Você está na página 1de 244

A NUMERICAL INVESTIGATION OF A

TWO-STROKE POPPET-VALVED
DIESEL ENGINE CONCEPT



Philip Robert Teakle BE MEngSc



Tribology and Materials Technology Group
Mechanical, Manufacturing and Medical Engineering
Queensland University of Technology



Submitted for the degree of Doctor of Philosophy
2004



v
Keywords
Two-stroke, poppet valves, KIVA, thermodynamic modelling, zero-dimensional
modelling, multidimensional modelling, engine modelling, scavenging models,
scavenging simulations.

Executive Summary
Two-stroke poppet-valved engines may combine the high power density of two -
stroke engines and the low emissions of poppet-valved engines. A two-stroke diesel
engine can generate the same power as a four-stroke engine of the same size, but at
higher (leaner) air/fuel ratios. Diesel combustion at high air/fuel ratios generally
means hydrocarbons, soot and carbon monoxide are oxidised more completely to
water and carbon dioxide in the cylinder, and the opportunity to increase the rate of
exhaust gas recirculation should reduce the formation of nitrogen oxides (NOx). The
concept is being explored as a means of economically modifying diesel engines to
make them cleaner and/or more powerful.

This study details the application of two computational models to this problem. The
first model is a relatively simple thermodynamic model created by the author capable
of rapidly estimating the behaviour of entire engine systems. It was used to estimate
near-optimum engine system parameters at single engine operating points and over a
six-mode engine cycle. The second model is a detailed CFD model called KIVA-
ERC. It is a hybrid of the KIVA engine modelling package developed at the Los
Alamos National Laboratory and combustion and emissions subroutines developed at
the University of Wisconsin-Madison Engine Research Center. It was used for
detailed scavenging and combustion simulations and to provide estimates of
emissions levels. Both models were calibrated and validated for four-stroke cycle
operation using experimental data. The thermodynamic model was used to provide
initial and boundary conditions to the KIVA-ERC model. Conversely, the
combustion simulations were used to adjust zero-dimensional combustion
correlations when experimental data was not available.

vi
Scavenging simulations were performed with shrouded and unshrouded intake
valves. A new two-zone scavenging model was proposed and validated using
multidimensional scavenging simulations.

A method for predicting the behaviour of the two-stroke engine system based on
four-stroke data has been proposed. The results using this method indicate that a
four-stroke diesel engine with minor modifications can be converted to a two-stroke
cycle and achieve substantially the same fuel efficiency as the original engine.
However, emissions levels can not be predicted accurately without experimental data
from a physical prototype. It is therefore recommended that such a prototype be
constructed, based on design parameters obtained from the numerical models used in
this study.
vii
Table of Contents
Keywords .............................................................................................................................................. v
Executive Summary ............................................................................................................................. v
Table of Contents ............................................................................................................................... vii
List of Figures....................................................................................................................................... x
List of Tables ..................................................................................................................................... xvi
Nomenclature .................................................................................................................................... xix
Statement of Original Authorship .................................................................................................... xx
Acknowledgements............................................................................................................................ xxi
Chapter 1 - Introduction...................................................................................................................... 1
1.1 BACKGROUND...................................................................................................................... 1
1.1.1 Two-stroke poppet-valved engines....................................................................................... 1
1.1.2 Outline of previous investigations ....................................................................................... 3
1.2 ROTEC ENGINE CONCEPT...................................................................................................... 4
1.3 OBJECTIVES ......................................................................................................................... 7
1.4 OUTLINE OF STUDY ............................................................................................................. 9
1.5 ORIGINAL CONTRIBUTIONS................................................................................................ 10
Chapter 2 - Literature Review.......................................................................................................... 11
2.1 INTRODUCTION .................................................................................................................. 11
2.2 OVERVIEW OF TWO-STROKE POPPET-VALVED ENGINE STUDIES ......................................... 11
2.2.1 Motives for study ............................................................................................................... 11
2.2.2 Description of prototypes/models ...................................................................................... 14
2.2.3 Simulation techniques........................................................................................................ 27
2.2.4 U-loop scavenging............................................................................................................. 29
2.2.5 Prototype performance ...................................................................................................... 36
2.2.6 Two-Stroke Poppet-Valved Diesel Emissions.................................................................... 41
2.2.7 Summary............................................................................................................................ 41
2.3 OVERVIEW OF DIESEL ENGINE EMISSIONS .......................................................................... 42
2.3.1 Diesel emissions formation................................................................................................ 42
2.3.2 Emissions regulations........................................................................................................ 44
2.3.3 Emissions control strategies .............................................................................................. 46
2.4 SCAVENGING ..................................................................................................................... 49
2.4.1 Fundamentals .................................................................................................................... 49
2.4.2 Scavenge pumps................................................................................................................. 50
2.5 ENGINE MODELLING........................................................................................................... 56
2.5.1 Thermodynamic modelling ................................................................................................ 56
viii
2.5.2 One-dimensional modelling................................................................................................57
2.5.3 Multidimensional modelling...............................................................................................57
2.6 DISCUSSION........................................................................................................................58
Chapter 3 - Thermodynamic model development............................................................................60
3.1 REQUIREMENTS ..................................................................................................................60
3.2 DESCRIPTION......................................................................................................................62
3.2.1 Basic assumptions ..............................................................................................................62
3.2.2 Numerical solver ................................................................................................................66
3.2.3 Gas properties ....................................................................................................................68
3.2.4 Heat transfer.......................................................................................................................74
3.2.5 Gas flow through valves and orifices .................................................................................75
3.2.6 Scavenging..........................................................................................................................77
3.2.7 Combustion.........................................................................................................................79
3.2.8 Turbocharging..................................................................................................................101
3.2.9 Charge air cooling ...........................................................................................................104
3.2.10 Mechanical friction.........................................................................................................107
3.2.11 Automated parametric investigation...............................................................................108
3.3 VALIDATION.....................................................................................................................110
3.3.1 Caterpillar SCOTE data...................................................................................................110
3.3.2 Caterpillar 3406E data.....................................................................................................117
3.4 DISCUSSION......................................................................................................................120
Chapter 4 - KIVA-ERC multidimensional model adaptation.......................................................123
4.1.1 KIVA package overview....................................................................................................123
4.2 ERC SPRAY AND COMBUSTION MODEL LIBRARY............................................................124
4.2.1 Turbulence........................................................................................................................125
4.2.2 Heat transfer.....................................................................................................................125
4.2.3 Atomisation and drop drag...............................................................................................126
4.2.4 Fuel/wall impingement .....................................................................................................129
4.2.5 Ignition .............................................................................................................................130
4.2.6 Combustion.......................................................................................................................130
4.2.7 NO
x
formation...................................................................................................................132
4.2.8 Soot...................................................................................................................................132
4.2.9 Computational meshes......................................................................................................134
4.3 VALIDATION.....................................................................................................................136
4.4 DISCUSSION......................................................................................................................142
Chapter 5 - Results............................................................................................................................144
5.1 SCAVENGING SIMULATIONS..............................................................................................144
5.1.1 Shroud geometry...............................................................................................................144
ix
5.1.2 Initial and boundary conditions....................................................................................... 147
5.1.3 Scavenging flow............................................................................................................... 149
5.1.4 Valve flow........................................................................................................................ 156
5.2 O-D SCAVENGING MODEL ............................................................................................... 160
5.2.1 Description ...................................................................................................................... 160
5.2.2 Comparison with KIVA-ERC calculations....................................................................... 164
5.3 SYSTEM SIMULATIONS ..................................................................................................... 177
5.3.1 Revision of combustion correlation constants ................................................................. 177
5.3.2 Two-stroke adaptation of Caterpillar SCOTE................................................................. 181
5.3.3 Addition of Reciprocating Air Pump................................................................................ 187
5.3.4 Addition of turbocharger ................................................................................................. 193
Chapter 6 - Discussion ..................................................................................................................... 199
6.1 THERMODYNAMIC MODEL PERFORMANCE ....................................................................... 199
6.1.1 Speed................................................................................................................................ 199
6.1.2 Flexibility......................................................................................................................... 201
6.1.3 Accuracy .......................................................................................................................... 201
6.2 SCAVENGING OF TWO-STROKE POPPET-VALVED ENGINES................................................ 204
6.3 PERFORMANCE OF TWO-STROKE POPPET-VALVED ENGINES............................................. 206
6.3.1 Numerical modelling procedure ...................................................................................... 206
6.3.2 System simulation results................................................................................................. 209
Chapter 7 - Conclusions and Further Work.................................................................................. 212
7.1 SUMMARY........................................................................................................................ 212
7.2 CONCLUSIONS.................................................................................................................. 212
7.3 FURTHER WORK............................................................................................................... 214
7.3.1 Improvements to the thermodynamic model .................................................................... 214
7.3.2 Improvements to multidimensional modelling................................................................. 215
7.3.3 Two-zone scavenging model ............................................................................................ 215
7.3.4 Further system studies ..................................................................................................... 215
References ......................................................................................................................................... 217
x
List of Figures
Figure 1-1: Two-stroke poppet-valved engine cycle.................................................... 2
Figure 1-2: Common scavenging arrangements........................................................... 3
Figure 1-3: Schematic of Rotec engine prototype. Cylinders may not be in this order
in a practical engine block due to balancing considerations. ............................... 5
Figure 1-4: Intake valve shroud. .................................................................................. 5
Figure 2-1: Toyota S-2 petrol-fuelled prototype (left) and cylinder detail (right)
(Nomura and Nakamura, 1993).......................................................................... 15
Figure 2-2: Toyota S-2D diesel-fuelled prototype (left) and details of the cylinder
head (right) (Nomura and Nakamura, 1993). ..................................................... 15
Figure 2-3: Ricardo "Flagship" engine prototype (Hundleby, 1990). ........................ 17
Figure 2-4: Shibaura/Honda water scavenging rig (Nakano et al., 1990). ................. 19
Figure 2-5: Shibaura/Honda petrol engine prototype (left) with intake port section
showing mask detail (right) (Nakano et al., 1990). ............................................ 19
Figure 2-6: Computational mesh of Huh et al. (1993). .............................................. 21
Figure 2-7: Water flow visualisation rig (left) and port layout (right) (Kang et al.,
1996)................................................................................................................... 22
Figure 2-8: Cylinder cross-sections and injector location and orientation in the
Loughborough study (Das and Dent, 1993). ...................................................... 24
Figure 2-9: Suzuki prototype cylinder head (Morita and Inoue, 1996). .................... 25
Figure 2-10: Some valve arrangements investigated by Yang et al. (1999). ............. 26
Figure 2-11: Toyota S-2 Petrol-Fuelled Engine Scavenging (Nomura and Nakamura,
1993)................................................................................................................... 31
Figure 2-12: Reported S-2 scavenging performance (Nomura and Nakamura, 1993).
............................................................................................................................ 31
Figure 2-13: Scavenging measurements of Sato et al. (1992).................................... 32
Figure 2-14: Cylinder geometry and symbol definitions (Yang et al., 1999). ........... 33
Figure 2-15: Calculated scavenging parameters for the cases in Table 2-9 (Yang et
al., 1999). Note that symbols appear to be missing or ambiguous for cases 34
and 35. ................................................................................................................ 35
Figure 2-16: Power density vs speed for two-stroke poppet-valved engine prototypes.
............................................................................................................................ 37
xi
Figure 2-17: Torque density vs speed for two-stroke poppet-valved engine
prototypes........................................................................................................... 37
Figure 2-18: Bsfc vs engine speed for two-stroke poppet-valved engine prototypes.38
Figure 2-19: Estimated losses vs speed for the Suzuki prototype and original engine
(Morita and Inoue, 1996). .................................................................................. 40
Figure 2-20: Average composition of PM from analysis of 16 heavy-duty
turbocharged diesel engines (Needham et al., 1991). ........................................ 43
Figure 2-21: Representation of 6-mode FTP simulation for a Caterpillar 3406E 500
hp engine (Montgomery, 2000, p. 48). Circle size indicates the weighting of
each mode. ......................................................................................................... 45
Figure 2-22: Typical injection strategy for a modern diesel fuel injector and problems
addressed (Caterpillar, 1998). ............................................................................ 46
Figure 2-23: Increases in injection pressure with engine model year (Tschoeke,
1999). ................................................................................................................. 47
Figure 2-24: Roots-type blower (Heywood, 1988). ................................................... 51
Figure 2-25: Vane compressor (Heywood, 1988)...................................................... 52
Figure 2-26: Screw compressor (Bosch, 1986).......................................................... 53
Figure 2-27: Spiral (scroll) supercharger (Bosch, 1986). .......................................... 54
Figure 2-28: Elements of a centrifugal compressor (Bosch, 1986). .......................... 55
Figure 2-29: Reciprocating piston supercharger. ....................................................... 55
Figure 2-30: Schematic of a thermodynamic model of a turbocharged , intercooled
four-cylinder engine with EGR. ......................................................................... 56
Figure 3-1: Generalised compressibility chart (Van Wylen and Sonntag, 1985, p.
682). ................................................................................................................... 64
Figure 3-2: Simplified flowchart for using VODE to solve an IVP. ......................... 68
Figure 3-3: Calculated specific internal energy vs temperature for Fuel B-air
mixtures at thermochemical equilibrium at various pressures........................... 70
Figure 3-4: Calculated specific ideal gas constant vs temperature for Fuel B-air
mixtures at thermochemical equilibrium at various pressures........................... 71
Figure 3-5: Calculated specific internal energy vs temperature at 30 bar (Data) and
equation 3-16 (Function). ............................................................................... 73
Figure 3-6: Calculated specific gas constant vs temperature at 30 bar (Data) and
equation 3-17 (Function). ............................................................................... 73
Figure 3-7: Caterpillar SCOTE intake and exhaust valve lift profiles....................... 76
xii
Figure 3-8: Assumed poppet valve discharge coefficient vs valve lift/diameter ratio,
based on Heywood and Sher (1999, pp. 187-8). ................................................ 76
Figure 3-9: Estimated variation of effective area for intake and exhaust valves. ...... 77
Figure 3-10: Caterpillar SCOTE apparatus (Montgomery, 2000, p. 23). .................. 85
Figure 3-11: Outline of the process for digitising graphs. ......................................... 88
Figure 3-12: Definitions of t
ID
and . CA = Crank Angle. ........................................ 89
Figure 3-13: Comparison of measured ignition delay and the ignition delay
correlation vs mean pressure. ............................................................................. 92
Figure 3-14: Comparison of measured ignition delay and the ignition delay
correlation vs mean temperature. ....................................................................... 93
Figure 3-15: Comparison of measurement and the correlation vs . ...................... 97
Figure 3-16: Comparison of measurement and the correlation vs t
ID
. .................... 97
Figure 3-17: Typical heat release rate diagram reported by Watson et al. (1980). Note
the difference between this and Figure 3-12. ..................................................... 98
Figure 3-18: Example of measured heat release rate and best fits for Equation 3-36
(Watson et al., 1980) and Equation 3-41 (labelled Modified). ....................... 99
Figure 3-19: Shape factor correlations - data points and linear regression are shown.
.......................................................................................................................... 100
Figure 3-20: Example of a compressor map. ........................................................... 101
Figure 3-21: Example of a turbine map (values on axes omitted to protect
manufacturers proprietary information).......................................................... 103
Figure 3-22: Estimated fmep and correlation........................................................... 108
Figure 3-23: Sample file specifying parameter values to be investigated. The first
column is the initial values, the second column is the final values, the third
column specifies the number of steps, and the fourth column describes the
parameters. ....................................................................................................... 109
Figure 3-24: Mode 3 (high load, mid speed). Case numbers refer to Table 3-8. ..... 112
Figure 3-25: Mode 5 (mid load, high speed) . Case numbers refer to Table 3-8. .... 113
Figure 3-26: Mode 6 (low load, high speed). Case numbers refer to Table 3-8. ..... 114
Figure 3-27: Other comparisons using different engine hardware. Case numbers refer
to Table 3-8. ..................................................................................................... 115
Figure 3-28: Cases 18-22 (75% load, 1600 rpm) (Kong, 2002). Case numbers refer to
Table 3-8. ......................................................................................................... 116
xiii
Figure 3-29: Cases 23-27 (25% load, 1690 rpm) (Kong, 2002). Case numbers refer to
Table 3-8. ......................................................................................................... 117
Figure 3-30: Schematic representations of models for Caterpillar 3406E simulations.
.......................................................................................................................... 118
Figure 4-1: Fuel injection and atomisation within the break-up length (Reitz and
Diwakar, 1987)................................................................................................. 127
Figure 4-2: 60 sector mesh of Caterpillar SCOTE cylinder. The mesh was supplied
by the Engine Research Center at the University of Wisconsin-Madison. ...... 135
Figure 4-3: Computational mesh of Caterpillar SCOTE engine at BDC. The four
valves and the mexican hat bowl-in-piston geometry can be discerned. The
mesh was supplied by the Engine Research Center at the University of
Wisconsin-Madison. ........................................................................................ 136
Figure 4-4: KIVA-ERC model results and experimental data for Mode 3 (75% load,
993 rpm) . Case numbers refer to Table 3-8. ................................................... 139
Figure 4-5: KIVA-ERC model results and experimental data for Mode 5 (57% load,
1737 rpm) . Case numbers refer to Table 3-8. ................................................. 140
Figure 4-6: KIVA-ERC model results and experimental data for Mode 6 (20% load,
1789 rpm) . Case numbers refer to Table 3-8. ................................................. 141
Figure 5-1: Cross-section through cylinder mesh showing a shroud on the intake
(left) valve. ....................................................................................................... 145
Figure 5-2: Cross-section of cylinder showing shroud geometries used in KIVA-ERC
simulations. The intake poppet valves are in the lower half of the cylinder. The
valve stems and shrouds appear white. ............................................................ 146
Figure 5-3: Definitions of shroud parameters. ......................................................... 147
Figure 5-4: Intake pressure vs crank angle for Cases 11 and 12.............................. 149
Figure 5-5: Scavenging efficiency versus delivery ratio for all cases with each
shroud. .............................................................................................................. 150
Figure 5-6: Caterpillar SCOTE cylinder sections. ................................................... 151
Figure 5-7: Gas velocity vectors in section A-A for Case 6, Shroud 1. The intake
valves are to the left. A small vortex is visible above the piston bowl. ........... 152
Figure 5-8: CO
2
mass fraction contours in section A-A for Case 6, Shroud 1. Dark
shading indicates low mass fraction. Intake valves are to the left. .................. 153
xiv
Figure 5-9: CO
2
mass fraction contours in sections B-B to E-E for Case 6, Shroud 1.
Dark shading indicates low mass fraction. The intake valves are to the left rear.
.......................................................................................................................... 154
Figure 5-10: Comparison of gas composition through sections B-B to E-E during
scavenging for Cases 1 (left column) and 9 (right column) with Shroud 4.
Adjacent images have approximately equal scavenging gas volumes in the
cylinder............................................................................................................. 155
Figure 5-11: Calculated intake valve effective areas for shroud 1 based on KIVA-
ERC results. Case numbers are indicated. The effective area estimated in Figure
3-9, reduced by the shroud area, is shown for comparison. ............................. 156
Figure 5-12: Calculated exhaust valve effective areas for shroud 1 based on KIVA-
ERC results. Case numbers are indicated. The effective area estimated in Figure
3-9, reduced by the shroud area, is shown for comparison. ............................. 157
Figure 5-13: Pressure contours in the intake port for Shroud 1, Case 5 at 202 deg
ATDC, showing the pressure gradient between the port boundary and the intake
valves due to flow deceleration (the ram effect). Zero-dimensional modelling
assumes constant pressure throughout the port. ............................................... 158
Figure 5-14: Pressure contours in the exhaust port for Shroud 1, Case 5 at 175 deg
ATDC. Here, the pressure difference between the region above the valve and
the port boundary appears to be due to the port geometry. .............................. 159
Figure 5-15: vs scavenging volume fraction for all cases with shroud 4.............. 161
Figure 5-16: Representation of two-stage scavenging process. ............................... 162
Figure 5-17: KIVA-ERC-calculated cylinder quantities for Case 4, Shroud 1 from
EVO to IVC...................................................................................................... 169
Figure 5-18: Zone 1 quantities calculated from Figure 5-17 and matched calibration
constants x = 0.1 and k = 8 10
-4
..................................................................... 170
Figure 5-19: Calculated exhaust gas purities and scavenging efficiencies for Case 4,
Shroud 1. .......................................................................................................... 170
Figure 5-20: Comparison of KIVA and two-zone scavenging model results for no
shroud on the intake valves. Case numbers refer to conditions described in
Table 5-2. Results for the perfect displacement and diffusion models are also
indicated. The x-axis is time from EVO in seconds. For each case x = 0.34 and k
=1.0................................................................................................................... 171
xv
Figure 5-21: Comparison of KIVA and two-zone scavenging model results for
shroud 1 on the intake valves. Case numbers refer to conditions described in
Table 5-2. Results for the perfect displacement and diffusion models are also
indicated. The x-axis is time from EVO in seconds. For each case x = 0.12 and k
= 7 10
-4
. ......................................................................................................... 172
Figure 5-22: Comparison of KIVA and two-zone scavenging model results for
shroud 2 on the intake valves. Case numbers refer to conditions described in
Table 5-2. Results for the perfect displacement and diffusion models are also
indicated. The x-axis is time from EVO in seconds. For each case x = 0.10 and k
= 8 10
-4
. ......................................................................................................... 173
Figure 5-23: Comparison of KIVA and two-zone scavenging model results for
shroud 3 on the intake valves. Case numbers refer to conditions described in
Table 5-2. Results for the perfect displacement and diffusion models are also
indicated. The x-axis is time from EVO in seconds. For each case x = 0.09 and k
= 8.5 10
-4
. ...................................................................................................... 174
Figure 5-24: Comparison of KIVA and two-zone scavenging model results for
shroud 4 on the intake valves. Case numbers refer to conditions described in
Table 5-2. Results for the perfect displacement and diffusion models are also
indicated. The x-axis is time from EVO in seconds. For each case x = 0.07 and k
= 9.0 10
-4
. ...................................................................................................... 175
Figure 5-25: Ignition delay vs mean temperature. ................................................... 179
Figure 5-26: Ignition delay vs mean pressure. ......................................................... 179
Figure 5-27: Comparison of KIVA-ERC results and the correlation vs ............ 180
Figure 5-28: Comparison of KIVA-ERC results and the correlation vs ignition
delay. ................................................................................................................ 180
Figure 5-29: Engine sub-unit comprising one reciprocating pump cylinder and two
engine cylinders. .............................................................................................. 188
Figure 5-30: Turbocharged and intercooled engine subunit. ................................... 194
Figure 5-31: Comparison of KIVA-ERC and thermodynamic models for the cases in
Table 5-11. ....................................................................................................... 198
Figure 6-1: Flowchart of two-stroke poppet-valved engine modelling process. The
boxes represent tasks and the arrows represent the flow of information from one
task to the next. ................................................................................................ 208
xvi

List of Tables
Table 2-1: Citations and author affiliations................................................................ 12
Table 2-2: Motives for investigation of two-stroke poppet-valved engines. ............. 13
Table 2-3: Toyota S-2 Engine Specifications (Nomura and Nakamura, 1993). ........ 16
Table 2-4: Ricardo Flagship Engine Specifications (Stokes et al., 1992). ............. 18
Table 2-5: Shibaura/Honda flow visualisation rig and engine specifications (Nakano
et al., 1990)......................................................................................................... 20
Table 2-6: KIVA-II engine model (Huh et al., 1993) and water flow visualisation rig
(Kang et al., 1996) specifications....................................................................... 23
Table 2-7: Loughborough University of Technology engine model specifications
(Das and Dent, 1993). ........................................................................................ 24
Table 2-8: Supercharger types selected for two-stroke poppet-valved engines......... 29
Table 2-9: Cases shown in Figure 2-15 (Yang et al., 1999)....................................... 34
Table 2-10: A comparison of scavenging methods (Abthoff et al., 1998; Knoll,
1998)................................................................................................................... 36
Table 2-11: Comparison of small engine performance (motorcycle and passenger
engine data from Heywood and Sher, 1999)...................................................... 39
Table 2-12: EU emissions standards for heavy-duty diesel engines. ......................... 45
Table 2-13: Summary of compressor characteristics. ................................................ 56
Table 3-1: "Fuel B" analysis results (Montgomery, 2000, p. 27) .............................. 69
Table 3-2: Calculated heat of combustion of Fuel B at 298.15 K using various fuel
models. O
2
data from Van Wylen and Sonntag (1985) p. 658, products internal
energy from thermochemical calculation using CEA (Gordon and McBride,
1994; McBride and Gordon, 1996). ................................................................... 81
Table 3-3: Summary of experimental apparatus (Montgomery, 2000)...................... 86
Table 3-4: Measured and estimated data for determining calibration constants for the
ignition delay correlation (Equation 3-32). ........................................................ 91
Table 3-5: Data used for calibration of correlation (Equation 3-40). ..................... 96
Table 3-6: Charge air cooler data for Caterpillar 3406E-475hp engine (Wright,
2001)................................................................................................................. 105
Table 3-7: Estimated charge air cooler quantities and parameters........................... 106
Table 3-8: Cases for calibration/validation of the O-D model................................. 110
xvii
Table 3-9: Experimental and calculated results. ...................................................... 111
Table 3-10: CAT3406E-500hp engine operating conditions. The first five rows are
from Wright (2001, p. 62). The remainder are assumed or estimated as
described in Section 3.2.9. ............................................................................... 119
Table 3-11: Comparison of measured and calculated data for CAT3406E engine
model without turbcharger or charge air cooler............................................... 119
Table 3-12: Comparison of measured and calculated data for CAT3406E engine
model with turbcharger and charge air cooler.................................................. 120
Table 4-1: Adjustable parameters in KIVA-ERC model, adapted from Hessel (2003)
and Kong (2002). ............................................................................................. 138
Table 5-1: Shroud parameters (see Figure 5-3 for definitions of symbols). ............ 147
Table 5-2: Scavenging simulation cases. ................................................................. 148
Table 5-3: Predicted air consumption for cases using KIVA-ERC and 0-D modelling.
.......................................................................................................................... 160
Table 5-4: Comparison of scavenging efficiencies calculated using the 0-D model
and KIVA-ERC. ............................................................................................... 176
Table 5-5: Turbulence and swirl values at IVC for the Shroud 1 cases listed in Table
5-2. ................................................................................................................... 178
Table 5-6: Predicted near-optimum valve timings, engine performance and
combustion parameters for Modes 1-6 with the Caterpillar SCOTE engine
running on a two-stroke cycle with Shroud 1 on the intake valves. Four-stroke
results (italicised) are shown for comparison................................................... 185
Table 5-7: Modal BSFCs for valve timings optimised for a 6-mode FTP cycle
approximation. (IVO=145, IVC=280, EVO=140, EVC=270ATDC). Four-
stroke baseline results are italicised. ................................................................ 187
Table 5-8: Predicted near-optimum valve timings, engine performance and
combustion parameters for Modes 1-6 for the arrangement in Figure 5-29 based
on the Cat SCOTE engine with Shroud 1 on the intake valves. Four-stroke
baseline results are italicised. ........................................................................... 190
Table 5-9: Modal BSFCs for parameters optimised for a 6-mode FTP cycle
approximation. Four-stroke baseline results are italicised. .............................. 192
Table 5-10: Predicted near-optimum valve timings, engine performance and
combustion parameters for Modes 1-6 for the arrangement in Figure 5-30 based
on the Cat SCOTE engine with Shroud 1 on the intake valves. ...................... 195
xviii
Table 5-11: Optimised parameters for a 6-mode FTP cycle approximation. Four-
stroke cycle baseline results are italicised........................................................ 197
Table 6-1: Variation of x with average shroud angle. .............................................. 205



xix

Nomenclature
ABDC After Bottom Dead Centre
AHRR Apparent Heat Release Rate
ATDC After Top Dead Centre
BBDC Before Bottom Dead Centre
BMEP Brake Mean Effective Pressure
BTDC Before Top Dead Centre
CA Crank Angle
CFD Computational Fluid Dynamics
EGR Exhaust Gas Recirculation
EVC Exhaust Valve Close
EVO Exhaust Valve Open
HC Unburned hydrocarbons
IMEP Indicated Mean Effective Pressure
IVC Inlet Valve Close
IVO Inlet Valve Open
NO
x
Oxides of nitrogen
ODE Ordinary Differential Equation
PM Particulate Matter
RMS Root Mean Square
ROI Rate of fuel injection
RSSV Rotatable Shrouded Scavenging Valve
SOI Start of injection
TDC Top Dead Centre
xx
Statement of Original Authorship
The work contained in this thesis has not been previously submitted for a degree or
diploma at any other higher education institution. To the best of my knowledge and
belief, the thesis contains no material previously published or written by another
person except where due reference is made.

Signed:

Date:

xxi
Acknowledgements
I would not have been able to undertake this project without the love, encouragement
and patience of my family, Margot, Alex, Matthew and Felicity. Many thanks for
allowing me to pursue this goal. Thanks also to my parents for their tremendous help
during this project, and for encouraging my curiosity long before I started.

I am grateful to my supervisor Associate Professor Doug Hargreaves for the
opportunity to undertake doctoral work with him, and for his support, encouragement
and advice.

Rotec Design Ltd provided not only an interesting problem to explore, but also
financial support and complete freedom to explore it. I trust that the knowledge and
computational tools and techniques that have been generated will be of value now
and in the future. My thanks go especially to Robert Rutherford, Paul Dunn and
Mark Stefl for assistance, ideas, information and criticism. Glenn OBrien not only
contributed an important suggestion, but made the whole experience much more
enjoyable.

The highlight of my study was the opportunity to visit in February 2002 the Engine
Research Center (ERC) at the University of Wisconsin-Madison. I am grateful to
Professor Patrick Farrell for hosting me and to students Mark Beckman, William
Church and Toshiyuki Hasegawa for their help and hospitality. Dr Song-Charng
Kong and Professor Rolf Reitz kindly allowed me to use the ERCs version of
KIVA, the ERCLIB emissions subroutines library and computational meshes of their
research engine: arguably the best multidimensional model available at the time of
writing to investigate the problem. Dr Kong and Dr Randy Hessel also gave much
advice on engine modelling, both while I was there and by email correspondence
when I returned to Australia. Dr David Montgomery kindly discussed with me the
experimental data he collected while a student at the ERC. I look forward to
returning to the ERC to present and discuss the results I obtained with its help.

The QUT Tribology and Materials Technology Research Concentration and the QUT
Grants-in-Aid Scheme provided financial assistance with my overseas study visit.
xxii

I was fortunate to have access to the QUT High Performance Computing Centres
Silicon Graphics Origin 3000 supercomputer and to the support of Bernadette
Savage, Dr Anthony Rasmussen, Dr Neil Kelson and Dr Mark Barry, who were
always there when I needed computing assistance.

Thanks to Chris Middlemass of Garrett Turbo, Honeywell International Inc., for the
proprietary turbocharger data I required to complete my model.

Finally, this project originated while I was working at Gilmore Engineers. I am
grateful to Dr Duncan Gilmore for the opportunity to work on a wide variety of
engineering problems that eventually led to this study.

Chapter 1 - Introduction
1.1 Background
1.1.1 Two-stroke poppet-valved engines
The combination of two-stroke cycle and poppet intake and exhaust valves is
unusual. Two-stroke engines typically have either intake and exhaust ports in the
cylinder wall or an intake port and exhaust poppet valve. Four-stroke engines usually
have poppet intake and exhaust valves.

One advantage of two-stroke engines is they can have a much greater maximum
power density (rated power output per unit displacement) than four-stroke engines.
This is due largely to the two-stroke cycle having one power stroke every engine
revolution, whereas the four-stroke cycle has one every two revolutions. A higher
power density is generally desirable because it generally allows smaller, lighter
engines for a given application. The greater number of firing strokes results in
smoother engine operation, an attractive feature in passenger vehicle applications.

Poppet valves have many advantages over cylinder ports. Port opening and closing is
governed by piston motion and is necessarily symmetrical about bottom dead centre
(BDC), whereas poppet valves can have asymmetric timing, which usually gives
better engine performance and allows supercharging. Ports cause bore distortion (and
therefore increased wear) through asymmetric liner temperatures. Oil tends to be
swept into ports by the piston rings, where it accumulates and contributes to
increased lubricant consumption and emission of unburned hydrocarbons. Ports
require an increase in cylinder spacing to accommodate them.

The main advantage that ports have over poppet valves is their simplicity and low
cost. However the disadvantages have meant that ported engines have not competed
successfully with the cleaner and more efficient poppet-valved engines, except in
those applications and markets where emissions and fuel efficiency are not strictly
controlled.

2
The two-stroke poppet-valved engine cycle is illustrated in Figure 1-1. Compression,
combustion and expansion all occur in a similar fashion to most two-stroke and four-
stroke engines. At the end of the expansion phase the exhaust valves open, starting
the blowdown period. The intake valves open shortly afterwards, starting the
scavenging period. Because of the close proximity of the intake and exhaust valves,
this arrangement is prone to short-circuiting of the fresh charge to the exhaust,
indicated by the dotted line in Figure 1-1(e). Various means of reducing the short-
circuiting of air from the intake to the exhaust and otherwise improving scavenging
have been tested; these will be discussed further in Chapter 2. The U-shaped
scavenge loop is sometimes called a reverse tumble, but is referred to as a U-
loop herein. The exhaust valves close, trapping the charge while fresh charge is still
being forced into the cylinder or is being blown back into the intake port. This is the
supercharging phase. Finally the intake valves close and the cycle begins again.
Figure 1-1: Two-stroke poppet-valved engine cycle.

Inlet
poppet
valve
Exhaust
poppet
valve
Piston
(a) Compression
(f) Supercharging
(d) Blowdown
(c) Expansion
(b) Combustion
(e) Scavenging
3
Note that the valve train must operate at twice the speed and the valve open periods
are reduced relative to a four-stroke engine. The consequences can be:
Increased valve train requirements
Increased valve train noise, wear and friction
Reduced maximum engine speed
Valve open times that are longer-than-optimum
Reduced maximum valve lift

Similarly, fuel injection also occurs twice as often.

