared to the acceleration period. During the deceleration period
(from point n), the engine torque output was quickly reduced and reaches negative value for a few
seconds and the engine speed is also reduced from 1500 to 700 rpm. In figure 3.5, the engine torque
output during cycle 5 (last cycle of Bag 1) is plotted against the engine speed. The CAI operating
regime was also superimposed on the graph to illustrate the details during transient operation. Starting
from idle at point (a), the trajectory only stayed in CAI domain for one second or so during
acceleration. It is also perceivable that when the engine was operated at cruising speed, a lot of mode
transition has to be made depending on the lower boundary of CAI regime. It can also be argued that
because it takes a few cycles for the CAI mode to stabilize, a brief mode transition is not desirable.
Furthermore, the trapped residual gas energy from a very lightly-loaded SI cycle may not be sufficient
to initiate CAI. Therefore, an appropriate algorithm is needed for the transition decision.
8
7
+bm ep(bar)
d
6
5
I-
0
4
3
E
In 2
-
SAE 2002-01-0420
n $ k
1
0
-1 0
so0
10
15 00
20 00
2500
-2
Speed (rpm)
Figure 3.5 Residence Time in CAI Regime
36
There are two types of mode transitions; the first one is from SI to CAI and the second one is
from CAI to SI. The transition from CAI to SI mode can be made relatively easily due to weaker
cycle-to-cycle coupling within SI mode. Maintaining the combustion process during mode transition
from CAI mode to SI mode is not very difficult. However, commanding a certain load or even
maintaining a constant load during a mode change from CAI to SI is challenging. There are a lot of
uncertainties involved during mode transition, especially in the airflow and residual mass fraction, that
extensive mapping is necessary to control the torque during mode transition. Mode transition from SI
to CAI mode is generally considered more difficult to make to due to significant cycle-to-cycle
coupling of CAI combustion. Without a transition strategy, a robust static CAI operating point does
not imply a robust SI to CAI transition at that point. The first CAI cycle is usually very different from
the rest of the cycle due to different residual gas mass fraction and temperature, fresh air mass fraction,
and fuel air equivalence ratio. The residual gas produced from the first CAI cycle has to be able to
maintain CAI combustion in the next cycle and the process needs to be repeatable.
In order to further analyze the characteristic of SI-CAI mode transition, there is a need to
define some desirable characteristics of a robust transition. In this study the following criteria are used
to define a robust transition strategy:
1.
The operating mode can be changed into CAI mode within 1 engine cycle.
In order to maximize the benefit of running in CAI mode, it is desirable to operate in CAI
mode as frequent as possible. Therefore a fast transition strategy needs to be devised. In a
conventional spark ignition engine, a throttle is used to control the engine load. As such when
the engine is operated at idle or low load SI mode, the residual gas temperature may not be hot
enough to initiate CAI combustion in a single cycle. The finite throttle movement rate also
makes it difficult to precisely control the airflow rate for the first CAI cycle. In this study the
engine load in SI condition is controlled by intake valve closing. The author believes that when
an engine is equipped with a fully flexible valve timings system, the engine load will mainly be
controlled with the valve timing even when it is operated in SI mode.
37
2.
No misfires
There should be no misfires during and after the mode transitions. Without a proper mode
transition strategy, it is very likely that the engine will have one or more misfiring cycles
before a stable CAI combustion can be achieved. Therefore a robust transition should always
be able to maintain combustion in CAI mode.
3.
Acceptable NIMEP variations.
The transition should be made smoothly that the passengers do not feel any sudden
acceleration or deceleration during mode transition.
4.
Acceptable pressure oscillations.
An uncontrolled mode transition often results in an early burning cycle for the first CAI cycle,
which exhibits pressure oscillation similar to knocking combustion. The pressure oscillations
can cause undesirable damage to the engine and high pitch noise. Since the transition needs to
be made frequently during regular driving pattern, it is necessary to reduce the occurance of the
pressure oscillations.
Chapter V mainly focuses on the phenomena encountered during mode transition both from SI
to CAI and vice versa. Since no experimental results were available in the literature when the project
started in 2002, some preliminaries experiments were conducted to investigate the constraints and
problems encountered during mode transition. The results were then analyzed using several simulation
programs and MATLAB subroutines. A mode transition strategy was then proposed using the insights
gain from preliminary experiments. The strategy was then implemented in the engine controller and
has been demonstrated to be capable of achieve a fast and robust mode transition. Experimental data
were then analyzed again to further improve our understanding of the phenomena encountered during
mode transition.
38
Chapter 4: Steady State and Transient Operation
Within CAI Regime
4.1
Introduction
The original goal of this section is to identify the physical constraints and phenomena
encountered during load transient within CAI regime. However, the scope of the project was
extended to include some important steady state phenomena within CAI regime. The inclusion of
these steady state phenomena is necessary to help one to understand the results of transient
experiments and simulations. For example, the effect of fuel air equivalence ratio (in steady
state) can also be used to understand the misfiring phenomena during transient experiments.
Simulations using Ricardo Wave cycle simulation program were conducted to interpret the
engine behavior. A very simple open loop controller based on fuel and valve timings
interpolation was built to investigate engine transient behaviors within CAI regime. No
optimization was conducted for the controller in this study.
4.'
Mapping of the CAI regime
There are two schools of thoughts regarding the proper fuel air equivalence ratio that
should be used for steady state mapping of CAI regime. One approach is to maintain
stoichiometric air fuel ratio. The other approach is to run the engine in fuel lean condition.
Running the engine in lean conditions has been shown to significantly improve the thermal
conversion efficiency due to higher specific heat ratio during expansion stroke. However,
running the engine lean depletes the oxygen storage capacity of a conventional 3-way catalyst
(CO, NOx, and HC), which can lead to high NOx emission level when the engine is switched
back into SI mode during high load operation. Running the engine at stoichiometric condition
eliminates this problem but faces penalty in the improvement of thermal conversion efficiency.
When the engine is run at lean conditions, 8% of improvement in fuel consumption can be
achieved compared to regular SI engine, while only 3~4% of improvement can be achieved when
running in stoichiometric condition [8].
Figure 4.1 illustrates the result of CAI operating regime mapping for this particular
experimental set-up. The goal of this mapping exercise is to provide initial and final valve timing
39
and fuel injection settings for the transient experiments. Thus, the CAI operating map presented
in Figure 4.1 is not a representation of the widest operating range possible with this experimental
set-up. Extensive CAI operating regime has been presented many times in the literature and is
not the focus of this project. The geometric compression ratio of the engine was set to 12.28.
This value was a compromise between the high compression ratio preference of a CAI engine
and knocking limit in SI engine. Experiments were conducted without a throttle plate to reduce
pumping losses. The intake air was heated to 120 'C to get a reasonable range of operation. The
experiments were conducted at 3 different fuel air equivalence ratio: 0.8, 0.9, and 1.0 at 1500
rpm. It should be noted that 0.8 is not the leanest operating limit. Leaner fuel air equivalence
ratio (as low as 0.7) can be used for limited number of valve timing combinations. The engine
speed was varied from 1250 to 1750 rpm. The shaded region represents the operating regime
published in SAE 2004-01-1898 by Honda at 11.2 compression ratio. The experimental data
resulted in a wider operating range than that of Honda due to the intake air heating.
5
2
=0.9
#=O.
8
These sets of data were at 1500 rpm, displaced for clarity
b00
1250
1750
1500
Engine speed (rpm)
2000
Figure 4.1 Map of CAI Operating Range
40
The fuel injection amount in these experiments was controlled using a PID controller to
keep a constant equivalence ratio. For constant equivalence ratio, the engine load was controlled
by the VVT system. Lower engine load was achieved by trapping more residual gas, which was
done by closing the exhaust valve earlier before piston TDC. With high residual gas mass
fraction, the charge could tolerate high effective compression ratio without resulting in excessive
pressure oscillations caused by rapid pressure increase. Therefore, the intake valve was closed
near BDC to maximize the effective compression ratio. Higher engine load was achieved by
trapping less residual/dilution. With less residual the heat release rate became faster and could
result in excessive pressure oscillations and ringing. Therefore retarded intake valve timing was
used to suppress the pressure oscillations by reducing the effective compression ratio and trapped
fresh air mass. Figure 4.2 and figure 4.3 illustrates an example of the actual intake and exhaust
valve timing used in the experiments.
50
0
4.5
4
0
3.5 0
0
-
190
200
210
data
linear fit
220
230
IVC (ATDCi)
Figure 4.2 IVC effect on Nimep
Stoichiometric, 1500 RPM, WOT, EVC varied
41
5
0
4.5
0
40
--4
O
data
quadratic
0
3.5
R 0
640
630
EVC (ATDCi)
650
Figure 4.3 EVC effect on Nimep
Stoichiometric, 1500 RPM, WOT, IVC varied
In general, the high load in CAI regime is limited by occurrence of excessive pressure
oscillations due to rapid heat release, which is followed by ringing noise similar to that of
knocking in SI engine. Figure 4.4 depicts pressure trace of four consecutive engine cycles at the
highest load used in this study. Pressure oscillations can be detected in the expansion stroke of
some the cycles, especially the ones that have early ignition timings. Post processing the data
indicates that the maximum pressure increase for this data set was 22 bar/CAD, which
corresponds to 198 bar/ms at 1500 rpm. The average pressure trace of 300 cycles is shown in
figure 4.5. There are no pressure oscillations in the average pressure trace. The maximum rate of
pressure increase for the average pressure trace is only 4 bar/CAD, which correspond to 36
bar/ms. It can be inferred from the figures that averaging the pressure data will smoothen the
pressure oscillations and that the rise time of the average pressure trace is also longer than the
rise time of individual pressure traces. These two facts can camouflage the presence of knock
and skew the high load limit to be higher than what it should be. The lowest load in CAI regime
42
is limited by cycle-to-cycle variations with eventual misfire in some of the cycle. Although the
COV of most of the experimental data were less than 5%, the absolute maximum COV used was
10%. Normally a COV larger than 10% indicates the presence of a misfiring cycle. The average
pressure trace in the low load condition is shown in Figure 4.6.
6 O.50 Pressure Oscillations
40
I
30
-
I
I.
I
I
I
I
I
I
I
I
20F
9,
9,
10
'i.
I
:500
320
340
360 380 400 420 440
Crank Angle (ATDCi)
460
480
Figure 4.4 Four Consectitive Pressur6 Trace when Penetrating High Load Limrit
Average Nimep 4.6 bar, 1500 RPM, Stoichiometric.
50
40P
30
I
20
10
0'
0
180
360
540
Crank Angle (ATDCi)
72 0
Figure 4.5 Average of 300 Cycles Pressure Trace at High Load Limit
Average Nimep 4.6 bar, 1500 RPM, Stoichiometric.
43
50
40
&
30
20
10
0
0
180
360
540
720
Crank Angle (ATDC i)
Figure 4.6. Average of 300 Cycles Pressure Trace at Low Load Limit
Average Nimep 3.5 bar, 1500 RPM, Stoichiometric
Simulation models were built to analyze the experimental pressure data. Several
important physical properties such as: residual gas mass fraction, in-cylinder temperature,
trapped charge mass, and heat release rate were estimated using appropriate models. The general
procedure to calculate those properties can be described as follows. The cylinder pressure trace is
fed into a 2- zone heat release models where the burn rate, location of 50% mass fraction burned,
and residual gas condition are calculated. The burned profile is then used to specify the wiebe
function, which is used to set the combustion process in the Ricardo Wave simulation program.
The Ricardo simulation program requires some predefined inputs such as: valve timings, fuel
injection or air fuel ratio, and manifold pressure. The calculated pressure from Ricardo Wave
results is compared with the experimental pressure trace until a satisfactory match can be
achieved. When the two pressure traces do not match, adjustment is made to the intake or
exhaust manifold pressure inputs and the simulation program is repeated. The intake and exhaust
manifold pressure need to be adjusted to match the actual pressure data when the intake and
exhaust valve were closed. When the engine is operated in CAI mode, the valve opening
44
duration for both intake and exhaust valve are very short compared to that of regular SI
combustion. Therefore, there is not enough time for the pressure in the manifold to equilibrate
and settle down to its steady state value. Hence, the in-cylinder pressure during the valve
opening period exhibits some dynamics, which is difficult
to be captured by the Wave
Simulation program. This phenomenon explains why the onset of CAI combustion also depends
on EVO timing. Using this phenomenon, residual gas mass fraction can be finely tuned by
advancing or retarding the EVO timings. However this phenomenon is highly correlated with
intake and exhaust port tuning, which is engine and experimental set-up specific. Therefore, in
these experiments EVO timing is mainly fixed at 145 degree after TDC firing, unless otherwise
specified. Figure 4.7 illustrates the flow chart of in-cylinder conditions estimations.
Experimental
Pressure Data
Two Zone Heat
Release Model
10~90% bum duration, CA50
Intake &Exhaust
Port Tuning
Ricardo Wave
Engine Cycle Simulation
Valve Timings
Fuel Air Ratio
No
Pressure Data =Sim. Result
Yes
Temperature Profile
Residual Fraction
Figure 4.7 Schematic of In-cylinder Condition Estimation Procedure
45
50
--
Simn
40
I
30
20
10
0
200
600
400
Crank Angle (ATDCi)
2500.
a)
2000
1500
1000
S
a)
5 00 1
H
0
180
)
180
360
540
72
0
180
360
540
72
0
360
540
7 20
400
350
300
250
200
150
100
0
0.9
0.8
0.7
CIS
0.6
0.5
0.4
0.3
Crank Angle (ATDCi)
Figure 4.8 In-cylinder Temperature, Residual Gas Mass Fraction, and
Charge Mass Estimation
46
Figure 4.8 illustrates an example of in-cylinder condition estimations. The experimental
in-cylinder pressure data matches well with the pressure data from the simulation result. The
peak cylinder temperature before the onset of auto-ignition ranges from 950~1050 K, while the
in-cylinder peak temperature due to combustion can reach 2400 K. It is interesting to notice that
the in-cylinder temperature during the recompression process can reach up to 1400 K. This high
recompression temperature can further oxidize any hydrocarbon left in the residual gas but it can
also cause some pumping loss due to heat transfer to the cylinder wall during the recompression
process. The in-cylinder charge mass indicates some fluctuation during the valve opening period.
This fluctuation is due to the manifold dynamics during the valve opening period. The incylinder residual mass fraction remains constant through most of the cycle. It reduces when the
intake valve is opened due to the induction of fresh air and increases as the combustion occurs.
The relevant in-cylinder residual mass fraction is the one just before combustion, which is equal
to the residual mass fraction at intake valve closing.