1.1.2 Outline of previous investigations
A number of leading engine manufacturers, including Toyota, Honda, Ricardo
Consultants and Suzuki built and tested two-stroke poppet-valved engine prototypes.
Toyota constructed both diesel-fuelled and petrol-fuelled versions, whereas the
remainder were exclusively petrol-fuelled. Reports on these tests appeared from 1989
to 1996.

Some studies concentrating on the scavenging behaviour of two-stroke poppet-
valved engines were published from 1993 to 1999, although one appeared in 1981.
Flow visualisation experiments and computational fluid dynamics (CFD) modelling
indicated the U-loop scavenging could achieve scavenging efficiencies generally
better than cross scavenging, similar to loop scavenging, but not quite as high as with
uniflow scavenging. These alternative scavenge loops are illustrated in Figure 1-2.

Figure 1-2: Common scavenging arrangements
(a) Cross scavenging (b) Loop scavenging (c) Uniflow scavenging
4
1.2 Rotec engine concept
This study was inspired by investigations undertaken by Rotec Design Ltd, a
company based in Brisbane, Queensland, Australia. They have developed a
reciprocating blower-scavenged two-stroke poppet-valved diesel engine concept. A
standard implementation using a four-cylinder engine block is represented
diagrammatically in Figure 1-3. Intake air is compressed by a turbocharger and
cooled by an intercooler, then fed to an intake manifold. This manifold distributes the
air to the reciprocating air pump cylinders. Each pump cylinder has an intake reed
valve. The air pump is external to the engine block and driven at twice the engine
speed by a toothed belt connected to the engine crankshaft. Outlet reed valves are
optionally placed at the pump cylinder exhausts to prevent back flow into the
cylinder. Each pump cylinder supplies air to a transfer manifold, which distributes it
to two engine cylinders that are 180 out of phase. In this way, one pump cylinder
alternately scavenges two engine cylinders. Each engine cylinder operates on the
cycle outlined in Figure 1-1. It is essentially identical to a four-cylinder engine,
except that the valve train operates at twice the speed, the valve timing and open
period is significantly altered and the fuel injector must inject fuel every engine
revolution rather than every second revolution. A shroud is usually placed on each
intake valve to reduce short-circuiting of air from the intake to the exhaust during
scavenging (Figure 1-4). The engine exhaust is then delivered to a turbocharger
turbine and exhausted. The air pump could conceivably be driven at three times the
engine speed so that a two-cylinder air pump could scavenge a six-cylinder engine
block.
5
Figure 1-3: Schematic of Rotec engine prototype. Cylinders may not be in this order in a
practical engine block due to balancing considerations.


Figure 1-4: Intake valve shroud.
Intake Exhaust
Compressor Turbine
Intercooler
Air pump
(2 x engine speed)
Engine
Reed valve
Poppet valve
Intake manifold Transfer manifold
Exhaust manifold
Shroud
Intake Exhaust
6

Rotec Design has suggested that emissions might be reduced relative to equivalent
four-stroke engines because of the increased air consumption of two-stroke engines.
Two-stroke engines induct air into each cylinder once every revolution, whereas
four-stroke engines induct air once every two revolutions. Thus a two-stroke engine
will consume air at nearly twice the rate as a four-stroke engine with the same
displacement, running at the same speed and having the same boost pressure. At the
same load, the two-stroke engine will therefore burn its fuel at nearly double the
air/fuel ratio as the four-stroke engine. An increase in air/fuel ratio generally results
in a reduction of partially oxidised emissions, such as unburned hydrocarbons,
carbon monoxide and particulate matter. However, this is often not true at high
air/fuel ratios where a significant portion of the fuel forms a very lean mixture with
the air and does not burn completely.

Additionally, exhaust gases could be cooled and mixed with the intake air, a
technique called exhaust gas recirculation (EGR) that is known to be very effective
in suppressing NO
x
formation at medium to high air/fuel ratios (discussed further in
Section 2.3.1.1). EGR reduces air consumption because it displaces intake air, and
excessive EGR causes unacceptable increases in other emissions.

Incomplete scavenging also reduces air consumption and increase the presence of
residual exhaust gases during combustion. This internal EGR might suppress the
formation of oxides of nitrogen (NO
x
). It is expected to be less effective than cooled
external EGR, but would avoid the relatively high capital and maintenance cost of
external EGR systems.

Finally, the charge should have greater turbulent kinetic energy in two-stroke engines
because of:
the reduced time for dissipation of this energy between intake valve closure
and fuel injection;
the greater rate at which the charge must be forced into the cylinder;
the presence of a turbulence-generating obstacle to short-circuit flow like a
shroud.

7
This higher turbulent kinetic energy has been linked to improved combustion and
emissions.

Rotec Design have built and tested prototypes based on three different engines. Two
of these engines were small (1.6 litre and 2.0 litre) four-cylinder indirect injection
automotive diesel engines. Exhaust measurements showed reductions in all
emissions except particulate matter. Optimisation of the prototypes has proved
difficult because of the problems of measuring in-cylinder processes, the expense of
making and modifying hardware, and the large number of parameters that have to be
optimised. These limitations suggest that a computational approach should be
introduced if possible into the prototype development and optimisation process.
Experiments would then be confined to validation and calibration of the
computational model and fine-tuning of prototypes. This would result in faster and
cheaper engine development. Because of the unusual cycle and unusual requirements
(such as the ability to model shrouded intake valves), commercial engine modelling
packages are not suitable. Computational models must be developed or adapted for
this purpose.

The prospect of a cleaner diesel engine cycle is of great interest, possibly more so
than that of an increase in power density. There is widespread opinion that current
engine technology will not be able to meet the stringent emissions targets due to be
phased in over the next few years. Should two-stroke operation allow considerable
reductions in some or all of the emissions formed within the cylinder, the
development and unit costs of future engines may be significantly reduced.
Additionally, the capital cost of producing these engines could be reduced because of
their similarity to current four-stroke engines. Finally, existing engines may be able
to be economically adapted to two-stroke poppet-valved operation. If this concept
were viable, this latter application would be the quickest and easiest to realise.
1.3 Objectives
As mentioned above, the potential advantages of two-stroke poppet-valved engines
are numerous. However, investigations to date have been limited to the construction
of a small number of prototypes adapted from research or production engines, and
very basic analytical and numerical investigations of parts of the system. Detailed
8
CFD simulations incorporating moving valves, which are a cheaper, faster way of
estimating in-cylinder processes than experimentation, were not readily available
until 1997 with the release of the KIVA 3V engine simulation program, after most of
these investigations were reported. An opportunity exists to apply this new
simulation tool and the increase in computing power to more systematically
investigate the two-stroke poppet-valved diesel engine concept.

The objectives of this study are to:
a) develop and validate flexible and inexpensive numerical tools for rapidly
simulating two-stroke poppet-valved engine systems;
b) use the tools to investigate scavenging and combustion processes in a two-stroke
poppet-valved engine cylinder;
c) undertake a parametric investigation of the fuel consumption and emissions
potentials for prototypes that might be constructed to continue this investigation.
Chapter 2 - Literature Review
2.1 Introduction
Since 1990, several studies of two-stroke poppet-valved engine prototypes have been
reported in the technical literature. This chapter begins with a summary of the
motives for these investigations. The investigations are then briefly described and the
outcomes are grouped under various sub-headings. Subsequent sections on diesel
emissions, scavenging and computational engine modelling establish some
background for the following chapters.
2.2 Overview of two-stroke poppet-valved engine studies
2.2.1 Motives for study
Previous studies of two-stroke poppet-valved engines have generally been stimulated
by the shortcomings of conventional two-stroke and four-stroke engines. Often there
is a desire to combine the advantages of each type of engine. The relatively few
experimental studies can be grouped by author affiliation as in Table 2-1, while the
stated motives for the studies are shown in Table 2-2.

12
Table 2-1: Citations and author affiliations.
Citation Author affiliations (abbreviation)
(Nomura and Nakamura,
1993)
Toyota Motor Company (Toyota)
(Hundleby, 1990), (Stokes et
al., 1992)
Ricardo Consulting Engineers (Ricardo)
(Sato et al., 1981), (Nakano et
al., 1990), (Sato et al., 1992)
Shibaura Institute of Technology, Honda R&D Co.
(Shibaura/Honda)
(Das and Dent, 1993) Loughborough University of Technology
(Huh et al., 1993), (Kang et al.,
1996)
Pohang Institute of Science and Technology,
Daewoo Motor Co., Korea Institute of Machinery
and Metals (Korean Studies)
(Morita and Inoue, 1996) Suzuki Motor Corporation (Suzuki)
(Yang et al., 1997), (Yang et
al., 1999)
Kanazawa University, Gunma University
(Kanazawa/Gunma)
(Rutherford and Dunn, 2001) Rotec Design Ltd (Rotec)

13
Table 2-2: Motives for investigation of two-stroke poppet-valved engines.

T
o
y
o
t
a

R
i
c
a
r
d
o

S
h
i
b
a
u
r
a
/
H
o
n
d
a

K
o
r
e
a
n

S
u
z
u
k
i

K
a
n
a
z
a
w
a
/
G
u
n
m
a

R
o
t
e
c

Higher torque and power density compared to
four-stroke engines





Reduction of emissions from lubricant
consumption relative to ported two-stroke engines



Asymmetric valve timing, unavailable in ported
engines



The ability to be supercharged, difficult in ported


engines



Smoother operation due to the increased number


of firing strokes





Reduction of bore distortion due to asymmetric
cylinder heating, giving better piston ring and
liner durability





The possibility of combining manufacture with
four-stroke engines





Closer cylinder spacing (smaller engine size)
relative to uniflow engines due to the absence of
ports


Lower mechanical friction losses relative to four-
stoke engines due to lower peak pressures and
lower speed operation for the same torque output

Reduction in emissions and fuel consumption due


to higher turbulent kinetic energy in the trapped
cylinder gas at the point of fuel injection and
more available air and/or EGR



14
Nakano et al. (1990) of Shibaura/Honda, claimed that ported two-stroke engines
would never succeed in automotive applications due to the excessive consumption of
lubricant inherent in piston-ported engines. Investigation of this alternative to non-
ported two-stroke concepts was therefore warranted. Nomura (1993) from the Toyota
Motor Company started from the premise that ported two-stroke lubrication in
general was unsatisfactory and saw the use of poppet valves as a method of
employing superior four-stroke lubrication techniques.

Hundleby (1990) of Ricardo Consulting Engineers claimed that a major barrier to
adoption of two-stroke engines was their dissimilarity to current production engines,
and that two-stroke poppet-valved engines may lead the way because of the reduced
investment required to test and produce the engines. While the performance of these
engines was potentially greater, it was at the expense of increased cost and
complexity, so it was suited to high-performance prestige cars. Interest in two-stroke
engines in general in the 1980s and 1990s was also sparked by the advent of direct
petrol fuel injection. Electronic fuel injection (EFI) means that the gas entering the
cylinder can be air, rather than a mixture of fuel and air. Thus should any inlet gas be
exhausted during scavenging, no unburned fuel will be taken with it. Additionally,
EFI allows a stratified charge to be formed in the cylinder. Thus lean overall air-fuel
ratios can be achieved, as long as the mixture in the vicinity of the spark is readily
ignitable. Lean burning helps reduce emissions and fuel consumption. The former is
largely through more complete oxidation of fuel, and the latter is largely through a
reduction in pumping work at part load.

More recent studies have tended to concentrate on the scavenging of cylinders with
poppet intake and exhaust valves.

2.2.2 Description of prototypes/models
2.2.2.1 Toyota Motor Corporation
Toyota created two engine prototypes called S-2 and S-2D for supercharged two-
stroke and supercharged two-stroke diesel, respectively. These engine are
discussed in Nomura and Nakamura (1993). The authors presented scavenging flow
patterns generated by a simulation that appears to use a 3-dimensional mesh and
15
moving valves. The specifications of the engines are shown in Table 2-3. The
engines are illustrated diagrammatically in Figure 2-1 and Figure 2-2.





Figure 2-1: Toyota S-2 petrol-fuelled prototype (left) and cylinder detail (right) (Nomura and
Nakamura, 1993).




Figure 2-2: Toyota S-2D diesel-fuelled prototype (left) and details of the cylinder head (right)
(Nomura and Nakamura, 1993).

This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
16
Table 2-3: Toyota S-2 Engine Specifications (Nomura and Nakamura, 1993).


A later article by Freudenberger (1995) discussed another prototype of the S-2
petrol-fuelled engine, stating that it displaced three litres, had double overhead cams
and four (rather than five) valves per cylinder.

This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
17
According to the Toyota Australia public relations department
1
, the S-2 engine had
emissions problems and a four-stroke S-4 engine is under development.

2.2.2.2 Ricardo Consulting Engineers
Hundleby (1990) and Stokes et al. (1992) described the development of a single-
cylinder two-stroke poppet-valved petrol-fuelled research engine. This engine first
ran in May 1989. The basic specifications reported in Stokes et al. (1992) are given
in Table 2-4. During experimentation, many of these specifications were varied. This
engine is illustrated in Figure 2-3.



Figure 2-3: Ricardo "Flagship" engine prototype (Hundleby, 1990).

1
From a telephone conversation with Mr Greg Storok of Toyota Australia Public
Relations on 10
th
July 2000.
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
18
Table 2-4: Ricardo Flagship Engine Specifications (Stokes et al., 1992).


2.2.2.3 Shibaura Institute of Technology and Honda R&D Co.
Nakano et al. (1990) first conducted scavenging flow visualisation experiments with
a single-stroke dynamic simulator adapted from a diesel engine (illustrated in Figure
2-4). Tracer gas was illuminated and filmed through a Pyrex window. This system
was also used for measurements of scavenging efficiency. In this case, this cylinder
was filled with CO
2
and scavenged with compressed air. The density of the
remaining CO
2
was analysed to obtain a volumetric scavenging efficiency.

This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
19

Figure 2-4: Shibaura/Honda water scavenging rig (Nakano et al., 1990).

A four-stroke petrol-fuelled motorcycle engine was then modified to two-stroke
poppet-valved operation. The modifications included reduction of the exhaust valve
lift to prevent contact between the intake and exhaust valves, and reduction of the
geometric compression ratio to 7.84 due to knocking (autoignition) phenomena at a
geometric compression ratio of 11. The engine and test equipment are illustrated in
Figure 2-5.


Figure 2-5: Shibaura/Honda petrol engine prototype (left) with intake port section showing
mask detail (right) (Nakano et al., 1990).

This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
20
The flow visualisation and engine specifications are shown in Table 2-5.

Table 2-5: Shibaura/Honda flow visualisation rig and engine specifications (Nakano et al., 1990).

A follow-on study (Sato et al., 1992) replaced the intake port deflector with shrouds
on the inlet valves. The inlet valves were prevented from rotating so that the shrouds
were always oriented in the most favourable direction.

They found that the shroud reduced the delivery ratio because it reduced the effective
intake area. However, the charging efficiency, trapping efficiency and overall engine
performance were improved. The prototypes were carburetted, however the authors
recommend using a direct injection system to reduce the loss of unburned fuel to the
exhaust ports.

This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
21
They also conducted flow visualisation experiments using water/dye systems to
qualitatively observe the scavenging flow pattern.

2.2.2.4 Korean studies
Huh et al. (1993) used a modified KIVA-II CFD code to simulate the scavenging of a
two-stroke poppet-valved engine cylinder. The computational mesh is shown in
Figure 2-6. The authors modified the KIVA-II code so that moving valves could be
approximated. The simulation is discussed in more detail in Section 2.2.3.


Figure 2-6: Computational mesh of Huh et al. (1993).

A follow-on study (Kang et al., 1996) used the same simulation code as discussed in
Huh et al. (1993) and a water flow visualisation rig (Figure 2-7). The simulation code
was used to model the water flow visualisation rig experiments. The rig had
stationary valves and a stationary piston, which compromised the realism of the
experiment. The authors aims were to obtain qualitative insights into U-loop
scavenging with and without intake valve shrouds.

The specifications of the modelled engine used in the study by Huh et al. (1993) and
the flow visualisation rig used by Kang et al. (1996) are presented in Table 2-6.


This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
22

Figure 2-7: Water flow visualisation rig (left) and port layout (right) (Kang et al., 1996).




This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
23
Table 2-6: KIVA-II engine model (Huh et al., 1993) and water flow visualisation rig (Kang et al.,
1996) specifications.


2.2.2.5 Loughborough University of Technology
Das and Dent (1993) also used KIVA-II with some modifications to study
scavenging, fuel spray development and combustion in a four-poppet-valved fuel-
injected spark ignition engine. The modifications to KIVA-II included new
subroutines to account for fuel spray/wall interaction and mixing-controlled
combustion. A one-dimensional model was used to estimate the pressure at the outlet
port. They used a relatively coarse 3-D cylinder mesh with 4000 cells. The
specifications of the engine they modelled are shown in Table 2-7.

This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
24
Table 2-7: Loughborough University of Technology engine model specifications (Das and Dent,
1993).

















Figure 2-8: Cylinder cross-sections and injector location and orientation in the Loughborough
study (Das and Dent, 1993).
This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
25

2.2.2.6 Suzuki Motor Corporation
Morita and Inoue (1996) adapted a 1.3 litre petrol-fuelled four cylinder four-stroke
engine to two-stroke poppet-valved operation. They reduced the compression ratio
from 11 to 8.7. The engine had a supercharger, intercooler and high-pressure fuel
injection. The engine is illustrated in Figure 2-9.

They also did a numerical scavenging simulation and experimentally measured the
tumble induced by the intake port.


Figure 2-9: Suzuki prototype cylinder head (Morita and Inoue, 1996).

2.2.2.7 Kanazawa and Gunma Universities
Yang et al. (1997) and Yang et al. (1999) concentrated on scavenging simulations
using fog-marked gas and CFD. The CFD simulations were based on the Flagship
engine geometry (Figure 2-3 and Table 2-4), but investigated the effects of changes
to the cylinder head and valves (Figure 2-10), port spacing, piston bore/stroke ratio,
valve timing and boost pressure.

This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
26

Figure 2-10: Some valve arrangements investigated by Yang et al. (1999).

The basic engine specifications and the parameters investigated by Yang are
presented in Section 2.2.4.

2.2.2.8 Rotec Design Ltd
Rotec Design Ltd (Rotec) has been experimenting with two-stroke poppet-valved
engines since the early 1990s. Their basic concept is illustrated in Figure1-3. Their
latest prototypes have been retrofitted indirect injection (IDI) diesel engines from
passenger vehicles.

The company claims reduced development cost and time compared with other
emissions reduction technologies such as NO
x
catalysts for diesels, low sulphur fuels
and regenerating particulate filters. In addition, the concept can be combined with
other emissions reducing technologies if and when they become available. Given
that the engine block and valve gear is largely the same as conventional four-stroke
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
27
engines, engine manufacturers would have smaller capital costs in retooling for this
sort of engine than for more radical concepts (Rutherford and Dunn, 2001).

2.2.3 Simulation techniques
Nomura and Nakamura (1993) presented scavenging flow patterns generated by a
simulation that appeared to use a 3-dimensional mesh and moving valves. These
features would have made it a sophisticated simulation for its time, however the
authors did not present any details apart from the images. No other simulation
techniques are mentioned, although they may have been used.

Hundleby (1990) reported that overall engine performance predictions were made
using a one-dimensional engine model. This type of model is discussed in more
detail in Section 2.5.2. One-dimensional models estimate gas parameters throughout
the engine, and can account for wave effects in pipes. The model was initially used to
optimise valve timing and overlap. The technique used is worth noting briefly as
optimising valve timing is a task to be performed in this study. The rated condition
was chosen for analysis, in this case 95 kW/litre at 5000 rpm. Valve timings were
varied until an optimum condition was found. Once valve timings were chosen, the
engine performance over the entire speed and load range was calculated. In the
present study, it is desired to optimise engine parameters for a standard engine cycle
comprising several load/speed conditions.

Hundleby modelled the scavenging flow using a two-dimensional CFD simulation.
No details were given of the results of the modelling. No indication was found as to
whether the CFD model was a transient or steady-state model.

Sato et al. (1981) and Nakano et al. (1990) employed a transparent cylinder to
observe scavenging flow. The same apparatus was used with gas sampling
techniques to measure scavenging efficiencies. The prototype engine had a number
of significant differences to the scavenging simulator, especially cylinder
dimensions, piston displacement, cylinder head and valve geometry, and the use of
valve shrouds on the simulator and deflectors on the prototype. The follow-up study
reported in Sato et al. (1992) employed a dye-marked water flow visualisation rig for
28
qualitative observations of scavenging.

Huh et al. (1993) and Kang et al. (1996) simulated cylinder scavenging using a
modified version of the KIVA-II CFD code. KIVA is described in more detail in
Section 2.5.3. The model used a three-dimensional mesh. Valves were simulated by
making valve-shaped blocks of cells into obstacle cells with zero fluid velocity on
the nodes. To represent valve movement, cells in the direction of the valve
movement were progressively converted from fluid cells to obstacle cells, while
those on the opposite surface were converted from obstacle cells back to fluid cells.
Air at a constant temperature and pressure was specified at the intake port boundary,
and a constant pressure exhaust boundary was specified.

Morita and Inoue (1996) used CFD to simulate the scavenging flow. The figures they
reproduced in their report showed simplified valve geometry without valve motion.
Piston motion was simulated, indicating that it was a transient simulation.

Yang et al. (1997) employed a flow visualisation rig to view scavenging flow in a
cylinder. Later, a numerical study was undertaken using KIVA 3V Release 2 (Yang
et al., 1999). The CFD model used constant pressure boundary conditions, a three-
dimensional mesh, realistic valve geometry, simulated moving valves and piston, and
had a total of 25,578 grid nodes. This model employed all of the features available to
CFD modellers for this type of problem and therefore represents the most realistic
numerical simulation of the problem prior to this present study.

Scavenge pumps
Section 2.4 discusses scavenge pumps more generally. This section briefly describes
the scavenge pumps used in the studies discussed above.

Toyota elected to use a helical-rotor Roots-type blower (Nomura and Nakamura,
1993). Hundleby (1990) compared the engine air requirement for full load and the
characteristics of centrifugal and positive displacement blowers. He concluded that
neither type of blower matched the engine air requirement satisfactorily when driven
directly by the engine. At high speeds, positive displacement pumps would barely
generate the desired boost pressure and would probably do so at low efficiency,
29
whereas centrifugal pumps would generate excessive boost at high speed and
insufficient boost at low speed. He concluded that a variable ratio drive was
necessary, probably coupled to a centrifugal compressor. It was noted that the
variable ratio drive could be used to regulate the air supply according to the engine
load over much of its load range, without throttling or bypassing air. The studies by
Hundleby and Stokes et al. did not actually use a blower, but used compressed air to
simulate a blower.

Nakano et al. (1990) employed a centrifugal blower with a variable-speed electric
drive to provide scavenging for their prototype. This was not intended to be the
system employed for future production engines.

Morita and Inoue (1996) used what they termed a Reshorm (assumed herein to be
mis-translated Lysholm, i.e. Roots-type) supercharger.

Rotec used a reciprocating piston air pump directly driven by the engine, as
described in Section 1.2.. Rotec intends the reciprocating pump to provide pulses of
air that are timed to maximise scavenging efficiency and reduce pump work.


The scavenge pump types are summarised in Table 2-8.

Table 2-8: Supercharger types selected for two-stroke poppet-valved engines.
Organisation Scavenge pump type Reference
Toyota Roots-type positive displacement Nomura and Nakamura (1993)
Ricardo Centrifugal with variable ratio
drive
Hundleby (1990)
Suzuki Roots-type positive displacement Morita and Inoue (1996)
Rotec Reciprocating piston positive
displacement
Rutherford and Dunn (2001)


30
2.2.4 U-loop scavenging
The close proximity of the intake and exhaust poppet valves, coupled with the
largely radial velocity distribution over the valve heads, means that a significant
fraction of the scavenge flow gets short-circuited from the intake to the exhaust,
without scavenging any residual gases in the cylinder. This reduces scavenging and
trapping efficiency, wasting pump work and decreasing charge purity. Various
methods were employed by researchers to reduce the scavenge flow.

Toyota elected to use a mask between the intake and exhaust valves for their S-2
prototype as shown in Figure 2-11 (Nomura and Nakamura, 1993). The bulk of the
intake flow was directed away from the exhaust valve, along the cylinder wall,
forming an effective scavenge flow. The reported scavenging efficiency (Figure
2-12) was very high, approximately the same as for uniflow scavenging. The
performance is compared with perfect displacement and perfect diffusion scavenging
models. Perfect displacement, in which the residual gas is displaced by the incoming
gas without mixing, is obviously preferable to perfect diffusion, in which the
incoming gas mixes thoroughly with the residual gas, because the amount of gas
required to achieve a given charge purity is minimised. In practical engine systems,
scavenging can be considered a mixture of displacement, diffusion and short-circuit
flow.

The diesel-fuelled S-2D had a flat cylinder head with vertical valves and no apparent
means of reducing the short-circuit flow. The intake port was quite vertical, which
may have improved scavenging by imparting a downwards momentum to the intake
flow. The scavenging performance of the S-2D was not discussed.

31

Figure 2-11: Toyota S-2 Petrol-Fuelled Engine Scavenging (Nomura and Nakamura, 1993).


Figure 2-12: Reported S-2 scavenging performance (Nomura and Nakamura, 1993).

Hundleby (1990) also used downward-directed inlet ports (Figure 2-3). The use of
shrouds on the valves or flow deflectors was avoided, as the authors felt that the
engine would be limited by its ability to breathe, and they did not wish to reduce
the valve effective area. This could have been offset by increasing the valve lift, but
the combination of high lift and high speed could have caused problems with valve
spring surge, valve bounce, noise, wear and vibration. A minimum of four valves per
cylinder was considered necessary for adequate gas exchange.

Nakano et al. (1990) sketched the scavenging flow with and without inlet valve
shrouds based on high-speed photography of flow visualisation experiments. They
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
32
noted that short-circuit flow was apparent without the shrouds, and that the
remainder of the scavenging flow took time to reach the bottom of the cylinder.
When a 90 shroud was added, short-circuit flow was almost absent and a U-shaped
scavenge loop was formed. Some residual fluid was trapped near the centre of a
vortex near the middle of the cylinder. Sato et al. (1992) measured scavenging
efficiency using gas sampling experiments (Figure 2-13). The results indicated that a
90 shroud was best, giving a scavenging efficiency better than that of perfect
diffusion.

Figure 2-13: Scavenging measurements of Sato et al. (1992).

Kang et al. (1996) used flow visualisation experiments and numerical simulations to
qualitatively investigate scavenging flow with shrouded intake valves. The authors
concluded that too large a shroud (>>90) caused a large vortex to form near the
middle of the cylinder. This was considered undesirable, as residual gases would be
trapped within it. Too small a shroud (<<90) allowed a significant fraction of the
intake flow to short-circuit straight to the exhaust valve.

Morita and Inoue (1996) used a fixed mask (Figure 2-9) very similar to that used in
Toyotas prototype. The scavenging performance was not reported.

Yang et al. (1999) used CFD to investigate many combinations of vertical and canted
valves, masks, intake valve shrouds, bore-stroke ratio, port spacing, valve timing and
boost pressure. The results are summarised in Table 2-9, Figure 2-14 and Figure
2-15.
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
33


Figure 2-14: Cylinder geometry and symbol definitions (Yang et al., 1999).

Figure 2-15 shows that scavenging efficiencies exceeding those calculated using the
perfect diffusion model were predicted for many configurations. The engine speed
being modelled was 5,000 pm. The delivery ratio and scavenging efficiency would
presumably have been increased at lower speeds if the boost pressure remained
constant, because of the greater scavenging time.

This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
34
Table 2-9: Cases shown in Figure 2-15 (Yang et al., 1999).

This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
35

Figure 2-15: Calculated scavenging parameters for the cases in Table 2-9 (Yang et al., 1999).
Note that symbols appear to be missing or ambiguous for cases 34 and 35.

Yang et al. concluded:
a) The optimum stroke/bore ratio was in the range 0.4 to 0.6. Most engines have a
ratio of approximately 0.8 to 1.1. The scavenging efficiency (the ratio of the
delivered mass retained to the total trapped mass) increased from 72% to 77%
when the stroke/bore ratio was decreased from 0.89 to 0.5.
b) The optimum shroud angle was in the range 69 to 108. This is consistent with
previous studies that suggested a shroud angle of approximately 90 was optimal.
c) A shroud turn angle of approximately 18 was better than 0.
d) Separating the valves improves performance without a shroud, because of the
lengthening of the short circuit; however it reduced performance with a shroud
because of interference between the inlet valve and cylinder wall.
e) Canted valves have better scavenging because the angled flow against the
cylinder wall is better converted to a smooth scavenging loop, whereas with
vertical valves the scavenging flow is normal to the cylinder walls and more of it
is converted to turbulent flow.

Abthoff et al. (1998) and Knoll (1998) briefly discussed the relative merits of port
loop scavenging, poppet-valve loop scavenging and uniflow scavenging. Their
conclusions are combined in Table 2-10.
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
36

Table 2-10: A comparison of scavenging methods (Abthoff et al., 1998; Knoll, 1998).


The claimed limited swirl capability is debatable, as adequate swirl is generated in
four-stroke poppet-valved engines and the orientation of intake shrouds can generate
significant swirl. Additionally, uniflow scavenging is still struck by problems with
lubricating oil consumption (and therefore the associated high emissions discussed
later in Section 2.2.6) because of the intake ports.

2.2.5 Prototype performance
One of the main reasons cited for the interest in two-stroke engines is their high
power and torque density (see Table 2-2). In this section, the demonstrated
performance (power/torque output, fuel consumption and emissions) of two-stroke
poppet-valved engine prototypes is briefly reviewed.

The power density of the prototypes was calculated from data published in the
references listed in Table 2-2 and shown in Figure 2-16. This figure shows a wide
variation, from a maximum of over 60 kW/litre for the Suzuki prototype at high
speed to approximately 20 kW/litre for the Shibaura/Honda prototype. Interestingly,
the Ricardo Flagship engine showed a general decline in power density with speed,
indicating progressively deteriorating performance with speed.

This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
37
0
10
20
30
40
50
60
70
0 1000 2000 3000 4000 5000 6000
Engine speed (rpm)
P
o
w
e
r

d
e
n
s
i
t
y

(
k
W
/
l
i
t
r
e
)
Suzuki
Toyota S-2
Toyota S-2D
Ricardo
Shibaura/Honda

Figure 2-16: Power density vs speed for two-stroke poppet-valved engine prototypes.

Torque density (proportional to brake mean effective pressure, bmep) for the
prototypes was calculated and is shown in Figure 2-17. The torque curves generally
show maxima at 2000-3000 rpm, except for Ricardos Flagship prototype, which
again showed a steady decline with speed.


Figure 2-17: Torque density vs speed for two-stroke poppet-valved engine prototypes.
0
40
80
120
160
200
0 1000 2000 3000 4000 5000 6000
Engine speed (rpm)
T
o
r
q
u
e

d
e
n
s
i
t
y

(
k
W
/
l
i
t
r
e
)
Suzuki
Toyota S-2
Toyota S-2D
Ricardo
Shibaura/Honda
38

The fuel efficiency of the prototypes is shown in Figure 2-18. They demonstrate a
marked sensitivity to operating speed. Generally, at high speeds the blower losses
become significant.

250
300
350
400
450
500
0 1000 2000 3000 4000 5000 6000
Engine speed (rpm)
b
s
f
c

(
g
/
k
W
h
)
Suzuki
Toyota S-2
Toyota S-2D
Ricardo
Shibaura/Honda

Figure 2-18: Bsfc vs engine speed for two-stroke poppet-valved engine prototypes.

To put these figures in context, Table 2-11 compares the two-stroke poppet-valved
engine performance with typical values for medium-sized two-stroke and four-
stroke engines.

39
Table 2-11: Comparison of small engine performance (motorcycle and passenger engine data
from Heywood and Sher, 1999).

Stokes et al. (1992) reported that very high BMEP (greater than 13 bar indicated)
was achieved at low speeds, however operation at high speed was prevented by the
onset of knock, even at low loads. The maximum speed was increased by the use of
high octane fuels, delaying the onset of knocking. One hypothesis for the early onset
of knock was elevated charge temperatures caused by the retention of residual gas.
The reverse tumble scavenge loop was thought to trap residuals in the core of the
induced vortex. It was thought that improvements in scavenging would reduce the
amount of trapped residuals and hence the onset of knock. Blower losses were
expected to increase rapidly with boost pressure, so that net output actually reduced
with increasing boost pressure above 1.3 bar. To achieve the target operating
conditions, trapping efficiency had to be increased, minimising pumping losses.
Successive improvements to test engines included increasing the ratio of areas of
exhaust ports to inlet ports, redesign of the inlet system to reduce short-circuiting and
This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
40
increase the tumble speed, increasing the valve overlap and delaying exhaust valve
opening to reduce the temperature of the trapped charge.

Nakano et al. (1990) noted insufficient cylinder charging at high speeds. The exhaust
valve lift was reduced to avoid clashing with the intake valve, and the intake valve
deflector (Figure 2-5) reduced the port flow area. Consequently, their external blower
could not supply enough air. The authors noted that improvements could be made if
boost pressure could be adjusted appropriately with speed (Ricardo addressed this
problem with a variable ratio centrifugal blower).

Morita and Inoue (1996) analysed the friction losses of the Suzuki four-stroke engine
and its two-stroke adaptation. Their results are reproduced in Figure 2-19. Note the
dominant effect of the scavenging blower. Obviously, a design challenge is to reduce
the contribution of the scavenge pump to the engine losses. The basic two-stroke
losses (neglecting the blower) are presumably slightly higher than the four-stroke
losses because of the higher mean effective pressure in the cylinder. This causes
greater rubbing friction between the piston and cylinder.


Figure 2-19: Estimated losses vs speed for the Suzuki prototype and original engine (Morita and
Inoue, 1996).


This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
41
A two-stroke poppet-valved engine prototype demonstrated a 71% increase in
continuous torque and power over the original Peugeot XUD11 engine at
approximately 1600 rpm (Gilmore, 1998).

2.2.6 Two-Stroke Poppet-Valved Diesel Emissions
Toyota claimed a reduction in both NO
x
and PM emissions with the S-2D prototype
(Nomura and Nakamura, 1993). These reductions were not quantified. Rotec has not
published the results of their testing.