The relation between in-cylinder residual mass fraction and Gimep is depicted in Figure
4.9. The residual mass fraction was mainly controlled by changing the EVC timings. IVC timing
was changed to tune the combustion phasing. It is clear from the figure that the torque output is a
strcng function of residual gas mass fraction. High residual mass fraction leads to low Gimep
while low residual mass fraction leads to high Gimep. The residual mass fraction varies from low
30% to mid 60%, while the Gimep ranges from 2.5 bar to 5 bar. The influence of air fuel
equivalence ratio is not as prominent as the residual mass fraction for this particular experiments.
In general leaner mixture tends to results in lower Gimep due to less fuel injection amount but
the change in combustion phasing tends to offset this influence. Figure 4.10 illustrates the
relation between- residual gas mass fraction and in-cylinder charge temperature at 60 degree
BTDC firing. The location of 60 degree BTDC is selected to represent the charge temperature
before the onset of auto ignition. Higher residual mass fraction results in higher in-cylinder
charge temperature due to more hot residual gas trapped in the cylinder. This effect overcomes
the reduction of the absolute residual gas temperature with high residual gas fraction. For the
same residual mass fraction, higher in-cylinder fuel air equivalence ratio results in higher charge
temperature due to hotter residual gas.
47
5.5
1.0
# 0.9
5
0
O #
4.5 i
0
3.5
31
}SJ
2.5
,.~
~
0.3
0.4
0.5
0.6
0.7
Residual Mass Fraction
Figure 4.9 Residual Mass Fraction vs Gimep
Inlet air 120 'C, 1500 rpm, Valve Timing varied
930 140
910
A-
-4
0
Q-)
A
890
ci
0
0
1.0
0.9
870
I
85 b3
N
M
0.4
0.5
0.6
0.7
Residual Mass Fraction
Figure 4.10 Residual Mass Fraction vs In-cylinder Charge Temperature
Inlet air 120 'C, 1500 RPM, Valve Timing Varied
48
4.3
Effects of Fuel Metering on Combustion Stability
Figure 4.11 illustrates the effect of fuel air equivalence ratio on the combustion process in
CAI mode. The goal of this set of experiments is to evaluate the lean and rich limit of
combustion in CAI mode. The experiments were conducted by keeping the valve's timings
constant (IVC at 211 ATDCi and EVC at 634 ATDCi) and varying the fuel air equivalence ratio
by changing the fuel injection amount. The throttle remained fully open. The fuel injection
amount was controlled using a PID controller. It can be inferred from Figure 4.11 that the
combustion in CAI mode is more tolerant to leaner mixture rather than richer mixture. Taking
the stoichiometric mixture as the point of reference; in the rich side, the engine started to misfire
when 0 exceeded 1.10 while in the lean side no missed cycles were detected until 0 becomes
leaner than 0.8. In fact, when the engine was run at a different set of valve timing configuration,
stable combustion with 4 equals to 0.73 could be achieved with Nimep coefficient of variation
kept less than 5%.
oil
.
-
8
..
a
Cycle-by-cy cle Gimep
*
Average Gimep
7
6
51
it
-
S
4
3
-
aa2
-
2
01
-1
0.8
0.9
1
1.1
1.2
Figure 4.11 Fuel Metering Effect on Combustion Stability
49
-
6.00 -
- 50.00
-+- Nimep (bar)
5.00 -
COV (%)
40.00
4.00
A30.00
a)
3.00
--
20.00
2.00
10.00
1.00
0.00
0 .70
0.80
0.90
1.00
1.10
1.20
000
.n
1.30
Figure 4.12 Fuel Equivalence Ratio Effect on Nimep and COV
Some cycle simulations have been conducted to investigate the in-cylinder
condition of the experimental data using Wave simulation program. There is a good match
between pressure calculated from the Ricardo Wave simulation program and the experimental
data (Figure 4.13). When COV was less than 5% (0=0.84,0.90,0.95,1.00,1.05), the average
of
300 cycles of pressure data is used as the benchmark for wave simulation. When the COV was
greater than 5% (0=0.79,1.10,1.15,1.20), a typical pressure trace from a single cycle, prior to
a
missed cycled, was used. This choice was made to investigate the in-cylinder condition that
would lead to a missed cycle. In this experiments the spark was kept at TDC so that even when
a
cycle failed to ignite in CAI mode, there could still be slow flame propagation (the residual mass
fraction in this case was low enough), which could lead to a new CAI cycle.
50
60
40
-
Wave Sim,
=0.79
20
60
60
+ Exp.Data
0w
0
360
-
401
540
Exp.Data
JWave Sim.
0=.9O
20
0
0
60
360
540
Exp.Data
Wave Sim,
=1.00
20
4=0.84
60
0
40
-180
60r
360
180
540
+ Exp.Data
-
Wave Sim
#=0.95
20
180
40
40
20
180
L.
+ Exp.Data
Wave Sim.
-
--0
360
180
540
+ Exp.Data
- Wave Sim.
40
4=1.05
20
K
F'
-Y80
0
An
180
360
540
+ Exp.Data
Wave Sim
40
*=1.10o
20
n
-180
-
0
180
360
540
180
0
40
r
-180
360
540
+ Exp.Data
Wave Sin
-
#=1.20
20
hi
Iq
- 80
60
-
0
'II-
180
360
540
Figure 4.13 Wave Simulation vs Experimental Pressure Data
51
3000
3000
2000
2000
=O.79
1000
1000
-?8 0
3000
0
2000
180
360
540
-Y80
3000i-
0
3000
180
360
540
-380
300011
0
2000
=1.oo
180
360
540
180
360
540
360
540
360
540
=1.05
1000
I
0
180
360
- 80
540
31
=1.1o
2000
I
0
180
1
=1.2O
2000
1000
1000.I
-0-80
0
0
I K=.95
1000
2000
30
-?80
3nnn
2000
=O.9O
10001
1000
=O.84
0
180
360
540
-?I
-80
0
180
Figure 4.14 Temperature Profile at Different Fuel Equivalence Ratio
52
ami
360
340
320
3
.1
1. 25
Figure 4.15 Equivalence ratio effect on total charge mass
245C
2400
(
2350
2300
2250
2200
U
275
1
1.25
Figure 4.16 Equivalence ratio effect on peak charge temperature
53
870.
8601
F-z
8501
U
K
-. 75
1
1.25
Figure 4.17 Equivalence ratio effect on charge temperature @30BTDC
013t5.
0.34
0
0.33
0.32
I-
0.31
0.3
0.29
IR
0
.5
1
1.25
Figure 4.18 Equivalence ratio effect on residual mass fraction
54
Figure 4.15 indicates that changing the fuel injection amount has no apparent
effects on total charge mass. Total charge mass is mainly a function of intake manifold pressure,
cylinder volume, and charge temperature at intake valve closing. The intake manifold pressure
was kept constant by operating wide open throttle. The cylinder volume was fixed by using
constant IVC in this particular experiment. The cylinder temperature was effected by the residual
gas temperature but since the inlet temperature was kept at 120 C, the overall charge temperature
at intake valve closing did not change that much (within 3%). Therefore there is no significant
effect of equivalence ratio on the total trapped mass in this case. However, Figure 4.16 shows
that the equivalence ratio influences the peak cylinder gas temperature. The maximum of peak
cylinder gas temperature occurs at stoichiometric mixture. Lean mixture has less fuel energy to
be converted into sensible energy compared to stoichiometric mixture. Richer mixture produces
carbon monoxide (CO) and hydrogen (H2); thus not all the chemical energy has been released. In
this case the combustion is not complete due to lack of oxygen. The residual mass fraction tends
to decrease with higher equivalence ratio for this particular experiment. While the exhaust valve
closing timing is fixed for this particular set-up, the residual gas mass fraction is influenced by
the exhaust pressure dynamics, which was created due to the short valve opening duration in CAI
mode. In this particular case, richer mixture tends to have later ignitioi timing and higher
pressure at exhaust valve open timing which means that more gases are pushed out during the
exhaust stroke. The residual gas mass fraction changes from 34.7% to 28.8% as more fuel was
injected. Figure 4.17 illustrates the effect of air fuel equivalence ratio on the charge temperature
at 30 degree BTDCf The temperature at this location is selected to represent the peak charge
temperature since at this location no oxidation has occurred yet. This temperature is critical to
assess the ignition timing of CAI engine. Since combustion in CAI mode is governed by
chemical kinetics, the peak compression temperature governs whether a mixture will ignite or
not. It is interesting to notice that the charge temperature at 30BTDC is not sensitive to change in
fuel injection amount between phi equals to 0.9 ~ 1.0. The peak charge temperature depends on 2
factors: initial charge temperature (Figure 4.19) and specific heat ratio of the charge (Figure
4.20). The initial charge temperature depends on the residual mass fraction and the residual gas
temperature. Initial charge temperature for stoichiometric mixture is the highest in this case due
to the high residual gas (exhaust temperature) for the stoichiometric mixture. The specific heat
ratio is greater for leaner mixture (more 02 and N2) instead of CO. There is a small reduction in
55
the peak charge temperature for mixture leaner than 0.9 phi due to the canceling effect of greater
gamma but lower initial charge temperature. The reduction in the peak charge temperature is
prominent for rich mixture due to the compounding effects of both the reduction of gamma for
richer mixture and lower initial charge temperature.
528
527
526
525
524
5231
522
521
520
1
F75
1
1.25
Figure 4.19 Fuel Injection Effect on Initial Charge Temperature
1.314
1.312
S1.31 F
0
o
0
0
1.308
1.306
1.304
ta)1.
302
1.3
-
1.298
1.291
0.7 5
1
1.25
Figure 4.20 Fuel Injection Effect on Average Gamma during Compression
56
Pumping Loss in CAI mode
4.4
4.4.1 Introduction
Pumping work in internal combustion engine corresponds to the work required to pull in
and push out the gases to/from the engine cylinder. Pumping work is always negative (the piston
needs to do work to the gas) in naturally aspirated engine but can be positive in a
turbocharged/supercharged engine. Spark ignition engine faces penalty in pumping work when
operating in light-load condition due to lower intake pressure. In order to reduce the air flow in a
Spark Ignition engine, the engine needs to be throttled, which creates restriction in engine
breathing and leads to additional work required to pull the fresh air into the cylinder. Diesel
engine, however, operates without a throttle. The load in diesel engine is controlled by changing
the fuel injection amount while maintaining roughly constant airflow. Therefore, diesel engine
faces smaller pumping loss compared to spark ignition engine. CAI engine can operate with and
without a throttle. However, in this project the throttle was kept wide open to reduce the
pumping loss.
There are several calculation conventions for the definitions of pumping work. Figure
4.21 shows a schematic of the P-V diagram when an SI engine is operated at light load condition.
1.
360 Degree Integration
This is the most commonly used convention to define pumping work. This
convention defines pumping work as
pdV
(TDC intake stroke until BDC
compression stroke; positive since dV is positive) and
d pdV
(BDC expansion
stroke until TDC intake stroke; negative since dV is negative). The result of this
integration is area B and C in figure 4.20. The gross indicated work (Gimep) is
defined as area (A+C), while the net indicated work (Nimep) is defined as area (A-B).
This method implicitly assumes that the gas exchange process has no influence on the
gross indicated work (Gimep). Since it is not the intention of the author to propose a
new method of defining a pumping work, this generally accepted definition of
pumping work is used throughout this thesis.
57
d
C
a
b
Volume
4.21 Schematic of P-V diagram in light load SI engine
2.
Area B calculation
The proponents of area B calculation argue that area C should not be included in
pumping work calculation since area C appeared as positive work during compression-expansion
stroke and negative work during the gas exchange process. They argue that since area C has no
influence on net indicated work calculation it should not be treated as pumping work. This
method fails to capture some gas exchange phenomena in a VVT equipped engine. For example
when the exhaust valve open timing is retarded so that the exhaust blow down process finished
late, the work done to push the exhaust gas out of the cylinder in the beginning of the exhaust
process will increase and yet, the size of area B will not be affected (the extra work was
subtracted from A). This convention makes it difficult to evaluate the improvement of the
combustion characteristics, which is represented by area A and C
58
3.
Modified 360 degree integration method.
Shelby et al [12], proposed a new method to calculate the pumping work for a VVT
equipped SI engine. They attempted to capture the whole gas exchange process by adding 2 extra
terms to the 360 degree integration method:
PMEPadj
Where
=
PMEP 3 60 + ICW + EVOexposs
PMEP 360 : PMEP as calculated by convention 1
ICW: Incremental Compression Work, which depends on Intake valve closing
EVOexp loss : Expansion loss due to early exhaust valve opening
While this definition takes into account of valve timing effects that occur in the compression and
expansion strokes by making adjustment to PMEP and IMEP, it requires some approximations in
calculating the ICW and EVOexposs It also assumes that the expansion stroke can be expanded
until piston BDC without altering the combustion process, which is not true in CAI mode.
4.4.2 Sources of Pumping Loss in CAI engine
The widely-accepted original 360-degree integration method is chosen in this study to
calculate the pumping loss in a CAI engine. With residual trapping method (early exhaust valve
closing), the pumping loss in CAI mode results from the work required for engine breathing and
the heat loss during the recompression process. Even at wide open throttle, the work required for
engine breathing is still influenced by pressure dynamics when the valves are open. In CAI
engine with residual trapping method, the valve opening duration is much shorter than regular
spark ignition engine; therefore the pressure does not have enough time to reach steady state
condition. The pumping loss due to engine breathing can be reduced by tuning the intake and
exhaust manifold. The pumping loss due to heat loss during the recompression process occupies
a significant portion in the total pumping loss in a CAI engine. With residual trapping method
the residual gas temperature can reach well above 1100 K during the recompression process.
This high residual gas temperature drives the heat transfer to the cylinder wall, which is
approximately at 450 K, and becomes the source of sensible energy loss. The heat loss can be
59
reduced by decreasing the residual gas temperature (running lean) and using a better insulation
for the cylinder liner.
50~
Nimep
M
eiN
0)
40
Pmep
Recompression
Pep
30 Recompression
20
Pmep
Breathing
Pm'ep
Breathing
10
Gimep
C
%M1
-
IVO
0
Uomm4
moso
<0*mmm
- -
,
360
180
540
EVC
720
Crank Angle (TDCi)
Figure 4.22 Schematic of Pmep Definition used in this study
-12
1ll
Recompression
Pumping Loop
8
6
Breathing
umping Loss
4
2
V
0
0
100
200
300
V
400
500
Volume (cc)
Figure 4.23 Illustration of two sources of pumping loss in CAI Combustion
60
Using the 360-degree integration convention, PMEP can be expressed mathematically
by the following expression:
TDC int ake
BDCcompression
fpdV /Vd
Pmep=
pdV /Vd
+
BDC exp ansion
TDC int ake
which can be further divided into pumping loss during the valves are open (due to engine
breathing) and when the valves are closed (due to heat loss).