2.2.7 Summary
Some broad conclusions can be drawn from the studies discussed in this section:
a) Two-stroke poppet-valved prototypes demonstrated significant torque and power
increases over the parent four-stroke engine at low to medium speeds (Hundleby,
1990; Nakano et al., 1990; Sato et al., 1992; Stokes et al., 1992; Freudenberger,
1995; Morita and Inoue, 1996; Gilmore, 1998).
b) Combustion problems occurred at higher speeds, possibly due to insufficient
charge flow into the engine and combustion gases trapped in the vortex caused by
the tumbling motion of the gas in the cylinder (Hundleby, 1990; Nakano et al.,
1990; Sato et al., 1992; Stokes et al., 1992; Morita and Inoue, 1996).
c) Losses using centrifugal blowers were unacceptably high at high speeds
(Hundleby, 1990; Nakano et al., 1990; Sato et al., 1992; Stokes et al., 1992;
Morita and Inoue, 1996). More efficient supercharging is required for satisfactory
fuel consumption.
d) Matching of the engine air demand and blower supply characteristics required
special techniques (Hundleby, 1990).
e) Acceptable scavenging, approaching that of Uniflow scavenging, could be
achieved at high speeds through the use of valve shrouds and canted valves (Sato
et al., 1992; Yang et al., 1999).

42
2.3 Overview of diesel engine emissions
2.3.1 Diesel emissions formation
Any diesel engine concept must be able to comply with stringent emissions
regulations. Therefore, an understanding of the fundamentals of diesel emissions,
regulations and control strategies is required.

Diesel exhaust is mostly a mixture of nitrogen, water, carbon dioxide and unburned
oxygen. Carbon dioxide is a greenhouse gas. The term pollutants or emissions
usually refers to the fraction of exhaust gas (less than 1%) made up mostly of (in
approximately decreasing order):
a) Nitrogen oxides (NO
x
mostly NO and NO
2
)
b) Carbon monoxide (CO)
c) Unburned hydrocarbons (HC sometimes further broken down to methane (CH
4
)
and non-methane hydrocarbons (NMHC))
d) Particulate matter (PM)

There are numerous other potentially harmful compounds emitted in much smaller
quantities, but these are not regulated presently. They include:
e) Nitrous oxide (N
2
O)
f) Sulphur dioxide (SO
2
)
g) Aldehydes
h) Polycyclic aromatic hydrocarbons (PAH)
i) Soluble organic fractions (SOF)
j) Dioxins
k) Metal oxides

The regulated emissions are discussed in greater detail in the following sections.

2.3.1.1 Nitrogen oxides (NO
x
)
NO
x
represents NO and NO
2
. NO is formed by the combination of atmospheric
nitrogen and oxygen at high pressures and temperatures, conditions found in diesel
engines during combustion. Once emitted into the air, NO readily oxidises to form
NO
2
, which is an irritant and an important component in smog formation.
43

2.3.1.2 Particulate matter (PM)
PM is also sometimes called soot. It is a complex product, defined as whatever is
caught in filters in diluted exhaust. Thus it is a mixture of solid particles, liquid
droplets and condensed vapours. It consists mainly of (Kittelson, 1998):
a) tiny carbon particles, either single or agglomerated, with heavy hydrocarbons
from the fuel and lubricant adsorbed onto the surface
b) Metal ash, from engine wear and corrosion and lubricant additives
c) hydrated sulphuric acid droplets, formed from sulphur in the fuel and lubricant
d) unburned heavy hydrocarbons, from the fuel or lubricant.


Figure 2-20: Average composition of PM from analysis of 16 heavy-duty turbocharged diesel
engines (Needham et al., 1991).

The carbon particles are thought to be formed in fuel-rich regions around the injected
fuel droplets and in cooler regions near the walls of the combustion chamber
(Horrocks, 1994). Part of the fuel spray usually forms a film on the cylinder wall or
the piston bowl, which also gives rise to a cool, fuel-rich region conducive to soot
formation.

This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
44
2.3.1.3 Unburned hydrocarbons (HC)
HC denotes unburned fuel or lubricant hydrocarbons in the gas phase. Those in the
liquid phase are counted as PM. These are formed in the same regions as PM. As HC
levels decrease, the contribution of fuel retained in the injector nozzle is becoming
significant, which has forced injector redesign in recent years.

2.3.1.4 Carbon monoxide (CO)
CO is present in very small amounts when fuel/air mixtures are at chemical
equilibrium; however, chemical kinetics is the dominant mechanism of formation of
CO in diesel engines. CO emissions from diesel engines are relatively low. Bowman
(1975) states that CO formation is one of the principal paths in hydrocarbon
combustion. It is thought to oxidise readily in the presence of hydroxyl radicals to
CO
2
. However, hydroxyl is thought to react preferentially with hydrocarbons, so CO
oxidation only occurs late in the diesel combustion cycle when the concentration of
HC is low.

The greatest problems for diesel engine manufacturers are NO
x
and PM. Emissions
of HC and CO are usually small compared with spark ignition engines and the
emissions regulations (Schindler, 1997). A well-known problem in diesel engine
design is the NO
x
PM tradeoff, in which most measures that reduce one tend to
increase the other (Han et al., 1996). Measures that reduce both are eagerly sought.

2.3.2 Emissions regulations
Two most widely adopted emissions standards are those of the European Union (EU)
and the USA. Compliance to standards from either place is acceptable for heavy
vehicles in Australia. EU standards for heavy duty diesel engines since 2000, for
example, use a combination of a 13-mode steady-state (ESC) test and 4-mode
transient (ELR) test. The EU Emissions Standards are summarised in Table 2-12 to
show the dramatic rate of improvement required in diesel engine emissions in the
recent past and near future.

45
Table 2-12: EU emissions standards for heavy-duty diesel engines.
Tier
Date &
Category
Test
Cycle
CO
(g/kWh)
HC
(g/kWh)
NOx
(g/kWh)
PM
(g/kWh)
Smoke*
(m
-1
)
1992
<85 kW
4.5 1.1 8.0 0.612
Euro I
1992
>85 kW
4.5 1.1 8.0 0.36
1996.10 4.0 1.1 7.0 0.25
Euro II
1998.10
ECE
R-49
4.0 1.1 7.0 0.15
Euro III 2000.10 2.1 0.66 5.0 0.10 0.8
Euro IV 2005.10 1.5 0.46 3.5 0.02 0.5
Euro V 2008.10
ESC &
ELR
1.5 0.46 2.0 0.02 0.5
*Measured as the extinction coefficient, which is the inverse of the path length over
which light intensity decreases by the factor of e (2.718).

Researchers often use simplified procedures that approximate the standard tests. One
such test is a 6-mode Federal Transient Procedure (FTP) simulation. An example of
test conditions corresponding to the 6-mode cycle for a Caterpillar 3046E 500 hp
heavy-duty diesel engine is shown in Figure 2-21.

Figure 2-21: Representation of 6-mode FTP simulation for a Caterpillar 3406E 500 hp engine
(Montgomery, 2000, p. 48). Circle size indicates the weighting of each mode.
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
46

2.3.3 Emissions control strategies
Diesel engines generally produce less NO
x
, HC and CO but more PM than petrol
engines. Petrol engine exhausts have a very low oxygen and PM content, which
allows effective catalytic conversion of NO
x
, HC and CO relatively easily. The same
technologies can not be used for diesel engines because of the high oxygen content
of the exhaust gas and the relatively high emissions of PM. In order to meet the strict
emissions requirements, a system approach has proved necessary, as no one method
has been found to be sufficient. Those diesel engine emission control options
relevant to the present study are discussed briefly below.

2.3.3.1 Electronic engine control
Electronic engine control presently allows control of coolant and lubricant
temperatures. Fuel injection timing is important, with earlier timing improving
efficiency and PM emissions but increasing NO
x
emissions, all due to the higher
average combustion temperature. The opposite is true for injection retardation.
Recently, fuel injection rate shaping has been made possible. The benefits of rate
shaping on emissions are discussed in Han et al. (1996). One scheme described by an
engine manufacture is illustrated in Figure 2-22.

Figure 2-22: Typical injection strategy for a modern diesel fuel injector and problems addressed
(Caterpillar, 1998).
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
47

2.3.3.2 Fuel Injection System Modifications
One of the striking trends in modern diesel engine fuel injection systems is the
increase in fuel injection pressures (Figure 2-23). Higher pressures have been found
to reduce PM and HC emissions, however this is at the expense of increased NO
x

emissions. In combination with other techniques such as injection rate shaping, split
injection and injection retardation, NO
x
can be reduced as well. Examples of new
fuel injection technologies allowing higher pressures over the entire engine speed
range and electronic control include Electronic Unit Injectors and Common Rail
injection systems.


Figure 2-23: Increases in injection pressure with engine model year (Tschoeke, 1999).

2.3.3.3 Exhaust Gas Recirculation (EGR)
The addition of cooled exhaust gases to the intake charge has been found to
significantly reduce NO
x
emissions at the expense of a small increase in PM and HC.
The reasons for this are the depression of the combustion temperature through the
increase of the specific heat of the charge and the dilution of the available oxygen
(Stokes et al., 1992; Ladommatos et al., 1996; Ladommatos et al., 1996;
Ladommatos et al., 1997; Ladommatos et al., 1997). If the recirculated exhaust gases
are cooled, more charge can be fitted into the cylinder and the initial charge
temperature is reduced (which reduces NO
x
emissions) (Ladommatos et al., 1998).
Cool EGR is therefore more effective than hot EGR or internal EGR caused by
incomplete scavenging. However, EGR equipment adds cost and complexity to
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
48
engines, and there are durability problems associated with handling corrosive exhaust
streams that have condensed water and reactive compounds.

2.3.3.4 Combustion Air Intake Improvements
Charge air cooling reduces NO
x
formation by reducing combustion temperatures (the
rate of NO
x
formation is sensitive to temperature). Charge air cooling has been
investigated by Dickey et al. (1998) and others. Charge air cooling can cause poor
starting and increased PM and HC in cold climates.

An increase in intake turbulence has been shown to reduce NO
x
emissions (Timoney
et al., 1997). The turbulence was generated by a shroud on the intake valve. Since a
shroud may be necessary to prevent short-circuiting of the charge in two-stroke
poppet-valved engines, it may have a beneficial side-effect in reducing emissions.

2.3.3.5 Exhaust aftertreatment
Aftertreatment refers to modification of the exhaust gas composition after it has left
the cylinder.

Diesel Oxidation Catalysts are effective in reducing CO and HC, however these are
already at low levels. They remove some of the HC adsorbed to PM, but do not
reduce the carbonaceous material, sulphates or ash.

Lean NO
x
Catalysts aim to reduce NO
x
despite the oxidising (lean) exhaust stream.
Sulphur in the fuel poisons some of the proposed catalysts, and none have been
shown to be both durable and effective.

Selective Catalytic Reduction (SCR) catalysts use urea or ammonia to reduce NO
x
.
They are used in stationary engines but are difficult to apply to automotive engines
where the exhaust gas composition and temperatures change very rapidly.

Diesel Particulate Traps collect PM. One type of system uses catalytic material on
the filter that causes regeneration (decomposition of the PM and removal of the
residue), in a continuous or periodic manner, during the regular operation of the
49
system. Another approach uses an electric heater or fuel burner to heat the filter and
regenerate the trap.

Despite the large amount of activity on diesel exhaust aftertreatment, especially in
the last ten years, only the Diesel Oxidation Catalyst has been commercially
available for use with commonly available fuels.
2.4 Scavenging
2.4.1 Fundamentals
Scavenging in two-stroke engines is analogous to the combined exhaust and
induction processes in a four-stroke engine. Since exhaust and induction are fairly
separate and simple processes in a four-stroke engine, they are relatively easy to
model accurately. This is obviously not the case in two-stroke engines.

Two-stroke engines require some means of creating a pressure drop between the
intake and exhaust ports to generate scavenge flow. Ported two-strokes often use
crankcase compression, which is simple but does not allow wet sump lubrication.
Multicylinder engines generally do not have a changing crankcase volume
(downward pistons are balanced by upward pistons) so external compression is
necessary. This does allow wet sump lubrication, which is generally superior in
terms of friction, probability of engine seizure, lubricant consumption and emissions
(Nomura and Nakamura, 1993).

After the blowdown phase the inlet valve opens and the fresh charge is forced in
under pressure. External compression is usually from a turbocharger and/or crank-
driven supercharger. When both are used, they may be in parallel or series (Heywood
and Sher, 1999). The residual gases are a combination of unburned air and
combustion products, mostly carbon dioxide and water vapour.

The incoming air partly displaces, partly mixes with and partly bypasses (by short-
circuiting) the residual gases. The upper limit of scavenging efficiency is bounded
by perfect displacement. Some mixing will occur. Although this dilutes the residual
gases, it also means some of the fresh charge is exhausted with the residual gases
50
and, conversely, some of the residual gases are retained when the exhaust valves
close. Short-circuiting merely absorbs pumping work.

The scavenging air supply may not be purely fresh air. Often, the charge contains
cooled exhaust gases in a process commonly referred to as exhaust gas recirculation
(EGR). EGR has been found to significantly reduce NO
x
emissions without
significantly increasing PM and unburned hydrocarbon emissions (see Section
2.3.3.3).

There are many quantitative measures of scavenging performance. Some of them are
defined below:

Delivery ratio r
d
=
density inlet pump volume swept
mass delivered



Equation 2-1
Scavenging efficiency
s
=
mass trapped
retained mass delivered


Equation 2-2
Trapping efficiency
t
=
mass delivered
retained mass delivered


Equation 2-3
Charging efficiency
c
=
density inlet pump volume swept
retained mass delivered



Equation 2-4
Volumetric efficiency
v
=
density inlet pump volume swept
mass trapped



Equation 2-5

2.4.2 Scavenge pumps
A number of compressor and blower types have been used as superchargers,
including roots blowers, sliding vane compressors, screw compressors, rotary piston
pumps, spiral-type superchargers, variable displacement piston superchargers, and
centrifugal compressors. With the exception of the centrifugal compressor, all are
positive displacement pumps that deliver a specific volume of air per revolution.

51
Since the volumetric efficiency is almost constant, the flow is proportional to the
supercharger or engine speed. Positive displacement devices can provide high boost
pressures without the need for high speed. This is an advantage in two-stroke
engines, where reasonable pressure is required for scavenging at all speeds. A
turbocharger alone would be unlikely to provide sufficient pressure at low speeds,
because little exhaust power would be available to drive the turbine. Positive
displacement superchargers are well suited for a mechanical connection with the
engine, such as through a gearbox or a belt/pulley drive. Centrifugal compressors
boost the pressure roughly proportionally to the square of the supercharger speed, so
they are better suited for coupling with high speed electric motors or to the engine
via a variable ratio gearbox (as in Ricardos Flagship prototype described in
Hundleby, 1990). Much of the following discussion is taken from Heywood (1988)
and Bosch (1986).

In a Roots-type or Lysholm supercharger (Figure 2-24), leakage between the rotors
as well as backflow from the receiver to the inlet side of the blower take place, thus
reducing its overall compression process. Roots blowers are used in applications
where the pressure ratio is rather low, typically in the range of 1.0-1.3. More losses
would be experienced at higher pressure ratios, where the use of Roots-type blowers
would be questionable.


Figure 2-24: Roots-type blower (Heywood, 1988).

Roots blowers have good volumetric efficiency (about 90%) as well as reasonable
mechanical efficiency (85%). However, their isentropic efficiency is usually less
than 65% and strongly contributes to the modest overall efficiency of about 55%

This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
52
In the vane compressor, thin vanes are housed in slots (Figure 2-25). They are flung
outwards by centrifugal force, trapping the charge in the cavity formed by the vanes,
the rotor and the housing. At low speeds the contact pressure is low and leakage can
be high. The rotor itself is mounted eccentrically in the housing. This causes
compression of the charge near the outlet.



Figure 2-25: Vane compressor (Heywood, 1988).

Heating results from the friction of the rotors against the housing. Unless this heat is
dissipated through cooling, it is transferred to the air thus decreasing its density and
increasing its volume. This reduces the compressor efficiency and adds to the engine
cooling system load.

The overall efficiency of the sliding vane compressor is only 40%. This is due to a
combination of low volumetric efficiency (85%), mechanical efficiency (about 65%),
and an isentropic efficiency of just 60%.

As the screw compressor (Figure 2-26) rotor turns, air is inducted through ports
arranged around the cylindrical housing and occupies the volume between two
consecutive screws and the housing. This air is delivered through a discharge port, as
shown in Figure 2-26. The delivery pressure is a function of the rotor speed and the
discharge port flow area. Screw compressors rotate at 3,000 to 30,000 rpm, and
generate substantial heat from friction between the rotor and the housing. Measures
are usually required to dissipate the heat to maintain the compressors mechanical
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
53
integrity. Screw compressors have high volumetric efficiencies as long as their
clearances are kept extremely small.


Figure 2-26: Screw compressor (Bosch, 1986).

The spirals in a Spiral (or Scroll) supercharger are arranged in a flat-sided casing
having a shaft that rotates eccentrically. Sandwiched in between the fixed spirals are
moving displacer walls attached to a disc that is connected to an eccentric pin roller
bearing (Figure 2-27). As the drive shaft rotates, the displacer performs an oscillating
circular motion of double eccentricity. Air entering the blower moves from one
working chamber to the next. The rotation of the eccentric and the rotor is through a
toothed belt.

This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
54


Figure 2-27: Spiral (scroll) supercharger (Bosch, 1986).

The overall efficiency of this supercharger is about 55%. Its isentropic efficiency is
68% and its volumetric efficiency is close to 90%. The speed range for this type of
supercharger is 0-13,000 rpm, and it can deliver up to 80 kPa boost pressure. The
spiral-type supercharger is also often referred to as a scroll-type supercharger.

The centrifugal compressor is normally used with an exhaust-driven turbine.
However, it can also be coupled to the engine crankshaft or driven independently by
a high speed electric or hydraulic motor. It usually consists of a single stage radial
compressor. The air is accelerated to a high speed and flows radially outward via a
stationary diffuser stage toward a volute as shown in Figure 2-28. The volute
converts the kinetic energy of the air to pressure. The centrifugal compressor is ideal
for providing high mass flow rates at a pressure ratio of less than 3.5, which is
desirable in internal combustion engine applications.

This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
55

Figure 2-28: Elements of a centrifugal compressor (Bosch, 1986).

Reciprocating piston compressors (Figure 2-29) are commonly used outside of
automotive applications. They are characterised by high efficiencies, low mass flow
rates and high pressure ratios.

Figure 2-29: Reciprocating piston supercharger.

The various types of compressors are summarised in Table 2-13.

One-way
valve
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
56
Table 2-13: Summary of compressor characteristics.
Roots Vane Screw Rotary
piston
Spiral Centrif-
ugal
Recip.
piston
Overall efficiency
(%)
55 40 ~55 ~55 ~55 ~60 ~65
Max pressure
ratio
1.3 N/A* 1.5 1.8 1.8 3.5 >10
Positive
displacement
Yes Yes Yes Yes Yes No Yes
* No data available

2.5 Engine modelling
2.5.1 Thermodynamic modelling
Thermodynamic engine models of the type also known as "zero-dimensional",
"lumped parameter" or "filling and emptying" models represent engine systems as a
number of control volumes that may be linked by elements such as pipes, orifices,
valves, compressors, turbines and heat exchangers. An example of a system
representing a turbocharged, intercooled four cylinder engine is illustrated in Figure
2-30.

Figure 2-30: Schematic of a thermodynamic model of a turbocharged , intercooled four-cylinder
engine with EGR.

The physical engine system can be reduced to a system of ordinary differential
equations using conservation of mass, the First Law of Thermodynamics
Compressor
Plenum
Throttle
valve
Plenum Plenum
Intake
poppet
valves
Exhaust
poppet
valves
Cylinders
EGR valve
Turbine
Waste gate
valve
57
(conservation of energy) and the perfect gas law (or other suitable relationship
between gas properties). Several sub-models regarding heat transfer, combustion, gas
flow etc. are required to complete the model. Given initial values of parameters in
each control volume, such as temperature, pressure and composition, the problem is
then an initial value problem, which can be solved numerically. The solution can
then be used to estimate the performance (power output, fuel consumption etc.) of
the engine. For steady-state calculations, the solution may be calculated over many
engine cycles to allow the effects of the initial conditions to diminish, as the solution
usually converges towards a periodic solution.

The advantages of this type of model are simplicity, minimum computational effort
and the ability (that it shares with one-dimensional modelling, discussed below) is
that it is suited to modelling entire engine systems.

2.5.2 One-dimensional modelling
In real engine systems there are spatial variations of gas properties within pipes and
manifolds. One-dimensional modelling use one-dimensional unsteady conservation
equations for mass, momentum and energy for gas flow in a duct. These equations
are solved by either the method of characteristics or finite difference procedures
(Heywood and Sher, 1999).

One-dimensional modelling can resolve ram effects (pressure gradients due to flow
acceleration) and pressure wave effects. It can also resolve gas concentration
gradients.

Drawbacks of models with one-dimensional elements include greatly increased
computational effort relative to thermodynamic models, and difficulties in
representing some wave behaviour such as shocks.

2.5.3 Multidimensional modelling
Neither thermodynamic nor one-dimensional models describe detailed behaviour
within control volumes such as engine cylinders. For many purposes, such as
obtaining overall performance estimates of conventional engine systems, this is not a
58
problem. However, where the gas flow, fuel spray formation, combustion or
emissions formation details need to be simulated, a multidimensional model is
required. Initially, thermodynamic models were adapted to this task by breaking up
control volumes into many zones (e.g. Hiroyasu and Nishida, 1989; Bazari, 1992).
Through the use of many semi-empirical correlations the number of zones could be
kept to a minimum and calculations could be done with only modest computing
power. As the computing power available to researchers increased, computational
fluid dynamics (CFD) models based on more fundamental principles became
popular. These models still require some semi-empirical correlations, as the
computational effort required to capture the very wide range of length scales and
time scales involved in engine processes is prohibitive. For example, fuel droplets
may have diameters of the order of microns, while the engine bore might be on the
scale of 100 mm a difference of five orders of magnitude. Presently a large grid has
of the order of 10
5
elements, which is far too coarse to resolve details at the droplet
scale.

A survey of emissions modelling literature shows that KIVA is the most widely used
CFD package for engine research. Commercial packages that have been used to
model internal combustion engines include FIRE, STAR CD and VECTIS. The latest
release of KIVA has the ability to model moving canted valves, spray formation and
evaporation, combustion and moving pistons (Amsden, 1999). Additional
subroutines have been developed at the Engine Research Center at the University of
Wisconsin-Madison which estimate wall heat transfer and the formation of NO
x
and
PM (Hampson and Reitz, 1995; Hampson et al., 1996). These have been used with
KIVA to explain such phenomena as the reduction in PM and NO
x
with split
injection schemes (Han et al., 1996). Other institutions have developed their own
subroutines for emissions formation (Beatrice et al., 1996).

2.6 Discussion
A review of engine modelling suggests that at least two-types of model are
necessary: thermodynamic models (possibly including one-dimensional elements) to
represent the behaviour of the overall engine system, and a CFD model to represent
details of the cyinder, valves and ports. The thermodynamic model would provide
59
boundary conditions for the CFD model, and the CFD model would be used to tune
submodels within the thermodynamic model.

Commercial thermodynamic engine models are unlikely to be able to model
scavenging through poppet valves satisfactorily since this is such an unusual case. A
purpose-built model would allow complete control over the simulation tasks.

Some consideration has been given to the validity or otherwise of a purely zero-
dimensional thermodynamic model in this study, since sufficient time is not available
to include a one-dimensional modelling capability. Wave and ram effects should
least affect low-speed engines with relatively short pipe lengths, especially
turbocharged heavy-duty diesel engines. For example, a pressure wave travels more
than ten metres in the time it takes for a heavy duty diesel engine at a high speed of
2000 rpm to complete one revolution. If the engine is turbocharged, pipe lengths on a
heavy-duty diesel engine are typically of the order of 1 metre, and the spatial
variation of gas properties should be small. Turbines and compressors also tend not
to transmit poppet valve-induced waves, although the turbocharger speed does vary
slightly in response to pulsations in the exhaust flow.

Comparisons of filling and emptying models and models with one-dimensional
elements indicate that for most engine simulations, especially those concerning
turbocharged engines, the filling and emptying models are sufficiently accurate
(Chen et al., 1992). However, during the course of this study the validity of the zero-
dimensional modelling will be assessed.
Chapter 3 - Thermodynamic model development
3.1 Requirements
After reviewing the literature on engine modelling for Section 2.5, it was apparent
that there were many ways that thermodynamic engine models had been
implemented. In order to assist in choosing between alternatives, it was helpful to
have a set of model requirements. Each alternative had pros and cons, the most
common trade-off being accuracy versus speed, another being generality versus
speed. Reference to the model requirements often clarified the decision.

Briefly, the model requirements were:
a) Speed. When evaluating relatively novel systems that are not yet well
understood, it is advantageous to be able to run many simulations in a short
period. For example, if the effects of nine interacting parameters are to be
investigated, and six values of each parameter are chosen, the number of runs
is 9
6
or about 530,000. Clearly, the investigation must be carefully designed
to minimise the number of cases investigated; however, a fast model will
allow the investigation of more cases and reduction of the chance of missing
a feature.
b) Flexibility. It was envisaged that not only the parameters of engine systems
would be varied, but the systems themselves. For example, model validation
would largely have to be done with a well-studied four-stroke engine, while a
two-stroke adaptation would be the main object of study. Various
turbocharger and/or supercharger characteristics (or no turbo- or
supercharger) might be examined. A major goal was not having to rewrite
any of the model when such a change was made. Another goal was the ability
for automatic optimisation. Equally important was knowing what was not
required to be changed. For example, it was envisaged that the model would
be used only for heavy-duty direct injection diesel engine simulations and not
indirect injection or spark ignition engines. Moreover, it was likely that most
of the study would be based on one particular engine block and fuel system,
so a completely generalised internal combustion engine model was not
necessary. In many cases, this allowed great simplifications to be made.
61
so a completely generalised internal combustion engine model was not
necessary. In many cases, this allowed great simplifications to be made.
c) Accuracy. The model should account for all of the major thermodynamic
processes occurring in the engine, using well-known submodels and semi-
empirical correlations. The method of solution should not introduce
significant errors.

There were many instances in which techniques had to be invented or adapted. For
example, there was not found in the literature any attempt to develop a zero-
dimensional two-stroke poppet-valved scavenging model. Also, the most common
heat release rate model could not be made to satisfactorily match experimental data
on the engine and had to be adapted.

To achieve the best accuracy, it was decided to validate the model using a well-
studied research engine. In reviewing the literature, the University of Wisconsin
Engine Research Center was one source notable for its many contributions to the
literature on engine modelling and engine analysis. During a study visit to the Engine
Research Center published data from their Caterpillar Single Cylinder Oil Test
Engine was obtained, much of it from a single source: a PhD thesis by David
Montgomery (Montgomery, 2000). The Engine Research Center also kindly supplied
a detailed KIVA mesh and some sample input files of the Caterpillar SCOTE engine.
This engine is a single-cylinder version of the Caterpillar 3406E 6-cylinder 14.6 litre
heavy-duty engine. Given the relative wealth of data on this engine that was suitable
for modelling and validating, and the interest in adapting four-stroke heavy duty
engines to two-stroke cycles, the logical choice of basic engines for this study were
the Caterpillar SCOTE and Caterpillar 3406E engines.

If the thermodynamic and KIVA models could be validated, they could then be used
to simulate the adaptation of the Caterpillar 3406E engine to two-stroke operation
and fulfil the goals of the study.

62
3.2 Description
3.2.1 Basic assumptions
The thermodynamic model was based on three fundamental principles:
Conservation of mass
Conservation of energy
Ideal gas equation of state

They can be expressed mathematically as follows:

=
i
i
dt
dm
dt
dm

Equation 3-1

( )

=
i
i
oi
dt
dQ
dt
dV
P
dt
dm
h mu
dt
d

Equation 3-2

mRT PV =
Equation 3-3

where:
m = mass of gas within the control volume
m
i
= mass flow entering from a separate control volume i
u = specific internal energy
h
oi
= stagnation enthalpy of flow to or from control volume i
P = absolute pressure
V = volume of the control volume
Q = heat transfer out of the control volume
R = specific ideal gas constant
T = gas temperature (absolute)
t = time

Note that implicit in Equation 3-2 is the assumption that gas properties are constant
throughout the control volume. This is sometimes far from the case, for example
when a control volume represents a diesel engine cylinder during the early stages of
combustion. One consequence is that this type of model is not suitable for
63
simulations of detailed in-cylinder processes. Provided that correlations are available
that make knowledge of these detailed processes unnecessary, the model should yield
a good approximation of the overall behaviour of the engine system. The advantage
of this assumption is once again simplicity and rapidity of calculation.

The conservation laws are incontrovertible, but the ideal gas assumption is an
approximation rather than a law. It is a very convenient approximation because of
its simplicity, and is very accurate when the intermolecular forces between gas
molecules are negligible (Van Wylen and Sonntag, 1985, p. 379). The
compressibility factor is a measure of how well a particular gas at a particular state
approximates ideal gas behaviour and is defined as follows:

mRT
PV
Z =
Equation 3-4

A compressibility factor of unity indicates ideal gas behaviour. At 1 bar and 300 K
the specific volume of nitrogen (R = 296.80 J/(kgK)) is 0.890205 m
3
/kg (Van Wylen
and Sonntag, 1985, p. 644), giving a compressibility factor of 0.9998. This indicates
that air, which is mostly nitrogen, at ambient conditions behaves very much like an
ideal gas. The worst case in terms of deviation from ideal gas behaviour occurs at the
maximum temperature and pressure. In a turbocharged diesel engine the maximum
pressure seldom exceeds 15 MPa and the maximum bulk gas temperature seldom
exceeds 2000 K. The compressibility factor at this state can be estimated using a
generalised compressibility chart, such as that shown in Figure 3-1. Once again
approximating air using the properties of nitrogen, the critical temperature and
pressure is 126.2 K and 3.39 MPa, respectively (Van Wylen and Sonntag, 1985, p.
650). The reduced temperature and pressure is defined as the ratio of the absolute
values and the critical values. Thus, the maximum reduced temperatures and
pressures are up to 15.8 and 4.4, respectively. Figure 3-1 indicates this gives a worst-
case compressibility factor of approximately 1.03. It can be concluded that the ideal
gas assumption is sufficiently accurate throughout the entire internal combustion
engine system, and that the gain in accuracy to be made in using a more complicated
equation of state would be very small.

64

Figure 3-1: Generalised compressibility chart (Van Wylen and Sonntag, 1985, p. 682).

The equivalence ratio of a fuel-air mixture is used here to describe the burnt gas
fraction. It can be written:

FAS m
m
a
f

=
Equation 3-5
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
65
where:
= equivalence ratio
m
f
= mass of burnt fuel in the mixture
m
a
= mass of air originally in the mixture
FAS = stoichiometric fuel-air ratio

If is zero, there is no burned fuel in the mixture. If is unity, the mixture is
comprised of the combustion products of a stoichiometric mixture of fuel and air. It
is possible to have greater than unity in which the mixture is comprised of the
combustion products of a richer-than-stoichiometric mixture of fuel and air.

For diesel engine modelling purposes, it is argued in Section 3.2.3 that the internal
energy of gases within the control volumes can be considered a function of
temperature and equivalence ratio, and that the gas constant can be considered a
function only of the equivalence ratio. Note that pressure is omitted. The effect of
unburned fuel vapour is also neglected. If u = u(T,) and P is replaced with (mRT/V)
(Perfect Gas Law), the First Law equation (Equation 3-2) can be rewritten:

= +

i
i
oi
dt
dQ
dt
dV
V
mRT
dt
dm
h
dt
dm
u
dt
d u
m
dt
dT
t
u
m


Equation 3-6

which may be rearranged to form a first order differential equation:

T
u
m
dt
dm
u
dt
d u
m
dt
dQ
dt
dV
V
mRT
dt
dm
h
dt
dT
i
i
oi


Equation 3-7

The rate of change of the equivalence ratio in the control volume can be obtained by
differentiating Equation 3-5:

|
|
.
|

\
|

=
dt
dm
m
dt
dm
m
m FAS dt
d
f
f
2
1

Equation 3-8
66

m
f
changes due to fuel burning within the control volume and due to burnt fuel-air
mixture being carried across the control volume boundaries. These can be expressed
as:

( )

+ =
i
i
i
f
dt
dm
FAS fuel
dt
mfb d
dt
dm

Equation 3-9

where:
mfb = mass fraction of the fuel charge that has been burnt
fuel = total mass of fuel injected into the control volume

i
= equivalence ratio of mass flow i entering or leaving the control volume

Combining Equation 3-8 and Equation 3-9 we get a first order ODE for the
equivalence ratio in each control volume:

( )
|
.
|

\
|
+ =

i
i
i
dt
dm
dt
dm
dt
mfb d
FAS
fuel
m dt
d

1

Equation 3-10

There is now a system of three first order ordinary differential equations (Equation
3-1, Equation 3-7 and Equation 3-10) for each control volume.

In order to be able to solve the equations, the internal energy must be expressed as a
function of the gas composition and state, and functions for mass and heat flow
across the control volume boundaries must be supplied. These are discussed in the
following sections.

3.2.2 Numerical solver
The system of first order ordinary differential equations (ODEs), when combined
with initial conditions, form an initial value problem (IVP). Most forms of this
problem are not able to be solved analytically. The best alternative is to obtain an
accurate approximation to the solution using numerical methods.

67
The program went through stages of using solvers created by the author based on
Euler and implicit improved Euler methods (Equation 3-11 and Equation 3-12) to
using packaged Runge-Kutta solvers in Mathcad and Fortran numerical libraries.

y
n+1
= y
n
+ hf(t
n
,y
n
)
Equation 3-11

h
y t f y t f
y y
n n n n
n n
2
) , ( ) , (
1 1
1
+ +
+
+
+ =
Equation 3-12

where:
dt
dy
y t f = ) , (
h = step size

It was recognised that the system of ODEs might be stiff for some problems and non-
stiff for others (especially when combustion and valve flow was not being modelled).
Stiffness in systems of ODEs appears from the mathematical literature to be
difficult to define precisely. Byrne and Hindmarsh (1975) noted that stiff systems of
ODEs have a Jacobian matrix (matrix of partial derivatives
y
f

) has one or more


eigenvalues whose real parts are negative and large in modulus. Another suggestion
due to Shampine (1994) was that stiff ODEs are those that are best solved with
backward difference or implicit methods. Equation 3-12 is an example of an implicit
formulation. The unknown y
n+1
appears on both sides of the equation and is not
defined explicitly on the right hand side. The equation must be solved iteratively,
which takes more time than explicit formulations, but for some problems this is more
than offset by the increased step size allowed for the same accuracy.

Finally, it was decided to use a public domain IVP solver called VODE (Byrne and
Hindmarsh, 1975; Brown et al., 1988). It has been regularly upgraded and tested for
nearly 30 years. The latest revision at the time of writing was in April 2002. It has
the option of efficient stiff and non-stiff solvers (the type of solver is selected by
setting a variable to the appropriate value). The program dynamically adjusts the
solution step interval and the order of the method used to keep the estimated error
68
within the user-defined bounds. Since stiff and non-stiff solvers are available, it is
possible to do a few test runs with each type and see which method works best, based
on the CPU time used.

To use VODE in solving an IVP, all that is required is to have a subroutine that
calculates f(t,y(t)), set the method variable to stiff or non-stiff, and call VODE for
every point at which the solution is desired.


Figure 3-2: Simplified flowchart for using VODE to solve an IVP.
Error handling options are not shown.

3.2.3 Gas properties
In order to solve Equation 3-1, Equation 3-7 and Equation 3-10, it is necessary to be
able to express the internal energy and specific ideal gas constant as functions of the
gas state. Initially, a simple correlation from (Krieger and Borman, 1966) for fuel
with the general formula C
n
H
2n
(i.e. alkenes) was used. However, the experimental
Initial conditions, method flag (stiff or non-stiff solver),
local and global error tolerances
Calculate derivatives
dt
dy

Set time for first solution point
Call VODE
End of solution interval?
Write solution point to file
Calculate and write
solution summary
End
Yes
No
Set time for next
solution point
69
data used to validate the model used a commercial diesel fuel denoted Fuel B with
the characteristics shown in Table 3-1. The C/H mass ratio of 7.25 means the average
formula was C
n
H
1.643n
, which would be expected to have a lower heat of combustion
than C
n
H
2n
.

Incidentally, the other commercial diesel fuel (Fuel A) analysed by Montgomery
had a slightly lower C/H ratio (6.869 versus 7.253). Fuel A was used on experiments
with multiple injection that were not used for model validation.

Table 3-1: "Fuel B" analysis results (Montgomery, 2000, p. 27)

Since the fuel properties were known, a simple correlation like Krieger and
Bormans could be generated to get the internal energy and specific gas constant as
functions of temperature, pressure and equivalence ratio. In this form, the ideal gas
equation has the form:

PV = m R(T,P,) T
Equation 3-13

Since P appears on both sides, it is an implicit equation that must be solved for P. It
would be far more convenient if the pressure variable could be neglected. To see
whether (or under what conditions) the pressure can indeed be neglected, the
following analysis investigates the effect of pressure on the internal energy and
specific ideal gas constant.

Thermochemical equilibrium calculations were performed using the NASA
Chemical Equilibrium with Applications (CEA) program (Gordon and McBride,
This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
70
1994; McBride and Gordon, 1996), . The Fuel B composition was assumed to be
C
n
H
1.6430n
N
0.00038n
O
0.00033n
S
0.00027n
, based on Table 3-1. The C, H and S weight
percentages do not quite sum to 100%, so the balance was assumed to be equal
weight percentages of nitrogen and oxygen, the two elements in addition to carbon,
hydrogen and sulphur that are found in appreciable quantities in assays of some
crude oils and fuel oils (Avallone and Baumeister, 1996, p. 7-11). The molar
composition of air was assumed to be (Gordon, 1982):
N
2
= 78.084%, O
2
= 20.9476%, Ar = 0.9365%, CO
2
= 0.0319%
This gives a stoichiometric air/fuel ratio of 14.253.

The results of the calculations are summarised in Figure 3-3 and Figure 3-4.

-4000
-3000
-2000
-1000
0
1000
2000
0 500 1000 1500 2000 2500 3000
Temperature (K)
S
p
e
c
i
f
i
c

i
n
t
e
r
n
a
l

e
n
e
r
g
y

(
k
J
/
k
g

o
r
i
g
i
n
a
l

a
i
r
)
1 bar
3 bar
10 bar
30 bar
100 bar
phi = 1
phi = 0

Figure 3-3: Calculated specific internal energy vs temperature for Fuel B-air mixtures at
thermochemical equilibrium at various pressures.
71
285
290
295
300
305
310
315
0 500 1000 1500 2000 2500 3000
Temperature (K)
S
p
e
c
i
f
i
c

i
d
e
a
l

g
a
s

c
o
n
s
t
a
n
t

(
J
/
k
g
*
K

o
r
i
g
i
n
a
l

a
i
r
)
1 bar
3 bar
10 bar
30 bar
100 bar
phi = 1
phi = 0

Figure 3-4: Calculated specific ideal gas constant vs temperature for Fuel B-air mixtures at
thermochemical equilibrium at various pressures.

A number of observations can be made:
Pressure becomes a significant factor above approximately 1600 K.
The difference between the internal energy and gas constant at pressures
between 10 bar and 100 bar up to 2200 K are very small.
The effect of pressure increases with equivalence ratio.
The variation in specific internal gas constant with temperature and pressure
is small (less than 1%) below 2200 K.

For diesel engine applications, the bulk gas temperature seldom exceeds 2200 K.
Bulk gas temperatures between 1600 K and 2200 K generally occur in diesel engine
cylinders early in the expansion stroke, when the gas is generally between 10 bar and
100 bar. As the gas is further expanded it cools and the composition is frozen.

The decision was made therefore to fit polynomial functions to the 30 bar pressure
lines. At low temperatures the 30 bar curve converges to all the other pressure
curves, while at high temperature the pressure is likely to be in the vicinity of 30 bar.
The gas constant was assumed to be a function of equivalence ratio only. These
simplifications obviously introduce small errors and limit the applicability of the
72
correlation to internal combustion engine applications. However, this does not
compromise the requirements, and enables a high-speed model.

Krieger and Borman (1966) used a function of the form:

u(T,) =
1
(T) -
2
(T)
Equation 3-14

where:

1
(T),
2
(T) = polynomial functions of temperature that are chosen to fit the
thermochemical calculations.

A 5
th
-order polynomial was fitted to the u(T,P=30 bar,=0) calculations in Figure
3-3, and a 4
th
-order function was fitted to u(T) = u(T,P=30 bar,=0) - u(T,P=30
bar,=1), using the least-squares method. The resultant expression for internal
energy per kilogram of air (not fuel-air mixture) was:

u(T,) = -2.898110
5
+ 6.423010
2
T + 8.837310
-2
T
2
+
3.353610
-5
T
3
2.176910
-8
T
4
+ 4.112310
-12
T
5
-
(3.02910
6
+ 1.374110
2
T - 4.148910
-1
T
2
+
2.322710
-4
T
3
- 4.836010
-8
T
4
) (J/kg air)
Equation 3-15

and the expression for the specific gas constant with respect to kilograms of original
air was:

R() = 287.05 + 17.52 (J/(kg air)K)
Equation 3-16

These functions are compared with thermochemical calculations in Figure 3-5 and
Figure 3-6

73
-4000
-3000
-2000
-1000
0
1000
2000
0 500 1000 1500 2000 2500 3000
Temperature (K)
S
p
e
c
i
f
i
c

i
n
t
e
r
n
a
l

e
n
e
r
g
y

(
k
J
/
k
g

o
r
i
g
i
n
a
l

a
i
r
)
Data, phi=0
Curve fit, phi=0
Data, phi=0.5
Curve fit, phi=0.5
Data, phi=1
Curve fit, phi=1

Figure 3-5: Calculated specific internal energy vs temperature at 30 bar (Data) and equation
3-16 (Function).
285
290
295
300
305
310
0 500 1000 1500 2000 2500 3000
Temperature (K)
S
p
e
c
i
f
i
c

i
d
e
a
l

g
a
s

c
o
n
s
t
a
n
t

(
J
/
k
g
*
K

o
r
i
g
i
n
a
l

a
i
r
)
Data, phi=0
Function, phi=0
Data, phi=0.5
Function, phi=0.5
Data, phi=1
Function, phi=1

Figure 3-6: Calculated specific gas constant vs temperature at 30 bar (Data) and equation 3-
17 (Function).

The agreement is obviously very good.for specific internal energy, and good for the
specific gas constant at low equivalence ratios or temperatures below about 2200 K.


74
3.2.4 Heat transfer
The simple heat transfer correlation due to Hohenberg (1979) is well-known and
matches experimental results reasonably well. The heat transfer to the walls, cylinder
head and piston is obviously difficult to measure accurately, so accurate calibration is
not possible. Fortunately, errors in the estimated heat transfer result in only relatively
small errors in the estimated engine performance (Assanis and Heywood, 1986).

Hohenbergs relationship takes the form:

( )
8 . 0
2
4 . 0 06 . 0
8 . 0
1
C v
T Vol
P C
h
p
cyl
+ =
Equation 3-17

where:
h = heat transfer coefficient (W/m
2
K)
P = cylinder pressure (bar)
Vol
cyl
= instantaneous cylinder volume (m
3
)
T = instantaneous bulk gas temperature (K)
p
v = mean piston speed (m/s)
C
1
,C
2
= calibration constants

( )
w
T T A h Q =
&

Equation 3-18

where:
Q
&
= total heat transfer rate (W)
A = heat transfer area (m
2
)
T
w
= wall temperature (K)

Hohenberg suggested values of 130 for C
1
and 1.4 for C
2
. The model was modified
slightly to allow the piston face to have a different temperature to the walls.

75
3.2.5 Gas flow through valves and orifices
Gas flow through poppet valves is estimated using compressible flow equations for
choked and unchoked flow as appropriate (Streeter and Wylie, 1983). For unchoked
flow, the equation is:

(
(
(

|
|
.
|

\
|

|
|
.
|

\
|

=

k
k
o
k
o
o o eff
P
P
P
P
k
k
P A
dt
dm
1
2
1
1
2
Equation 3-19

whereas for choked flow:

1
1
1
2

+
|
.
|

\
|
+
=
k
k
o
o eff
k RT
k
P A
dt
dm

Equation 3-20

Flow is choked if:

1
1
2

|
.
|

\
|
+

k
k
o
k P
P

Equation 3-21

where:
k = ratio of specific heats =
( ) T u
R

+ 1 (from the ideal gas equation)
A
eff
= effective flow area = discharge coefficient flow area
= density
u = specific internal energy
subscript o indicates upstream conditions

The poppet valve lift profile of the Caterpillar Single Cylinder Oil Test Engine
(SCOTE) was used in the model. This is shown in Figure 3-7. No adjustment was
made to the maximum lift to compensate for changes in the poppet valve acceleration
as discussed in Hundleby (1990) as it was not known whether valve train speed was a
limiting factor.
76
0
4
8
12
16
0 1
V
a
l
v
e

l
i
f
t

(
m
m
)
Intake
Exhaust
OPEN CLOSE

Figure 3-7: Caterpillar SCOTE intake and exhaust valve lift profiles.

The valve lift and valve diameter (Table 3-3) was used to calculate the valve curtain
area. In the absence of flow data for the Caterpillar SCOTE poppet valves, the
discharge coefficient was assumed to follow a relationship similar to that reported in
Heywood and Sher (1999, pp. 187-8). The relationships used are reproduced in
Figure 3-8.

0.4
0.5
0.6
0.7
0.8
0 0.1 0.2 0.3 0.4
Valve lift/diameter
C
D
Intake
Exhaust

Figure 3-8: Assumed poppet valve discharge coefficient vs valve lift/diameter ratio, based on
Heywood and Sher (1999, pp. 187-8).

77
The valve curtain areas were multiplied by the discharge coefficient to obtain
functions of valve effective area, shown in Figure 3-9.

0
0.0005
0.001
0.0015
0.002
0.0025
0 1
E
f
f
e
c
t
i
v
e

a
r
e
a

(
m
2
)
Intake
Exhaust
OPEN CLOSE

Figure 3-9: Estimated variation of effective area for intake and exhaust valves.

Reed valves were simply treated as one-way orifices. The effective area of the reed
valve block was assumed constant. More complex models were available (e.g. Fleck
et al., 1997) but the effect of reed valve design on engine performance was not
intended to be a variable in this study. Instead, the model can be thought of as
idealised reed valves, permitting flow through the maximum effective area of the
reed block in one direction only.

3.2.6 Scavenging
The simplest models are perfect displacement and perfect diffusion, respectively.
The former assumes that the incoming gas volume displaces an equal volume of
residual gases. There is no mixing. Perfect displacement assumes that the incoming
gas instantaneously and completely mixes with the residual cylinder gases. Each
model represents an extreme case, whereas the actual scavenging behaviour is
expected to involve some displacement, some mixing and some short-circuit flow
(not accounted for in either model).

78
In four-stroke engines the perfect diffusion model is acceptable because there is
relatively little valve overlap, therefore displacement and short-circuiting are
minimal. Instead, residual gases are well-mixed with the fresh charge during the
induction stroke, so the perfect diffusion model is most appropriate.

As it happens, perfect diffusion is the default behaviour of the thermodynamic
model, as all quantities including species concentration are always averaged
(mixed) over the cylinder volume. During scavenging, combustion is neglected,
and the change of gas composition can be expressed as (from Equation 3-10):

=

i
i
i
dt
dm
dt
dm
m dt
d

1

Equation 3-22
where:
= burnt gas fraction in the engine cylinder
m = total mass of cylinder gas

i
= burnt fraction of gas entering from control volume i
m
i
= mass entering from control volume i

Until a more accurate model was developed for scavenging through poppet valves,
inspection of scavenging simulations for poppet-valved engines (e.g. Figure 2-12,
Figure 2-13 and Figure 2-15) showed that the behaviour was reasonably
approximated by perfect diffusion. No single-zone models were found in the
literature that specifically addressed the scavenging of two-stroke poppet-valved
engines. This is perhaps not surprising given the relatively small amount of work on
the subject. The best scavenging data is from Yang et al. (1999), but that was for one
particular engine at one particular speed. Other scavenging models were examined.

The development of a scavenging model for this project is described in Section 5.2.

To investigate scavenging through the poppet valves, a KIVA 3V model of a
Caterpillar SCOTE cylinder, including valves and intake and exhaust ports was used.
The KIVA program and the model is described more fully in Chapter 4. Obtaining a
79
correlation for scavenging through the poppet valves became a major part of this
project, and is discussed in Section 5.1.

3.2.7 Combustion
3.2.7.1 Calculation of heat release
The heat of combustion of fuel is measured by reacting a mixture of fuel and oxygen
at a certain temperature in a sealed constant volume bomb. Heat is transferred from
the bomb to the surroundings until the products reach the temperature of the original
mixture. The process may be expressed in terms of the First Law of
Thermodynamics as:

Q + mu
R
= mu
P

Equation 3-23

where:
Q = the quantity of heat transferred from the bomb calorimeter
m = mass of reactants and products
u
R
= the specific internal energy of the reactants
u
P
= the specific internal energy of the products

The internal energy of the products can be calculated using the CEA thermochemical
equilibrium program (see Section 3.2.3), and the internal energy of the oxygen can
be calculated from thermodynamic tables. If the fuel is liquid, its internal energy is
calculated using the following relationship:

+ =

P
h h m U f
o

Equation 3-24
where:
U = internal energy
m = mass
h
f
= enthalpy of formation at reference conditions
h = difference in enthalpy between the current state and the reference conditions
P = pressure
= fuel density
80

so that Equation 3-23 may be rewritten as:

| |
2
2
O
fuel
O
fuel
f
P
fuel
fuel air
fuel
RT h
m
m
P
h h u
m
m m
m
Q

(

+
+
=

o

Equation 3-25

where the subscripts denote the properties of the fuel, oxygen or products.

The internal energy of the reaction products can be calculated using the CEA
program (the correlation developed in Section 3.2.3 is for fuel-air mixtures and not
applicable to fuel-oxygen mixtures). Those for oxygen can be found the same way or
using thermodynamic tables such as Van Wylen and Sonntag (1985, p. 658). Diesel
fuel, being a complex mixture of hydrocarbons with carbon numbers ranging on
average from about C
13
to about C
21
(Avallone and Baumeister, 1996, p. 7-12), is
sometimes approximated by another compound or by a fictitious compound with
properties similar to that of diesel fuel. Some approximations that have been used are
n-dodecane C
12
H
26
(e.g. Borman and Johnson, 1962), n-tetradecane C
14
H
30
(e.g.
Fuchs and Rutland, 1998), and fictitious compounds df2 and di (Amsden, 1993,
p. 36). df2 is a fictitious compound with the heat of formation of diesel, the latent
heat of vaporisation of n-hexadecane and the enthalpy of n-dodecane. di uses
diesel properties compiled from a number of sources and has the chemical formula
C
13
H
23
.

Calculations assumed a temperature of 298.15 K. At this temperature the fuel is
nearly all liquid. The vapour fraction can be neglected. The properties of n-
tetradecane, and a modified n-tetradecane with the enthalpy of formation adjusted
to give good agreement with the measured heat of combustion were assumed. For
comparison, results using no fuel model were also calculated. Tetradecane data were
obtained from the National Institute of Standards and Technologys Chemistry
WebBook (NIST, 2003). The fuel composition for the CEA calculations was assumed
to be C
n
H
1.6430n
N
0.00038n
O
0.00033n
S
0.00027n
, and the oxygen/fuel ratio was assumed to be
stoichiometric. The results of the calculations are summarised in Table 3-2.
81

Table 3-2: Calculated heat of combustion of Fuel B at 298.15 K using various fuel models. O
2

data from Van Wylen and Sonntag (1985) p. 658, products internal energy from
thermochemical calculation using CEA (Gordon and McBride, 1994; McBride and Gordon,
1996).
Products
u
P
with H
2
O
(l)
(MJ/kg) -10.6969
u
P
with H
2
O
(g)
(MJ/kg) -10.1436
Reactants
h
O2
(J/kg) 0.0
R
O2
(J/kgK) 259.84
Fuel model C
14
H
30
Mod. C
14
H
30
No model
h
f
(MJ/kg)
-2.033 -0.728 0
Fuel h (MJ/kg) 0 0 0
Heat of combustion
fuel
m
Q
with H
2
O
(l)
(MJ/kg)
43.69 44.94 45.73
Discrepancy with measured
gross heat of combustion
-3.0% -0.2% 1.6%
fuel
m
Q
with H
2
O
(g)
(MJ/kg)
41.32 42.56 43.35
Discrepancy with measured
net heat of combustion (%)
-2.7% +0.2% 2.1%
Note: O
2
/Fuel mass ratio = 3.29867

The modified tetradecane has a calculated enthalpy of formation higher than that
of n-tetradecane, which is consistent with Fuel B being an unsaturated hydrocarbon.
Saturated hydrocarbons generally have lower enthalpies of formation than
unsaturated hydrocarbons.

In the thermodynamic model, the internal energy of the products is for fuel-air
mixtures, not fuel-oxygen, and it is estimated using a polynomial. However, the net
heat of combustion should still be very close to the measured value of 42.47 MJ/kg.
This can be easily verified. If the reference temperature is 298.15 K, the pressure is
82
low and the reactant and product temperatures are also 298.15 K, then Equation 3-23
can be expressed as:

( ) ( ) ) , ( ,
f fuel R air P air
h m T u m T u m Q + =
Equation 3-26
where:

R
,
P
= the fuel-air equivalence ratios of the reactants and products, respectively

Dividing both sides by m
fuel
:
f
fuel
air
fuel
h
u
m
m
m
Q
+


Equation 3-27

Using Equation 3-5:
f
fuel
h
u
FAS m
Q
+

1

Equation 3-28

Evaluation of Equation 3-28 for all values of between 0 and 1 and using the
enthalpy of formation of modified tetradecane gives a net heat of combustion of
42.53 MJ/kg fuel, which is 0.14% from the measured value.

The actual energy balance used in the thermodynamic model during combustion is as
follows (from Equation 3-6):

( )
dt
dQ
dt
dV
V
mRT
v
P
T T c h ROI
dt
d u
m
dt
dT
t
u
m
fuel
ref inj p f

|
|
.
|

\
|
+ + + =

2
2


Equation 3-29
where:
ROI = rate of injection of fuel
c
p
= specific heat at constant pressure of liquid fuel
v
fuel
= the injection velocity of the fuel
T
inj
= the temperature of the injected fuel
T
ref
= the reference temperature on which the enthalpy of formation of the fuel is
based

83
Note that the heat of formation, if it is for the liquid state, accounts for the heat of
vaporisation.

In the model, the specific heat of liquid tetradecane was used, being 2200 J/(kgK)
(NIST, 2003).

The injection velocity was calculated from the rate of injection, which for the
Caterpillar SCOTE engine was found to agree generally with:

ROI = 0.051 + 6.8*10
-4
(-
inj
)
Equation 3-30
where:
ROI = rate of injection (kg/s)
= engine speed (rad/s)
= engine crank angle (radians)

inj
= start of injection angle (radians)

This expression forms a ramped injection pulse, which is consistent with the standard
Caterpillar electronic unit injector. This has a rocker arm which follows a constant
ramp rate cam (Montgomery, 2000, p. 27).

The injection velocity is calculated from:

v
fuel
ROI/(
fuel
.N
noz
./4d
noz
2
)
Equation 3-31
where:

fuel
= fuel density
N
noz
= the number of injector nozzle holes
d
noz
= the injector hole diameter
Note that N
noz
= 6 and d
noz
= 0.214 mm for the Caterpillar SCOTE engine
(Montgomery, 2000, p. 28).

3.2.7.2 Heat release rate correlation
A commonly used and very simple combustion correlation is that of Watson et al.
(1980). The ignition delay (interval between start of fuel injection and start of
84
combustion) is calculated from the average bulk cylinder temperature and pressure
during the ignition delay period. Two burning modes are assumed: premixed
(causing an initial spike in the heat release rate data) and diffusion-controlled (see
Figure 3-12 for definitions). The proportion of premixed fuel combustion and
diffusion-controlled combustion was correlated to the equivalence ratio and the
ignition delay. Finally, the fuel combustion rate was modelled as a linear
combination of two functions that fitted the shapes of their heat release rate data. The
relationships proposed by Watson et al. were:

3
2
1
exp
) , (
A
m
m
m m ID
P
T
A
A
P T t
|
|
.
|

\
|
=
Equation 3-32

6 5
4
0 . 1 ) , (
A
ID
A
tr ID tr
t F A t F =
Equation 3-33

| |
2
1
1 1 ) (
p
p
C
C
p
M =
Equation 3-34

| |
2
1
exp 1 ) (
d
C
d d
C M =
Equation 3-35

) ( ) 1 ( ) ( ) (
d p t
M M M + =
Equation 3-36
where:
t
ID
= ignition delay period
T
m
, P
m
= mean bulk temperature and pressure during ignition delay period
= proportion of fuel mass in premixed combustion
F
tr
= trapped equivalence ratio
= proportion of total combustion duration
M
p
= Cumulative mass of fuel burnt in premixed combustion
M
d
= Cumulative mass of fuel burnt in diffusion combustion
M
t
= Cumulative total mass of fuel burnt
A
1
A
6,
C
p1
, C
p2
, C
d1
, C
d2
= constants selected to fit experimental data for a
particular engine
85

3.2.7.3 Experimental data
The engine selected for calibration of the correlations was the University of
Wisconsin-Madisons Caterpillar Single Cylinder Oil Test Engine (Caterpillar
SCOTE), a single-cylinder version of Caterpillars production six-cylinder 3406E
14.6 litre heavy duty direct injection diesel engine. This engine has various ratings
from 355 hp and 1800 rpm to 550 hp and 2100 rpm (Caterpillar, 2002). The data
from the Caterpillar SCOTE engine used to find constants for equations 3-21 to 3-25
were reported in Montgomery (2000). The experimental apparatus is summarised in
Figure 3-10 and Table 3-3.


Figure 3-10: Caterpillar SCOTE apparatus (Montgomery, 2000, p. 23).

This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
86
Table 3-3: Summary of experimental apparatus (Montgomery, 2000).

The engine operating conditions for which the calibration data was gathered is
summarised in Table 3-8. The cases shown in the table represent three modes of the
six-mode FTP simulation: mode 3 (high load, mid speed), mode 5 (medium load,
high speed) and mode 6 (low load, high speed). They are all of the runs reported by
Montgomery for which the engine apparatus was consistent. In other runs, the fuel
This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
87
injection equipment was altered, and this was shown to adversely affect the
calibrations.

The fuel injection rate and cylinder pressure were measured. From the pressure trace,
the heat release rate was inferred. All three quantities were plotted versus crank
angle. Rather than try to locate the original electronic data, the plots were digitised
using a scanner at 150 dpi and saved in bitmap format. The bitmap was converted to
a matrix using a built-in Mathcad function. Each value in the matrix represents the
shade of grey of a particular pixel. Since the traces were usually between three and
five pixels wide, a program was written to convert the trace to an average trace just
one pixel wide. The scale of the bitmap was then used to convert the pixel position to
a value on the x- and y-axes. In summary, the graphical information was digitised
and converted to vectors of x values and corresponding y values. The process is
illustrated in Figure 3-11. The digitised information was then in a form that was
amenable to automated analysis.

88

Step 1: The plotted data (from Montgomery, 2000) is scanned.


Step 2: The required data is isolated (in this case its the apparent heat release rate).
XY
0 1
0
1
2
3
4
5
6
7
8
9
10
-30 -2.60710
-4
-29.869 -2.60710
-4
-29.739 -2.60710
-4
-29.608 -2.60710
-4
-29.477 -2.60710
-4
-29.346 -2.60710
-4
-29.216 -2.60710
-4
-29.085 -2.60710
-4
-28.954 -2.60710
-4
-28.824 -2.60710
-4
-28.693 -2.60710
-4
=

Step 3: The bitmap is analysed to determine the middle of the trace. The axis scales are measured and
the position of the trace pixels (non-white) are converted to x- and y-values (the first eleven pixels are
shown above)
30 20 10 0 10 20 30 40 50
0.01
0
0.01
0.02
0.03
0.04
0.05
XY
1
XY
0

Step 4: The result is graphed and checked against the original plot for accuracy.

Figure 3-11: Outline of the process for digitising graphs.
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
89

The apparent heat release rate (AHRR) plots were used to measure the ignition delay
and premixed burn fraction. The ignition delay was measured from the elbow in the
injection rate plot, i.e. where the injection rate starts levelling off after its initial rapid
rise, to the bottom of the initial dip in the AHRR plot. After this point the AHRR
trace starts rising due to combustion of the fuel. The dip may be partly due to the
AHRR algorithm not accounting correctly for the heat transfer of the compressed
charge through the walls, and it may be partly due to the absorption of heat from the
charge in the evaporation of the fuel spray. The premixed burn fraction was
estimated by measuring the area under the first peak in the AHRR plot. Since the
AHRR data was normalised, the total area under the AHRR curve by definition
equals unity. The area under the first peak is therefore the fraction of heat released in
premixed burning. If the heat of combustion is assumed roughly constant, that area is
also the proportion of fuel consumed in premixed burning. These definitions of the
ignition delay period and premixed burn fraction are illustrated in Figure 3-12.


Figure 3-12: Definitions of t
ID
and . CA = Crank Angle.


t
ID

=
2
1
CA
CA
dCA AHRR
CA
1
CA
2
90
3.2.7.4 Ignition delay correlation
In order to estimate the mean temperature and pressure during the ignition delay
period, a Mathcad worksheet was used that simulated induction and compression in
the Caterpillar SCOTE engine. Equations 3-1, 3-2 and 3-3 were integrated using a
built-in adaptive Runge-Kutta scheme. Gas flow through the poppet valves was
modelled as described in Section 3.2.5, and heat transfer to the cylinder walls was
modelled as described in Section 3.2.4. The boundary conditions were obtained from
Table 3-8. In this way, the temperature and pressure history of the cylinder during
induction and compression was estimated. The mean temperature and pressure
during the ignition delay period was found using:

+
=
ID
ID
t SOI
SOI
t m
dt t T T ) (
1

Equation 3-37

+
=
ID
ID
t SOI
SOI
t m
dt t P P ) (
1

Equation 3-38

where:
SOI = the time at the start of injection

The data obtained from the digitised graphs, and the estimated mean temperatures
and pressures during the ignition delay are shown in Table 3-4.

91
Table 3-4: Measured and estimated data for determining calibration constants for the ignition
delay correlation (Equation 3-32).
Case T
m
(K) P
m
(bar) t
ID
(ms)
1 892 6.82 0.44
2 895 6.97 0.42
3 886 6.57 0.42
4 892 6.82 0.42
5 877 6.26 0.46
6 899 7.73 0.45
7 912 8.00 0.38
8 893 7.57 0.48
9 908 7.96 0.40
10 885 7.34 0.54
11 899 7.73 0.45
12 905 5.22 0.64
13 897 4.97 0.70
14 904 5.24 0.67
15 898 5.17 0.69
16 902 5.10 0.60
17 905 5.22 0.60
18 941 8.08 0.6
19 944 8.28 0.4
20 940 8.26 0.5
21 927 7.95 0.5
22 903 7.29 0.6
23 884 5.67 0.8
24 890 5.87 0.7
25 888 5.91 0.7
26 876 5.68 0.8
27 854 5.24 0.9


Calibration constants for Equation 3-32 were fitted to the data using the least-squares
method. Owing to the complicated nature of the data, the result was dependent on the
initial guesses of the constants. A short program was written to vary all of the initial
guesses by small steps as follows:
A
3
= 0.5, 0.72.5
A
2
= 1000, 110011000
92
A
1
was estimated assuming that the ignition delay period was 0.5 ms, the mean
pressure was 65 bar and the mean temperature was 900 K. A
1
was then varied from
half this value to 1.5 times the value in ten steps. This reduced the number of
iterations and focussed the search on that part of the parameter space where the
solution was known to lie.

The results were:
A
1
= 2.33
A
2
= 2230
A
3
= 0.94

This is reasonably similar to the values calculated by Watson et al (1980), who
derived A
1
= 3.45, A
2
= 2100 and A
3
= 1.02.

The results are plotted against the original data in Figure 3-13 and Figure 3-14.
0
0.2
0.4
0.6
0.8
1
40 50 60 70 80 90
Mean pressure P
m
(bar)
I
g
n
i
t
i
o
n

d
e
l
a
y

(
m
s
)
Measured
Correlation

Figure 3-13: Comparison of measured ignition delay and the ignition delay correlation vs mean
pressure.
93
0
0.2
0.4
0.6
0.8
1
840 860 880 900 920 940 960
Mean temperature T
m
(K)
I
g
n
i
t
i
o
n

d
e
l
a
y

(
m
s
)
Measured
Correlation

Figure 3-14: Comparison of measured ignition delay and the ignition delay correlation vs mean
temperature.

The agreement can be seen to be reasonably good and the trends are predicted
correctly. The Hardenberg and Hase (1979) correlation for ignition delay was
examined, but generally overpredicted the ignition delay and did not match the
experimental data as well as Watsons correlation.

The use of the mean temperatures and pressures complicated the implementation of
this model. Equation 3-37 and Equation 3-38 require pressure and temperature be
known as functions of time. The subroutine kept an array with the temperature and
pressure history. The numerical solver was known to jump forwards or backwards in
time to satisfy the user-supplied error tolerance, so the history array had to be
carefully maintained so that only those values prior to the current time were in the
array. If the correlation was an explicit function of temperature and pressure, the task
would have been much easier.

In the subroutine, ignition was assumed to have started when the following condition
was satisfied:
94
( )
1
,

i i i ID
i
P T t
t

Equation 3-39
where:
i = time steps between SOI and the present time
t = length of time step i
T
i
,P
i
= average temperature and pressure in time step i

3.2.7.5 Premixed combustion fraction correlation
The correlation for suggested by Watson et al. (Equation 3-33) could not be made
to fit the measured data well, or predict the measured trends. A new correlation was
therefore proposed. It seemed reasonable that the proportion of fuel in premixed
combustion should depend mostly strongly on the proportion of fuel that had been
injected up until ignition. Otherwise, there was the absurd possibility of more fuel
undergoing premixed combustion than had been injected to that point. The other
factor that was thought to influence the amount of premixed combustion was the
ignition delay period. A short ignition delay should not allow much time for
premixing. The correlation proposed was of the form:

6 5
4
) , (
A
ID
A
ID
t A t =
Equation 3-40
where:
= (amount of fuel injected up to ignition)/(total amount of fuel to be injected)

When comparing Equation 3-40 with Equation 3-33, note that the 1 has been
replaced with , which sets the upper bound for . The trapped equivalence ratio has
also been replaced by .

Values for were calculated from Montgomerys data (Montgomery, 2000) and
Kongs data (Kong, 2002), which included the rate of injection. This data is
summarised in Table 3-5. The calibration constants were determined using a least
squares method. The results were:
95
A
4
= 0.41
A
5
= 1.50
A
6
= 1.17

A comparison of the data and the correlation is shown in Figure 3-15 and Figure
3-16. Again, the agreement with the data and the trends are good.



96
Table 3-5: Data used for calibration of correlation (Equation 3-40).
Case Measured proportion of fuel
injection prior to ignition,
Measured proportion of fuel
in premixed combustion,
1 0.101 8.0
2 0.108 8.1
3 0.103 8.5
4 0.104 8.0
5 0.113 8.5
6 0.171 9.1
7 0.152 7.2
8 0.202 10.2
9 0.161 8.1
10 0.246 11.5
11 0.194 9.4
12 0.571 28.4
13 0.590 31.3
14 0.549 29.2
15 0.626 31.5
16 0.512 28.1
17 0.508 28.0
18 0.288 8.8
19 0.176 9.5
20 0.176 9.7
21 0.233 11.1
22 0.288 13.5
23 0.986 27.2
24 0.907 40.1
25 0.907 40.4
26 0.986 45.5
27 1.0 66.9

97
0
20
40
60
80
0 20 40 60 80 100
Proportion of fuel injected before ignition ()
P
r
o
p
o
r
t
i
o
n

o
f

f
u
e
l

i
n

p
r
e
m
i
x
e
d

c
o
m
b
u
s
t
i
o
n

)
Measured
Correlation

Figure 3-15: Comparison of measurement and the correlation vs .