Pmep
= Pmepbreathing +
Pmeprecompression
Pmep during recompression is due to the heat transfer from in-cylinder charge to the cylinder
wall. The rate of heat loss to cylinder wall can be expressed as:
Q = h A (Tin-cylinder - Twali)
h = 2.94 x B('
x P 0.8 x TEMP (-50) x W0 8
h: heat transfer coefficient, based on modified woschni in W/m2 K
B: engine bore in m
TEMP:charge temperature in K
P: in-cylinder pressure in kPa
W: average gas velocity (m/s)
Tin-cylinder : In-cylinder gas temperature in K
Twall: Wall temperature
A: Surface area (piston, cylinder liner, and cylinder head)
Figure 4.24 illustrates the Pmep during steady state mapping of CAI operating regime at
1500 rpm. In general it can be seen that leaner mixture leads to lower pmep due to less
Pmerecompression.
Leaner mixture results in lower peak burned gas temperature (Figure 4.16),
which reduces the residual gas temperature during the recompression process and hence less heat
loss during the recompression process. The same figure also shows that as Gimep is reduced
from its maximum value at a constant phi, pmep is increased initially before it becomes flat and
slightly decreases again. When 0 is kept at a constant value, the load (Gimep) is changed by
changing the trapped residual gas amount. Closing the exhaust valve early, leads to more trapped
61
----------------
residual and lower engine load. However, since the peak trapped residual gas temperature is
higher, one would expect a monotonous increase in pmep as Gimep is reduced, which contradicts
the experimental result. Therefore in order to explain this phenomenon, several wave simulations
were conducted. The results for 0 equals to 0.9 are shown in Figure 4.25. Figure 4.25 indicated
that pmep recompression was indeed increased as the load was reduced but the breathing pmep
was also reduced. This is due to the fact that lower engine load would require less fresh air to be
inducted into the cylinder and less exhaust gas needs to be expelled from the cylinder.
0.7
Y
e
0.6
*
Phi 0.80
Phi 0.90
Phi 1.00
0.5
0
0
0
0.41
e ***0
0.31
0.2
5
0
3
3.5
4
4.5
5
Gimep (bar)
Figure 4.24 Experimental Data of PMEP during steady state mapping of CAI regime
62
0.5
0Pmep
0
0.40.
0. 2
0.1
3
3.5
4
4.5
Gimep (bar)
5
Figure 4.25 Wave simulation results, 1500 rpm, phi 0.9
4.5
Heat Release Analysis and Combustion Phasing Characteristics.
An equivalent of 2-zone heat release model was created using several FORTRAN
subroutines to investigate the combustion phasing characteristics and assess the residual gas
conditions. The model assumes:
1. There are 2 separate zones of mixture: unburned and burned mixture.
2. There is no heat loss from the unburned mixture and the unburned gas composition is
frozen during compression.
3. The unburned mixture is compressed isentropically.
4. The burned gas is always in equilibrium.
5. There is heat transfer from the burned gas to cylinder wall and the heat transfer
coefficient is based on modified Woschni.
63
The inputs to the model are experimental pressure data, initial charge temperature, total trapped
mass, and initial charge composition. The inputs are calculated using wave simulation program.
The model outputs include: burned gas mass fraction profile, cumulative heat loss, and residual
gas state. Examples of the program outputs are shown in Figure 4.26 and Figure 4.27.
1.2
1
7:
CA50
0.8
06
0
0.4
0.2
C
-0.
'
0
320
340
360
380
400
420
Crank Angle (ATDCi)
Figure 4.26 Fuel Mass Fraction Burned Profile
0. 14 ,
0.121
0.11
0.08
0.06
0.04
0.02
0
00 (
- .
-- 00
320
340
360
380
400
420
Crank Angle (ATDCi)
Figure 4.27 Normalized heat release rate
Figure 4.26 illustrates the CA50 (the crank angle location where 50% of the fuel energy has been
released), which is commonly used to indicate the combustion timing. The combustion duration
64
is commonly measured using 10-90% burned duration. When the engine is operated in CAI
mode, the combustion duration is very short (less than 10 CA degree at 1500 rpm) even with
high level of dilution.
5.41
0
5.2k
0
5F
0
o3
(U
-0
4.61
W4
2 .4
0
AA
A
4.2-
Gross
A
Net
Al
41
3.
0
60
380
375
370
365
CA50 Location
Figure 4.28 Location of CA50 vs IMEP
1500 RPM. Stoichiometric. NO 84. IVC'217. EVO 505. EVC 636 ATDCi
4.8
Gross
o
Net
Lin e
4.6
0
.0(U
o
0
o40
--
.-.
0 0
110o
4.4
W.. 4.2
0
go
0
0
41
0
3.84
0
170
375
380
385
390
CA50 Location (ATDCi)
Figure 4.29 Location of CA50 vs IMEP
1500 RPM, Stoichiometric, IVO 83, IVC 229, EVO 505, EVC 637 ATDCi
65
Figure 4.28 and 4.29 illustrates the influence of CA50 location on engine torque
output (IMEP). Figure 4.28 indicates that when combustion occurs too early, there will be
reduction in IMEP. When CA50 is located between 372 to 378, there is very little variations in
ensemble average IMEP. In fact, the vertical variation of IMEP (constant CA50) is more
prominent than horizontal/ensemble IMEP variation in this case. As CA50 becomes very
retarded the imep becomes smaller again. This phenomenon implies the presence of an optimum
combustion timing, around which the imep is insensitive to combustion timing (similar to MBT
timing in spark ignition engine). The vertical imep variation is caused by cycle-to-cycle variation
in fuel, air, and residual mass; while the horizontal imep variation is due to random mixing of the
charge, which creates thermal and charge composition stratification within the cylinder. The
dotted line in Figure 4.29 shows 2 consecutive cycle, where a late burning cycle is followed by
early burning cycle due to cycle-to-cycle coupling when the engine is operated in CAI mode.
7C,
d
data 1
- line ar
o
601
0
0
501
"0
401
U
en-4
301
20'
360
365
370
375
380
385
390
CA50 (ATDCi)
Figure 4.30 Correlation between CA50 and Peak Cylinder Pressure
66
Figure 4.30 shows the correlation of peak cylinder pressure to the CA50 in CAI
mode. The figures shows very clearly that early combustion phasing is associated with high peak
cylinder pressure. Since GIMEP is the integration of p dV over the compression and expansion
stroke, high pressure in the compression stroke as caused by early combustion phasing leads to
reduction of GIMEP. A more important question is whether the insensitivity of GIMEP to CA50
location over a certain range is caused by the same mechanism as encountered in Spark Ignition
engine. Figure 4.31 illustrates the relative cumulative heat loss at different combustion phasing.
The cycle with late combustion phasing tends to exhibit lower cumulative heat loss, which
compensate for lower peak cylinder pressure. Therefore JIMEP is not always sensitive to the
combustion phasing.
20C,
C4,
150
C
-
---
CA50@372
CA50@375
CA50@378
CA50@387
1001
Le ss
Heat loss
U
50
I
ni
:00
350
400
450
Crank Angle (ATDCi)
Figure 4.31 Relative Cumulative Heat Loss
67
1 Cum.H eat
Release
I
0.8
Sensible Energy
0.6
Gum.
5-
0.4
Indi cat
ed Wort
. ----
390
420
450
0.2
0
-0.2)q
3 0
360
480
Crank Angle (ATDCi)
Figure 4.32 Details of Energy Release Curve
The details of fuel energy released are shown in Figure 4.32. The fuel energy released is
distributed into change in sensible energy of mixture, cumulative heat loss, and cumulative
indicated work. Figure 4.33 compares each of these terms for an early burning cycle (CA50 at 12
ATDCf) and a late burning cycle (CA50 at 18 ATDCf). The cumulative heat release and
cumulative indicated work are about the same for both cases. However the late burning cycle
exhibits less cumulative heat loss and larger change in sensible energy. Larger change in sensible
energy leads to hotter exhaust gas temperature. This phenomenon explains the occurrence of
early burning cycle after late burning cycle as shown in Figure 4.29. When the engine is operated
in CAI mode, a modulation in CA50 location (late burn - early burn - late burn - early burn, so
on) is often observed. Therefore to a certain degree, CAI combustion has a self-correcting
method to settle down into a stable operating condition. The author believes that this
phenomenon is caused by the change in the sensible energy as shown in figure 4.33.
68
Normalized Cum. Heat Loss
Normalized Cum. Heat Release
2
1. .
.
-
0.25I
C
1
Early Burning Cycle
Late Burnine Cycle
0.2-
0.8
E"
()
Ul 0.6
- Early' BUring Cycle
U) 0.4
"Late Burning Cycle
=3
0.15[
ci)
0.1
LL
0.2
0.051
C
0
-0.; I
35
40
45
50
350
400
450
500
E30
Crank Angle (ATDCi)
450
420
390
360
480 500
Crank Angle (ATDCi)
Normalized Sensible Energy
Normalized Cum. Indicated Work
11
I
r%
-
2 0.5
:
Late burnin1g Cycle
0.4
0 .3
-
Early Burning Cycle
0.8
,, % e
E)
Early Burning Cycle
Late Burning Cycle ,
t
LU 0.6
,
LL
> 0.2.
- 0.4
0.1
LU
0.2
0
-nIL
'30
LU
360
390
420
450
Crank Angle (ATDCi)
480 500
r~I
A
130
360
390
420
450
480 500
Crank Angle (ATDCi)
Figure 4.33 Comparison between Late and Early Burning Cycle
69
4.6
Spark Assisted Controlled Auto Ignition.
One interesting phenomena encountered during mapping of CAI regime is the occurrence
of spark assisted auto-ignition. In a spark assisted CAI, the combustion process is started with a
spark discharge by the spark plug which initiate flame propagation from the spark plug region.
The unburned mixture is further compressed until auto-ignition occurs. This process is very
similar to auto ignition of end gas in regular spark ignition engine but since the mixture (end gas)
is highly diluted with residual gas, the pressure oscillations are minimal.
The spark timing
controls the auto-ignition timing in this process. There is often a limited range of spark timing
where spark assisted controlled auto ignition can occur. When the spark is too retarded, autoignition will not occur and the combustion progresses as regular spark ignited combustion.
Consider an engine without throttle but equipped with a fully flexible valve timing
mechanism (current experimental set-up). The geometric compression ratio of the engine is set at
12.3. Two different cases were evaluated. In the first case, the intake valve timing was set at 80
degree before TDC for a low effective compression ratio. The exhaust valve closing was initially
set at piston TDC and gradually advanced toward piston BDC to trap more residual gas. The fuel
air equivalence ratio was fixed at stoichiometric value with a PID controller. As the residual gas
amount becomes larger, the combustion becomes unstable. In this case, the compression ratio
was not high enough to initiate auto-ignition and the engine misfired when the residual mass
fraction was too large for flame to propagate or the flame propagation was too slow to compress
the charge temperature to its auto-ignition temperature. In the second case, the intake valve
closing time was set at 10 degree after BDC. The exhaust valve was closed at 100 degree before
TDC (a lot of residual) and the engine was running in CAI mode. The exhaust valve was then
retarded to trap less residual and increase the engine load. When the residual gas mass fraction
was in the low 30% by mass fraction or low 40% by volume (through post-processing the data),
the ignition timing starts to affect the combustion process. The engine misfired when the spark
was turn off. In this particular experimental set-up, the region where the spark started to affect
the combustion process was very close to the high load limit where pressure oscillations could be
observed.
70
4CIi
U
I
I
Auto-igiution
30
a)
20
a)
.
. .
-- - - - - - - --
- - - --.
. . . ...... ----- - --- - --- -
...................
......
.........
..
.....
...-- - -
................................
10
.. ......
. ...........
............
. ............
......... ......
.... ....................
0
-10"
0
I
U
I
180
360
540
720
Crank Angle (CA)
Figure 4.34 An Example of Spark Assisted CAI during mode transition
1500 RPM, Spark at 20 BTDC
4
S
40
I
I
U
.......................................
...........................................
30
a)
3U,
a)
3-
20
....
.....
. ...... .............. .............
.. ....... ...........
......... .........
10
rr '
0
I
180
360
540
720
Crank Angle
Figure 4.35 An Example of Spark Assisted CAI during steady state testing
1500 RPM, Spark at 20 BTDC, Residual Mass Fraction (Wave) 24%
71
Figure 4.34 illustrates an example of spark assisted CAI combustion. In this figure, there
are two distinct regions of combustion. The first one is due to spark ignited flame propagation.
The spark was set at 20 degree BTDC. As the flame propagates, the unburned mixture is
compressed and its temperature is increased. In this case the unburned mixture auto-ignited about
25 degrees ATDC. The auto-ignition timing can be inferred from the pressure curve; very rapid
pressure increase can be identified to mark the auto-ignition. Figure 4.35 illustrates another
example of spark assisted CAI combustion. However unlike Figure 4.34, the pressure trace in
this figure is very similar to a regular CAI combustion. There was no clear flame propagation
region that clearly marked the spark assisted CAI combustion. However, when the spark was
turn off the engine started to misfire, which clearly showed the requirement of spark assistance to
initiate combustion.
There are two potential use of spark assisted CAI combustion. The first one is to extend the
operating regime of CAI combustion. It is possible to use the spark assisted CAI combustion to extend
the high limit of CAI operating range. With this particular experimental set-up, the spark assisted CAI
combustion can be used to increase the highest load achieved by 0.3 bar (the map shown in Figure 4.1
does not include the spark assisted CAI). The spark has no effect on the combustion process when the
residual mass fraction is too high. The second potential use of spark assisted CAI combustion is to
help the mode transition process between SI and CAI mode. In Chapter 5, successful mode transition
within a single cycle was demonstrated without going through the spark assisted CAI combustion.
However, when the engine is only equipped with variable cam phasing mechanism (not with a fully
flexible valve train system), it is possible that the engine has to be operated in spark assisted CAI
mode before it can be switched into pure CAI mode. When large change in valve timings can not be
made within a single step or can not be changed independently, it would require several cycles before
the transition to CAI mode can be made. In this case, some adjustments to the last few SI cycles need
to be made before the transition into CAI mode can be achieved.
72
4.7
Oxidation in the recompression process.