0
20
40
60
80
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Ignition delay t
ID
(ms)
P
r
o
p
o
r
t
i
o
n

o
f

f
u
e
l

i
n

p
r
e
m
i
x
e
d

c
o
m
b
u
s
t
i
o
n

)
Measured
Correlation

Figure 3-16: Comparison of measurement and the correlation vs t
ID
.

3.2.7.6 Combustion rate shape parameter correlation
The heat release rate correlation proposed by Watson et al. (1980) did not match the
shapes of the heat release rates measured from the Caterpillar SCOTE engine. When
attempts were made to fit the shape parameters to the data, large residual errors
98
resulted. Often the resultant shapes bore little qualitative resemblance to the data.
The resemblance could be greatly improved if the start of diffusion burning occurred
near the conclusion of premixed combustion. It is as though the fuel-air mixture
ignites and consumes all of the mixed fuel and air, and then all that remains is
diffusion-controlled combustion. In most of the heat release data from the Caterpillar
SCOTE engine the heat release almost returns to zero after the initial peak, which is
quite different to the heat release profiles reported by Watson et al. (compare Figure
3-12 and Figure 3-17. The differences could be due to the presumably much higher
injection pressures in the Caterpillar SCOTE engine. Watsons correlation assumes
that both premixed and diffusion combustion commence at approximately the same
time.



Figure 3-17: Typical heat release rate diagram reported by Watson et al. (1980). Note the
difference between this and Figure 3-12.

Equation 3-36 was modified to reflect a delayed start of diffusion combustion as
follows:

) ( ) (
p t
M M = if <
d

) ( ) 1 ( ) ( ) (
d d p t
M M M + = if
d

Equation 3-41
where:

d
= start of diffusion burning

This introduces another parameter into the correlation, but after investigation it was
decided that this drawback was much more than offset by the improvement in the
predicted heat release rate.

99
Again, a least squares method was used to fit the parameters to the data in cases 1-17
(referring to Table 3-8). Cases 1-17 were used because they were from the same
source (Montgomery, 2000) and using the same equipment. The data was therefore
likely to be most consistent. Data from other sources had differences in engine
hardware which made correlating the data difficult. A comparison of the data and
fitted correlation is shown in Figure 3-18.

Watson et al.
Modified
Measurement

Figure 3-18: Example of measured heat release rate and best fits for Equation 3-36 (Watson et
al., 1980) and Equation 3-41 (labelled Modified).

The shape parameters were then examined to see whether they could be correlated to
engine operating conditions. The best correlations that could be found are shown in
Figure 3-19. Note that C
p2
was assumed to be 5000.


100
0 0.2 0.4 0.6 0.8 1
1.5
2
2.5
3

(a) C
p1
vs
10 14 18 22 26
10
20
30
40

(b) C
d1
vs
-1.07

-0.90

2.2 2.6 3 3.4 3.8
1
1.4
1.8

(c) C
d2
vs
-0.33

-0.40

2 2.2 2.4
0
0.02
0.04
0.06

(d)
d
vs C
p1

Figure 3-19: Shape factor correlations - data points and linear regression are shown.

It can be seen that the diffusion burning shape parameters C
d1
and C
d2
are reasonably
well correlated. The other parameters are more weakly correlated. Fortunately, they
vary over narrow range and have only small influence over the overall shape.

The expressions for the rate shape constants used in the model were:
C
p1
= 2.15 + 0.10
C
p2
= 5000
C
d1
= 0.079 + 1.47
-1.07

-0.90

C
d2
= -0.061 + 0.53
-0.33

-0.40

d
= -0.14 + 0.081 C
p1


How well all of these correlations agree with the measurements is shown in Section
3.3.

101
3.2.8 Turbocharging
3.2.8.1 Compressor
The turbocharger subroutine assumes a compressor map is available. The turbo speed
and the stagnation pressure ratio across the compressor are known, and a double
interpolation is done to obtain the mass flow rate and the isentropic efficiency. A
typical compressor map is shown in Figure 3-20.


Figure 3-20: Example of a compressor map.

The isentropic efficiency is used to calculate the compressor outlet temperature
according to (Joyce, 1999):
|
|
|
|
|
|
.
|

\
|
+

|
|
.
|

\
|
=

1
1
1
1
2
1 2

P
P
T T
Equation 3-42
102
where:
T = temperature
P = pressure
= ratio of specific heats
= isentropic efficiency
subscript 1 indicates inlet conditions, subscript 2 indicates outlet conditions.

The rate of work done by the compressor is calculated using:
( )
1 2
h h m W = &
&

Equation 3-43
where:
W
&
= rate of work done by the compressor
m& = mass flow rate
h = enthalpy

3.2.8.2 Turbine
The turbine mass flow parameter and isentropic efficiency were assumed functions
of the turbine pressure ratio:

|
|
.
|

\
|
=
2
1
P
P
f
P P
T T m
ref
ref
&

Equation 3-44
and
|
|
.
|

\
|
=
2
1
P
P
f
Equation 3-45
where:
T
ref
, P
ref
= reference temperature and pressure specified by the turbine manufacturer.

A turbine map which is amenable to this treatment is shown in Figure 3-21.
103

Mass flow
parameter
Pressure ratio

Figure 3-21: Example of a turbine map (values on axes omitted to protect manufacturers
proprietary information).

Fifth-order polynomials were fitted to the turbine mass flow parameter and efficiency
data. Using polynomial functions obviously simplifies and speeds up the procedure
greatly, relative to interpolating map data. Additionally, it is a more robust method,
less susceptible to failure due to excursions outside mapped regions. These are
common especially in the initial conditions before the modelled engine behaviour
settles down to a periodic function.

The turbine efficiency is used to calculate the turbine outlet temperature as follows
(Joyce, 1999):
(
(
(

|
|
|
.
|

\
|

|
|
.
|

\
|
+ =

1 1
1
2
1
1 2

P
P
T T
Equation 3-46

The rate of work done on the turbine is calculated similarly to Equation 3-43.

3.2.8.3 Turbocharger speed
The turbocharger acceleration is given by:
( )
c t
c t m
I I
W W
+
+
=

& &

Equation 3-47

Efficiency
104
where:
= angular acceleration of turbocharger

m
= mechanical efficiency of turbine-compressor connection
= angular speed of turbocharger
I = moment of inertia
subscripts t and c indicate the turbine and compressor, respectively.

The turbocharger speed at time step i is simply:
i i i
t + =
1

Equation 3-48
where:
t = the duration of the time step

In implementing this procedure in a subroutine, a history of the turbocharger speed at
recent time steps had to be stored. The numerical solver sometimes stepped forwards
and backwards several steps, so the history array had to be large enough to
accommodate the largest steps backwards. An array size of ten steps has always
proved to be sufficient.

A compressor and turbine map for a 475 hp Caterpillar 3406E engine was kindly
supplied by Mr Chris Middlemass of Garrett Turbo, Honeywell International Inc.
(Middlemass, 2002).

3.2.9 Charge air cooling
The intercooler subroutine in this model assumed constant heat exchanger
effectiveness, pressure loss factor and ambient temperature. The effectiveness and
pressure loss factor are defined as follows (Joyce, 1999):

amb in
out in
T T
T T

=
Equation 3-49

105
where:
= effectiveness
T
in
= temperature of the charge entering the charge air cooler
T
out
= temperature of the charge leaving the charge air cooler
T
amb
= ambient temperature

2 2
2
m
P
A
f
&

=


Equation 3-50
where:
f = pressure loss factor
A = flow area
= charge density
P = pressure drop across the charge air cooler
m& = mass flow rate through the charge air cooler

The charge air cooler effectiveness and pressure loss factor for a Caterpillar 3406E
engine were estimated as follows. Data for a Caterpillar 3406E475hp charge air
cooler at three operating conditions were found in Wright (2001). The data are
replicated in Table 3-6.

Table 3-6: Charge air cooler data for Caterpillar 3406E-475hp engine (Wright, 2001).
This table is not available online.
Please consult the hardcopy thesis
available from the QUT Library
106
If a compressor inlet temperature and pressure of 293.8 K and 96.8 kPa, respectively,
and an ambient pressure of 101.3 kPa are assumed, the ambient temperature can be
estimated from:

1
1
2
1 2

|
|
.
|

\
|
=
P
P
T T Equation 3-51

This gives an ambient temperature of approximately 298 K.

The compressor outlet temperatures can be estimated from Equation 3-42, which
then allows estimation of the density of the flow from the ideal gas equation
(Equation 3-3).

From these calculations, the charge air cooler effectiveness and pressure loss factor
can be estimated. The results are shown in Table 3-7.

Table 3-7: Estimated charge air cooler quantities and parameters.
Condition 1 Condition 2 Condition 3
Compressor outlet
temperature (K)
419.7 425.2 430.0
Compressor outlet
density (kg/m
3
)
2.156 2.216 2.261
Charge air cooler
effectiveness
0.844 0.814 0.822
Charge air cooler
pressure loss factor (m
-4
)
1.23210
5
1.27010
5
1.37810
5


The mean effectiveness and pressure loss factor of the charge air cooler over the
three conditions were 0.827 and 1.29310
5
, respectively. These values were used in
the charge air cooler subroutine.

107
3.2.10 Mechanical friction
Several simple correlations reported in Stone (1992) have been tried. The one used in
this model has the form:

p
v c
b
P
a fmep + + =
max

Equation 3-52
where:
fmep = friction mean effective pressure
P
max
= maximum cylinder pressure
p
v = mean piston speed
a, b, c = calibration constants

Up to this point, the model is sufficiently well-developed to estimate the indicated
mean effective pressure (imep) and P
max
for the Caterpillar SCOTE engine. The
difference between the imep and the measured brake mean effective pressure (bmep)
can be assumed to be the fmep.

The calibration constants from these data were determined to be:
a = 1.23
b = 193.7
c = 0.063

The correlation is compared with the data inferred from Cases 1-17 (Table 3-8) in
Figure 3-22.
108

1000 1400 1800
2.1
2.2
2.3
2.4
Correlation
Data
Fmep
(bar)
Speed
(rpm)

Figure 3-22: Estimated fmep and correlation.

3.2.11 Automated parametric investigation
The initial intention was to develop an automatic optimisation scheme. Optimisation
generally involves maximising or minimising an objective function, subject to
constraints. For example, if minimum fuel consumption was desired, the problem
might be stated:

f(X) = bsfc
where:
X = {x
1
, x
2
x
n
} which is an array of parameters such as start of
injection, boost pressure, inlet valve open timing etc.
bsfc = brake specific fuel consumption
Find X which minimises f(X)
subject to p constraints:
g
i
(X) = 0, i = 1, 2m
h
j
(X) < 0, j = m+1, m+2p
Equation 3-53

109
Constraints could include maximum cylinder pressure and maximum exhaust,
coolant and component temperatures.

It is helpful in optimisation to have some understanding of the behaviour of the
objective function. The simplest way to achieve this is to evaluate the objective
function over a range of all the input parameters. The parameter space can be limited
if it is known in advance that there is little merit in searching outside certain
boundaries. For parameters such injection timing, the range can be known reasonably
well in advance.

The parameters to be searched are put in a table. The minimum and maximum values
are entered, as well as the number of steps. An objective function and constraints
must be built into the output subroutine. The program then automatically evaluates
the objective function for every combination of the parameters. The minimum or
maximum is noted, as well as violation of any constraints.

Optimisation file
110.0 110.0 1 | EVO (deg ATDC)
270.0 270.0 1 | IVC (deg ATDC)
30.0 30.0 1 | IVO - EVO (deg)
20.0 20.0 1 | IVC - EVC (deg)
350.0 355.0 3 | SOI (deg ATDC)
194.0 194.0 1 | Air pump bore (mm)
-180. -120. 7 | Air pump phase (deg)

Figure 3-23: Sample file specifying parameter values to be investigated. The first column is the
initial values, the second column is the final values, the third column specifies the number of
steps, and the fourth column describes the parameters.

110
3.3 Validation
3.3.1 Caterpillar SCOTE data
Table 3-8: Cases for calibration/validation of the O-D model.
Case
number
Engine
speed
(rpm)
Fuel rate
(kg/hr)
Start of
injection
(ATDC)
Intake
temp (K)
Intake
pressure
(kPa)
Exhaust
pressure
(kPa)
EGR rate
(%)
Ref.
1 993 6.276 -5.5 301 161 137 0 1
2 993 6.330 -3.5 301 161 137 0 1
3 993 6.288 -7.5 301 161 137 0 1
4 993 6.306 -5.5 301 161 137 0 1
5 993 6.312 -9.5 301 161 137 0 1
6 1737 6.930 1.5 305 184 181 0 1
7 1737 6.936 -3.5 305 184 181 0 1
8 1737 6.876 2.5 305 184 181 0 1
9 1737 6.972 -0.5 305 184 181 0 1
10 1737 6.876 3.5 305 184 181 0 1
11 1737 7.002 1.5 305 184 181 0 1
12 1789 3.846 -5.5 301 121 135 0 1
13 1789 3.732 -9.5 301 121 135 0 1
14 1789 3.810 -3.5 301 121 135 0 1
15 1789 3.786 -1.5 301 121 135 0 1
16 1789 3.816 -7.5 301 121 135 0 1
17 1789 3.828 -5.5 301 121 135 0 1
18 1600 8.064 -7.0 2
19 1600 8.064 -4.0 2
20 1600 8.064 -1.0 2
21 1600 8.064 2.0 2
22 1600 8.064 5.0
Conditions at 144 BTDC:
Pressure = 212.1 kPa
Temperature = 361.4 K
Burned gas fraction 0.007
2
23 1690 3.054 -9.0 2
24 1690 3.054 -6.0 2
25 1690 3.054 -3.0 2
26 1690 3.054 0.0 2
27 1690 3.054 3.0
Conditions at 144 BTDC:
Pressure = 148.4 kPa
Temperature = 334 K
Burned gas fraction 0.007
2
28 1737 6.804 -7.5 306 181 191 18.3 1
29 1737 6.798 -4.5 309 181 191 19.4 1
30 1737 6.786 -1.5 310 181 191 19.6 1
31 1737 6.810 -1.5 305 181 191 0 1
32 1737 6.834 -5.0 305 181 191 0 1
33 1737 6.870 -7.5 305 181 191 0 1
34 1789 3.84 -5.5 310 103 115 26.9 1
35 750 0.522 -8.0 299 100 100 0 3
36 953 2.022 -0.5 302 108 112 0 3
37 1657 10.140 7.5 313 239 220 0 3
Reference 1: Montgomery, D.T. (2000) University of Wisconsin-Madison PhD Thesis
Reference 2: Kong, S-C (2002) Personal communication.
Reference 3: Wright, C. C. (2001) University of Wisconsin-Madison MSc. Thesis
111
Table 3-9: Experimental and calculated results.
Ignition
delay
(ms)
Premixed
burn fract.
(%)
Air flow
rate
(kg/min)
Maximum
pressure
(MPa)
Brake
power (kW)
BSFC
(g/kWh)
Ref.
Exp Cal
Exp Cal Exp Cal Exp Cal Exp Cal Exp Cal
1* 0.44 0.52 8.0 8.8 1.77 2.16 10.3 10.6 29.6 29.5 212 212 1
2* 0.42 0.50 8.1 8.4 1.77 2.16 9.4 9.8 29.0 29.5 218 215 1
3* 0.42 0.54 8.5 9.4 1.76 2.15 11.0 11.5 30.0 29.8 209 211 1
4* 0.42 0.51 8.0 8.7 1.77 2.16 10.2 10.6 29.6 29.7 213 212 1
5* 0.46 0.58 8.5 10.2 1.75 2.16 11.7 12.4 30.3 30.1 208 210 1
6* 0.45 0.45 9.1 10.4 4.20 4.16 8.1 8.1 28.4 27.9 244 251 1
7* 0.38 0.42 7.2 9.4 4.16 4.16 8.4 8.7 30.3 29.6 229 234 1
8* 0.48 0.47 10.2 11.1 4.17 4.16 8.1 8.1 27.6 27.6 249 249 1
9* 0.40 0.43 8.1 9.6 4.20 4.16 8.1 8.1 29.1 29.0 240 241 1
10* 0.54 0.49 11.5 11.9 4.21 4.16 8.1 8.1 26.9 27.3 256 252 1
11* 0.45 0.45 9.4 10.4 4.17 4.16 8.1 8.1 28.4 28.5 247 246 1
12* 0.64 0.65 28.4 27.0 3.15 2.8 6.3 6.4 12.7 12.9 303 299 1
13* 0.70 0.70 31.3 30.9 3.15 2.8 7.0 7.3 12.7 12.4 294 302 1
14* 0.67 0.65 29.2 27.0 3.15 2.9 6.2 6.1 12.4 12.5 306 305 1
15* 0.69 0.67 31.5 28.3 3.15 2.8 5.8 5.8 11.9 12.2 317 311 1
16* 0.60 0.67 28.6 28.3 3.14 2.8 6.7 6.9 12.7 12.8 300 298 1
17* 0.60 0.65 28.1 27.0 3.13 2.8 6.5 6.4 12.7 12.8 301 300 1
18 0.6 0.41 8.8 7.5 - - 11.1 10.1 - - - - 2
19 0.4 0.39 9.5 7.0 - - 9.7 8.9 - - - - 2
20 0.4 0.39 9.7 7.1 - - 9.1 8.4 - - - - 2
21 0.5 0.42 11.1 7.8 - - 8.6 8.4 - - - - 2
22 0.6 0.48 13.5 9.4 - - 8.6 8.4 - - - - 2
23 0.8 0.67 27.2 30.2 - - 7.9 7.5 - - - - 2
24 0.7 0.63 40.1 27.3 - - 7.5 6.9 - - - - 2
25 0.7 0.62 40.4 26.9 - - 7.1 6.5 - - - - 2
26 0.8 0.66 45.5 29.4 - - 6.1 6.0 - - - - 2
27 0.9 0.74 66.9 36.0 - - 6.1 6.0 - - - - 2
28 0.05 0.47 - 11.1 3.32 3.30 9.3 9.6 31.1 29.1 219 233 1
29 0.10 0.44 - 10.1 3.26 3.23 8.7 8.6 29.6 28.4 230 239 1
30 0.15 0.44 - 10.1 3.22 3.21 8.4 8.1 27.9 27.5 244 246 1
31 0.09 0.43 - 9.8 4.07 4.07 8.6 8.2 28.8 28.0 236 243 1
32 0.14 0.43 - 9.9 4.08 4.07 8.9 8.8 29.6 29.0 231 235 1
33 0.06 0.46 - 10.6 4.08 4.07 9.4 9.6 30.3 29.7 227 231 1
34 0.57 0.80 28.0 38.2 1.74 1.73 6.1 6.0 13.0 12.5 297 306 1
35 1.24 0.88 77.0 65.0 ? 1.0 5.6 5.8 0.3 0.2 1700 2781 3
36 0.91 0.75 75.9 32.5 ? 1.3 6.8 5.8 7.6 7.4 266 274 3
37 0.50 0.44 13.1 7.6 ? 5.1 10.0 10.5 38.4 40.9 264 248 3
*These experiments were used for model calibration. Other experiments used
different injectors, and in cases 25-34 there may have been additional differences.



112
The calculated and experimental pressure and heat release rate curves are shown.

Experimental
Calculated

-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(a) Case 1
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(b) Case 2
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(c) Case 3
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(d) Case 4
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(e) Case 5

Figure 3-24: Mode 3 (high load, mid speed). Case numbers refer to Table 3-8.
113
Experimental
Calculated

-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)
(a) Case 6
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(b) Case 7
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(c) Case 8
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(d) Case 9
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(e) Case 10
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(f) Case 11
Figure 3-25: Mode 5 (mid load, high speed) . Case numbers refer to Table 3-8.

.
114
Experimental
Calculated

-1
0
1
2
3
4
5
6
7
8
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(a) Case 12
-1
0
1
2
3
4
5
6
7
8
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(b) Case 13
-1
0
1
2
3
4
5
6
7
8
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(c) Case 14
-1
0
1
2
3
4
5
6
7
8
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(d) Case 15
-1
0
1
2
3
4
5
6
7
8
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(e) Case 16
-1
0
1
2
3
4
5
6
7
8
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(f) Case 17
Figure 3-26: Mode 6 (low load, high speed). Case numbers refer to Table 3-8.

Note: the experimental pressure trace in this table was scaled down to give approximately the same
pressure at 30 BTDC as the calculation. It is believed that the original pressure data was incorrectly
calibrated because the reported data was inconsistent with calculations based on the reported intake
pressure, the reported air flow rate, and the reported pressure data for other runs with similar intake
pressures. Scaling the reported pressure data by a factor of 0.77 brought the peak pressure and the air
consumption into line with the reported intake pressure and other reported results. Pressure data in
other tables were not scaled.


115
Experimental
Calculated

-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(a) Case 34, Mode 6 with high pressure injector
and 26.9% EGR
-1
0
1
2
3
4
5
6
7
8
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.04
0
0.04
0.08
0.12
0.16
0.2
0.24
0.28
0.32
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(b) Case 35, Mode 1 (idle)
-1
0
1
2
3
4
5
6
7
8
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.04
0
0.04
0.08
0.12
0.16
0.2
0.24
0.28
0.32
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(c) Case 36, Mode 2 (low load, low speed)
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50 60
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(d) Case 37, Mode 4 (high load, mid speed)
Notes:
1. 100
2 % 2 %
2 % 2 %
%

=
ambient exhaust
ambient inlet
CO CO
CO CO
EGR
2. For Fuel B composition, air consumption total intake (1 EGR fraction)
Figure 3-27: Other comparisons using different engine hardware. Case numbers refer to Table
3-8.



116
Experimental
Calculated



Figure 3-28: Cases 18-22 (75% load, 1600 rpm) (Kong, 2002). Case numbers refer to Table 3-8.




This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
117
Experimental
Calculated



Figure 3-29: Cases 23-27 (25% load, 1690 rpm) (Kong, 2002). Case numbers refer to Table 3-8.


3.3.2 Caterpillar 3406E data
Wright (2001) published some data for a production CAT3406E-500hp engine. The
data represents measurements at maximum load and high speed for the engine
system including turbocharger and charge air cooler. The operating conditions are
reported in Table 3-10 and the calculated and experimental results are compared in
Table 3-11 and Table 3-12. The first table represents modelling of the six cylinder
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
118
engine block without the turbocharger and charge air cooler (Figure 3-30a). The
boundary conditions were estimated from Wrights reported data. The second table
represents calculated results for the complete model, including turbocharger and
charge air cooler (Figure 3-30b).

Note that the turbocharger data is for a CAT3406E-475hp engine that has a rated
speed of 1800 rpm. This is the only turbocharger data that could be obtained for this
study. It was expected to be similar to the CAT3406E-500hp turbocharger.

(a) Model for Table 3-11.



(b) Model for Table 3-12.

Figure 3-30: Schematic representations of models for Caterpillar 3406E simulations.
Plenum Plenum
Intake
poppet
valves
Exhaust
poppet
valves
Cylinders
Compressor
Plenum
Plenum Plenum
Intake
poppet
valves
Exhaust
poppet
valves
Cylinders
Turbine
Plenum
Charge
air
cooler
119

Table 3-10: CAT3406E-500hp engine operating conditions. The first five rows are from Wright
(2001, p. 62). The remainder are assumed or estimated as described in Section 3.2.9.
Engine speed (rpm) 1700 1900 2100
Fuel rate (kg/min) 1.241 1.254 1.233
SOI ( ATDC) -7.1 -10.6 -15.7
Assumed compressor intake
pressure (kPa)
96.8 96.8 96.8
Assumed compressor intake
temperature (K)
293.8 293.8 293.8
Estimated intake manifold
pressure (kPa)
268.5 258.7 246.6
Estimated intake manifold
temperature (K)
320 322 317
Estimated exhaust manifold
pressure (kPa)
225.3 234.6 241.9
Brake power (kW) 364 365 359


Table 3-11: Comparison of measured and calculated data for CAT3406E engine model without
turbcharger or charge air cooler.
Engine speed (rpm) 1700 1900 2100
Exp. Calc. Exp. Calc. Exp. Calc.
Brake power (kW) 364 355 365 359 359 347
Air flow (kg/min) 35.2 34.7 37.2 36.8 38.4 38.9

120
Table 3-12: Comparison of measured and calculated data for CAT3406E engine model with
turbcharger and charge air cooler.
Engine speed (rpm) 1700 1900 2100
Exp. Calc. Exp. Calc. Exp. Calc.
Brake power (kW) 364 346 365 349 359 336
Air flow (kg/min) 35.2 34.8 37.2 37.0 38.4 38.0
Turbo speed (1000 rpm) 85.5 87.2 85.8 87.1 85.4 85.9
Turbine pressure ratio 2.23 2.67 2.32 2.76 2.38 2.76
Compressor pressure ratio 2.88 2.96 2.79 2.86 2.68 2.70
Turbine efficiency % 72.6 69.7 69.8 69.2 67.8 69.3
Compressor efficiency % 76.2 74.7 76.3 74.5 76.0 73.7
Intercooler pressure loss (kPa) 10.4 9.6 11.0 11.1 11.7 12.2
Intercooler temp. drop (K) 112 114 107 109 106 101


3.4 Discussion
The thermodynamic model has been calibrated for the Caterpillar SCOTE and
3406E-500hp engines at the University of Wisconsin-Madison Engine Research
Center, based on data published by Montgomery (2000) and Wright (2001).

The agreement with the Caterpillar SCOTE calibration cases (cases 1-17 in Table
3-9) is clearly very good, with most quantities being within 5% of experimental
values. There is generally an underprediction of pressure dip due to fuel injection.
This could be remedied by increasing the fuel enthalpy of formation. This would also
bring the overall pressure rise due to combustion more into line with the
experimental data. However, it would also reduce the net heat of combustion, and
presently this agrees very well with the Fuel B analysis (Table 3-1).

As expected, the more that the engine hardware departs from that used in the
calibration runs, the poorer the agreement. However, the trends are still reflected in
the calculations. The difference between the experimental and calculated pressure
traces in cases 34-37 are thought to be errors in the reported data rather than errors in
the calculation. Wave effects are unlikely owing to the relatively short intake ports
121
and low speeds in cases 35 and 36. Also, there is generally excellent agreement
between measurement and calculation during the compression strokes in the other
cases that are at similar speeds.

Similarly, the discrepancies between the reported and calculated air flow rates in
cases 1-5 and 12-17 are difficult to explain, because all of the other results are in
such good agreement. The experimental results could be in error, because a small
error in the charging efficiency should be magnified in the pressure trace. Sometimes
the agreement is excellent, as in cases 6-11 and 28-34.

In Section 3.2.3 it was argued that the maximum cylinder temperatures seldom
exceeded 2000 K and the pressure was also high, so that dissociation could be
neglected. Reviewing all of the cases, the maximum temperatures were calculated for
cases 1-5, in which the maximum bulk gas temperature was 2060 K at a pressure of 9
MPa (90 bar). Looking at Figure 3-3 and Figure 3-4, the errors in the specific internal
energy and specific gas constant are very small, much less than 1%. In all of the
other cases, including the Caterpillar 3406E simulations, the maximum bulk gas
temperature was less than 1800 K. The decision to simplify the gas property
calculations by neglecting the effects of pressure (i.e. dissociation) would appear to
be justified, provided future simulations do not exceed these temperatures.

The model generally under-predicted the brake power output of the Caterpillar
3406E engine (Table 3-11 and Table 3-12). In the case of Table 3-11, in which the
engine block only was modelled, the major source of error could be the friction
correlation, which was unchanged from the Caterpillar SCOTE simulations. The
parasitic losses in the Caterpillar SCOTE engine are expected to be greater per unit
displacement than in the Caterpillar 3406E engine, because the single cylinder has to
drive all of the ancillary equipment, such as the fuel system, valve train, oil and
coolant pumps etc. Additionally, errors in the estimated boundary conditions would
have an effect.

In the case of Table 3-12, in which the entire engine was modelled, including charge
air cooler and turbocharger, the errors listed above would have been compounded by
the differences between the correct turbocharger and the turbocharger for which data
122
was available. Nevertheless, the errors are relatively small, considering the model
was calibrated solely on a single-cylinder version of the engine. (The charge air
cooler model was calibrated on the Caterpillar 3406E data).

Comparison of the simulations and experimental data give good confidence that the
model will give useful results for the predicted behaviour of engine systems based on
the Caterpillar 3406E engine. The accuracy of the adaptation of the model to two-
stroke operation is obviously contingent upon the development of a suitable model
for two-stroke poppet-valved cylinder scavenging.
Chapter 4 - KIVA-ERC multidimensional model
adaptation
4.1.1 KIVA package overview
KIVA was selected for the multidimensional modelling phase of this study for
several reasons, including:

a) It contains specialised engine modelling features such as moving pistons and
valves and modelling of diesel spray formation, evaporation, ignition and
combustion.
b) The source code comes with the package, allowing the calculation methods to
be known exactly, enabling troubleshooting of difficulties and modifications
to suit requirements.
c) Independently-developed subroutines, such as those developed at the
University of Wisconsin-Madison Engine Research Center (ERC), are readily
available. These provide alternatives to the original subroutines for the
estimation wall heat transfer and the formation of NOx and PM.
d) Low cost, no yearly maintenance fees.

Another advantage, not known at the time of selection, was the availability of
meshes, experimental data and input files for the Caterpillar SCOTE engine, which
was the engine used to validate many of the ERC subroutines.

KIVA is a package that numerically solves the unsteady equations of motion of a
turbulent, two- or three-dimensional, chemically reactive mixture of ideal gases with
a single-component vaporising fuel spray (Amsden et al., 1989). The computational
domain is divided into hexahedral finite volumes. The vertices of the mesh may
move with time. The gas flow is solved using an arbitrary Lagrangian-Eulerian
(ALE) method. During the Lagrangian phase, the vertices are assumed to move with
the fluid velocity (i.e. there is no convection across the fluid boundaries). In the
Eulerian phase, the flow field is frozen, the vertices are moved to new user-specified
positions, and the flow field is remapped to the new mesh by convecting material
across the cell boundaries.
124

Extensive use of implicit formulations is used to enhance the overall stability of the
solver, which generally allows the use of larger time steps based on the desired
accuracy of the solution. The main time step size is calculated from several accuracy
conditions related to mesh size, mesh distortion limits, rate of chemical heat release
and rate of mass and energy exchange with the fuel spray. There are also time steps
for convection subcycles, and their size is determined by the Courant stability
condition which relates the mesh size and fluid velocity relative to the mesh.

KIVA is described in detail in Amsden et al. (1989); Amsden (1993); Amsden
(1997) and Amsden (1999).

4.2 ERC Spray and Combustion Model Library
The finest practicable mesh for engine simulations cannot adequately resolve many
important phenomena, such as gas turbulence, spray formation, combustion and
emissions formation. Semi-empirical submodels are used to estimate these processes
occurring within each small finite volume element. The relatively small scale of
these elements allows far greater reliance on the fundamental physics than the single-
zone thermodynamic model.

The University of Wisconsin-Madison Engine Research Center (ERC) has developed
a library of submodels for use with KIVA called the ERC Spray and Combustion
Model Library (ESC-lib). A license was granted for the use of the library in this
project. The submodels are based largely on experimental work using the Caterpillar
SCOTE and similar engines. ESC-lib should therefore be particularly suitable for this
study.

The subroutines are compiled with KIVA and complement or replace the standard
submodels that come with the KIVA package. KIVA-ERC is used to distinguish
KIVA compiled with ESC-lib from the standard KIVA release. A brief description of
each ESC-lib submodel is included in the following sections.
125
4.2.1 Turbulence
A modified renormalised group (RNG) k- model is used. This was first developed
for KIVA by Han and Reitz (1995) at the ERC. It has been incorporated as a standard
option in KIVA since 1997, and is not actually a part of the ESC-lib. This model has
been shown to predict large-scale flame structures, and therefore the in-cylinder
temperature field and NOx formation, better than the original k- turbulence model.
Features of this submodel include the accounting for the effects of compressibility
and spray-turbulence interaction.

4.2.2 Heat transfer
Han and Reitz (1997) developed a gas-wall convective heat transfer model through a
temperature wall function. This allows reasonable accuracy with relatively coarse
grids. It is based on the one-dimensional energy conservation equation. Variations in
gas density and the turbulent Prandtl number in the boundary layer are accounted for.
The expression for heat transfer is:

( ) 513 . 2 ln 1 . 2
ln
+

=
y
T
T
T u c
q
w
p
w


Equation 4-1

y u
y

=
Equation 4-2
2
1
4
1
k C u

= Equation 4-3
where:
q
w
= wall heat flux
= density
c
p
= specific heat at constant pressure
T, T
w
= flow temperature, wall temperature
y = distance from the wall
= molecular viscosity
C

= RNG k- turbulence model constant (value = 0.0845)


k = turbulent kinetic energy
126

4.2.3 Atomisation and drop drag
Except where noted, the rate of injection was calculated using Equation 3-30. The
validation runs used the measured injection rate shapes. KIVA automatically scaled
the user-specified rate shapes give the user-specified quantity of fuel injected per
cycle. The injection velocity is calculated from the injection rate, the nozzle diameter
and the nozzle discharge coefficient. The latter was set to a nominal value of 0.7.
This is the default value, and is suggested in Han et al. (1996) based on experimental
results.

The physical fuel is emitted from the nozzle as a continuous jet. The jet is discretised
in KIVA as a series of fuel parcels or blobs. The initial diameter of the blobs is
assumed to be the same as the effective diameter of the nozzle:

D noz eff
C d d =
Equation 4-4
where:
d
eff
= effective diameter
d
noz
= nozzle diameter
C
D
= nozzle discharge coefficient

Liquid fuel jets are observed to have a liquid core for a certain length, called the
break-up length. Surrounding the core are liquid fuel droplets that have been stripped
off the liquid jet. The droplet stripping occurs because of the high relative velocities
at the liquid-gas interface.


127

Figure 4-1: Fuel injection and atomisation within the break-up length (Reitz and Diwakar,
1987).

This process is modelled using the concept of Kelvin-Helmholtz instability, in which
initial perturbations in the liquid jet grow at a rate that is a function of the
wavelength. The maximum growth rate and its corresponding wavelength are
estimated. The radius of the droplets stripped off the blobs is assumed to be
proportional to this wavelength. The diameter of the original blob is reduced as fuel
droplets are stripped off. The rate of change of the parent blob is assumed to follow
the rate equation:
( )
0
0 0
r r
r r
dt
dr


Equation 4-5

=
0 1
726 . 3 r B

Equation 4-6
( )
0 0 0
r B B r =
Equation 4-7
This figure is not available online.
Please consult the hardcopy thesis
available from the QUT Library
128
where:
r
0
= radius of parent blob
r = droplet radius
= break-up time
= maximum wave growth rate
= wavelength corresponding to
B
0
, B
1
= constants

This model is detailed by Reitz and Diwakar (1987).

Beyond the break-up length, the fuel is a spray consisting of many small droplets
with fuel vapour and entrained gas. In this regime, Rayleigh-Taylor instability is
assumed to occur as well as the Kelvin-Helmholtz instability described above.
Rayleigh-Taylor instability is also a surface wave instability caused by an
acceleration perpendicular to the interface between two fluids of different densities.
If the growth time of the Rayleigh-Taylor waves exceeds a certain break-up time, the
droplet disintegrates. The Rayleigh-Taylor break-up mechanism competes with the
Kelvin-Helmholtz break-up mechanism beyond the break-up length.

Incidentally, KIVA models droplet collision and dispersion as described in Amsden
et al. (1989). If two spray particles, which represent groups of droplets, are in the
same computational cell, then they either collide or not according to a random
process following a Poisson distribution. If they collide, they either coalesce or graze
according to an impact parameter that is based on the droplet sizes and Weber
number. The Weber number is a dimensionless parameter related to the ratio of the
droplet inertia and surface tension forces. Droplet dispersion due to turbulence is
estimated by imposing a randomly-fluctuating velocity following a Gaussian
distribution with a mean square deviation proportional to the specific turbulent
kinetic energy.

Droplet drag was initially modelled in KIVA by assuming the droplets were
spherical. Liu and Reitz (1993) observed that droplets in air with a large relative
velocity were flattened so that they had a higher drag coefficient than spheres. A
129
spring-mass analogy is used to estimate the distortion of a droplet (the Taylor
Analogy Break-up model described in Amsden et al., (1989)). The drag coefficient is
then approximated as a function of a droplet distortion parameter. This model is
summarised in Kong et al. (1995).

4.2.4 Fuel/wall impingement
Fuel droplets may impact on liner or piston surfaces, particularly where high-
pressure injectors and bowl-in-piston geometries are used. When this happens,
droplets may break up, suddenly vaporise, form a thin liquid film, rebound or slide
along the surface (Kong et al., 1995).

The impingement submodel considers only rebounding or sliding phenomena.
Droplets rebound if:

0
2
80
d V
We
We
n
i
i
=
<

Equation 4-8
where:
We
i
= incident Weber number
= droplet density
V
n
= droplet velocity normal to the surface
d
0
= droplet diameter
= surface tension

If this condition is not satisfied, droplets are assumed to slide.

After impingement, droplets have been observed to be highly distorted and tend to
break-up sooner. This is modelled by reducing the break-up time constant B
1
. This
also has the effect of bringing the post-impingement droplet sizes into line with
experiment.

130
4.2.5 Ignition
ESC-lib uses a multistep kinetics combustion model called the Shell combustion
model. This model was proposed by Halstead et al. (1977) to account for cool flame
and two-stage ignition phenomena that are observed during the autoignition of
hydrocarbons. Eight reactions represent the transition from fuel through intermediate
species to oxidised products. One reaction in particular appears to govern the
transition from cool ignition to hot ignition. The pre-exponential constant of this
reaction rate is adjusted to match experimental ignition delay periods (Kong et al.,
1995).

4.2.6 Combustion
The Shell ignition model is considered suitable for low temperature chemistry. Once
the temperature reaches 1000-1100 K, a combustion model is activated. The
combustion model uses a characteristic time formulation that has been used in spark
ignition combustion modelling. Spark ignition engines often have a homogeneous
premixed charge, in which the rate of combustion is governed by laminar and
turbulent flame speeds. Diesel combustion is obviously very different, with fuel
evaporating from many small droplets, resulting in a very heterogeneous mixture.
However, combustion in both cases is assumed to be dominated by laminar
phenomena initially, then increasingly governed by turbulence (Kong et al., 1995).

Seven species are considered: fuel, oxygen, nitrogen, carbon dioxide, carbon
monoxide, hydrogen and water. Six species are considered reactive (all but nitrogen).
The rate of change of mass fraction of species m is formulated as:

c
m m m
t
Y Y
dt
dY *
=
Equation 4-9
131
where:
Y
m
= mass fraction of species m
Y
m
* = thermodynamic equilibrium mass fraction of species m
t
c
= characteristic time to reach equilibrium

The characteristic time is assumed the same for all species, and is a combination of a
laminar timescale and turbulent timescale:
t
c
= t
l
+ ft
t

Equation 4-10
where:
t
l
= laminar timescale
t
t
= turbulent timescale
f = delay coefficient, controlling the significance of turbulence

The laminar timescale is derived from experiments, and the turbulent timescale is
proportional to the eddy turnover time:

k
C t
t 2
=
Equation 4-11
where:
C
2
= 0.1 if the RNG k-e model is used.

The coefficient f is calculated by:
1
1
1

=
e
e
f
r

Equation 4-12
where:
2
2 2 2
1
N
H CO O H CO
Y
Y Y Y Y
r

+ + +
=
Equation 4-13
132

r indicates the completeness of the fuel-air reaction, varying from 0 (no combustion
products) to 1 (all combustion products).

4.2.7 NO
x
formation
The well-known extended Zeldovich mechanism is used. This consists of the
following series of reactions:
O + N
2
NO + N
Equation 4-14
N + O
2
NO + O
Equation 4-15
N + OH NO + H
Equation 4-16

Heywood (1988) expressed the rate of formation of NO as a single equation:

[ ]
[ ][ ]
[ ] [ ][ ] ( )
[ ] [ ] [ ] ( ) OH k O k NO k
N O K NO
N O k
dt
NO d
f f b
f
3 2 2 1
2 2
2
2 1
1
1
2
+ +

=
b
f
b
f
k
k
k
k
K
2
2
1
1
=
Equation 4-17
where:
k = reaction rate constant
subscripts f and b indicate forward and reverse reaction directions
square bracket indicate molar (or volumetric) concentrations.

The rate constants recommended by Bowman (1975) are used. The resultant NO is
multiplied by 1.533 to account for the transformation of NO to NO
x
in the tailpipe
and atmosphere.

4.2.8 Soot
Soot production is modelled as two separate processes: soot formation from fuel and
soot consumption through oxidation. Soot formation is modelled as an Arrhenius rate
equation:

133
fv
sf
sf sf
M
RT
E
P A M

= exp
2
1
&

Equation 4-18
where:
sf
M
&
= rate of soot formation
A
sf
= pre-exponential constant
P = pressure
E
sf
= the apparent activation energy
R = the molar ideal gas constant
T = temperature
M
fv
= fuel vapour mass

The Nagle and Strickland-Constable (1962) model is used for soot oxidation. It is
based on measurements of the oxidation of graphite. The soot surface is assumed to
have reactive A sites and less reactive B sites. Some A sites are converted to B
sites. The reaction rate is given by:

( ) X P K X
P K
P K
R
O B
O Z
O A
tot
+
+
= 1
1
2
2
2

Where X is the proportion of A sites:
B T O
O
K K P
P
X
+
=
2
2

Equation 4-19

P
O2
= Oxygen partial pressure
K = various experimentally-determined rate constants

The soot mass oxidation rate is:

tot s
s s
C
so so
R M
d
MW
C M

6
=
&

Equation 4-20
134
where:
MW
C
= the molar weight of carbon

s
= soot density
d
s
= nominal soot particle diameter
C
so
= constant

The parameter A
sf
is used to tune the model to experimental results.

4.2.9 Computational meshes
Two computational meshes representing the Caterpillar SCOTE engine were kindly
made available by the Engine Research Center at the University of Wisconsin-
Madison. The smaller mesh was a 60 sector mesh. The Caterpillar SCOTE injectors
have six equispaced nozzles, and when the valves are closed the combustion chamber
therefore has six-fold symmetry. Modelling just one-sixth of the cylinder allows
large reductions in computational memory and CPU time requirements, and/or a
reduction in the mesh cell size and increase in accuracy. The mesh is shown in
Figure 4-2. It has nearly 19,000 cells and 21,000 vertices.

135

Figure 4-2: 60 sector mesh of Caterpillar SCOTE cylinder. The mesh was supplied by the
Engine Research Center at the University of Wisconsin-Madison.

For intake and scavenging simulations it was necessary to model the entire cylinder,
the valves and a substantial portion of the intake and exhaust ports. Fortunately, such
a mesh of the Caterpillar SCOTE engine existed. It is shown in Figure 4-3, and has
approximately 150,000 cells and 160,000 vertices. KIVA by default allows up to
50,000 vertices, so modifications had to be made to the source code to allow arrays
of sufficient size. Files containing the valve lift profiles were bundled with the
model. These valve lift profiles were used in the zero-dimensional simulation
described in the preceding chapter.

136

Figure 4-3: Computational mesh of Caterpillar SCOTE engine at BDC. The four valves and the
mexican hat bowl-in-piston geometry can be discerned. The mesh was supplied by the Engine
Research Center at the University of Wisconsin-Madison.

4.3 Validation
Validation of the KIVA-ERC multidimensional model was done using the same
experimental data reported by Montgomery (2000) as was used to validate the zero-
dimensional model. Specifically, the model results were compared to cases 1-17
described in Section 3.3.1. The procedure for validation was as follows:

a) For each case, the zero-dimensional thermodyncamic model was run through
ten cycles. The cylinder pressure, temperature and burnt gas fraction at IVO
(25 BTDC) were noted. The burnt gas fraction was converted to mass
fractions of oxygen, nitrogen, carbon dioxide and water.

137
b) A KIVA-ERC simulation using the full mesh, the initial conditions estimated
by the zero-dimensional model and the boundary conditions reported by
Montgomery was run until IVC (217 ATDC).

c) A KIVA-ERC simulation using the sector mesh was run from IVC until just
before SOI. The initial cylinder temperature, pressure, burnt gas fraction,
swirl ratio and turbulence parameters at IVC were obtained from the previous
KIVA-ERC full mesh simulation. The temperature and pressure were very
close to those predicted by the 0-D simulation. KIVA-ERC predicted a
slightly higher burnt gas fraction, perhaps because it could account for the
piston geometry, which traps residual gases.

d) KIVA-ERC simulations using the sector mesh were continued from the
previous step, but with a finer time step. This was required to model the
injection and combustion processes. This step was repeated several times,
with small adjustments made to model parameters until satisfactory
agreement between calculation and experiment was reached.

Despite the complexity of the KIVA-ERC model, only a few parameters are adjusted
to account for variations in fuel properties and engine hardware. All but one are
associated with ESC-lib submodels. They are summarised in Table 4-1. The
parameters are adjusted in the same order as the combustion process; for example,
the ignition delay parameter is adjusted before the combustion parameters, and the
combustion parameters are adjusted before the emissions parameters. The parameters
distant and cnst22 are only adjusted if a satisfactory match cannot be achieved
by adjusting the first three parameters alone.







138
Table 4-1: Adjustable parameters in KIVA-ERC model, adapted from Hessel (2003) and Kong
(2002).
Parameter
name
Effect Typical value
or range
Model
value
af04 A larger value speeds up ignition,
reducing the ignition delay period.
10
4
5.010
5
1.2510
5

denomc A larger value increases the proportion
of premixed combustion.
0.76810
10
0.76810
10

cm2 Turbulent timescale constant. A larger
value slows diffusion burning.
0.1 5.0 0.25
distant Initial break-up length 1.4 2.0 1.9
cnst22 Droplet break-up time constant before
impingement. Smaller value makes
smaller droplets which vaporise faster.
10 - 60 40
asf A larger value increases soot oxidation
rate, reducing the total soot produced
- 250
rsc Affects NO
x
formation 1.45 2.0 1.47

The model results are compared with Montgomerys data in Figure 4-4, Figure 4-5
and Figure 4-6.














139
Experimental Calculated
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(a) Case 1
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(b) Case 2
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(c) Case 3
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(d) Case 4
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(e) Case 5

3
5 4 1
2
1 3 2
4
5
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60
NOx (g/kg fuel)
S
o
o
t

(
g
/
k
g

f
u
e
l
)
Measured
KIVA

(f) Soot vs NO
x
emissions for Cases 1-5
Figure 4-4: KIVA-ERC model results and experimental data for Mode 3 (75% load, 993 rpm) .
Case numbers refer to Table 3-8.
140
Experimental Calculated
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(a) Case 6
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(b) Case 7
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(c) Case 8
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(d) Case 9
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(e) Case 10
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(f) Case 11
7
9
10
11
8
6
11 6
7
9
8
10
0
0.2
0.4
0.6
0.8
0 10 20 30 40 50 60
NOx (g/kg fuel)
S
o
o
t

(
g
/
k
g

f
u
e
l
)
Measured
KIVA

(g) Soot vs NO
x
emissions for cases 6-11
Figure 4-5: KIVA-ERC model results and experimental data for Mode 5 (57% load, 1737 rpm) .
Case numbers refer to Table 3-8.
141
Experimental Calculated
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(a) Case 12
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(b) Case 13
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(c) Case 14
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(d) Case 15
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(e) Case 16
-2
0
2
4
6
8
10
12
14
-30 -20 -10 0 10 20 30 40 50
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
n
o
r
m
a
l
i
s
e
d
)

(f) Case 17
12
14
17
16
13
15
15
13
16
17
14
12
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60
NOx (g/kg fuel)
S
o
o
t

(
g
/
k
g

f
u
e
l
)
Measured
KIVA

(g) Soot vs NO
x
emissions for cases 12-17
Figure 4-6: KIVA-ERC model results and experimental data for Mode 6 (20% load, 1789 rpm) .
Case numbers refer to Table 3-8.
142

4.4 Discussion
The cases used in the validation of the KIVA-ERC model range from high load, low
speed operating conditions to low load, high speed conditions. The essential
combustion phenomena, such as ignition delay, premixed combustion (which causes
the initial spike in the heat release rate curve), diffusion-controlled combustion are
well-predicted over all operating conditions. This indicates that the KIVA-ERC
turbulence, spray formation, ignition and combustion submodels, all of which affect
these results, are reasonably accurate. The model constants were generally at their
default values. Deviation from these values caused deterioration in the model
predictions.

There is often a slight divergence of the pressure traces during combustion with the
experimental pressure traces falling below the calculated pressure traces (see Figure
4-4, Figure 4-5 and Figure 4-6). The reason for this is not clear. The experimental
heat release rate is calculated from the pressure trace, and since the heat release rate
shapes are similar, the pressure traces should also be similar.

The emissions trends are generally well-predicted, although there are errors in the
absolute values. The model constants were equal for all cases. If the model was
matched for each mode, the emissions predictions would be much more accurate.
There is also some uncertainty in the emissions measurements. The magnitude of this
uncertainty can be estimated from by the difference in measured emissions for the
pairs of cases 1 and 4, 6 and 11, and 12 and 17, in which the engine operating
conditions are almost identical.

Since the model will be used in the following sections to predict the behaviour of a
novel engine system for which there is no experimental data, the emissions models
cannot be calibrated. Therefore, there cannot be much confidence in the predicted
emissions levels. This means that the model can not reliably indicate whether the
two-stroke cycle engine will emit higher or lower levels of pollutants than the four-
stroke cycle engine. The only way to improve the predictions would be to improve
the emissions models (Kong, 2003), which is outside the scope of this study.
143
However, since the novel engine system notionally uses the same hardware (intake
ports, valves, piston and cylinder note however that the intake valves may be
shrouded), the model can be expected to provide useful prediction of the charging
and combustion phenomena and emissions trends. Charging phenomena represents
scavenging and supercharging in the case of the two-stroke cycle engine, and the
heat release rate is required to estimate fuel efficiency. The emissions models can be
used to estimate what effect a particular change may make on the emissions levels.

Chapter 5 - Results
5.1 Scavenging simulations
5.1.1 Shroud geometry
In Section 2.2.4 the review of prior research indicated that an intake valve shroud
subtending an angle of approximately 90 was optimum. Shrouds smaller than this
allowed too much short-circuiting of the scavenging flow from the intake directly to
the adjacent exhaust valves without the desired scavenging of residual gases, while a
larger shroud excessively impeded the intake flow and established a vortex near the
middle of the cylinder that trapped residual gases. Four shrouds were investigated
using KIVA-ERC and the mesh of the entire Caterpillar SCOTE cylinder (Figure 4-
3). The first shroud was shaped and located based on the results of previous studies.
Subsequent shrouds were adjustments to the first shroud aimed at reducing the short-
circuit flow. It was expected that the largest (final) shrouds would give the best
scavenging efficiency, but at the expense of considerable intake flow restriction.

The pre-processor file used to generate the mesh was unavailable, so the mesh
specification file itself (ITAPE17) was modified to reflect the addition of a shroud.
ITAPE17 is described in Amsden (1993) and Amsden (1997). The shroud was
created by firstly identifying a group of cells that approximated the desired shroud
shape. The cell numbers and orientation (each cell has front, left, bottom etc.
faces) were determined, then the F flag was changed from 1.0 to 0.0, indicating
that the cells are deactivated. The boundary conditions BCL, BCF and BCB for the
shroud cells and some of their neighbours had to be altered to reflect the presence of
the shroud surface. FV and IDREG for the shroud cells were unchanged because the
shrouds were always only one cell thick. Shroud vertices were given an IDFACE
value of 1 so that they behaved like valve stem vertices, i.e. they had the same
velocity as the valve, but did not actually change their position with each time step.
As the valve moved down and up, the top of the shroud maintained its position just
inside the intake port, but the rest of the shroud grew and shrank to maintain a
continuous barrier from the top of the shroud to the upper surface of the valve. This
was done automatically by KIVA-ERC, using the same process that it uses for valve
stems. The way that the valve shroud grows and shrinks as the valves move is
145
illustrated in Figure 5-1, which is a vertical cross section of the cylinder mesh
through an intake and exhaust valve.


Figure 5-1: Cross-section through cylinder mesh showing a shroud on the intake (left) valve.

Simulations were performed with no shroud and four different shroud geometries.
The shroud geometries studied are specified in Figure 5-2, Figure 5-3 and Table 5-1.

Shroud
146

(a) Shroud 1

(b) Shroud 2


(c) Shroud 3


(d) Shroud 4
Figure 5-2: Cross-section of cylinder showing shroud geometries used in KIVA-ERC
simulations. The intake poppet valves are in the lower half of the cylinder. The valve stems and
shrouds appear white.
147

Figure 5-3: Definitions of shroud parameters.
Intake valve diameter = 46.7 mm, exhaust valve diameter = 41.8 mm, bore = 137.2 mm

Table 5-1: Shroud parameters (see Figure 5-3 for definitions of symbols).

1
(deg)
1
(deg)
2
(deg)
2
(deg)
Shroud 1 110 6 100 -2
Shroud 2 135 18 122 -13
Shroud 3 135 18 147 -1
Shroud 4 170 0 165 9


5.1.2 Initial and boundary conditions
Twelve cases were run with each valve shroud type, as well as with no shroud. The
case conditions are outlined in Table 5-2.

1
61
61
74
148
Table 5-2: Scavenging simulation cases.
Case
Speed
(rpm)
P
in

(kPa)
T
in

(K)
P
ex

(bar)
Valve
timing*
1 993 150 310 100 A
2 993 200 310 100 A
3 993 200 310 150 A
4 993 250 310 150 A
5 1789 200 310 100 A
6 1789 250 310 100 A
7 1789 300 310 200 A
8 1789 350 310 200 A
9 1789 350 360 200 A
10 1789 250 310 100 B
11 1789
Function
1**
310 100 A
12 1789
Function
2**
310 100 A
* A: EVO = 100, IVO = 130, EVC = 250, IVC = 270 ATDC
B: EVO = 150, IVO = 150, EVC = 260, IVC = 270 ATDC
** Function 1: ( )

< < |
.
|

\
|

+
=
otherwise
IVC CA IVO if
IVO IVC
IVO CA
CA P
in
100
sin 200 100

Function 2: ( )

< <
|
.
|

\
|

=
otherwise
IVC CA IVO if
IVO IVC
IVO CA
CA P
in
300
sin 100 300


In order to investigate the scavenging behaviour over a wide range of conditions, the
cases covered low and high engine speeds and a wide range of intake and exhaust
pressures. A substantially different valve timing and higher intake gas temperature
were each tried to see whether they had significant effects on the scavenging
behaviour.

All of the cases except 11 and 12 have constant intake port pressures. Case 11 has an
intake pressure function that increases from 100 kPa at IVO to 300 kPa at maximum
intake valve lift, then returns to 100 kPa at IVC. Case 12 has a pressure function that
149
decreases from 300 kPa at IVO to 200 kPa at maximum intake valve lift, then returns
to 300 kPa at IVC. They are illustrated in Figure 5-4.

Figure 5-4: Intake pressure vs crank angle for Cases 11 and 12.

The simulations were undertaken with a maximum time step of 10 s. The initial
cylinder conditions were estimated by the thermodynamic model described in
Chapter 3, assuming perfect diffusion scavenging. The swirl ratio at EVO was set to
zero. This is expected to introduce small errors into the results. One alternative is to
allow the simulation run for one whole cycle before EVO, however this would
increase the run time for each of the sixty cases from approximately three days to
approximately nine days and was considered too expensive for the small
improvement in accuracy.

5.1.3 Scavenging flow
The KIVA-ERC code was modified to write data useful for measuring scavenging
parameters to a file. The data include:

0
100
200
300
400
0 100 200 300
Crank Angle (degrees ATDC)
I
n
t
a
k
e

p
o
r
t

p
r
e
s
s
u
r
e

(
k
P
a
)
Case 11
Case 12
150
the cumulative mass of gas delivered through the intake valves
the total instantaneous cylinder gas mass
the instantaneous mass of CO
2
in the cylinder
the cumulative mass of gas exhausted through the exhaust valves
the instantaneous mass fraction of CO
2
being exhausted

When used in conjunction with the regular KIVA-ERC output, many performance
measures can be calculated, such as the delivery ratio, scavenging efficiency,
retaining efficiency, exhaust gas purity and valve effective area.

KIVA-ERC runs failed (due to tinvrt overflow) for Cases 1 and 2 with Shrouds 1, 2
and 3. Restarts were attempted with different initial conditions and reduced time step
sizes, however these failed as well. The reason(s) for the failed runs could not be
determined during the course of the study. 54 of the 60 runs completed successfully.

To compare the performance of the shrouds, the scavenging efficiency was plotted
against the delivery ratio. The delivery ratio was defined as the cumulative mass
entering the cylinder through the intake valves between IVO and IVC normalised by
the mass of air occupying the displaced volume of the cylinder at atmospheric
pressure and 298 K. The results for all successful cases are shown in Figure 5-5.

0
0.2
0.4
0.6
0.8
1
0 1 2 3 4
Delivery Ratio
S
c
a
v
e
n
g
i
n
g

E
f
f
i
c
i
e
n
c
y
No shroud
Shroud 1
Shroud 2
Shroud 3
Shroud 4

Figure 5-5: Scavenging efficiency versus delivery ratio for all cases with each shroud.
151

Figure 5-5 shows that not having a shroud results in a higher delivery ratio but
markedly lower scavenging efficiency that having a shroud. It also shows that as the
shroud size is increased, both the delivery ratio and scavenging efficiency are
decreased. The reduction in scavenging efficiency with increasing shroud size is
most probably due to the reduction in delivery ratio.

Several graphical representations of scavenging simulations are shown in the
following figures. The sections or cutplanes correspond to those shown in Figure 5-6.
Generally, the figures show a favourable scavenging flow pattern. Fresh charge is
directed down the back and sides of the cylinder. At the piston face the flow is
diverted towards the centre and front of cylinder. Finally, the flow rises up the centre
and front of the cylinder, efficiently pushing the residual gases out the exhaust
valves. Small short circuit streams form around and between the pair of intake valve
shrouds and flow towards the exhaust valves.

Inspection of gas velocity vectors (Figure 5-7) shows that only a small vortex is
formed near the piston surface. Examination of the residual gas fraction contours
(Figure 5-8) shows that little residual gas is trapped in the vortex. In fact, the
contours do not form a closed loop around the vortex, which would indicate
significant trapping of residual gases due to recirculation.


Figure 5-6: Caterpillar SCOTE cylinder sections.

152
(a) 178 ATDC

(b) 194 ATDC
(c) 206 ATDC

(d) 223 ATDC
Figure 5-7: Gas velocity vectors in section A-A for Case 6, Shroud 1. The intake valves are to the
left. A small vortex is visible above the piston bowl.

153
(a) 178 ATDC (b) 194 ATDC
(c) 206 ATDC (d) 223 ATDC
Figure 5-8: CO
2
mass fraction contours in section A-A for Case 6, Shroud 1. Dark shading
indicates low mass fraction. Intake valves are to the left.
154


(a) 178 ATDC

(b) 194 ATDC

(c) 206 ATDC

(d) 223 ATDC
Figure 5-9: CO
2
mass fraction contours in sections B-B to E-E for Case 6, Shroud 1. Dark
shading indicates low mass fraction. The intake valves are to the left rear.

155

Case 1, 160 ATDC

Case 9, 160 ATDC

Case 1, 170 ATDC

Case 9, 180 ATDC

Case 1, 180 ATDC

Case 9, 190 ATDC

Case 1, 200 ATDC

Case 9, 220 ATDC
Figure 5-10: Comparison of gas composition through sections B-B to E-E during scavenging for
Cases 1 (left column) and 9 (right column) with Shroud 4. Adjacent images have approximately
equal scavenging gas volumes in the cylinder.
156
5.1.4 Valve flow
The 1-D compressible gas flow equations described in Chapter 3 were used to
calculate the effective valve areas from KIVA-ERC simulations. Recall that for the
0-D simulation, the valve effective areas were estimated using the valve curtain areas
and a relationship for discharge coefficient versus valve lift reported by Heywood
and Sher (1999, pp. 187-8) and represented in Figure 3-8. The estimated valve
effective areas versus valve lift were plotted in Figure 3-9.

Using the KIVA-ERC-calculated valve flow rates, valve curtain area, intake port and
cylinder pressures and gas temperature and composition, the effective valve areas in
the KIVA-ERC simulations were calculated and compared with the 0-D estimates in
Chapter 3. Note that the 0-D estimates were adjusted to account for the presence of
the shroud. This was done by simply reducing the effective area by the same
proportion as the reduction in the valve curtain area, so 90 shrouds would reduce the
curtain area and effective area by 25%. The results for several cases are shown in
Figure 5-11 (intake valves and Shroud 1) and Figure 5-12 (exhaust valves).


0 0.004 0.008 0.012 0.016
0
0.001
0.002
0.003
0.004


Figure 5-11: Calculated intake valve effective areas for shroud 1 based on KIVA-ERC results.
Case numbers are indicated. The effective area estimated in Figure 3-9, reduced by the shroud
area, is shown for comparison.

6
5
8
11
12
7
9
0-D
estimate
Error!
Referenc
E
f
f
e
c
t
i
v
e

a
r
e
a

(
m
2
)

Time (s)
157
0 0.004 0.008 0.012 0.016
0
5
.
10
4
0.001
0.0015
0.002
0.0025


Figure 5-12: Calculated exhaust valve effective areas for shroud 1 based on KIVA-ERC results.
Case numbers are indicated. The effective area estimated in Figure 3-9, reduced by the shroud
area, is shown for comparison.

Examining results for the shrouded intake valves (Figure 5-11), the KIVA-ERC
valve effective areas rise more slowly than the 0-D estimate. There is a knee in the
curves at an effective area of about 0.001 m
2
, approximately the same value as the
knee in the 0-D estimate. Shortly after this point, in all cases but case 11, the KIVA-
ERC effective areas climb much higher than the 0-D estimate, then briefly return to
approximately the 0-D estimate.

The differences between the KIVA-ERC effective areas and the 0-D estimate can be
partly attributed to gas dynamics. The effective area is based on the ratio of the
intake port boundary pressure and the bulk cylinder pressure. When the intake flow
is suddenly accelerated, the pressure at the valve deviates from that at the port
boundary (the ram effect described in Heywood and Sher (1999, p. 193) and
others).

The peaks in Figure 5-11 which apparently approach infinity coincide with rapid
deceleration of the intake flow. These peaks are greatly magnified because they
occur when the pressure ratio is close to unity, and small differences in the flow rate
yield large differences in the calculated effective area. This hypothesis is supported
6
5
8
11
12
7
9
0-D
estimate
Error!
R f
E
f
f
e
c
t
i
v
e

a
r
e
a

(
m
2
)

Time (s)
158
by case 11, in which the intake port boundary pressure rises and falls approximately
with the valve lift. The intake flow is more constant in this case, and the KIVA-ERC
effective area is approximately the same as the 0-D estimate for most of the valve
open period. KIVA-ERC-calculated pressures in the intake and exhaust ports were
examined at times when the calculated effective area differed from the 0-D
estimation. Representative results are shown in Figure 5-13 and Figure 5-14. They
clearly demonstrated significant differences between the pressures near the upper
surfaces of the intake and exhaust valves and the imposed, constant pressure at the
port boundaries. This is consistent with reasoning based on gas dynamics.



Figure 5-13: Pressure contours in the intake port for Shroud 1, Case 5 at 202 deg ATDC,
showing the pressure gradient between the port boundary and the intake valves due to flow
deceleration (the ram effect). Zero-dimensional modelling assumes constant pressure
throughout the port.
159

Figure 5-14: Pressure contours in the exhaust port for Shroud 1, Case 5 at 175 deg ATDC. Here,
the pressure difference between the region above the valve and the port boundary appears to be
due to the port geometry.

Examination of Figure 5-12 shows that the KIVA-ERC effective areas vary
significantly from the 0-D estimate, and vary significantly from case to case. The
maximum KIVA-ERC effective areas are approximately half of the 0-D estimate,
and some cases exhibit dips in the KIVA-ERC effective area at maximum valve lift.
The maximum discharge coefficient is therefore approximately 0.3 0.4, which is
much less than the experimentally-determined values reported by Heywood (1989).

The discrepancies in the intake and exhaust valve effective areas may be attributed to
the relative coarseness of the mesh on the scale of the valves and the inability of
KIVA-ERC to accurately model complex flow details that influence the effective
area, especially flow separation.

The KIVA-ERC simulations generally predict lower air consumption than the 0-D
simulations. The results for some cases with Shroud 1 are tabulated below.
Agreement is within 20%. Note that the KIVA-ERC results are sensitive to the initial
conditions that were estimated using 0-D modelling, whereas the 0-D simulations
had run over at least ten cycles and had converged to a solution.
160

Table 5-3: Predicted air consumption for cases using KIVA-ERC and 0-D modelling.
Case 3 4 5 6 7 8 9 10
KIVA
(kg/min)
4.9 8.0 7.8 8.1 8.9 12.4 11.7 6.7
0-D
(kg/min)
6.0 8.9 7.1 10.5 9.1 11.8 14.1 8.1

Experimentation is necessary to determine whether real air consumption more
closely matches the 0-D model or KIVA-ERC model.
5.2 O-D Scavenging model
5.2.1 Description
Examination of graphical representations of the KIVA-ERC scavenging simulations
in the previous section revealed geometrical similarities in the scavenging flow
entering the cylinder. For example, Figure 5-10 shows the concentration of CO
2
at
four horizontal cross-sections through the cylinder at four points in time for cases 1
and 9. These cases represent different speeds (993 rpm vs 1789 rpm) and different
scavenging pressures (150 kPa vs 350 kPa), yet the scavenging flow appears similar
in each case. This similarity suggested a simple scavenging model in which the only
independent variable was the ratio of the instantaneous volume of fresh charge in the
cylinder to the total instantaneous cylinder volume. This parameter was called the
scavenging volume fraction. One sensitive measure of scavenging behaviour is the
exhaust gas purity, (Heywood and Sher, 1999). The exhaust gas purity is defined as
the mass fraction of fresh charge in the exhaust gas at any given instant. Plots of
versus scavenging volume fraction for all cases using shroud 4 are summarised in
Figure 5-15.

161

0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Scavenging volume fraction
1
2
3
4
5
6
7
8
9
10
11
12


Figure 5-15: vs scavenging volume fraction for all cases with shroud 4.

Figure 5-15 shows that while there are similarities between the various cases, there is
also significant variation that renders it inadequate. Examination of these plots and
graphical output shows an apparent two-stage process occurring. The first stage is
dominated by the exhaust of residual gases and the development of short-circuit
flow. At a scavenging volume fraction of between 0.4 and 0.6 a relatively sharp
transition begins to a second stage, in which the exhaust gas is a mixture of residual
and intake gas, as well as the short circuit flow. The situation is represented
graphically as in Figure 5-16.

162
Figure 5-16: Representation of two-stage scavenging process.

The model assumes that at IVO the cylinder contents, which are residual gases, are
contained in zone 1. As intake gases enter the cylinder, a portion is short-circuited to
the exhaust port at a rate
sc
m& that varies with time, and the remainder go into a
second zone called zone 2. Some of the gas in zone 1 enters and mixes with the gas
in zone 2 at a time-varying rate
12
m& . The gas flowing through the exhaust valve is a
mixture of the short-circuit flow and zone 1 gas. The transition to Stage 2 occurs
when the mass in zone 1 is entirely depleted. The exhaust flow is then a mixture of
the short-circuit flow and zone 2 gas.