During a failed transient (during load change within CAI regime or mode transition), it is
possible to have further combustion in the recompression process. One sign of such presence is a
negative pmep (to the piston during the pumping loop) due to the higher re-expansion pressure
(over the recompression value) resulting from the combustion process (depends on the timing of
the combustion). Figure 4.36 shows an example of the combustion in the recompression process
following misfire/incomplete combustion in SI mode. Combustion in the recompression process
occurs due to the presence of fuel and air in the recompression process, where the temperature is
still high (can reach above 1200 K). Combustion in the recompression process should generally
be avoided due to the following reason:
1.
It indicates an abnormal main combustion process.
The combustion in the recompression process implies that some of the fuel was
pushed out of the cylinder during the main exhaust stroke event. This means some of
the fuel was directly pushed out to the exhaust system without releasing the fuel
energy.
2. It can alter the intake process.
While a small amount of oxidation in the combustion process can help to promote the
auto-ignition, but an excessive combustion in the recompression process can lead to
engine misfire in the next engine cycle. The high pressure associated with combustion
in the recompression process could alter the intake airflow to the cylinder. This will
cause blow back into the intake port and make it difficult to adjust the fuel/air ratio
properly.
While a large amount of oxidation in the recompression process is detrimental to the combustion
of the next cycle, a small amount of oxidation in the recompression process is normal, following
a late burned cycle. Figure 4.37 illustrates the oxidation in the recompression process following a
late burned cycle.
Some researchers are trying to use the high residual gas temperature during the
recompression process by intentionally inject a very small amount of fuel into the cylinder
73
during the recompression process (using direct injection). They claimed that this dual injection
strategy can be used to expand the operating range of CAI regime.
50
Recompression
combustion
Recompression~
combustion
40
30
0
1
20
-
10
0
-10
720
0
1440
2160
2880
3600
4320
5040
Crank Angle
Figure 4.36 Recompression Combustion
19
1810
7
s-p
6[
5-4
15
14[
13
40
42
44
46
Volume (cc)
48
Figure 4.37 Oxidation during normal combustion in CAI mode
74
4.8
Engine Transient Response to Step Change in CAI mode
4.8.1
Fuel metering effects
Figure 4.38 illustrates the engine response when fuel injection was suddenly changed.
The fuel injection amount was step changed from 15.2 mg up to cycle number 87 to 12.8 mg
starting from cycle number 88. The exhaust lambda sensor indicated that the fuel air equivalence
ratio changed from stoichiometric to 0.9 in 2~3 engine cycles. The Gimep and Nimep slightly
decreased as the engine operate in lean condition. This reduction is not proportional to the
reduction of fuel injection. While the fuel injection amount was reduced by 15.7%, the reduction
in Gimep was only 9%, which indicated that CAI engine had higher thermal conversion
efficiency at lean condition. This higher thermal energy conversion efficiency comes from better
combustion phasing and lower heat loss for lean operation. In general the system reaches new
steady state response within 2 engines cycle at 1500 rpm (approximately 0.16 s) for a small
perturbation in fuelling. This step change is also reversible (the reverse is true for an increase in
fuelling).
Injected Fuel Mass x 0.5
Gimep
.4
1.2
2
Exhaust
-1.0
0
-0
----
,0
80
90
100
110
120
.8
130
Figure 4.38 Step Change in Fuel Injection Amount
75
4.8.2
Step change in Intake Valve Closing time
Figure 4.39 indicates the engine response for a step change in the intake valve closing
time when the IVC is moved closer to piston BDC. The intake valve was moved from 209
ATDCi to 191 ATDCi. There were no significant differences in engine torque output and fuel air
equivalence ratio before and after the step change. However, there was a noticeable increase in
the peak cylinder pressure after the step change due to higher effective compression ratio of the
charge. The CA50 location was also moved toward TDCf (became more advanced). Figure 4.39
shows that the engine reaches new steady state within 4 engine cycles (0.32 s) at 1500 rpm.
o20
x
IVC - 190
15
&
Gimep
10
ifi
S021
Exhaust p
0.)
100
110
105
115
120
Cycle#
Figure 4.39 Engine Response for a step change in IVC
76
NMRPq
"I
IMPR!, M flp? - Rql ' "M 1 111 '"'M
M14r 1 , 1 1 !'-
111IN".'
-'R
-
- ,
'-
" " ......
!j'JI 1 !jIF.I"'11 I
."'N qm jI" I RIO
I, Ill. . 'J'191 .. 1.
M
WfM
P"PqW
-
..
------ ----- --
.------- ---
20
15
1VC - 190
U4
10
5
CA50-360
0
-10
kn10
5
100
105
110
115
120
Cycle#
Figure 4.40 Engine Response for a step cfiange in IVC
4.8.3
Step change in Exhaust Valve Closing
Figure 4.41 illustrates the effect of advancing exhaust valve closing from 636 to 627
ATDCi. Closing the exhaust valve early results in more residual gas amount trapped. Since the
intake valve timing was fixed, the cylinder volume was fixed and therefore there was a reduction
in the fresh air that was inducted into the cylinder. This effect can be seen from the exhaust
lambda sensor, which indicated that the mixture becomes richer (from 0.9 $ to stoichiometric)
even though the fuel injection amount was kept constant. The Nimep and Gimep remained
constant before and after the step change. When the EVC timing was further advanced to 620
ATDCi, the engine started to misfire due to excessive residual and fuel rich mixture trapped
inside the cylinder. In general CAI engine tends to misfire with rich mixture (see section 4.3).
Figure 4.41 illustrates the details during the step change. The peak cylinder pressure was reduced
and the location of CA50 became slightly retarded. These phenomena are due to the dilution
effect of having more trapped residual inside the cylinder. For fixed fuel injection amount, IVC,
77
and EVO; advancing the exhaust valve closing results in more trapped residual gases, which can
lead to higher fuel air equivalence ratio, more dilute mixture, and higher/lower initial charge
temperature (depends on the new equivalence ratio). Therefore, the effect of EVC
on
combustion timing can be two fold. When the engine is operated at the lean limit, advancing
EVC may cause the combustion timing to become more advanced due to the increase of fuel air
equivalence ratio and initial charge temperature. However advancing the exhaust valve closing
time from stoichiometric will result in rich mixture, which will retard the ignition timing and
cause misfire in some of the cycles. The engine reaches the new steady state condition in 4-5
engines cycles at 1500 rpm.
x08
EVC-630
Gimep
- 6
-
.. Exhaulst#
2
.
-g
goJ
0
100
90E
120
110
1:30
Cycle#
Figure 4.41 Engine Response for a Step Change in EVC timing
4.8.4
Effect of step change in EVO
Figure 4.43 illustrates the results of retarding the EVO timing by 20 crank angle degrees.
The EVO timing influence CAI combustion through its effects on the pressure dynamics during
exhaust valve open period. Since the valve opening duration in CAI engine is very short (about
120 CA), there is not enough time for the pressure to stabilize in the exhaust system. Therefore
78
99M. "I IF, Wil '!Mmm
:-P W-' W-T-9-
-
-.-
lw"-N - I
1
9.
4, FPO M"11" III Rol IMI"j? 1,- el- - 1 m 10 '.
' IIF"
11MO-W 111 19. 9 1
.., I I
N x"9MONIFT16"
"9
"
' P
P0 Opq R!'111 "IMPIr
the EVO timing can influence the amount of residual gas trapped. In this particular case, more
residual was trapped with later EVO timing due to not enough time to push the exhaust gas out
15a
Q
$-
101
6e
5
CA50 - 360
.0
01
EVC - 630
U
130
120
110
100
Cycle#
Figure 4.42 Effect of EVC timing
7
CD
0D
EVO 525
EVO 505
61
GEmep
5
4
I-
3
2
1
Exhausf
0 -e
-1'
C
5
10
15
20
25
30
35
Cycle#
Figure 4.43 Effect of retarding EVO
79
60
50
40
30
rJ
201
Late EVO-
10
01
Early EVO
-10'
0
180
360
540
720
Crank Angle (ATDCi)
Figure 4.44 Effect of retarding EVO
of the cylinder. The pressure at exhaust valve closing is larger for later EVO timing case and
therefore more residual is trapped at EVC. Because IVC timing was fixed, the amount of fresh
air is approximately constant and therefore, richer mixture ($ changed from 0.98 to 1.04) is
observed after the exhaust valve timing is retarded. The pumping loss was bigger in this
particular case due to higher exhaust pressure at EVC as can be seen in figure 4.44. The system
reaches new steady state condition in 4 engine cycles.
80
4.9
Open Loop Load Control
In order to investigate engine response to a change in engine load, two operating
conditions (points) were selected from Figure 4.1. One to represents the low load condition (IVC
191, EVC 614, EVO 525, # 0.8) and one for high load (IVC 213, EVC 636, EVO 510, # 1.0).
The COV of both operating points were below 3%. These 2 points were chosen to be close to the
low load and high load limit but still have a low Nimep COV. This section aims to clarify the
following questions:
1. What is the time scale for load change in CAI mode?
2. Is the load change reversible in both directions (increasing and decreasing)?
3. What are the physical phenomena that need to be considered during a load change?
Figure 4.45 illustrates a large step change from high load to low load. The engine was run at the
high load operating point for some time before the valve timing and fuel injection settings were
changed to the low load operating point. The sequence of the change is as follows: the intended
fuel amount for cycle i is injected during the expansion stroke of cycle i-i (port fuel injection),
the exhaust valve open and close timings were changed to the low load condition, before finally
the intake valve open and close timings were changed.
4.9.1
Large Step Change from High Load to Low Load
-e-
Gimep
5
4.54
erma Stabilization
3.5
3
2.5
0
50
100
150
200
Cycle#
Figure 4.45 Large Step Change from High Load to Low Load
81
ycle-I
a
45
-0
Cycle 1
Cycle
40
I
351
I
I-
*
I
*
I
U,
U,
J
30
I
I
I
I
5,
I
I,
I,
a,
I,
Ug
Ig
Ig
1/
251
-
1
2
3
4
- - 5
6
7
-8
9
10
4
20 b
&I
0
340
380
360
Crank Angle (ATDCi)
400
Figure 4.46 Details of pressure trace before and after step change
It can be seen from figure 4.45 that there are two distinct time scales when the load is changed.
Within 4 engine cycles (0.32 s at 1500 rpm), the Gimep dropped from 5.3 bar to 3.2 bar. Figure
4.46 illustrates the ignition timing during the transition and it can be seen that there was no
significant difference in ignition timing after cycles# 4. The first cycle after the transition
showed an early ignition timing due to the hotter residual gas produced at high load condition.
While there is no significant difference in ignition timing after cycle#4, the Gimep decreased
slowly toward a new steady state condition. This slow decaying movement took more than 50
engines cycles and was associated with thermal stabilization of the cylinder wall. Since there was
a large difference in the charge temperature before and after the step change, the cylinder wall
temperature changed accordingly.
Wave simulation was conducted to investigate the shorter time scale. Figure 4.47 shows
the pressure, in-cylinder temperature, and residual mass fraction during the load change. The
results are summarized in figure 4.48.
82
-
Data
-
Wave + Wiebe
6C
4C
2C
3
S2
C2C
0
300CIs-m
1000
2000
3000
a
a
4
5
4000
200C
S
H
1Oo
0
as
4
a
0
1000
2000
3000
4000
a
a
a
a
a
1000
2000
3000
4000
1
0.51
'0
Crank Angle (CAD)
Figure 4.47 Wave Simulation Results for High Load to Low Load Step Change
It can be inferred from Figure 4.47 and Figure 4.48 that the smaller time scale was associated
with emptying and filling of the residual gas. It took about four engine cycles to reach the steady
state residual gas condition, which was equal to the time scale required to reach a new stable
ignition timing as shown in Figure 4.48.
83
750
0
0
(0
60
E
740
2400
730
2300
2200
720
I
710
2100
U
2000
700
-1
Steady Stalte
0
1
2
3
4
1900
Steady State
1i on
5
0
1
Cycle#
2
3
4
5
Cycle#
0.7
0.65LL~ 0.6 .
U-
U')
0.55-
CU,
0.5.
70
CU
U)
Steady State
0.450.40.35-
0.31
'
0
'
s
1
2
3
4
5
Cycle#
Figure 4.48 Summary of Wave Simulation
Two methods were tested to predict the ignition timing of a CAI engine. The first one was based
on the kinetics mechanism from Lawrence Livermore National Library for PRF fuel. It includes
1034 species and 4238 reactions. The linear algebraic differential equations were solved using a
DAEPAC solver, which reduced calculation time significantly. The inputs to the model were:
initial temperature, pressure, and species mass fraction. These inputs were calculated using Wave
simulation. The output of the simulation was the start of combustion timing. In this case, start of
combustion was defined as the time when 50% of the fuel has been converted to other species.
This usually corresponds to the timing where a noticeable pressure increase could be observed.
84
The second method was based on the knock integral method, which was developed by the HCCI
consortium kinetics team. This model was a semi-empirical method developed originally for isooctane. This method, also called livengood integral method, defines the start of combustion as
the timing when
It = I
ivc
f
or
28460
r = 2.25 x 10~
X
p-1 X0-0.87x z 0
1.54
2
xe
RT
,r: ignition delay (ms)
P: Pressure (atm)
0: Fuel Equivalence Ratio
X02:
Mol fraction of 02 (%)
R: universal gas constant (cal/mol.K)
Figure 4.49 shows the ignition timing estimation results using these 2 methods. While both
methods were able .to simulate the ignition timing behavior during transient, the LLNL model
was able to capture the change in ignition timing for the first cycle after transition more
accurately that the knock integral method.
375.
CA50 from data + heat
release analysis
---
Q
370
H
365
- v +
Wave +\
Livengood
360
%%
-%,%-%
Q
355
350[
~AA'-I
-1
0
1
Cycle#
2
3
Figure 4.49 Ignition timing estimation
85
4.9.2
Large step change from low load to high load
The reverse step change was conducted to investigate the engine behavior during large
step change from low load to high load transient. Figure 4.50 indicated that the first cycle after
step change always misfired even when the spark was kept on at TDC. Figure 4.51 shows the
pressure trace of a few cycles after the transition. The first cycle after the transition indicated a
complete misfire (negative Gimep), while the second cycle indicated incomplete combustion in
spark ignition mode. Cycle#3 is a knocking CAI cycle due to the hot residual produced from
cycle#2. Late spark ignition combustion is associated with high exhaust gas temperature. The
CAI combustion is maintained after cycle#4 and gradually settles down to a new steady state
value.