In order to complete the model, functions for
sc
m& and
12
m& were assumed.
Examination of KIVA-ERC simulations indicated that
sc
m& was a approximately a
constant fraction of the intake flow
i
m& , or:

>
=
otherwise m
m x m if m x
m
e
i e i
sc
&
& & &
&
Equation 5-1
where:
x = proportionality constant
e
m& = total exhaust flow

Inlet Exhaust
Stage 1 Stage 2
Zone 2 Zone 1
Zone 2
e
m
1
&
sc
m&
i
m&
e
m
2
&
i
m&
sc
m&
12
m&
163
One slight difficulty with Equation 5-1 is that the exhaust gas composition is a
function of the exhaust gas rate, however the exhaust gas rate is a function of the
exhaust gas composition (see Equation 3-19, Equation 3-20 and Equation 3-21). The
system of equations Equation 3-19, Equation 3-20, Equation 3-21 and Equation 5-1
must therefore be solved. A simple iterative scheme was used to achieve this.

It was assumed that the volumetric rate of mixing of zone 1 into zone 2 was
proportional to the speed of the engine, reflecting the variation of the level of
turbulence in the cylinder with engine speed. The mass rate of zone 1 entering zone 2
can therefore be expressed:

1 12
k m = &
Equation 5-2
where:
k = proportionality constant
= engine speed

1
= zone 1 density

The composition of zone 1 is constant, always being entirely composed of residual
gases. The rate of change of mass in zone 1 is:

12 1 1
m m m
e
& & & + =
Equation 5-3
where:
sc e e
m m m & & & =
1

Equation 5-4

The rate of change of temperature in zone 1 can be derived by combining Equation
3-3, the perfect gas relation, and Equation 3-7, conservation of energy, and assuming
the burned gas fraction is constant:

( )
P
R
T
u
T R
T
T
& &
) (
1
,
1 1
1
1 1

+ |
.
|

\
|

=
Equation 5-5
where:
164
1
T = temperature of the gas in zone 1
R = the specific gas constant (Equation 3-16) of the gas in zone 1
u = the specific internal energy (Equation 3-15) of the gas in zone 1
P =cylinder pressure

There is no need to explicitly determine the gas state in zone 2 if the bulk cylinder
gas state (zone 1 + zone 2) is being calculated as described in Section 3. When zone
1 is exhausted, zone 2 simply becomes the bulk cylinder gas.

5.2.2 Comparison with KIVA-ERC calculations
The model was compared with the sixty runs discussed in Section 5.1. The output of
each run was processed using Mathcad to generate functions of time of the following
quantities:

Cylinder pressure P(t)
Mass of cylinder contents m(t)
Cumulative mass flow through the intake valves m
i
(t)
Cumulative mass flow through the exhaust valves m
e
(t)
Bulk cylinder gas temperature T(t)
Mass of CO
2
in the cylinder mCO
2
(t)
Mass fraction of CO
2
flowing past the exhaust valves mfCO
2e
(t)

The exhaust gas purity is defined as the mass fraction of fresh charge in the exhaust
gas at any instant (Heywood and Sher, 1999), so:
| |
r f e
mfCO t mfCO t t mfCO
2 2 2
) ( 1 ) ( ) ( + =
Equation 5-6
where:
mfCO
2f
= mass fraction of CO
2
in fresh charge
mfCO
2r
= mass fraction of CO
2
in residual gases

Rearranging:
r f
r e
mfCO mfCO
mfCO mfCO
t
2 2
2 2
) (

=
Equation 5-7
165

Since mfCO
2f
= 0 for all KIVA cases and mfCO
2r
=
) 0 (
) 0 (
2
mCO
m
where IVO occurs at t
= 0:
) 0 (
) 0 (
) ( 1 ) (
2
2
mCO
m
t mfCO t
e
=
Equation 5-8

The scavenging efficiency is defined as the mass of fresh charge retained in the
cylinder divided by the mass of the cylinder contents:

) (
) ( ) ( ) (
) (
0
t m
dt t m t t m
t
t
e i
sc

=
&

Equation 5-9

Substituting the expression for exhaust gas purity (Equation 5-8):

) (
) (
) 0 (
) 0 (
) ( 1 ) (
) (
0 2
2
t m
dt t m
mCO
m
t mfCO t m
t
t
e e i
sc
|
|
.
|

\
|

=
&

Equation 5-10

Integrating and simplifying:

) (
) 0 (
) (
1 ) 0 ( ) ( ) (
) (
2
2
t m
mCO
t mCO
m t m t m
t
e i
sc
|
|
.
|

\
|
+
=
Equation 5-11

Since m(t) = m
i
(t) m
e
(t) + m(0):

) 0 (
) (
) (
) 0 (
1 ) (
2
2
mCO
t mCO
t m
m
t
sc
=
Equation 5-12

166
The scavenging efficiency that would have been achieved if the residual cylinder
gases were displaced by the intake gases with no mixing or short-circuiting (the
perfect displacement model) was calculated using:

<
=
otherwise
t m t m if
t m
t m
t
i
i
disp sc
1
) ( ) (
) (
) (
) (
.

Equation 5-13

The scavenging efficiency that would have been achieved if the intake gas instantly
mixed with the cylinder contents (the perfect mixing model) was also calculated.
With perfect mixing, the exhaust gas purity is equal to the scavenging efficiency, or:

) ( ) ( t t
sc
=
Equation 5-14

So, from Equation 5-9:

) (
) ( ) ( ) (
) (
0
t m
dt t m t t m
t
t
e sc i
sc

=
&

Equation 5-15

Differentiating:

) (
) (
) ( ) ( ) (
) (
) ( ) ( ) (
) (
2
0
t m
t m
dt t m t t m
t m
t m t t m
t
t
e sc i
e sc i
sc
&
&
& &
&


Equation 5-16

Simplifying:

) (
) (
) (
) (
) ( ) ( ) (
) ( t m
t m
t
t m
t m t t m
t
sc e sc i
sc
&
& &
&

=
Equation 5-17

Rearranging:
167

| |
) (
) ( ) ( ) ( ) (
) (
t m
t m t m t t m
t
e sc i
sc
& & &
&

=


Equation 5-18

Since ) ( ) ( ) ( t m t m t m
e i
& & & = and replacing
sc
with
sc.diff
to avoid confusion:

( ) ) ( 1
) (
) (
) (
. .
t
t m
t m
t
diff sc
i
diff sc
=
&
&
Equation 5-19

This ODE was solved using Mathcads built-in 4
th
-order Runge-Kutta solver to
obtain
sc.diff
(t).

The delivery ratio is defined as the mass of fresh charge delivered divided by a
reference mass, usually the displacement volume the fresh charge density at
ambient conditions:

g 89 . 2
) (
m kg 185 . 1 L 44 . 2
) (
3
IVC t m IVC t m
i i
=
=

=
=


Equation 5-20

When comparing the 0-D model to KIVA-ERC calculations, the KIVA-derived valve
flow rates, cylinder masses and pressures were used. For stage 1, when there are two
zones in the cylinder, the following expressions were used:
Zone 1 mass:

( )dt m m m t m
t
sc e

+ =
0
12 1
) ( & & &
Equation 5-21
where:
m
sc
is defined in Equation 5-1
m
12
is defined in Equation 5-2

For Stage 1, when m
1
(t) > 0:

168
e
sc
m
m
t
&
&
= ) (
Equation 5-22

and from Equation 5-16 and Equation 5-22:

) (
) ( ) ( ) ( ) (
) (
t m
t m t t m t m
t
sc sc i
sc
& & &
&


=
Equation 5-23

For Stage 2, when m
1
(t) has reached zero:

( )
e
e sc sc
m
m m
t
&
& &
2
1
) (

+
=
Equation 5-24

and from Equation 5-16 and Equation 5-24:

( )
) (
) ( ) ( 1 ) ( ) (
) (
2
t m
t m t m t m t m
t
sc e sc sc i
sc
& & & &
&


=
Equation 5-25

To illustrate the application of these relations, take for example Case 4 (referring to
the conditions in Table 5-2) with shroud 1 (described in Table 5-1). The quantities
calculated by KIVA-ERC are shown in Figure 5-17.

169
0 0.014 0.029
0
0.5
1


Figure 5-17: KIVA-ERC-calculated cylinder quantities for Case 4, Shroud 1 from EVO to IVC.

The data represented in Figure 5-17 and several guesses of the calibration constants x
and k were then tried until satisfactory matches of the exhaust gas purity and
scavenging efficiency were achieved for all cases using a particular shroud. The
calculated zone 1 quantities and the relevant equations used are shown in Figure
5-18. The calculated exhaust gas purities and scavenging efficiencies are shown in
Figure 5-19.

Pressure 10
-6
Pa
Temperature 10
-3
K
Mass 10
-1
kg
Time (s)
Cumulative intake
mass 10
-1
kg
CO
2
mass 10
-2
kg
Cumulative exhaust
mass 10
-1
kg
170
0 0.0084 0.0168
0
0.5
1


Figure 5-18: Zone 1 quantities calculated from Figure 5-17 and matched calibration constants
x = 0.1 and k = 8 10
-4
.


0 0.015 0.029
0
0.5
1


Figure 5-19: Calculated exhaust gas purities and scavenging efficiencies for Case 4, Shroud 1.

Time (s)
Temperature 10
-3
K
(Equation 5-5)
Mass 10
-1
kg
(Equation 5-21)
sc
m& kg/s
(Equation 5-1)
1
m& kg/s
(Equation 5-3)
12
m& kg/s
(Equation 5-2)
Time (s)

0-D
(Equation 5-24)

KIVA
(Equation 5-7)

sc.disp
(Equation 5-13)

sc.0-D
(Equation 5-25)

sc.KIVA
(Equation 5-12)

sc.diff
(Equation 5-19)
171
This procedure was repeated for all sixty combinations of cases and shrouds (Figure
5-20 to Figure 5-24).

sc.0-D

sc.KIVA

0 0.015 0.029
0
0.5
1

(a) Case 1

0 0.015 0.029
0
0.5
1

(b) Case 2

0 0.014 0.029
0
0.5
1
(c) Case 3

0 0.014 0.029
0
0.5
1

(d) Case 4

0 0.008 0.016
0
0.5
1

(e) Case 5

0 0.008 0.016
0
0.5
1
(f) Case 6

0 0.008 0.016
0
0.5
1

(g) Case 7

0 0.008 0.016
0
0.5
1

(h) Case 8

0 0.008 0.016
0
0.5
1
(i) Case 9
0 0.005 0.01
0
0.5
1

(j) Case 10
0 0.008 0.016
0
0.5
1

(k) Case 11

0 0.008 0.016
0
0.5
1
(l) Case 12
Figure 5-20: Comparison of KIVA and two-zone scavenging model results for no shroud on the
intake valves. Case numbers refer to conditions described in Table 5-2. Results for the perfect
displacement and diffusion models are also indicated. The x-axis is time from EVO in seconds.
For each case x = 0.34 and k =1.0.

sc.diff

sc.disp

172

sc.0-D

sc.KIVA




No data


(a) Case 1



No data


(b) Case 2

0 0.015 0.029
0
0.5
1
(c) Case 3

0 0.015 0.029
0
0.5
1

(d) Case 4

0 0.008 0.016
0
0.5
1
(e) Case 5

0 0.008 0.016
0
0.5
1
(f) Case 6

0 0.008 0.016
0
0.5
1

(g) Case 7

0 0.008 0.016
0
0.5
1
(h) Case 8

0 0.008 0.016
0
0.5
1
(i) Case 9
0 0.005 0.01
0
0.5
1

(j) Case 10

0 0.008 0.016
0
0.5
1
(k) Case 11

0 0.008 0.016
0
0.5
1
(l) Case 12
Figure 5-21: Comparison of KIVA and two-zone scavenging model results for shroud 1 on the
intake valves. Case numbers refer to conditions described in Table 5-2. Results for the perfect
displacement and diffusion models are also indicated. The x-axis is time from EVO in seconds.
For each case x = 0.12 and k = 7 10
-4
.

sc.diff

sc.disp

173

sc.0-D

sc.KIVA




No data


(a) Case 1



No data


(b) Case 2

0 0.015 0.029
0
0.5
1
| |
(c) Case 3

0 0.015 0.029
0
0.5
1
| |

(d) Case 4

0 0.008 0.016
0
0.5
1
(e) Case 5

0 0.008 0.016
0
0.5
1
(f) Case 6

0 0.008 0.016
0
0.5
1

(g) Case 7

0 0.008 0.016
0
0.5
1
(h) Case 8

0 0.008 0.016
0
0.5
1
(i) Case 9
0 0.005 0.01
0
0.5
1

(j) Case 10

0 0.008 0.016
0
0.5
1
(k) Case 11

0 0.008 0.016
0
0.5
1
(l) Case 12
Figure 5-22: Comparison of KIVA and two-zone scavenging model results for shroud 2 on the
intake valves. Case numbers refer to conditions described in Table 5-2. Results for the perfect
displacement and diffusion models are also indicated. The x-axis is time from EVO in seconds.
For each case x = 0.10 and k = 8 10
-4
.

sc.disp

sc.diff
174

sc.0-D

sc.KIVA




No data


(a) Case 1



No data


(b) Case 2

0 0.015 0.029
0
0.5
1
(c) Case 3

0 0.015 0.029
0
0.5
1

(d) Case 4

0 0.008 0.016
0
0.5
1
(e) Case 5

0 0.008 0.016
0
0.5
1
(f) Case 6

0 0.008 0.016
0
0.5
1

(g) Case 7

0 0.008 0.016
0
0.5
1
(h) Case 8

0 0.008 0.016
0
0.5
1
(i) Case 9
0 0.005 0.01
0
0.5
1

(j) Case 10

0 0.008 0.016
0
0.5
1
(k) Case 11

0 0.008 0.016
0
0.5
1
(l) Case 12
Figure 5-23: Comparison of KIVA and two-zone scavenging model results for shroud 3 on the
intake valves. Case numbers refer to conditions described in Table 5-2. Results for the perfect
displacement and diffusion models are also indicated. The x-axis is time from EVO in seconds.
For each case x = 0.09 and k = 8.5 10
-4
.

sc.disp

sc.diff
175

sc.0-D

sc.KIVA


0 0.015 0.029
0
0.5
1

(a) Case 1

0 0.015 0.029
0
0.5
1
(b) Case 2

0 0.015 0.029
0
0.5
1
(c) Case 3

0 0.015 0.029
0
0.5
1

(d) Case 4

0 0.008 0.016
0
0.5
1
(e) Case 5

0 0.008 0.016
0
0.5
1
(f) Case 6

0 0.008 0.016
0
0.5
1

(g) Case 7

0 0.008 0.016
0
0.5
1
(h) Case 8

0 0.008 0.016
0
0.5
1
(i) Case 9
0 0.005 0.01
0
0.5
1

(j) Case 10

0 0.008 0.016
0
0.5
1
(k) Case 11

0 0.008 0.016
0
0.5
1
(l) Case 12
Figure 5-24: Comparison of KIVA and two-zone scavenging model results for shroud 4 on the
intake valves. Case numbers refer to conditions described in Table 5-2. Results for the perfect
displacement and diffusion models are also indicated. The x-axis is time from EVO in seconds.
For each case x = 0.07 and k = 9.0 10
-4
.

sc.diff

sc.disp

176
The results are summarised in Table 5-4.

Table 5-4: Comparison of scavenging efficiencies calculated using the 0-D model and KIVA-
ERC.

No shroud
x = 0.34
k = 1
Shroud 1
x = 0.12
k = 7.010
-4

Shroud 2
x = 0.10
k = 8.010
-4

Shroud 3
x = 0.09
k = 8.510
-4

Shroud 4
x = 0.07
k = 9.010
-4

Case
(Table 5-2)
0-D KIVA 0-D KIVA 0-D KIVA 0-D KIVA 0-D KIVA
1 0.78 0.78 - - - - - - 0.90 0.88
2 0.81 0.81 - - - - - - 0.92 0.92
3 0.72 0.71 0.87 0.86 0.85 0.85 0.85 0.84 0.84 0.83
4 0.77 0.76 0.91 0.90 0.89 0.90 0.89 0.89 0.88 0.88
5 0.61 0.61 0.77 0.77 0.74 0.75 0.74 0.74 0.72 0.73
6 0.55 0.56 0.74 0.75 0.76 0.78 0.77 0.78 0.76 0.77
7 0.51 0.51 0.68 0.68 0.65 0.66 0.64 0.64 0.62 0.63
8 0.58 0.57 0.74 0.74 0.71 0.72 0.71 0.71 0.69 0.69
9 0.60 0.60 0.78 0.78 0.75 0.76 0.74 0.74 0.73 0.73
10 0.55 0.55 0.64 0.63 0.62 0.61 0.62 0.60 0.60 0.58
11 0.58 0.59 0.78 0.78 0.76 0.76 0.75 0.76 0.75 0.74
12 0.68 0.69 0.82 0.82 0.79 0.80 0.79 0.79 0.77 0.78

Given the complex nature of the in-cylinder flow field, the widely-varying engine
speeds, intake pressures valve timings and exhaust back pressures modelled, the
agreement is very good in all cases. This gives confidence that other results within
the wide range of parameters in Cases 1 to 12 will be similarly accurate. This model
was incorporated in the zero-dimensional model described in Chapter 3. The zero-
dimensional modelling results presented from this point on used this model, except
where noted otherwise.


177

5.3 System simulations
5.3.1 Revision of combustion correlation constants
In the absence of experimental heat release rate data from two-stroke poppet-valved
heavy-duty diesel engines, KIVA-ERC simulations were used. Eighteen cases were
run, representing three engine systems at each of the six modes of the FTP
approximation. The three engine systems were:
The Caterpillar SCOTE adapted to a two-stroke cycle
A two-cylinder engine based on the SCOTE cylinder with a reciprocating air
pump
A turbocharged and intercooled version of the system above

Conditions at IVC were estimated with the thermodynamic model using the
combustion correlation constants estimated in Section 3.2.7.

The SCOTE 60 sector mesh described in Section 4.2.9. (especially Figure 4-2) was
used. The simulations were run from IVC until EVO. The injection rate shape was
calculated from Equation 3-30. The adjustable KIVA-ERC model constant values
were those reported in Table 4-1.

Additionally, the initial swirl ratio and turbulence parameters were estimated from
the KIVA scavenging simulations reported in Section 5.1 using the full SCOTE
mesh. The calculated values of the parameters at IVC are shown in Table 5-5. They
are reasonably insensitive to engine operating conditions, and mean values were used
for all simulations using the SCOTE 60 sector mesh in this and subsequent sections
for consistency. The initial swirl ratio was assumed to be 0.39, the turbulent kinetic
energy parameter tkei was set at 6.6 and the length scale was 1.33 cm.

For four-stroke simulations, tkei is usually between 0.1 and 1.6 (Hessel et al., 2003).
The higher value in the two-stroke cases reflects the higher intake gas flow rate and
the turbulence-generating effect of the intake valve shroud.

178
Table 5-5: Turbulence and swirl values at IVC for the Shroud 1 cases listed in Table 5-2.
Case
Turbulent kinetic
energy (10
6
cm
2
/s
2
)
Turbulent kinetic
energy ratio
Turbulence length
scale (cm)
Swirl
ratio
1 - - - -
2 - - - -
3 1.02 6.7 1.40 0.43
4 1.15 7.5 1.44 0.30
5 3.10 6.3 1.36 0.38
6 3.01 6.1 1.47 0.33
7 2.57 5.2 1.30 0.33
8 2.96 6.0 1.34 0.37
9 3.12 6.3 1.33 0.41
10 4.23 8.5 0.96 0.21
11 3.35 6.8 1.44 0.85
12 3.26 6.6 1.29 0.28
Mean - 6.6 1.33 0.39

The procedure for determining the correlation constants was identical to that
described in Section 3.2.7.

The ignition delay constants (Section 3.2.7.4) used are listed below.
A
1
= 0.00385
A
2
= 6580
A
3
= 0.50

Previously, the constants were 2.33, 2230, 0.94. The higher apparent activation
energy (A
2
) indicates a decline in the effect of physical processes such as evaporation
and mixing, perhaps due to greater turbulence, and are closer to value determined for
premixed fuel-air mixtures (Heywood, 1989, p. 544).

The correlation is compared with the KIVA-ERC results in Figure 3-13 and Figure 3-
14..

179
0
0.5
1
1.5
2
800 900 1000 1100
Mean temperature T
m
(K)
I
g
n
i
t
i
o
n

d
e
l
a
y

(
m
s
)
KIVA-ERC
Correlation

Figure 5-25: Ignition delay vs mean temperature.
0
0.5
1
1.5
2
20 30 40 50 60 70
Mean pressure P
m
(bar)
I
g
n
i
t
i
o
n

d
e
l
a
y

(
m
s
)
KIVA-ERC
Correlation

Figure 5-26: Ignition delay vs mean pressure.

The values used for the premixed combustion correlation (Section 3.2.7.5) were:
A
4
= 0.41
A
5
= 1.50
A
6
= 1.17

180
A comparison of the data and the correlation is shown in Figure 3-15 and Figure 3-
16..

0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1

KIVA-ERC
Correlation

Figure 5-27: Comparison of KIVA-ERC results and the correlation vs .
0
0.2
0.4
0.6
0.8
1
0 0.5 1 1.5 2
Ignition delay t
ID
(ms)

KIVA-ERC
Correlation

Figure 5-28: Comparison of KIVA-ERC results and the correlation vs ignition delay.



181
The expressions for the rate shape constants used in the model were:
C
p1
= 1.90
C
p2
= 5000
C
d1
= 0.079 + 1.47
-1.07

-0.90

C
d2
= -0.061 + 0.53
-0.33

-0.40

d
= -0.14 + 0.081 C
p1


The diffusion burning rate shape parameters C
d1
and C
d2
reflect faster mixing due to
the increased turbulence in two-stroke cycle operation.

5.3.2 Two-stroke adaptation of Caterpillar SCOTE
The 0-D model described in Chapter 3 with the two-zone scavenging model
developed in Section 5-2 and the multidimensional model described in Chapter 4
were used to predict the performance of the Caterpillar SCOTE (described in Figure
3-10 and Table 3-3) if it was converted to a two-stroke cycle. If this were to be
performed experimentally, the following modifications would be required:

a) The valve gear speed would have to be doubled, so that the camshafts rotated
at the engine speed, rather than at half the engine speed.
b) The valve timing would have to be altered. In practice, new camshafts would
have to be manufactured.
c) Shrouds would have to be attached to the intake valves. Some means would
be required to prevent the shrouds from rotating. Valves are usually allowed
to rotate to allow even wear. Kang et al. (1996) describes one method of
preventing shrouds from rotating while allowing the valves to rotate.
d) The fuel pump would have to be doubled in speed to allow injection once per
engine revolution. If the same injection characteristics were required, the fuel
pump cam would have to be altered to allow for the doubling of the pump
speed.
e) A pressure differential between the intake and exhaust ports is required for
scavenging. Constant intake pressures of either 150 kPa or 200 kPa were used
for initial studies. Later, the boost pressure was supplied by reciprocating air
pumps and turbochargers which would better represent real engine systems.
182

For the numerical simulation, the following simple changes needed to be made to the
Caterpillar SCOTE input files for the thermodynamic model:

f) The cycle flag was changed from 4.0 (4-stroke) to 2.0 (2-stroke).
g) The fuel mass injected per cylinder per cycle was reduced to maintain the
same power output.
h) The scavenging model flag was changed from 0.0 (perfect diffusion model,
suitable for 4-stroke calculations) to 1.0 (two-zone scavenging model,
suitable for 2-stroke poppet-valved engines). The scavenging model constants
x and k was adjusted to be consistent with multidimensional calculations.
i) The valve timings were altered.
j) The intake and exhaust pressures were altered to give satisfactory scavenging
flow.
k) The combustion correlation constants were altered to better estimate the
ignition delay, proportion of fuel in premixed combustion and the heat release
rate.

For the simulations, Shroud 1 (referring to Table 5-1) was used, because it provided
good scavenging with the least valve flow obstruction. The valve effective area was
assumed to be due to the shroud was assumed to be proportional to the reduction in
valve curtain area, or 29%.

Valve timing was explored parametrically, with engine speed, fuel rate, injection
timing, intake pressure and temperature and exhaust pressure held constant. The
following valve timing parameters were explored:

EVO ( ATDC) = 90, 95 160
IVC ( ATDC) = 240, 245 270
IVO EVO ( CA) = 0, 5 15
IVC EVC ( CA) = 0, 5 25

The total number of combinations of these parameters is 1,512.
183

The function to be maximised was simply the reciprocal of the brake specific fuel
consumption (BSFC). This was so as to find the parameters giving the most fuel
efficient system. In practice, emissions must also be taken into account. However,
since there is currently no accurate means of predicting emissions in novel engine
systems, emissions could not be included in the merit function.

There were additional constraints placed on the merit function:

l) The predicted maximum burnt gas fraction was not to exceed 0.7, to ensure
adequate available oxygen for high combustion efficiency and low emissions.
m) The maximum cylinder pressure was not to exceed 15 MPa to avoid
overstressing the engine, as suggested by Montgomery (2000).
n) The root-mean-square difference between the predicted cylinder mass, burnt
gas fraction and temperature at one degree intervals for two consecutive
cycles at was to be less than 1%, indicating convergence towards a solution.
RMS values much less than this were possible with the perfect diffusion
scavenging model, however the two-zone scavenging model introduced some
instability to the numerical solution, and values much smaller than 1%
excluded too great a proportion of results, even when the calculation was set
to run over 100 engine cycles.
o) The predicted premixed fuel combustion fraction was to be less than 100%.
Physically, values exceeding 100% are impossible, and predicted values
greater than this were assumed to indicate that conditions were such that the
ignition model was no longer valid and/or that conditions were not suitable
for diesel autoignition.

The calculations with the 1,512 parameter combinations for each mode took less than
two hours CPU time on a Silicon Graphics Origin 3000 computer. Approximately
20% of the combinations violated one of the four constraints or caused the ordinary
differential equation integrator to fail.

Mechanical valve train considerations were considered outside the scope of this
study. It is noted that a combination of high valve lift, short valve open period and
184
high engine speed could lead to unacceptable wear, spring surge and other
unacceptable phenomena.

Combustion and emissions calculations were performed using KIVA-ERC and the
SCOTE 60 sector mesh described in Section 4.2.9 (especially Figure 4-2). The
simulations were run from IVC until EVO. Conditions at IVC (initial temperature,
pressure and gas composition) were taken from the 0-D results. The injection rate
shape was calculated from Equation 3-30.. The adjustable KIVA-ERC model
constant values were those reported in Table 4-1.

The results are summarised and compared with 4-stroke baseline SCOTE data in
Table 5-6. Note that the 4-stroke emissions values are not measured values but were
calculated using KIVA-ERC. This was done to provide a more apples with apples
comparison that might give more insight as to whether emissions would be increased
or decreased. It should be noted that the two-stroke cases are possibly sufficiently
different from the four-stroke cases that the KIVA-ERC model constants should be
adjusted and direct comparisons are not necessarily valid.
185
Table 5-6: Predicted near-optimum valve timings, engine performance and combustion
parameters for Modes 1-6 with the Caterpillar SCOTE engine running on a two-stroke cycle
with Shroud 1 on the intake valves. Four-stroke results (italicised) are shown for comparison.
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6
Input parameters
Speed (rpm) 750 953 993 1657 1737 1789
Intake air pressure
(kPa, absolute)
150 150 150 200 200 200
Intake air
temperature (K)
298 298 298 298 298 298
Exhaust pressure
(kPa, absolute)
100 100 100 100 100 100
Fuel rate
(g/cyl/cycle)
0.013 0.034 0.104 0.102 0.063 0.033
SOI ( ATDC) 352.0 359.5 354.5 367.5 361.5 354.5
Valve timing
IVO ( ATDC) 120 150 140 140 155 130
IVC ( ATDC) 290 280 270 270 290 290
EVO ( ATDC) 120 150 140 140 140 130
EVC ( ATDC) 290 275 260 255 280 285
Performance
Brake Power (kW) 0.5
(0.3)
7.9
(7.6)
31.3
(29.6)
38.5
(38.4)
30.0
(28.4)
12.7
(12.7)
BSFC (g/kWh) N/A 246
(266)
198
(212)
232
(264)
219
(244)
278
(303)
Max. press. (MPa) 3.3
(5.6)
5.4
(6.8)
9.5
(10.3)
5.3
(10.0)
6.6
(8.1)
6.5
(6.3)
Ignition delay (ms) 4.2
(1.24)
1.83
(0.91)
0.69
(0.44)
0.60
(0.50)
1.03
(0.45)
1.12
(0.64)
Premixed burn
fraction (%)
99
(77)
93
(76)
29
(8)
28
(13)
77
(9)
80
(28)
Max. burnt gas
fraction
0.11
(0.14)
0.23
(0.35)
0.62
(0.69)
0.67
(0.47)
0.40
(0.39)
0.22
(0.32)

On a practical engine system, the valve timing would be constant over all speeds and
loads. To estimate these parameters, a 6-mode FTP cycle approximation similar to
186
that used by Montgomery and Reitz (1996) and Montgomery (2000) was used. From
the former reference the mode weightings were calculated to be approximately:
Mode 1 46.5%
Mode 2 12.0%
Mode 3 7.1%
Mode 4 10.8%
Mode 5 12.9%
Mode 6 10.7%

The optimum parameters were those that minimised the cycle BSFC. The cycle
BSFC for each combination of parameters was evaluated as follows:

( )

|
|
.
|

\
|

=
modes
mode
mode
mode
modes
mode mode
rate Fuel
rate Fuel
BSFC Cycle
W
BSFC
W

Equation 5-26
Where W
mode
is the weighting for each mode.

Every combination of valve timings was evaluated and the optimum was found to be:
IVO = 145 ATDC
IVC = 280 ATDC
EVO = 140 ATDC
EVC = 270 ATDC
The cycle BSFC is 252 g/kWh and the modal BSFCs are shown in Table 5-9.

Note that the cycle BSFC for the baseline case is 272 g/kWh, the cycle NO
x
is
7.4 g/kWh and the cycle PM is 0.079 g/kWh.

187
Table 5-7: Modal BSFCs for valve timings optimised for a 6-mode FTP cycle approximation.
(IVO=145, IVC=280, EVO=140, EVC=270ATDC). Four-stroke baseline results are italicised.
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6
Input parameters
Speed (rpm) 750 953 993 1657 1737 1789
Intake air pressure
(kPa, absolute)
150 150 150 200 200 200
Intake air
temperature (K)
298 298 298 298 298 298
Exhaust pressure
(kPa, absolute)
100 100 100 100 100 100
Fuel rate
(g/cyl/cycle)
0.013 0.034 0.104 0.102 0.063 0.033
SOI ( ATDC)
352.0 359.5 354.5 367.5 361.5 354.5
Performance
Brake power (kW) 0.1 7.8 31.2 37.9 29.6 12.1
BSFC (g/kWh) N/A 249
(266)
198
(212)
236
(264)
222
(244)
294
(303)
Max. press. (MPa) 6.0
(5.6)
6.3
(6.8)
9.3
(10.3)
4.9
(10.0)
6.7
(8.1)
7.2
(6.3)
Ignition delay (ms) 1.51
(1.24)
1.33
(0.91)
0.93
(0.44)
1.22
(0.50)
0.63
(0.45)
0.61
(0.64)
Premixed burn
fraction (%)
89
(77)
86
(76)
41
(8)
73
(13)
37
(9)
33
(28)
Max. burnt gas
fraction
0.08
(0.14)
0.22
(0.35)
0.67
(0.69)
0.73
(0.47)
0.37
(0.39)
0.21
(0.32)
NO
x
(g/kg fuel) 0.69
(69)
13
(64)
86
(45)
-*
(15)
5.5
(13.8)
21
(22)
Soot (g/kg fuel) 0.47
(0.16)
0.23
(0.098)
0.016
(0.18)
-
(0.17)
0.27
(0.47)
0.25
(0.54)
*No ignition predicted by KIVA-ERC.

5.3.3 Addition of Reciprocating Air Pump
The Caterpillar SCOTE engine model was adapted to incorporate a reciprocating air
pump, similar to that described in Figure 1-3. Assuming the air pump operates at
twice the engine speed and each pump cylinder scavenges two engine cylinders, a
six-cylinder engine can be considered as three sub-units in parallel, each sub-unit
188
comprising one pump cylinder and two engine cylinders (Figure 5-29). Therefore,
only one of these sub-units need be modelled, saving much computational effort.

Figure 5-29: Engine sub-unit comprising one reciprocating pump cylinder and two engine
cylinders.

The addition of a reciprocating air pump adds at least two major parameters to the
system: the pump displacement and the pump phase. Other minor parameters
included the transfer manifold volume (assumed to be 2 litres) and the reed valve
effective area (assumed to be a generous 410
-3
m
2
, or more than twice the effective
area for two intake valves).

The work required to drive the pump was calculated by integrating the pump
pressure with respect to pump cylinder volume. Pump friction was estimated using
Equation 3-52, reduced by a factor of ten to account for the use of light metals, fewer
and less stiff piston rings, and thin-walled components. The latter is due to the
relatively light loads the pump components experience. A large bore/stroke ratio was
assumed for the air pump to keep piston velocities, inertial forces and friction to a
minimum. The air pump stroke was assumed to be fixed at half the engine stroke, or
82.55 mm. The assumed bore size therefore determined the pump displacement. The
parameter space searched was:

Air pump
(2 x engine speed)
Engine
Reed valve
Poppet valve
Intake manifold Transfer manifold
Exhaust manifold
189
EVO ( ATDC) = 110, 120 160
IVC ( ATDC) = 250, 260 300
IVO EVO ( CA) = 0, 10 30
IVC EVC ( CA) = 0, 10 30
SOI ( ATDC) = -15, -10 0
Air pump bore (mm) = 155.2, 174.6 232.8 (corresponding to a range of
64% 144% of the engine cylinder displacement)
Air pump phase ( ATDC with one engine cylinder at TDC) = 180, 190
230

This totals 69,120 combinations. For each mode, the calculations for ten engine
revolutions took approximately 50 hours CPU time on a Silicon Graphics Origin
3000 computer.