Q1
9
0
V
e-
5
41
I
Gimep-
11
0
-1
40
45
50
55
Cycle#
Figure 4.50 Large Step Change from Low Load to High Load
86
80v
60/0.
401
2
20;
01
-20'
C
200
600
400
Crank Angle
601
U
50
4
6-
40
~0
30'
201
10
01
-10
-20
1
C
200
400
600
Crank Angle
Figure 4.51 Details of Low Load to High Load Transient
87
Wave and LLNL kinetics simulations were conducted to clarify several phenomena observed in
the experimental data. Specifically, explanations for the following phenomena were investigated:
1.
Why did the first cycle, completely misfire
2.
Why did the second cycle show a combustion characteristic similar to that of SI mode
3.
What lead to CAI combustion in the third cycle.
Figure 4.52 illustrates the simulation results. The first cycle after transient misfired due to low
residual gas temperature (can not sustain CAI combustion) and high residual fraction (can not
initiate SI combustion). The burned gas residual gas mass fraction is about 44%, which is too
dilute for SI engine combustion. Hence, the first cycle completely misfire. The second CAI cycle
has a residual fraction of 55%, which consist of burned gas residual (from combustion in
cycle#O) and unburned gas residual (from cycle#1). Only 24% of the charge mass in the second
cycle was burned gas residuals. Therefore cycle#2 was a late burning SI engine. CAI combustion
was not achieved in cycle#2 due to low residual gas temperature (cycle#1, misfire). Cycle#3 is
an early burned CAI combustion with a lot of pressure oscillation due to the hot residual gas
produced from cycle#2 (late SI combustion).
0.
740
Steady State
2
U
700-
2
0.5
~0.45-
680
6
660-
0.4
640-
0.35-
620
*
Steady State
0.3-
0.25
600
580
-1
y
0.55
Burned Gas Res
0.2
0
1
2
Cycle#
2
01
3
Cycle#
Figure 4.52 Wave Modeling
When the results from Wave simulation programs were fed to LLNL kinetics mechanism and the
knock integral method, both methods estimated misfire for both cycle# L and cycle#2, which was
consistent with the experimental results.
88
4.9.3
Load modulation by Fuel Metering and Valve Timing Interpolation
The misfire phenomena when a large step change from low load to high load operation
was made prompted a question of how fast can one change the engine load. The simplest open
loop control strategy, one which is based on fuel metering and valve timing interpolation, was
designed and implemented to investigate the engine response during low load to high load
transient. The goal of the experiments was not to find an optimized control strategy within CAI
regime but rather to understand the physical phenomena and investigate the order of magnitude
of the time scale required for low load to high load transient under the simplest control strategy.
Figure 4.53 illustrates the implementation of this open loop control strategy with
modulation period of 60 cycles at 1500 rpm. In this case, the valve timings and fuel injection
amount were linearly interpolated from the high and low load operating settings. The high and
low load operating settings were the same as those in section 4.7.1 and 4.7.2. The figure
indicates that there was no misfire when the modulation period was set at 60 and 30 cycles
(Figure 4.54). But when the modulation period was reduced to 14 cycles, some of the cycles
started to misfire. Figure 4.55 and Figure 4.56 showed that altfidugh the actual misfiring cycle
was random, it always happened during the up transition (when the engine load was changed
toward higher load). This phenomenon was caused by the colder residual gas temperature
trapped from previous cycle when the engine load was changed from low load to high load. In
summary, this phenomenon is similar to the one described in section 4.7.2, only to a lesser
degree.
89
6
GIMEP
5
4
3
NiMEP
I
1
P1
Fue
.
mas
0
-1
0
100
50
150
Cycle no.
200
250
300
Figure 4.53 Modulation Period 60 cycles (4.8 s at 1500 rpm)
E
6
GIMEP
51
a)
4
0~
3
C.
E
L
wl
Cl
NIMEP
2
1
0
-1I '
0
50
100
150
Cycle no.
200
250
300
Figure 4.54 modulation period 30 cycles (2.4 s at 1500 rpm)
90
6
GIMEP
c,)
E 5
4
C.)
;3
E
NIMEP
Fuel mass x 10
2
I
-I
11
100
50
150
Cycle no.
oo
300
250
Figure 4.55 Modulation Period 14 cycles (1.12 s)
-
FumGIMEP'
I.
Fuel mass
C
I
-
x 10**
NIMEP
NI-
r
_-2
- 3'
170
175
180
185
190
195
200
Cycle no.
Figure 4.56 Details of failed transient when modulation period was 14 cycles
91
Chapter 5: Mode Transition Management between
Spark Ignition and CAI mode
5.1
Early work on SI to CAI mode transition
The early work on SI to CAI mode transition tried to identify the physical constraints and
evaluate proper mode transition strategy through conducting some experimental work. The
project was started in September 2001, when no technical paper addressing the mode transition
issue was available in the literature. Initial attempts to reach CAI combustion was conducted by
manually adjusting the cam timing since the VVT system was not available. The geometric
compression ratio of the engine was set at 16 to allow for high compression pressure and high
charge temperature. The engine is operated in Spark Ignition mode to provide hot residual gas
for CAI combustion by retarding the spark timing to 10 degrees ATDC. This late spark timing
setting was necessary to avoid knocking in SI mode. Since the valve timings could not be
changed dynamically (on-the-fly), the transition to CAI mode was achieved by changing the fuel
injection amount (with the throttle kept at WOT). Figure 5.1 and Figure 5.2 illustrates the details
of the transition. The valve timings were set at: jIVO 703 ATDCi, IVC 174 ATDCi, EVO 476
C,, -100
E
80
Fuel pulse width x 10
7 60~
C.
Peak pressur
0
Z
Cn
CO
40
20GIMENx10
50
100
Cycle#
150
200
Figure 5.1 Mode Transition by Fuel Metering
1500 RPM, 0=1 in SI, 0 = 0.7 in HCCI, Inlet Air Temperature 114 'C
92
ATDCi, EVC 672 ATDCi. These valve timings were based at the 1 mm valve lift point. The
transition is far from optimum. Starting from cycle# 24 the fuel injection amount was manually
reduced, which caused the GIMEP to become smaller (the engine was still operating in SI
mode). Because of the lean condition, the cycles prior to transition exhibited alternate cycles of
strong-bum/partial burn behavior. As matter of fact, cycle#130 (just before first HCCI cycle)
was almost a complete misfire (GIMEP close to zero). This initial experiment highlighted the
importance of managing the transition trajectory to avoid partial bum/misfire, which can affect
drivability and HC emission during a typical trip. It should also be noted that because the valve
timing can only be changed for a limited range (single-shaft-fixed-cam-lobe system) and that the
timing could not be changed dynamically, it was extremely difficult to optimize the transition.
Therefore all the subsequent efforts were focused on transitioning with a VVT system.
An electromechanical VVT system was later received through donation from Daimler
Chrysler Corporation. With the VVT system the mode transition can be achieved by changing
the valve timings, fuel metering, and throttle position. Figure 5.2 depicts early attempts to clarify
the phenomena encountered during mode transition (non-optimized).
The valve timing settings
60
30
50
20
10
40
0
30
330
360
390
420
450
(ATDCi)
No recompressionAnge
20104
20
0
2.4
10
2.5
2.6
Crank Angle / 10000
2.7
300
330
360
390
420
450
Crank Ange (ATDCi)
Figure 5.2 Non-synchronized mode transition
1500 RPM, MAP 0.8 bar, C.R. 14.2, Spark Timing 20 BTDCf (both before and after transition)
93
were: SI IVO 15, IVC 260, EVO 505, EVC 685 ATDCi and CAI IVO 84, IVC 208, EVO 505,
EVC 636 ATDCi. The throttle was fixed before and after the transition (the throttle was removed
for the subsequent sections). The spark was kept at 20 degrees BTDC even after the transition. In
this transition, the first intended CAI cycle was a knocking cycle. Pressure data analysis of the
first CAI cycle indicated unacceptable pressure oscillations. This knocking phenomenon was
caused by improper sequencing of the valve timing change (the valve timing change was not
synchronized, which result in random sequencing of the actual valve timing change). In this
particular case, the intake valve timing was advanced prior to advancing the exhaust valve
timing. Thus, the phenomena encountered during the first CAI cycle was that of spark ignited
combustion with a high compression ratio engine (14.2). Therefore the first intended CAI cycle
was, in fact, a knocking SI cycle. This phenomenon led to the conclusion that it is extremely
critical to control the sequencing of the valve timing change during a mode transition.
Therefore the controller was modified to allow for synchronized valve timing, fuel injection, and
throttle position change (the sequence of the variables changes can be consistently controlled) for
subsequent experiments.
The engine behavior during transition was further investigated after the engine controller
was modified to allow for synchronized change in valve timing, fuel injection, .nd throttle
position. Some of the issues are illustrated in Figure 5.3, which shows a non-robust transition. By
simultaneous switching of the valve timing, fuel injection, and throttle position, CAI combustion
was achieved in the first intended CAI cycle. However CAI combustion was not achieved for
cycle#2 and cycle#4. Thus a stable CAI operating point does not imply a robust transition at the
same speed and load. In this experiment the spark was kept on during the transition so that the
failed CAI cycles could still progress into a late spark ignited cycle to provide hot residual gas
for the following cycles. Figure 5.3 also illustrates the difficulty in maintaining stoichiometric
mixture during mode transition. This difficulty arose from the fact that the intake valve
closing timing, exhaust valve closing timing, and the throttle position were changed during the
transition. All these three parameters can influence the airflow into the cylinder. Figure 5.4
illustrates one of the first successful SI-CAI mode transition experiments. The valve timing
settings in this figure was different from those of Figure 5.3. All intended CAI cycles were
successful (without misfire or late spark ignited cycle). The throttle position was moved from 70
degree to zero degree (WOT) in three engine cycles. However, the difficulty in maintaining the
94
stoichiometric fuel air mixture was still presence although the excursion from stoichiometric was
much less in this case. Further experiments involving simultaneous opening the throttle and
changing the valve timing and fuel injection amount were conducted. It was found that it was
very difficult to achieve fast (in a single cycle) and smooth (constant Nimep) SI-CAI transition,
when the starting point was from idle/low load SI condition (MAP of 0.3-0.4 bar). This problem
was caused by:
1.
Difficulty in maintaining stoichiometric mixture.
Uncertainty in throttle position combined with the dynamic of the manifold pressure
during mode transition made it extremely difficult to predict the correct amount of fuel
injection required for maintaining constant air fuel ratio during mode transition.
2.
Low residual gas temperature.
The exhaust gas temperature of a spark ignition engine is lower in idle condition
compared to high load condition. Therefore the charge temperature was not hot enough to
initiate CAI combustion in the first CAI cycle.
140
120
100
Sl VHCCl
80-
80 -P
MAP(kpa)
-
60
-MAP
--
40
IVC 30
5 abdc
EVO 45 45 bbdcEVC 20 70 btdc
throttle
Partial SI
First CAI Cycle
IVO 20 80 atdc
(bar)
Int valve (V)
Exh valve
(bar)
*100 (V)
Exh Phi*1O0
700 00=WOT
angle
combustio
20-
-06
5o
100
1500
2 000
25'00
jb 0
Crank angle (deg)
350
4b _
(1500 rpm,
4500
5000
I50 BTDC
spark)
Figure 5.3 Non-robust SI-CAI Transition
95
120
100
80--
MAP(kpa)
60-40 -
--
IVO
IVC
EVO
EVC
P (bar)
Int valve (V)
Exh valve (V)
MAP*100 (bar)
Exh Phi*100
throttle
I st HCCI cycle
Sl
20
30
45
20
HCCI
95 atdc
10 abdc
45 bbdc
90 btd
700 00=W T
angle
20-
0-
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
Crank angle (deg)
Figure 5.4 R6bust SI-CAI'Transition with throttle movement
3.
Low peak cylinder pressure.
CAI combustion is governed by chemistry, which favors high pressure. The low
manifold pressure resulted in very low peak cylinder pressure, which induced longer
ignition delay period.
Some of these difficulties can be mitigated by controlling the engine load by using late/early
intake valve closing to control the engine load in spark ignition mode.
5.2
Late IVC as a Load Control Mechanism in SI mode
In a dual mode SI-CAI engine, the engine can always be able to be operated in SI mode.
The operating regime of SI mode encompasses the operating regime of CAI mode. One possible
practical implementation of this technology relies on the availability of a fully flexible variable
96
jN1!qR1R5q1' POP'91 R".1-
NOT,
411 1 5 wm!j I IN,1 .111. m 1 .11
Q I I R NLR
. 111 q -P
P
Pon 1"01.
IR
P.F
valve timing system to control the engine load. When such a system is available, it is very likely
that the engine will be run without mechanical throttle to minimize the pumping loss even when
it is operated in SI mode. Two strategies based on early intake valve closing and late intake valve
closing have been proposed in the literature [13]. Similar improvement in fuel economy over
regular gasoline SI engine has been reported for both strategies. This study focuses on the late
IVC strategy due to difficulty in controlling the VVT system with early IVC method. There is
not enough time for the intake valve to travel from a completely close position to completely
open and back to completely close position again with the early IVC strategy.
Figure 5.5 depicts the P-V diagram when the engine load is controlled with late IVC
strategy. The actual compression process occurs when the piston has already moved passed half
of the displacement volume. Using this strategy the operating regime in SI mode has been
mapped to find the MBT spark timing. The results are shown in Figure 5.6. The intake air was
heated to 120 'C and the throttle was removed from the intake system. The fuel air equivalence
ratio was kept at stoichiometric so that a conventional 3-way catalyst can still be used. Closing
the intake valve late effectively results in:
1. Reduction on the amount of air trapped in the cylinder.
2. Reduction of the effective geometric coinpression ratio
25
20
15
10
5
0
0
100
200
300
400
500
Volume (cc)
Figure 5.5 P-V diagram of throttleless SI engine with late IVC
IVO 10, IVC 655, EVO 495, EVC 715 ATDCi, Spark 30 'BTDC
97
8.00,
IVC266 ATDCi
A1TD
-IVC248
IVC280 ATO
-VC295
ATDO
____
IC305 ATMO
Later
I
___
"'N4
IVC
-40.00 -35.00 -30.00 -25.00 -20.00 -15.00 -10.00
-5.00
0.00
5.00
Spark (ATDC)
Figure 5.6 Spark Sweep for SI Engine with late IVC
1500 RPM, 120 'C Inlet Air Temperature, Geometric Compression Ratio 12.28
Therefore, late fVC leads to lower engine load. Lower engine load can *still be achieved by
closing the intake valve timing even later but was not attempted in the experiments due to the
limitation of CAI operating regime (the goal of this experiments were to provide initial/final
operating condition for constant load experiments). The highest operating point (knock-limited)
was lower than that of a typical spark ignition engine (up to 10 bar Nimep) due to the higher
geometric compression ratio (12.3) and the inlet air heating (120 'C) required to achieve a
reasonable range of operating regime for a CAI engine.
98
5.3
CAI to SI Mode transition.
5.3.1
CAI to SI Mode transition phenomena.
The transition from CAI to SI mode can be made without much difficulty due to lower
cycle-to-cycle coupling when the engine is operated in SI mode. Misfire is not a problem in
transitioning from CAI to SI mode. However, maintaining the engine load constant during the
transition requires some tuning. Figure 5.7 illustrates an example of CAI to SI transition when
the valve timing settings were changed directly into SI settings and the fuel injection amount was
not optimized correctly.
18
16
14
ta
12-
CAI
SI
010
Gimep
0
2.
n
WO
34
|102
First SI Cycle
C>
10
20
Cycle#
Nimep
30
40
Figure 5.7 Non-robust CAI to SI Mode transition
The first SI cycle was a weak cycle due to weak mixture strength (exhaust lambda sensor
indicated a phi of 0.8). Therefore a proper mode transition strategy needs to be devised to
maintain the load constant. It can also be seen from Figure 5.7 that the Nimep for the last CAI
cycle was larger than the Gimep. This is phenomena was caused by the fact that the first SI
cycles has no requirement for residual trapping as required in CAI cycle. In fact, no residual
should be trapped for the first SI cycle. The exhaust valve closing timing for the last CAI cycles
has to be set at the regular SI EVC timing, which creates a temporary positive work gain for the
last CAI cycle.
99
5.3.2
CAI to SI mode transition strategy.
The following strategy for CAI to SI transition is proposed for SI engine with late IVC strategy.
For the last CAI cycle:
-
Exhaust valve closing timing should be adjusted to the same value as the corresponding
SI exhaust valve closing timing.
-
Intake valve closing timing should still be maintained at CAI IVC timing.
Fuel injection should be adjusted to maintain stoichiometric mixture for the first SI cycle.
For operation of SI engine with late intake valve closing strategy, this means more fuel
need to be injected than steady state SI operating point (see discussion in the next
section)
For the first SI cycle:
-
The intake valve closing timing should be adjusted to regular SI with late IVC timing.
The spark timing should be set to the MBT timing.
All the subsequent cycles should be set at normal SI with late IVC settings.
5.3.3
CAI to SI mode transition demonstration.
Figure 5.8 illustrates the implementation of such a strategy. There was no significant drop
in Gimep or Nimep. The fuelling for the last CAI cycle was adjusted to maintain a stoichiometric
mixture for the SI cycle. The actual fuel injection amount was significantly higher
(approximately 23.5 mg/cycle) compared to steady state SI condition (approximately 15
mg/cycle). This additional fuel is required to compensate for the lack of fuel vapor in the intake
port when the engine is run in CAI mode. When the engine is run in steady state with late IVC
strategy in SI mode, the fuel that gets delivered into the cylinder consists of:
a.
The fuel vapor pushed back from the last cycle, plus
b.
The fuel vapor injected for the current cycle, minus
c.
The fuel vapor pushed out for the current cycle (due to late IVC timing).
100
. I- --- -_1
,
In steady state run, quantity (a) is equal to quantity (c) so the amount of fuel that gets burned is
equal to the amount of fuel injected for the current cycle. However, during CAI to SI mode
transition, quantity of (a) is much less (the intake valve timing for CAI cycle is close to BDC and
therefore there are little fuel vapor pushed back) than quantity (c). Hence, the fuel injected (b)
during mode CAI to SI mode transition needs to be increased substantially to compensate for (c).
25
SI
CAI
20
10b
1-
e,
10
Gmep
.
5
Nimep
0
0
10
20
Cycle#
30
40
Figure 5.8 Successful CAI to SI transition
Figure 5.9 illustrates the details of mode transition from CAI to SI mode. As can be seen
in Figure 5.7 and Figure 5.8, the last CAI cycle's Nimep was larger than its Gimep. This
phenomenon is caused by the fact that there is no residual gas trapping requirement for the last
CAI cycle (Figure 5.9). Some Ricardo Wave simulations were conducted to analyze the incylinder condition during the transition. The results are shown in Figure 5.10. It can be seen that
the residual mass fraction and the in-cylinder charge temperature are much lower in Spark
Ignition mode.
101
50
Last CAI cycle
40 ..............................
............
30
0
20
-----
- -
-------
- .--.-.----- . --------. No residual trapping .requirement
10
-
-S-.-.---
0 ....
-10
--..
360
0
720
1080
1440
1800
2160
Crank Angle (CA)
Figure 5.9 Details of CAI to SI mode transition
40
750
30
70a,
30-
650-
25
-
20
-550
15.
-
500-
10
-
600-
0
1
Cycle#
-
450
2
400
-1
0
1
2
Cycle#
Figure 5.10 Ricardo Wave Simulation for CAI-SI Mode Transition Results
102
5.4
SI to CAI mode transition
This section is devoted to clarify the phenomena encountered during SI to CAI mode
transition, devise a mode transition strategy, and implement that transition strategy. The premise
is to achieve fast and smooth transition from a SI operation point to a corresponding CAI
operation point at the same speed and load, which for our purpose, was specified by the Net
Indicated Mean Effective Pressure (NIMEP). The engine speed was kept constant with the
assumption that if mode transition can be achieved within 1 engine cycle, the inertia of the
system would keep the engine speed constant for a short period of time.
5.4.1
SI to CAI mode transition phenomena
Initial experiments to identify the phenomena during mode transition were conducted
with the valve timings and fuel injection settings directly changed from that of SI settings to CAI
settings. Because the transition was not optimized, the transitions were non-robust in most of
these cases.
5.4.1.1 Failed Transition
Figure 5.11 illustrates an example of a failed SI to CAI transition. While the first CAI cycles was
0
Fuel
6-
SI
4-
CAI
Gimep
0-
0
50
60
70
80
Cycle#
Figure 5.11 Failed SI-CAI mode transition
103
successful, the second and subsequent cycles misfired. The valve timing and the fuel injection
settings for the CAI cycles were taken from one of the steady state CAI valve timing settings
with minimum COV, which means that with a proper transition strategy, the CAI cycles can be
sustained. Figure 5.12 and Figure 5.13 showed the details of a failed SI to CAI transition.
80
/-1sl CAI cycle
70
60
Recompression
Combustion
50
40
-
A
.is
-
.
30
-j
20
10
0
-
-10
_
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
Crank Angle
Figure 5.12 Detailed of failed SI-CAI transition
When the transition was made with direct jump in the valve and fuel injection settings, it was
very common for the first CAI cycles to indicate knock-like, pressure oscillations due to the
very rapid pressure increase. It is also very common for the second cycle to misfire due to the
low thermal environment following the pressure oscillations in the first CAI cycles. Figure 5.14
shows that the ignition timing for the first CAI cycles was very advanced and there were
significant pressure oscillations for the first CAI cycles. The second CAI cycles misfired due to
the low residual gas temperature produced from the first CAI cycle. It is generally accepted that
the pressure oscillation in CAI engine can break the thermal boundary layer in the cylinder. This
phenomenon can promote heat transfer to cylinder walls. Therefore, the trapped residual gas
104
following such a pressure oscillation is not hot enough to promote auto-ignition. While the
second cycle failed to initiate CAI combustion, slightly positive Gimep was observed in figure
5.13. This phenomenon was caused by the fact that the spark was kept on at TDC after the
3
z
Girnep
'I
-,4
2
.1
Ist CAI
cycle
6
0 -1 I
F
Nimep
0
10
5
Cycle#
Figure 5.13 Gimep and Nimep of failed SI-CAI transition
3
8C
I-
-o 20
1
6C
0)
10
*A-
I-
0
4C
-1
/
2C
180
360
7:
540
Crank Angle (ATDCi)
50
9
8
40
1-
C
20
0)
-2
30
10
0
0
180
360
540
Crank Angle (ATDCi)
7'2
-10
C0 200k
4(0
600
Crank Angle (ATDCi)
Figure 5.14 Details of Failed Transition
105
transition. The residual mass fraction for the second CAI cycle was about 40%, which resulted in
a very slow flame propagation. It could also be inferred that the combustion for the second CAI
cycle was incomplete when exhaust valve was opened. Therefore combustion was detected in the
recompression process due to the high hydrocarbon level in following an incomplete
combustion. This combustion in the process resulted in hot residual gas which could be used to
initiate CAI cycle for cycle#3. However, the high pressure resulted from the combustion in the
recompression significantly alter the air flow when the intake valve was open. Significant back
flow occurred, which reduced the fresh charge mass fraction induced into the cylinder for
cycle#3. Therefore, the residual mass fraction was too high and the mixture was too rich to
initiate CAI combustion. Neither could the mixture initiate SI combustion. This resulted in
negative Gimep for the third cycle. The Nimep was larger than Gimep due to the combustion in
the recompression process from cycle#2. Subsequent cycles experienced abnormal combustion
due to erratic residual mass fraction, residual gas temperature, and in-cylinder fuel air
equivalence ratio that resulted from misfire in cycle#3.
5.4.1.2 Non-robust transition
The non-robust transition is the transition in which one or more cycles misfire before
settling into CAI mode. This is the most commonly encountered transition when no transition
strategy was employed (transition was made through direct jump in valve timing and fuel
injection settings). Figure 5.15 and Figure 5.16 illustrate an example of this kind of transition.
The first CAI cycle was a successful CAI cycles but the second CAI cycles failed to auto-ignite
near top dead center. Because the spark was kept on at piston TDC, the second CAI cycle was a
late burn SI cycle, which created hot residual gas for the third CAI cycles. The third CAI cycle
was an early CAI cycles and failed to produce residuals hot enough for cycle#4. The process
repeated itself until cycle#6 which produced residual gas at appropriate temperature for cycle#7.
Cycle#7 and the subsequent cycles were successful CAI cycles.
106
r%
Gimep
4
0)
S
3
CAI
2F
---- - ------
1
(
0
80
60
40
20
100
Cycle#
Figure 5.15 Non-robust SI-CAI Transition
60
603
50
40
co30
20
~10
4
5
1
20-
ox
I
-20
0
180
360
7?0
540
Crank Angle (ATDCi)
9
-I
60
0
-10
8
a
a
2
180
360
540
Crank Angie (ATDCi)
7
-
720
2C
2n
2
-0
180
360
540
720
Crank Angle (ATDCi)
Figure 5.16 Details of non-robust SI-CAI transition
107
Wave plus Wiebe function and LLNL kinetics simulations were conducted to analyze why the
second CAI cycle misfired. The results are shown in Figure 5.17. The simulation results
indicated that the engine misfired in the second CAI cycle due to the low in-cylinder
temperature. At this kind of temperature the LLNL kinetics mechanism also predicted an engine
misfire.
35
30
0
25
20
15-
101
0
2
Cycle#
680
0%
660
640
C
620
600
LLNL misfire
580
560
54cr
0
1
2
Cycle#
Figure 5.17 Wave and Kinetic Simulation Results for SI-CAI mode transition
108
In summary the phenomena during SI-CAI mode transition with direct valve timing and fuel
injection settings change can be described as follows:
-
Knowing the initial set of valve and fuel injection amount (for SI mode) and final set of
valve and fuel injection amount (for CAI) mode does not guarantee a successful mode
transition.
-
The sequence of the valve timing and fuel injection timing change were important. The
exhaust valve closing timing should be changed before the intake valve closing timing
was changed. Otherwise, the high effective compression ratio without any dilution level
can lead to severe knock in SI mode.
-
Non-robust transition was the most frequently encountered mode of transition when no
transition strategy was employed.
-
The ignition timing of the first CAI cycle after mode transition from SI mode is likely to
occur too advance and exhibits excessive pressure oscillation, which results in cold
residuals and often lead to misfire in the second CAI cycle.
-
Some of the non-robust transitions can eventually develop into a stable CAI combustion
when the spark was kept on. When the spark was kept at TDC, lAte SI cycles can develop
and provide hot residual for the subsequent cycles.
-
Spark timing may be used to assist transition when the residual fraction is not too high
(high load). Spark timing does not have any influence on mode transition at low load
limit, where high amount of residual gases are present.
-
Keeping the fuel air equivalence ratio constant during mode transition is a challenging
task. Direct jump in fuel injection settings resulted in fuel rich cycle for the first few CAI
cycles.
109
5.4.2
SI to CAI mode transition strategy.
The premise of this section is to develop a mode transition strategy that would allow for
smooth and fast transition between SI and CAI mode. The transition should be made when the
engine is already operated in CAI regime since it is extremely difficult to predict the driver
intention in real practice. The SI operating regime encompasses the CAI operating regime.
Therefore the engine can always be operated in SI mode.
When the engine is equipped with a fully flexible valve system, the mode transition can
be achieved by controlling the valve timings and the fuel injection amount. The following
general strategy is proposed. Prior to the transition, the engine should be operated at WOT with
load controlled with intake valve closing (early or late). This will eliminate the extensive
mapping required when the transient involve throttle opening. Because CAI combustion is
controlled by chemical kinetics, it is very important that the in-cylinder fuel air equivalence ratio
be precisely controlled. Opening the throttle will take several engine cycles (approximately 0.25
s) during which precise amount of fresh air induced into the cylinder is uncertain and the peak
cylinder pressure is not high enough. Therefore ensuring smooth mode transition with the throttle
movement over a wide range of operating point is a very challenging task. Arguably, when a
fully flexible valve train system is available, the engine will get better fuel economy with
throttleless operation even when it is operated in SI mode. Therefore this project assumes that the
engine is already operated with late IVC strategy in SI mode. The following strategy is proposed
for such an engine:
-
The last SI cycle (cycle 0):
a. The exhaust valve closing time is adjusted to CAI EVC timing to trapped hot residual
gas. The hot residual gas is necessary to promote auto-ignition in the first CAI cycle.
b. The IVC timing is still kept at IVC timing of an SI engine with late IVC strategy.
c. Spark was still on at MBT timing.
d. The fuel injection (to be delivered into the cylinder in the next cycle) should be adjusted
to maintain stoichiometric mixture for the first CAI cycles. This normally means that the
fuel injection needs to be reduced compared to steady state CAI fuel injection amount
(see discussion in the next section). When in doubt it is better to have a slightly lean
mixture compared to rich mixture due to greater tolerance in CAI mode for lean mixture
(see Chapter 4).
110
The first CAI cycle (cycle# 1)
-
The spark timing is moved to TDC to distinguish pure CAI cycle from spark assisted CAI
a.
cycle. Spark-assisted CAI may be used to help mode transition when fully flexible valve
timing is not available (when the transition can not be made in a single cycle). With
current geometric compression ratio (12.28), spark assisted CAI can only be used for
cases where residual mass fraction is lower than 35%.
The intake valve closing time should be adjusted between late IVC SI mode and steady
b.
state CAI mode to tune the ignition timing of the first CAI cycle. Without this
adjustment, the first CAI cycles tends to exhibit excessive pressure oscillation, which can
lead to misfire in the second CAI cycle. This IVC timing reduces the effective
compression ratio and therefore retards the auto-ignition timing.
Fuel injection (to be delivered in the second CAI cycle) is set to regular CAI fuel
c.
injection amount.
EVO timing is kept symmetric with EVC around TDC to minimize pumping loss
d.
-
The second CAI cycle (cycle#2)
a.
IVC timing is set at regular CAI IVC timing.
b.
All valve timings, fuel injection, and spark timing are set at regular CAI setting
condition.
The sensitivity of NIMEP history to valve timing programming in the SI to CAI mode
transition is shown in Figure 5.18. In this figure the EVC timing profile was fixed. The IVC
timing profile was varied as shown. There were substantial NIMIEP fluctuations in the first few
CAI cycles if the valve timing programming was not optimal. Although no misfire was detected
for the first CAI cycles, the engine could still misfire. In one case, the engine can misfire even
though the spark ignition was on, since the level of residual was too high for a SL cycle. The best
profile was found to be a transition in two cycles, with the valve timing went to an intermediate
ill
value in the first cycle and then to the final value in the second cycle. The implementation of
such a strategy is shown in figure 5.19.
300
2 50
200
0
800
--....-...-...-...-
2
4
6
8
10
4
6
8
10
6
4
Cycle Number
8
10
700 -.--.----
W
600
ia__-A_
0
2
0
2
10
I
Figure 5.18 Sensitivity of NIMEP history to valve programming in SI to CAI mode transition
112
120
100
80
;4x
d
S 60
P CAI
cycle
Pressure
Early EVC
40
20
0
-20
0
Early IVC
4000
2000
6000
8000
10000
Crank Angle (ATDCi)
Figure 5.19 An example of implementation of mode transition strategy
5.4.3
SI to CAI mode transition strategy implementation.
Figure 3.3 illustrates that during a FTP test, mode transition is likely to occur at an engine
speed lower than 1750 rpm. Two engine speeds, 1250 and 1750 rpm were selected to represent
the low and high load condition. Figure 5.20 and Figure 5.21 showed a robust transition. The
valve timings and fuel injection amount are shown on Figure 5.20. The EVC timing of the last SI
cycle was changed to 610 ATDC to trapped residual. The first CAI IVC timing was set at 217
ATDC to reduce the effective compression ratio (normal CAI operation IVC is at 182 ATDC).
The fuel injection amount was significantly reduced (5.5 mg instead of 13.4 mg) to maintain
stoichiometric mixture. The reason for this requirement of reduction in fuelling is as follows. In
the SI cycle, because of the late IVC timing, a substantial amount of the fuel vapor was pushed
back to the intake port. This fuel was re-inducted into the cylinder in the next cycle. Therefore,
the fuel vapor in a particular cycle comprised of the following:
a. the fuel vapor pushed back from the last cycle, plus
b. the fuel vapor injected for the current cycle, minus
c. the fuel vapor pushed out for the current cycle.
113
A1 4
Valve timing(o atdc exhaust)
EVO EVC
IVO Fuel(mg)
Cycle IVC
1%A
15
286
505
705
720
16
505
286
110
610
17
110 413A
505
610
A 2
18
110 / 13A
505
610
19,... 182
505
110 13A
610
2Z
1) 0.
CAI
Glimep
SI
8
CO
6-
IVC adjusted to
F
minimize knock
4
2
0
0
i;
Nimep
5
10
15
20
25
30
Cycle#
Figure 5.20 An example of robust SI-CAI transition
5
35
-I
Cycle
,- Cycle 2
Cycle 3
Cycle 4
Cycle 5
--- d Cycle 6
5C
45
4C
a)
35
a)
I-
Fuelling is
reduced to
avoid rich
mixture
Cycle 7
I st
Cycle 8
Cycle 9
3C
F.
25
2
350
360
370
390
380
Crank Angle (ATDCi)
400
Figure 5.21 Details of robust transition with slight pressure oscillations
114
In steady state operation, the amount in (a) equaled that in (c); the net effect was that the fuel
vapor in the trapped charge was equal to the amount of fuel injected. However, in mode
transition, the IVC timing remained late in the last SI cycle so that the amount (a) was there for
the first CAI cycle. Then since the IVC was changed to early for this first cycle, (c) was no
longer there. Thus the fuel vapor in the charge for the first CAI cycle comprised the amount of
(b)- the first cycle injected fuel, and the extra enrichment (a) from the last cycle since there was
no part (c) to cancel part (a). To eliminate this phenomenon, the fuel injection for the first CAI
cycle needs to be substantially reduced as seen in Figure 5.20. There was a sudden dip in Nimep
for the last SI cycle due to the residual trapping requirement for CAI cycle. This dip can usually
be recovered when transitioning back into SI mode. There was also a slight dip in Gimep for the
first CAI cycle. This phenomenon was due to the reduction of the amount of fresh air trapped
inside the cylinder (slightly retarded IVC) for the first CAI cycle. The ignition timing stabilized
after 2 engine cycles. There was no significant difference in the ignition timing after cycle#3 in
CAI mode. However, there were slight pressure oscillations during the first 2 CAI cycles but
they could be improved with better tuning of first cycle IVC timing and fuel metering.
16 atValve
16
timing(o atdc exhaust)
3
14
Cycle IVC
EVO
EVC
16
278
606
706
720
16
278
505
624
105
505
624
106 f16.2
50
603
606
624
624
17
12
18
19-...
10
8
e
Gimep
6
gi
IVC adjusted to
minimize knock
*Nimep
0
10
(.5
106 '16.2
106 15.2
/-
Fuelling is
reduced to
-A&A
Af
AAAA*.A&Aj&&&&SJ%~wJAAAavoid
A af
rich
mixture
4.
52
203
IVO Fuel(mg)
1.62
20
30
40
Cycle#
Figure 5.22 An example of robust SI-CAI transition
115
Cycle 1
Cycle 2
-45--
Cycle 3
Cycle 4
Cycle 5
-Cycle 6
-- - Cycle 7
Cycle 8
Cycle 9
--
40
-
35
30
25
20
400
390
380
370
360
Figure 5.23 Detailed pressure trace following a mode transition from SI to CAI
750
35
3-'700
25
650
20
600
15
550
50Q
10H
-1
1
0
Cycle#
2
-:1
1
0
2
Cycle#
Figure 5.24 Wave simulations results for SI-CAI mode transition
116
Figure 5.22 and 5.23 illustrates an optimized implementation of the transition strategy. There
were no pressure oscillations detected in the pressure trace. The ignition timing for the first and
second cycles were similar to those of the subsequent cycles. The transition to CAI mode was
made within 1 engine cycle. Wave simulation results for analyzing in-cylinder condition are
shown in Figure 5.24. Comparing the first and second CAI cycle, it can be seen that while the
first CAI cycle has a smaller residual fraction, the compression temperature is almost the same
for both cases. That is due to the fact that the exhaust gas temperature in SI mode is hotter than
the exhaust gas temperature in CAI mode. Figure 5.25, 5.26, 5.27, and 5.28 illustrate the
implementation of the transition strategy at 1250 rpm low load, 1250 rpm high load, 1750 rpm
low load, and 1750 rpm high load.
S
1dA-
12
10
S
ihIImIHhEhi~IbINIIHIuIHhuIHWI*
Ginep-
8
4-
6
~0
4
* O"W"
eng_::IF
S
0
a
10
t
Nirnep
- MENNONNNNA
- - --20
30
.
40
Cycle#
Figure 5.25 SI-CAI Transition, Low load 1250 RPM
117
I.
216
-
, -
.3
'I
14
12
.1.
.A.
Gimep
88
,~6
A-
4
e 2
}1im
-0
0
10
20
30
40
Cycle#
xl4rm
Figure 5.26 SI-CAI Transition, High Load, 1250 RPM
12
-qqpp
-
10
t14#
Gimep.
61
4
2
Nimept
0> C0
10
30
20
40
Cycle#
Figure 5.27 SI-CAI Transition, Low Load, 1750 RPM
118
Em, I'MR1,93 TRIR-
1-1-mmo,"INWIP-A IN
-1111PRII
IWWIWIPI III IIIIII OR RIPM."I
16
C)
14F
12
U-
.a
t10i
Gimep I
Reim e w"
4
-)
t
2
Nimep
0'
0
10
20
30
a
40
Cycle#
Figure 5.28 SI-CAI Transition, High Load 1750 RPM
119
Chapter 6: Summary and Conclusion
Experiment and modeling have been performed to identify the constraints, document the
phenomena, and devise a transient strategy for a dual mode SI/CAI engine. The content of the
project can be divided into two themes. The phenomena encountered within CAI regime and the
phenomena encountered during SI-CA
mode transition. The phenomena encountered within
CAI regime can be further divided into steady state phenomena and transient phenomena.
6.1
Within CAI regime
6.1.1
Steady state phenomena
The following phenomena were observed during steady state operation within CAI regime:
1
Fuel air equivalence ratio influence.
With the valve timings fixed (no throttle), the fuel injection amount was varied to
achieve fuel air equivalence ratio of 0.8 ~ 1.2. Starting from the stoichiometric
mixture, the engine has more limited tolerance to fuel rich condition. The engine
started to misfire when the fuel air mixture was set at 1.1 or richer. On the lean side,
the engine did not start to misfire until fuel air equivalence ratio was leaner than 0.8.
While the actual rich and lean limit is influenced by many other factors (valve
timings, geometric compression ratio, fuel properties, etc.), the tolerance to the fuel
rich condition is generally very narrow. This phenomenon is due to the compounding
effect of lower exhaust gas temperature and lower specific heat ratio for fuel rich
mixture. Therefore it is important to keep the fuel air mixture from going rich during
transient / mode transition operation.
2.
Pumping loss in CAI mode.
Even with WOT operation, the pumping loss in CAI mode is not negligible. The pmep
(using the conventional 360-degree integration method) of the engine ranges from 0.3
~ 0.5 bar. Analyzing this pumping loop closer, this pumping loss can further be
divided into two parts. The first part is due to engine breathing (inducing/pushing the
charge into/out of the cylinder) and the second part is due to heat loss in the
120
recompression process. The heat loss in the recompression process can be reduced by
using the residual reinduction method but this method will result in lower residual gas
temperature.
3.
Combustion phasing.
When the engine is set up for optimum combustion phasing (for Gimep), the Gimep
values are not sensitive to the fluctuation of the 50% fuel mass fraction burned
location. This phenomenon is similar to MBT spark timing phenomenon in spark
ignition engine. The indicated work output is equal to total heat release minus total
heat loss minus sensible energy difference between EVO and IVC. An early
combustion cycle has a higher heat loss but lower exhaust internal energy while the
reverse is true for a late combustion cycle.
4.
Spark Assisted Controlled Auto Ignition
The combustion in CAI mode can be influenced by spark when the residual mass
fraction is not high enough. In this study, the spark timing was found to influence the
onset of combustion when the residual mass fraction was lower than 35% (high
engine load). When the engine is operated at low load (with residual fraction higher
than 65%), the spark timing had no influence on the combustion timing.
5.
Oxidation in the recompression process.
Small amount of oxidation during recompression was detected following a late
combustion cycle. This oxidation was caused by the presence of unburned or partially
oxidized mixture at the high temperature encountered in the recompression process
(higher than 1100 K).
6.1.2 Transient Phenomena
The following phenomena were observed during transient operation within CAI regime:
I.
For small perturbation in fuel injection amount, intake valve timing, and exhaust
valve timing, the ignition timing can stabilize within 3~4 engine cycles at 1500 rpm.
This time scale correspond to the setting time for in-cylinder residual gas mass
fraction and in-cylinder charge temperature.
121
2.
For small perturbation, the effects of changing the controller variables are reversible.
The effects of retarding the IVC, EVC, EVO timing can be reversed by advancing the
parameters.
3.
For very large step change in load, thermal stabilization of the cylinder wall becomes
more prominent. The time scale required can be longer than 5 s to reach a new steady
state condition in this case.
4.
For a very large step change, the engine load can be changed from high load to low
load within a single cycle without engine misfire but the reverse is not true. When the
step is too large, the engine misfired during low load to high load transition. The
residual gas temperature at low load condition was much lower compared to residual
gas temperature at high load condition.
5.
A simple open loop controller (non optimized) was designed to investigate the time
scale required for low load to high load transition without misfire. The controller was
based on linear interpolation of fuel injection amount and valve timing settings.
Under this control scheme, the minimum time scale required for successful low load
to high load transient was about 1 s (modulation period 20 cycles at 1500 rpm). It is
not the focus of this study to design an optimized controller for CAI operation. An
example of such a controller can be found in another doctoral thesis ("Closed-Loop,
Variable-Valve-Timing Control of a Controlled-Auto-Ignition Engine", Dr.Jeff
Matthews, MIT Mechanical Engineering Department, September 2004)
6.2
SI-CAI Mode Transition
6.2.1
Transition from SI mode to CAI mode
The following constraints have been identified:
I.
During a Federal Test Procedure, the engine can be run in CAI mode for about 40%
of the time (the number can reach 70% for EPA high way test). About 40~50% of the
operating points within CAI regime involve mode transition. Therefore it is important
to manage mode transition properly.
2.
It is extremely difficult to predict the driver intention during real driving cycle.
Therefore the transition into CAI mode should be made after the engine enters the
CAI operating boundary.
122
3.
A lot of the mode transitions occur during low load condition, therefore the number
of transition required during a FTP test highly depends on the lower CAI operating
boundary.
4.
The throttle response for a regular SI engine was not fast enough to ensure smooth
and fast transition within one engine cycle. Therefore it is preferable for the engine
load to be controlled by intake valve timing closing even when it is operated in SI
mode.
The following phenomena were observed for SI to CAI mode transition with direct jump in
fuel injection and valve timing setting:
1.
The sequence of the valve timing and fuel injection settings needs to be synchronized
to avoid abnormal combustion during mode transition.
2.
Mode transition from SI to CAI mode can be categorized into three different types.
Failed transition, non-robust transition, and robust /successful transition. In a failed
transition the engine failed to develop into stable CAI mode. In a non-robust
transition, there may be one or more misfiring cycle before the engine stabilize into a
stable CAI mode. In a robust transition, every intended CAI cycles are successful
without excessive pressure oscillations and excursion in engine load.
3.
The non-robust transition is the most commonly encountered type of SI to CAI mode
transition due to excessive pressure oscillation in the first CAI cycle (leads to colder
mixture) and to the difficulty in maintaining acceptable fuel air equivalence ratio
during mode transition.
4.
The first and second cycle after mode transition exhibit different ignition timing
compared to the steady state value. It takes 3-8 cycles for the ignition timing to
stabilize. This phenomenon corresponds to the settling time of residual mass fraction
and in-cylinder charge temperature.
5.
The last SI cycle exhibits a dip in Nimep value due to the residual gas trapping
requirement for CAI cycle. This accounting dip is recovered during CAI-SI transition.
The following strategy was proposed and implemented during SI-CAI mode transition:
1.
The valve timing and fuel injection amount should be changed in a synchronized manner.
The sequence of the valve timing change should be EVC-IVO-IVC. The fuel injection
into the intake port should be made prior to mode transition to CAI mode.
123
2.
The IVC timing for the first CAI cycle should be set to an intermediate value between the
steady state CAI IVC timing and steady state SI IVC timing to reduce the pressure
oscillations. The residual gas trapped from SI cycles is normally hotter than the residual
gas trapped during CAI combustion
3.
The EVC timing for the last SI cycle should be set to the steady state CAI EVC timing to
trap residual gas. The EVC timing has two opposing effects on ignition timing; retarding
the ignition timing (dilution) and advancing the ignition timing (higher total charge
temperature)
4.
The fuelling for the first CAI cycle needs to be significantly reduced to maintain
stoichiometric mixture and avoid rich mixture during SI to CAI mode transition. When
the engine is operated in steady state, the amount of fuel injected for a particular cycle is
equal to the amount of fuel that get burned for that cycle. However this is not true during
mode transition since almost no fuel vapor is pushed back in the intake stroke (IVC
timing for CAI mode is close to BDC while IVC timing for SI mode is well after BDC).
Therefore, the fuel that gets burned in the first CAI cycle is the fuel injected for that
particular cycle plus fuel vapor from previous SI cycle. The addition of the fuel vapor
from the previous SI cycle makes it necessary to reduce the amount of fuelling for the
first CAI cycle to maintain constant fuel air ratio.
5.
A controller based on this strategy has been designed and SI to CAI mode transitions
have been demonstrated to achieve fast and smooth transition at four different speed and
load points representing low and high load and low and high speed.
6.2.2
Transition from CAI mode to SI mode.
The phenomena encountered during CAI to SI mode transition can be described as follow:
1.
It is easy to maintain combustion in SI mode since there is weaker cycle-to-cycle
coupling in Spark Ignition mode.
2.
The sequence of the valve timing change is important. The exhaust valve timing should
be changed before the IVC timing. Otherwise, the large amount of residual gas will lead
to weak spark ignition cycle.
124
3.
Maintaining the fuel air equivalence ratio constant is a challenging task. Direct jump in
fuel injection setting often results in lean first SI cycle, which lead to a weak SI cycle or
misfire.
4.
The last CAI cycle's Nimep is higher than its Gimep. This phenomenon is due to the fact
that the EVC timing was changed to a much later timing since there is no residual
trapping requirement for the first SI cycle.
The following strategy is proposed and demonstrated:
1.
The valve timing change should be synchronized. The order in which the valve timing are
changed should be EVC-IVO-IVC. The EVC timing for the last CAI cycle should be set
at steady state SI EVC timing.
2.
The fuel injection amount for the first SI cycle should be significantly increased to
maintain constant fuel air equivalence ratio. This increment in fuel injection amount is
necessary to compensate for the fuel vapor that is pushed back in the first SI cycle due to
late IVC timing.
3.
A controller based on this strategy has been designed and implemented to achieve smooth
and fast CAI to SI mode transition. No significant dip is *observed in Gimep of the first SI
cycle.
125
Appendix
Two-zone Heat Release Mode
TWO ZONE HEAT RELEASE MODEL
Integrating the unsteady energy equation (first law thermodynamics for a closed system,
a.k.a., no mass flux across the system boundary) as applied to the charge from a
reference point * where there is negligible heat release, to crank angle 0 (before the
exhaust valve open), the energy equation can be described as:
0
IV
[m(1- x)eu +mxeb]-m*e = -(QL +pd )d6
do
.
(Al) (x is the burned gas mass fraction)
Equation Al, can be rearrange as follows:
1 0
dV
(QL+ p
) do+(eu - eu)
m .
do
(eu - eb
(A2)
QL: heat loss, which can be calculated from Woshni correlation
P : Pressure is know from experimental data
V: Cylinder volume (known)eu and eb :specific internal energy of the unburned and burned gas
m: guess initial value which need to be validated from the Wave output
Isentropic process for the unburned gas:
Y-1
Tu = Turef (-)
(A3)
Pref
Since the unburned gas composition is assumed to be frozen, the internal energy (eu)
can be calculated at each time step.
From the volume constraint:
PV~lR
V=(-
x)RT+xRT,
(A4)
The are two unknown parameters: eb and Tb which can be calculated by simultaneously
solving eqn A2 and A4 at each time step. The burned gas composition is assumed to be
in chemical equilibrium at each time step at Tb and P. The burned gas species
considered are: H2, 02, H20, CO, C02, OH, H, 0, N2 , N, NO, NO 2, CH 4.
126
References
1.
Onishi,S., Hong Jo, S., Shoda, K., Do Jo, P. and Kato, S., "Active Thermo-Atmosphere
Combustion (ATAC) - A New Combustion Process for Internal Combustion Engine,",
SAE Technical Paper 790501 (1979).
2.
Noguchi, M., Tanaka,Y., Tanaka, T., and Takeuchi, Y.,"A Study on Gasoline Engine
Combustion by Observation of Intermediate Reactive Products during Combustion,",
SAE Technical Paper 790840 (1979).
3.
Najt, P., and Foster,D., "Compression-Ignited Homogeneous Charge Combustion", SAE
Technical Paper 830264
4.
Aoyama, T., Hattori, Y., Mizuta,J., Sato,Y.,"An experimental Study on Premixed Charge
Compression Ignition Gasoline Engine", SAE Technical Paper 960081
5.
Thring, R., H., "Homogeneous Charge Compression Ignition (HCCI) Engines,", SAE
Technical Paper 892068
6.
Takeda, Y., Keiichi,N., and Keiichi,N.," Emission Characteristics of Premixed 'Lean
Diesel Combustion with Extremely Early Staged Fuel Injection ", SAE Technical Paper
961163
7.
Lavy,J., Dabadie,J., AngelbergerC., Duret,P.,Willand, J., Juretzka, A.,Schaflein, J.,
MaT.,
Lendresse,Y. ,SatreA.,Schulz,C.,
KramerH.,
ZhaoH.,and
Damiano,
L.,"Innovative Ultra-low NOx Controlled Auto-Ignition Combustion Process for
Gasoline Engines: the 4-SPACE Project", SAE Technical Paper 2000-01-1837
8.
Yang,J., Culp,T., and KenneyT.,"Development of a Gasoline Engine System Using
HCCI Technology - The Concept and the Test Results", SAE Technical Paper 2002-012832
9.
LawD., Allen, J.Kemp, D. and Williams, P., " 4-Stroke Active Combustion (Controlled
Auto Ignition) Investigation Using a Single Cylinder Engine with Lotus Active Valve
Train
(AVT)", International
conference
on 21 " century emissions technology,
C588/006/2000, ImechE, 2000
127
10.
Koopmans, L., Strom,H., Lundren,S., Backlund,O., and Denbratt, I.,"Demonstrating a SIHCCI-SI Mode Change on a Volvo 5-Cylinder Electronic Valve Control Engine", SAE
Technical Paper 2003-01-0753
11.
Zhao,H., LiJ., Ma,T., and Ladommatos, N., "Performance and Analysis of a 4-stroke
Multi-cylinder Gasoline Engine with CAI Combustion", SAE Technical Paper# 2002-01 0420
12.
Shelby,M., Stein,R., and Warren,C.,"A New Analysis Method for Accurate Accounting
of SI Engine Pumping and Indicated Work", SAE Technical Paper# 2004-01-1262
13.
PodnarD.J., and Kubesh,J.T.," Characteristics of a Small Engine Equipped with an
Electromagnetic Valve Actuation System ", SAE Technical Paper 981908
14.
Office of Transportation Technologies, "Homogeneous Charge Compression Ignition
(HCCI) Technology - A Report to the U.S. Congress", April 2001
15.
Fuerhapter, A., Unger,E., Piock,W.F., and Fraidl,G.K., "The new AVL CSI Engine HCCI Operation on a Multi Cylinder Gasoline Engine", SAE Technical Paper# 2004-010551
16.
Bradley, D.,Morley,C., and Walmsley,H.L.,"Relevance of Research and Motor Octane
Numbers to the Predictian of Engine Autoignition", SAE Technical Paper 2004-01-1970
17.
Ohyama, Y.,"Simultaneous Control of Air/Fuel Ratio and Intake, Exhaust Valve Timing
for HCCI Operation", SAE Technical Paper 2003-01-1084
18.
Wolters, Y., Salber, W., Geiger, J., Duesmann, M., and Dilthey,J., "Controlled Auto
Ignition Combustion Process with an Electromechanical Valve Train", SAE Technical
Paper# 2003-01-0032
19.
Agrell, F., Angstrom, H.E., Eriksson, B., Wikander, J., and Linderyd, J., "Integrated
Simulation and Engine Test of Closed Loop HCCI Control by Aid of Variable Valve
Timings", SAE Technical Paper# 2003-01-0748
20.
Zhao,H., Peng, Z., Williams, J., Ladommatos, N., "Understanding the Effects of
Recycled Burnt Gases on the Controlled Autoignition (CAI) Combustion in Four-Stroke
Gasoline Engines", SAE Technical Paper# 2001-01-3607
21.
Koopmans, L., Denbratt, I., "A Four Stroke Camless Engine, Operated in Homogeneous
Charge Compression Ignition Mode with Commercial Gasoline", SAE Technical Paper#
2001-01-3610
128
22.
Koopmans, L., Backlund, 0., Denbratt, I., "Cycle to Cycle Variations: Their Influence on
Cycle Resolved Gas Temperature and Unburned Hydrocarbons from a Camless Gasoline
Compression Ignition Engine", SAE Technical Paper# 2002-01-0110
23.
Law,D., Allen, J., Chen, R., "On the Mechanism of Controlled Auto Ignition", SAE
Technical Paper# 2002-01-0421
24.
Law,D., Kemp, D., Allen, J., Kirkpatrick, G., and Copland, T., "Controlled Combustion
in an IC-Engine with a Fully Variable Valve Train", SAE Technical Paper# 2000-010251
25.
Chen, R., Turner, J., Blundell, D., "The Transition between Controlled Auto Ignition and
Spark Ignition", SAE Technical Paper# 2004-01-0939
26.
Urata, Y., Awasaka, M., Takanashi, J., Kakinuma, T., Hakozaki, T., and Umemoto,
A.,"A Study of Gasoline-Fuelled HCCI Engine Equipped with an Electromagnetic Valve
Train", SAE Technical Paper# 2004-01-1898
27.
Xu, H., Rudolph, S., Liu, Z., Wallace, S., Richardson, S., Wyszynski, M., Megaritis, A.,
"An Investigation into the Operating Mode Transitions of a Homogeneous Charge
Compression Ignition Engine Using EGR Trapping", SAE Technical Paper# 2004-011911
28.
Kumano, K., and Iida, N., "Analysis of the Effect of Charge Inhomogeneity on HCCI
Combustion by Chemiluminescence Measurement", SAE Technical Paper# 2004-011902
29.
Tsurushima, T., Kunishima, E., Asaumi, Y., Aoyagi, Y., Enomoto, Y., "The Effect of
Knock on Heat Loss in Homogeneous Charge Compression Ignition Engines", SAE
Technical Paper# 2002-01-0108
30.
Heywood, J.B., Internal Combustion Engine Fundamentals, McGraw-Hill Book Co.,
1988
31.
Milovanovic, N., Chen, R., and Turner, J., "Influence of the Variable Valve Timing
Strategy on the Control of a Homogeneous Charge Compression (HCCI) Engine", SAE
Technical Paper# 2004-01-1899
32.
Christensen, M., HCCI Combustion - Engine Operation and Emission Characteristics,
Doctoral Thesis, Division of Combustion Engines, Department of Heat and Power
Engineering, Lund Institute of Technology
129
33.
Haraldsson, G., Tunestal, P., Johansson, B., Hyvonen, J., "HCCI Combustion Phasing
with Closed-Loop Combustion Control Using Variable Compression Ratio in a Multi
Cylinder Engine", SAE Technical Paper# 2003-01-1830
34.
Agrell, F.,
Angstrom, H.E, Eriksson, B., Wikander, J., Linderyd, J.,"Integrated
Simulation and Engine Test of Closed Loop HCCI Control by Aid of Variable Valve
Timings", SAE Technical Paper# 2003-01-0748
35.
Ogink, R., Golovitchev,V., "Gasoline HCCI Modeling: An Engine Cycle Simulation
Code with a Multi-Zone Combustion Model", SAE Technical Paper# 2002-01-1745
36.
Kaahaaina, N.B., Simon, A.J., Caton, P.A, Edwards, C.F., "Use of Dynamic Valving to
Achieve Residual Affected Combustion", SAE Technical Paper# 2001-01-0549
37.
Stanglmaier, R.H., and Roberts, C.E., "Homogeneous Charge Compression Ignition
(HCCI): Benefits, Compromises, and Future Engine Applications ", SAE Technical
Paper# 1999-01-3682
38.
Aceves, S.M., Flowers, D.L, Westbrook, C.K., Smith, J.R., Pitz, W., Dibble, R.,
Christensen, M., Johansson, B., "A multi-zone model for prediction of HCCI combustion
and emissions", SAE Technical Paper# 2000-01-0327
39.
Urushihara, T., Hiraya, K., Kakuhou, A., Itoh, T., "Expansion of'HCCI Operating Region
by the Combination of Direct Fuel Injection, Negative Valve Overlap and Internal Fuel
Reformation", SAE Technical Paper# 2003-01-0749
130