190
Table 5-8: Predicted near-optimum valve timings, engine performance and combustion
parameters for Modes 1-6 for the arrangement in Figure 5-29 based on the Cat SCOTE engine
with Shroud 1 on the intake valves. Four-stroke baseline results are italicised.
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6
Input parameters
Speed (rpm) 750 953 993 1657 1737 1789
Intake air pressure
(kPa, absolute)
100 100 100 100 100 100
Intake air
temperature (K)
298 298 298 298 298 298
Exhaust pressure
(kPa, absolute)
100 100 100 100 100 100
Fuel rate
(g/cyl/cycle)
0.015 0.037 0.118 0.100 0.077 0.041
Near-optimised
parameters

IVO ( ATDC) 170 180 170 150 160 150
IVC ( ATDC) 280 270 270 280 280 280
EVO ( ATDC) 160 160 150 130 140 150
EVC ( ATDC) 280 270 250 250 260 270
SOI ( ATDC) 350 355 355 355 355 350
Air pump bore
(mm)
155.2 155.2 194.0 194.0 174.6 155.2
Air pump phase
( ATDC)*
200 180 170 170 170 180
Performance
Brake Power (kW) 1.4
(0.6)
14.7
(15.2)
51.9
(59.2)
80.9
(76.8)
63.6
(56.8)
27.8
(25.4)
BSFC (g/kWh) N/A 287
(266)
230
(212)
246
(264)
253
(244)
317
(303)
Air pump loss
(kW)
0.6 1.1 3.5 8.6 6.1 3.5
Max. press. (MPa) 3.1
(5.6)
4.8
(6.8)
7.3
(10.3)
7.4
(10.0)
6.0
(8.1)
4.7
(6.3)
Ignition delay (ms) 3.49
(1.24)
0.97
(0.91)
0.39
(0.44)
0.24
(0.50)
0.29
(0.45)
0.53
(0.64)
Premixed burn
fraction (%)
98
(77)
74
(76)
14
(8)
6
(13)
8
(9)
25
(28)
Max. burnt gas
fraction
0.16
(0.14)
0.41
(0.35)
0.69
(0.69)
0.70
(0.47)
0.65
(0.39)
0.47
(0.32)
*When one engine cylinder is at TDC.


191

Every combination was evaluated, and the optimum values were:
IVO = 160 ATDC
IVC = 270 ATDC
EVO = 130 ATDC
EVC = 250 ATDC
Pump bore = 194 mm (pump/engine cylinder displacement ratio = 1.00)
Pump phase = 190 ATDC when one engine cylinder is at TDC.
Cycle BSFC = 312 g/kWh
Cycle NO
x
= 9.9 g/kWh
Cycle PM = 0.25 g/kWh

Modal data are shown in Table 5-9.

192
Table 5-9: Modal BSFCs for parameters optimised for a 6-mode FTP cycle approximation.
Four-stroke baseline results are italicised.
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6
Input parameters
Speed (rpm) 750 953 993 1657 1737 1789
Intake air pressure
(kPa, absolute)
100 100 100 100 100 100
Intake air
temperature (K)
298 298 298 298 298 298
Exhaust pressure
(kPa, absolute)
100 100 100 100 100 100
Fuel rate
(g/cyl/cycle)
0.017 0.037 0.118 0.100 0.077 0.041
Injection timing
SOI ( ATDC) 360 355 355 355 355 355
Performance
Brake power (kW) 0.2 12.4 51.7 80.5 59.5 18.8
BSFC (g/kWh) N/A 341
(266)
231
(212)
247
(264)
270
(244)
469
(303)
Air pump work
(kW)
1.8 2.9 3.0 9.3 10.4 11.3
Max. press. (MPa) 4.7
(5.6)
5.4
(6.8)
7.2
(10.3)
7.7
(10.0)
7.0
(8.1)
6.0
(6.3)
Ignition delay (ms) 0.83
(1.24)
0.54
(0.91)
0.40
(0.44)
0.22
(0.50)
0.22
(0.45)
0.25
(0.64)
Premixed burn
fraction (%)
64
(77)
27
(76)
14
(8)
5
(13)
4
(9)
0
(28)
Max. burnt gas
fraction
0.12
(0.14)
0.26
(0.35)
0.70
(0.69)
0.66
(0.47)
0.50
(0.39)
0.27
(0.32)
NO
x
(g/kg fuel) 28
(69)
54
(64)
40
(45)
25
(15)
28
(13.8)
36
(22)
Soot (g/kg fuel) 0.23
(0.16)
0.25
(0.098)
1.12
(0.18)
1.03
(0.17)
0.85
(0.47)
0.54
(0.54)

The results indicate that the best efficiency is associated with the smallest air pump
required to obtain sufficient air to avoid violating the maximum burned gas fraction
constraint. Modes 2, 5 and 6 clearly suffer because the air pump is larger than
necessary for these engine operating conditions. It would be desirable in a practical
implementation to have a small air pump over all engine operating modes. One
method of realising this is the use of a turbocharger and intercooler. At high speeds
and loads this combination delivers compressed, cooled air to the air pump, so that
the air pump has to deliver less air volume.
193


5.3.4 Addition of turbocharger
The 0-D engine model was further modified to reflect the addition of a Caterpillar
3406E turbocharger and intercooler (described in Sections 3.2.8 and 3.2.9). Because
only two engine cylinders were being modelled, the mass flow rate parameters on the
turbine and compressor maps were reduced to one-third their original values. Thus
the system approximated a turbocharged six-cylinder Caterpillar 3406E engine that
had been retrofitted to two-cycle operation. It was recognised that there were some
important differences: for example the amplitude and frequency of the flow
pulsations in the intake and exhaust manifolds of the two-cylinder model would be
different from a six-cylinder model, but the savings in computational time and effort
would be worth a small loss of accuracy.

A large compressor receiver (20 litres) was added to reduce the amplitude of pressure
oscillations due to the air pump, and therefore the tendency of the compressor to
surge. This also allowed more rapid convergence to a periodic solution. It was
subsequently noted that Heywood and Sher (1999, p. 439) state that a relatively large
compressor receiver is required when a turbocharger compressor is coupled with a
reciprocating scavenge pump, which confirmed the observations of the behaviour of
the numerical solution.

The turbocharged engine model is represented in Figure 5-30. The parameters
searched were:
EVO ( ATDC) = 120, 130 170
IVC ( ATDC) = 240, 260 300
IVO EVO ( CA) = 0, 10, 30
IVC EVC ( CA) = 0, 10, 30
SOI ( ATDC) = -15, -10 0
Air pump bore (mm) = 155.2, 174.6
Air pump phase ( ATDC with one engine cylinder at TDC) = 100, 120
260

194
This totals 1,944 combinations. For each mode, the calculations for two hundred
engine revolutions took up to 50 hours CPU time on a Silicon Graphics Origin 3000
computer.

The results are summarised in Table 5-10.


Figure 5-30: Turbocharged and intercooled engine subunit.

Air pump
(2 x engine speed)
Engine
Reed valve
Poppet valve
Intake manifold Transfer manifold
Exhaust manifold
Intake
Compressor Turbine
Intercooler
195
Table 5-10: Predicted near-optimum valve timings, engine performance and combustion
parameters for Modes 1-6 for the arrangement in Figure 5-30 based on the Cat SCOTE engine
with Shroud 1 on the intake valves.
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6
Input parameters
Speed (rpm) 750 953 993 1657 1737 1789
Intake air pressure
(kPa, absolute)
100 100 100 100 100 100
Intake air
temperature (K)
298 298 298 298 298 298
Exhaust pressure
(kPa, absolute)
100 100 100 100 100 100
Fuel rate
(g/cyl/cycle)
0.015 0.037 0.118 0.100 0.077 0.050
Near-optimised
parameters

IVO ( ATDC) 180 180 180 160 160 160
IVC ( ATDC) 280 280 260 280 280 280
EVO ( ATDC) 170 170 170 150 150 150
EVC ( ATDC) 280 280 250 270 270 280
SOI ( ATDC) 355 360 360 355 355 355
Air pump bore
(mm)
155.2 155.2 155.2 155.2 155.2 155.2
Air pump phase
( ATDC)*
180 180 180 180 180 180
Performance
Brake Power (kW) 1.4 14.3 63.8 82.1 60.6 32.0
BSFC (g/kWh) N/A 294
(266)
221
(212)
242
(264)
265
(244)
336
(303)
Air pump loss
(kW)
0.4 1.2 1.7 7.2 7.7 7.1
Max. press. (MPa) 3.0
(5.6)
4.3
(6.8)
7.7
(10.3)
9.9
(10.0)
8.8
(8.1)
5.6
(6.3)
Ignition delay (ms) 2.28
(1.24)
1.22
(0.91)
0.29
(0.44)
0.19
(0.50)
0.22
(0.45)
0.30
(0.64)
Premixed burn
fraction (%)
95
(77)
83
(76)
8
(8)
3
(13)
3
(9)
3
(28)
Max. burnt gas
fraction
0.19
(0.14)
0.42
(0.35)
0.69
(0.69)
0.53
(0.47)
0.44
(0.39)
0.44
(0.32)
Turbo speed (rpm) 28,000 33,200 57,600 78,000 73,700 69,800
*When one engine cylinder is at TDC.

The best parameters for a 6-mode FTP cycle approximation are:
196
IVO = 170 ATDC
IVC = 260 ATDC
EVO = 170 ATDC
EVC = 250 ATDC
Pump bore = 155.2 mm (pump/engine cylinder displacement ratio = 0.64)
Pump phase = 180 ATDC when one engine cylinder is at TDC.
Cycle BSFC = 290 g/kWh
Cycle NO
x
= 6.1 g/kWh
Cycle Soot = 0.17 g/kWh

197
Table 5-11: Optimised parameters for a 6-mode FTP cycle approximation. Four-stroke cycle
baseline results are italicised.
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6
Input parameters
Speed (rpm) 750 953 993 1657 1737 1789
Intake air pressure
(kPa, absolute)
100 100 100 100 100 100
Intake air
temperature (K)
298 298 298 298 298 298
Exhaust pressure
(kPa, absolute)
100 100 100 100 100 100
Fuel rate
(g/cyl/cycle)
0.015 0.037 0.118 0.100 0.077 0.050
Injection timing
SOI ( ATDC) 360 355 360 355 355 355
Performance
Brake power (kW) 0.7 13.5 63.6 80.0 57.6 29.6
BSFC (g/kWh) N/A 314
(266)
221
(212)
249
(264)
278
(244)
362
(303)
Air pump work
(kW)
0.7 1.6 1.5 9.4 9.8 8.7
Max. press. (MPa) 3.8
(5.6)
5.1
(6.8)
7.8
(10.3)
11.3
(10.0)
10.2
(8.1)
7.4
(6.3)
Ignition delay (ms) 0.97
(1.24)
0.48
(0.91)
0.27
(0.44)
0.15
(0.50)
0.17
(0.45)
0.20
(0.64)
Premixed burn
fraction (%)
73
(77)
18
(76)
7
(8)
1
(13)
0
(9)
0
(28)
Max. burnt gas
fraction
0.15
(0.14)
0.42
(0.35)
0.69
(0.69)
0.46
(0.47)
0.39
(0.39)
0.35
(0.32)
Turbo speed (rpm) 27,600 32,500 60,800 76,900 72,600 66,500
NO
x
(g/kg fuel) 33
(69)
33
(64)
18
(45)
19
(15)
20
(13.8)
17
(22)
Soot (g/kg fuel) 0.19
(0.16)
0.30
(0.098)
0.50
(0.18)
0.69
(0.17)
0.66
(0.47)
0.70
(0.54)

The intercooled turbocharger reduced the estimated pump work in all modes, despite
the increased backpressure.

Again, it is recognised that the emissions estimates are possibly inaccurate as the
KIVA-ERC model constants, tuned for four-stroke cycle operation, are likely to
require adjustment for the very different conditions associated with two-stroke cycle
operation.
198
The predicted cylinder pressure and heat release rates for the operating conditions in
Table 5-11 using both thermodynamic and KIVA-ERC models are compared in
Figure 5-31.

KIVA 0-D
0
2
4
6
330 340 350 360 370 380 390 400 410
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
0
200
400
600
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
J
/
d
e
g
)
(a) Mode 1
0
4
8
12
330 340 350 360 370 380 390 400 410
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
0
100
200
300
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
J
/
d
e
g
)
(b) Mode 2
0
4
8
12
330 340 350 360 370 380 390 400 410
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
0
100
200
300
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
J
/
d
e
g
)
(c) Mode 3
0
4
8
12
330 340 350 360 370 380 390 400 410
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
0
100
200
300
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
J
/
d
e
g
)
(d) Mode 4
0
4
8
12
330 340 350 360 370 380 390 400 410
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
0
100
200
300
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
J
/
d
e
g
)
(e) Mode 5
0
4
8
12
330 340 350 360 370 380 390 400 410
Crank angle (deg ATDC)
P
r
e
s
s
u
r
e

(
M
P
a
)
0
100
200
300
H
e
a
t

r
e
l
e
a
s
e

r
a
t
e

(
J
/
d
e
g
)
(f) Mode 6
Figure 5-31: Comparison of KIVA-ERC and thermodynamic models for the cases in Table 5-11.

The agreement between the two models in terms of pressure, ignition delay,
premixed burn spike and heat release rate shape is reasonably good. The
thermodynamic model was not calibrated for just these cases, but for the cases
described in Table 5-7 and Table 5-9, otherwise better agreement could have been
achieved.
Chapter 6 - Discussion
6.1 Thermodynamic model performance
The aim of the thermodynamic model was to accurately and rapidly model a wide
variety of two-stroke and four-stroke engine systems. Recalling Section 3.1, the
requirements were:

Speed
Flexibility
Accuracy

Each of these will be reviewed briefly in turn.

6.1.1 Speed
The combination of the thermodynamic approach, the simplest practicable sub-
models (described in Section 3.2), an efficient stiff ODE solver and the use of the
Fortran programming language resulted in an efficient and fast engine modelling
program. Simple models representing engine systems like the Caterpillar SCOTE
(Figure 3-10) took approximately six seconds to complete one hundred engine
cycles. More complex systems, like a turbocharged, intercooled, two-cylinder engine
with single-cylinder reciprocating air pump (Figure 5-30) took between thirty and
sixty CPU seconds to complete one hundred engine cycles. These runtimes allow
many thousands of parameter combinations to be investigated in a reasonable amount
of time.

The runtimes could be further improved by testing for convergence to a periodic
solution. In this way, the program would run only as many cycles as is needed to
converge within a certain tolerance. The maximum number of cycles would also be
limited in case the results converged too slowly or not at all.

Convergence can be accelerated by modifying the behaviour of some of the
submodels, especially the two-zone scavenging model. The abrupt transition from
zone 1 to zone 2 does not occur at precisely the same time on successive cycles
200
because of slight differences in conditions at the beginning of each cycle. Changes in
the timing of the transition cause differences in the conditions at the beginning of the
next cycle, which affect the timing of the next transition, and so on. A more gradual
transition would reflect reality better (see Figure 5-20 to Figure 5-24) and probably
increase the rate of convergence of solutions. The cost may be one or more additional
model parameters. The perfect diffusion scavenging model, which had no zone
transitions, converged noticeably more quickly than the two-zone model. Typically
ten or twenty engine cycles were required for convergence to close tolerances,
whereas approximately ten times as many cycles are required for the two-zone
model. However, the two-zone model does appear to give much more accurate
scavenging efficiencies (see Table 5-4).

Another sub-model that limited the speed of convergence was the turbocharger sub-
model. A reduction in the turbo moment of inertia could accelerate convergence to
an equilibrium speed. On the other hand, the cyclic turbo speed fluctuations would
also be magnified. Further investigation might suggest a reasonable compromise
between speed and accuracy.

Little attempt at parallelisation of the code has been made to date. When used to
explore parameter spaces, many slightly different problems are solved independently
of each other, which suggests that efficient parallelisation is possible.

In the present study, the parameter space was divided into an n-dimensional regular
grid and a merit function was evaluated at each grid point. This large mass of data is
useful for optimisation over several operating modes, when operating conditions may
not be optimised for each mode. Other optimisation schemes for individual modes
may be used, and it may be possible to adapt these to multi-modal optimisation.
These schemes include univariate searching, factorial analysis, Bayesian techniques,
expert systems, Response Surface Methods and Genetic Algorithms. These are
discussed briefly in (Montgomery, 2000). The application of Genetic Algorithms to
engine optimisation with KIVA-ERC and experimentation is described in (Senecal et
al., 2000). These methods aim to reduce the number of parameter combinations that
require evaluation to approach an optimum.

201

6.1.2 Flexibility
The thermodynamic model is able to represent almost any combination of
turbocharger, intercooler, valve, orifice, diesel engine cylinder, air pump cylinder
and plenum with no alterations in the source code. In practice, changes were often
made to the output subroutine to obtain the desired results.

In order to minimise the user input, some program values such as constants related to
fuel-air mixture properties were embedded as parameters in the Fortran source code.
In future, many of these values may be made more accessible to the user by
incorporating them in the model input file, or by creating a separate input file with
seldom-adjusted model parameters.

The program design has been influenced by both the KIVA engine modelling
program and the nature of the Fortran 77 programming language, with its text-based
input and output files, driver routine and subroutines. There is undoubtedly scope for
incorporation of features in the C, Fortran 90 and 95 languages.

During validation of the program, input files representing several engine systems
were created. It was found that the fastest way to generate new engine models was to
adapt an existing input file that most closely resembled the new engine system. In
this way, new input files took only a few minutes to create. Most of the time taken in
developing new input files was in doing a few runs to make sure the initial conditions
in the engine system were reasonable. Good initial conditions improved the
reliability and speed of convergence of the numerical integrator towards a periodic
solution.

6.1.3 Accuracy
At all stages of the development of the thermodynamic, parallel versions of the entire
model or submodels were developed using Fortran and MathCAD, a mathematical
software package. MathCADs symbolic language, built-in equation-solving and
graphing features allowed the rapid generation and investigation of prototype
submodels. MathCADs compact symbolic representations also provided a good
202
model for the subsequent Fortran implementation it allowed concentration on the
function of the subroutine without much of the distraction of how to implement it in
a programming language. Drawbacks with MathCAD include slow execution of
programs and the lack of a built-in stiff ODE solver that can automatically generate
approximate Jacobian matrices (user-supplied Jacobians are impractical for this
application). A built-in 4
th
-order Runge-Kutta solver was used to generate
approximate solutions.

Once a prototype subroutine or submodel had been developed and tested in
MathCAD, a Fortran implementation was developed. The behaviour of the Fortran
version was compared with the MathCAD prototype, a process that caught errors in
both the MathCAD and Fortran versions. Since all parts of the Fortran program were
compared with a relatively independent MathCAD version, any remaining
programming errors are expected to be few or minor.

The VODE ODE solver was usually used with absolute and relative tolerances of
10
-6
. Smaller tolerances resulted in longer runtimes, fewer errors in the submodels
(possibly due to smaller time steps and consequently smaller changes in values
between successive steps) and fewer failures with the VODE integrator.

The thermodynamic model was validated against published experimental data
reported by Montgomery (2000) and Wright (2001). The former reference reported
detailed measurements from a Caterpillar Single Cylinder Oil Test Engine (SCOTE)
based on the Caterpillar 3406E heavy-duty diesel engine. The latter reference
reported measurements taken from a production version of the six-cylinder
Caterpillar 3406E engine. The agreement with the Caterpillar SCOTE calibration
cases was good, with the calculated brake power and brake specific fuel consumption
being within 5% of experimental data (Table 3-9). There were sometimes greater
discrepancies in other estimated values, such as the ignition delay, premixed burn
fraction and air flow rate. Some of the discrepancy was attributed to limitations in the
semi-empirical correlations used to model the combustion behaviour. Some
discrepancies, such as those between the reported and calculated air flow rates in
cases 1-5 and 12-17, were difficult to explain, because all of the other results for
these cases were in good agreement. A small error in the charging efficiency should
203
be magnified in the pressure trace. One explanation could be errors in the
measurements.

In Section 3.2.3 is was argued that the maximum cylinder temperatures seldom
exceeded 2000 K and the pressure was also high, so that dissociation could be
neglected. Reviewing all of the cases, the maximum temperatures were calculated for
mode3 in Table 5-11, in which the maximum bulk gas temperature was 1955 K at a
pressure of 7 MPa (70 bar). Looking at Figure 3-3 and Figure 3-4, the errors in the
specific internal energy and specific gas constant are very small, much less than 1%.
The decision to simplify the gas property calculations by neglecting the effects of
pressure (i.e. dissociation) would appear to be justified.

The thermodynamic model assumed a constant 100% combustion efficiency.
Heywood (1989, p. 509) argued from the maximum allowable emissions that less
than two percent of the fuel energy left the cylinder in the form of hydrocarbons, soot
and carbon monoxide, so the assumption of complete combustion was a good
approximation. This reasoning appears sound. KIVA-ERC sometimes predicts
combustion inefficiencies greater than 2% (Hessel, 2003b) under operating
conditions which would suggest that this figure is unreasonably high based on
Heywoods reasoning.

Improvements in accuracy could be achieved by giving to the ODE integrator all
integration tasks in all submodels. Presently, only the equations for mass transfer,
energy transfer and conservation of burned gas fraction are integrated using VODE.
All other integration tasks used the approximation:

( )
1
1
1

+
i i
i
i i
t t
dt
dX
X X
Equation 6-1

where:
X is any quantity
t is time
i (subscript) denotes the current time step
204

In practice, VODE occasionally stepped backwards in time when its error estimate
exceeded the user-defined tolerances, so an array of previous values and the step
times had to be maintained to avoid these errors being incorporated into the solution.
Making VODE do all of the integration tasks would mean that all integrations would
be subjected to error estimations and would influence the integrator step sizes.
However, this would reduce the modularity of the thermodynamic model code and
may significantly increase the computational effort required to do a simulation.

Examination of the gas flow rates through the intake valves (Section 5.1.4) suggests
that gas dynamic effects are significant for even relatively short pipe runs. Rapid
acceleration and deceleration of the gas flow induces pressure gradients that cause
significant deviations from the zero-dimensional treatment. One such implementation
is detailed in Zhu and Reitz (1999).

6.2 Scavenging of two-stroke poppet-valved engines
The KIVA-ERC scavenging simulations support conclusions in previous studies
reported in the literature review that the optimum shroud size is in the vicinity of
100. Shrouds much larger than this reduced the delivery ratio and the scavenging
efficiency.

The smallest shrouds modelled had an average size of 105 and significantly
improved the scavenging efficiency over having no shrouds at all. The scavenging
efficiency approximated that of a uniflow engine. It was significantly better than the
perfect diffusion model, which justified the effort in finding a better scavenging
model. There was no evidence found for this shroud of significant vortex formation
that might trap gases in the centre of the cylinder. This was used to explain
combustion instabilities in spark-ignition prototypes.

The scavenging model described in Section 5.2 represents a novel contribution to
engine modelling. It was inspired by the examination of KIVA-ERC scavenging
simulations and by the three-zone model of Benson and Brandham (1969). The
exhaust gas purity predicted by the KIVA-ERC simulations showed a definite two-
205
stage process, which suggested two zones. This is in contrast to the assertion in
Heywood and Sher (1999, p. 130) that the exhaust gas purity curve is always a
sigmoid (S-shaped) for all types of scavenging systems. The model was generalised
to allow variations in the cylinder volume and pressure and the intake gas
temperature and pressure. The model has just two calibration constants that are easy
to determine from experiments or CFD simulations. The agreement between the two-
zone model and KIVA-ERC simulation for a given shroud geometry was excellent
over a wide range of engine speeds, intake and exhaust pressures and valve timings.
This suggests that the two-zone model is a robust and accurate contribution to the
modelling of poppet-valved two-stroke engines. The model may also be useful for
other scavenging systems.

The model calibration constant x represents the proportion of intake flow short-
circuited to the exhaust port. As would be expected, the magnitude of x was inversely
proportional to the size of the shroud. Larger shrouds would be expected to shield the
exhaust valves more effectively, reducing the value of x. This is reflected in the
results of the optimisation in Section 5.2.2, reproduced in Table 6-1. Short-circuit
flow was observed from simulations to occur largely around and between the intake
valve shrouds. Optimum shroud turn angles would minimise x. As observed, x was
fairly constant over a wide range of boost and exhaust pressures.

Table 6-1: Variation of x with average shroud angle.
No shroud
0
x = 0.34
Shroud 1
105
x = 0.12
Shroud 2
128
x = 0.10
Shroud 3
138
x = 0.09
Shroud 4
168
x = 0.07

The calibration constant k represents the volumetric rate of mixing of the residual gas
and the fresh charge normalised against the engine speed. It varied with shroud size
but was almost independent of engine speed or inlet and exhaust pressures and
temperatures. The cylinder, valve and shroud geometry would be expected to
influence the mixing rate within the cylinder. It is recommended that further work
include a more detailed physical interpretation of k.

206
Minor improvements to the two-zone scavenging model may be possible. One that
would yield a smoother and more realistic transition from zone 1 to zone 2 without
introducing new calibration constants would be to assume two-way mixing between
the two zones. This would then resemble some features of the Streit-Borman
scavenging model (Streit and Borman, 1971), except with short-circuit flow added.
Presently, zone 1 (the residual gas zone) donates its contents at a fixed volumetric
rate to zone 2 (the fresh charge zone). This would also perhaps aid convergence to a
cyclic solution.

6.3 Performance of two-stroke poppet-valved engines
6.3.1 Numerical modelling procedure
The procedure used in this study to model a relatively novel engine system relied on
both detailed experimental measurements and results obtained from detailed
simulations based as much as possible on fundamental physical processes. The
process of modelling two-stroke poppet-valved engines is illustrated in Figure 6-1.

The general approach was to start from well-studied physical systems and extrapolate
in small steps towards the engine systems of interest. When empirical data was not
available, for example for scavenging and combustion behaviour in two-stroke
poppet valved engine cylinders, KIVA-ERC was used as the next best option. Its
substantial basis in physical fundamentals and the relatively fine resolution of the
computational mesh are expected to yield results of sufficient accuracy for the
purposes of this investigation.

The approach in this study was to use the thermodynamic model to estimate initial
conditions for the 3D KIVA-ERC model. The advantage of this approach was that
the calculations can be performed relatively quickly. One limitation was that the
thermodynamic model does not provide information on the spatial distribution of gas
composition and temperatures at IVC. Since simulations showed the residual gas and
fresh charge were mixed fairly quickly within the cylinder, the distribution of gas
species and temperatures at injection was thought in most cases to be fairly uniform
(see for example Figure 5-8d, Figure 5-9d and Figure 5-10, which show fairly
uniform distributions well before IVC substantial further mixing would occur
207
during compression and the ignition delay period). One possibility for further work
would be to investigate the significance of the initial residual gas distribution on
predicted emissions levels. As mentioned in Section 5.1.2, the calculations would
require use of the 360 mesh and would have to encompass at least one full engine
cycle, increasing the run time for from approximately three days to approximately
nine days.
208
Figure 6-1: Flowchart of two-stroke poppet-valved engine modelling process. The boxes
represent tasks and the arrows represent the flow of information from one task to the next.

Development/ calibration/ validation of a thermodynamic
model of a four-stroke engine system (Chapter 3)
Four-stroke cycle gas exchange simulation
using a full-cylinder KIVA-ERC model
with valves and intake ports (Section 4.3)
Calibration/validation of combustion and emissions
model using KIVA-ERC and 60 sector cylinder
mesh (Section 4.3)
Initial
conditions
Initial
conditions
Turbulence and
swirl parameters
Preliminary modelling of two-stroke poppet-valved
engine system (Section 5.3.1)
Two-stroke cycle scavenging simulations using a
full-cylinder KIVA-ERC model with valves and
intake ports (Section 5.1)
Development and validation of two-
zone scavenging model (Section 5.2)
Results
Calibration
constants
Combustion and emissions modelling using KIVA-
ERC and 60 sector mesh (Section 5.3.1)
Turbulence and
swirl parameters
Initial
conditions
Calibration
constants
Performance estimations using
thermodynamic model of two-
stroke poppet-valved engine
system (Sections 5.3.2-4)
Combustion model
constants
Scavenging model
Initial
conditions
Combustion and emissions
modelling using KIVA-
ERC and 60 sector mesh
(Section 5.3.2-4)
Estimate of
system behaviour
Estimate of combustion
and emissions
209
6.3.2 System simulation results
As discussed in Chapter 4, KIVA-ERC seemed better able to predict trends in
emissions rather than absolute values in the absence of empirical data. If a genuine
emissions prediction capability for novel engine systems is required, then better
emissions models need to be developed, which is a significant undertaking (Kong,
2003). The thermodynamic model has no emissions prediction capability. The
limited emissions prediction capability and the relatively good BSFC prediction
capability (Section 3.3) suggested an approach whereby engine system parameters
were optimised for fuel efficiency alone. Should physical engine systems be
constructed, these parameter combinations could be used as a starting point. Once
emissions have been measured, they could be compared with target levels and
computer models could be calibrated. Further experiments and computer simulations
could then be used to refine engine system parameters.

Comparisons between four-stroke and two-stroke engine systems assumed
substantially the same engine hardware was used in each case. The engine cylinder,
piston, injector, valves (excepting the addition of a shroud for two-stroke operation)
and intake ports were the same for all cases. Additionally, the engine operating loads
and speeds were the same for each case. This approach was taken for the following
reasons:

Computational model constants calibrated for the four-stroke cycle are more
likely to apply to the two-stroke cases.
The number of parameters that need to be optimised are limited to a
manageable set.
Any future experimentation is likely to start off with this approach, as has
past experimentation (Section 2.2.2).
There is commercial interest in adapting four-stroke engines to two-stroke
cycles, especially if fuel efficiency or emissions levels can be improved.
It has been hypothesised that operating a two-stroke poppet-valved diesel
engine at loads similar to that of a four-stroke engine of the same
displacement could result in low emissions (see Section 1.2).

210
The disadvantages of this approach are:
Significant advantages may be overlooked by restricting the number of
dimensions in the parameter space.
The 6-mode FTP approximation is of limited value to non-vehicle
applications, such as aeronautical engine or stationary generator applications.
This approach did not investigate the potential for improved power density,
since the power levels were the same as for the four-stroke engine. An engine
system optimised for a certain cycle is not likely to perform as well on a
different cycle.

Future studies could expand understanding of two-stroke poppet-valved engine
systems. A large proportion of this study was devoted to development of the unique
tools and techniques required to simulate these systems.

The thermodynamic and KIVA-ERC models predict good fuel economy from a two-
stroke adapted Caterpillar Single Cylinder Oil Test Engine (SCOTE, described in
Figure 3 10 and Table 3 3). The thermodynamic modelling predicts better fuel
economy than the baseline engine (Table 5 6 and Table 5 7) and a 7% reduction in
BSFC over a 6-mode FTP cycle approximation. However, the parasitic scavenging
pump losses were not accounted for in this case and the baseline engine was
optimised for emissions and fuel economy rather than fuel economy alone. This
numerical exercised indicated that such an apparatus should not be difficult to
construct and that the scavenging and combustion behaviour should yield good
performance. Such an apparatus would be an ideal test bed for experimentally
investigating emissions from two-stroke poppet-valved diesel engines.

When the scavenging pump work is accounted for as in Sections 5.3.3 and 5.3.4, the
predicted cycle BSFC is 15% greater than for the baseline case without a
turbocharger and 7% greater when a turbocharger is added. The most efficient
operating modes were the high load cases (modes 3, 4 and 5 in the 6-mode FTP cycle
approximation). This is perhaps because the scavenging pump load, which was
mainly a function of speed, was proportionally less in these cases.

211
Two-stroke poppet-valved engines have been shown to operate at idle speeds
approximately half that of equivalent four-stroke engines (Rutherford and Dunn,
2002). If the turbocharged system idle speed is reduced from 750 rpm to 450 rpm,
the cycle BSFC is reduced from 290 g/kWh to 280 g/kWh, which is close to the
baseline four-stroke engine cycle BSFC of 272 g/kWh.

Modelling showed that a Caterpillar 3406E turbocharger could be used without
modification on an engine retrofitted to two-stroke cycle operation, significantly
simplifying and reducing the cost of the retrofit. Additionally, the turbocharged
system required a considerably smaller air pump than the unturbocharged system.
The best turbocharged parameter combination had an air pump cylinder displacement
64% of the engine cylinder displacement, whereas the ratio for the unturbocharged
system was 100%. This could reduce the cost and mass of a retrofitted engine while
improving performance.

The predicted two-stroke emissions levels generally showed no clear advantage over
four-stroke operation; however the engine parameters were not optimised for
emissions. The question of whether two-stroke poppet-valved diesel engines can run
more cleanly than four-stroke equivalents without an excessive fuel consumption
penalty can presently only be addressed with the assistance of experimentation. Of
particular interest is whether the internal EGR (unscavenged residual gases) would
suppress NO
x
and the high turbulence intensity would reduce PM, CO and HC
emissions.

Some of the best-case valve motion may be difficult to achieve while retaining
adequate durability. The analysis of the turbocharged system (Section 5.3.4)
suggested intake and exhaust valve open periods of 90 CA and 80 CA respectively
while assuming the same lift profile and maximum lift as the four-stroke cases which
had valve open periods of 242 and 235. The valve acceleration is therefore increased
by (24290)
2
and (23580)
2
or approximately 7-fold and 9-fold, respectively.
Nevertheless, it is useful to know what the ideal case is.


Chapter 7 - Conclusions and Further Work
7.1 Summary
The performance of a four-stroke direct-injection diesel engine converted to a two-
stroke cycle and scavenged by a reciprocating air pump has been estimated using a
combination of four-stroke experimental data, multidimensional modelling of the
cylinder, valves and intake ports, and zero-dimensional thermodynamic modelling of
the entire engine system.

A fast, flexible zero-dimensional model was developed, calibrated and validated for
four-stroke cycle operation using experimental data. The model is robust and capable
of automatically evaluating tens of thousands of simulations of relatively complex
engine systems in a few days. The model is therefore useful gaining an
understanding of novel engine systems with complex behaviour. It was used to
estimate near-optimum engine system parameters at single engine operating points
and over a six-mode engine cycle. The model was also used to provide boundary
conditions to a multidimensional engine model for more detailed simulations.

The multidimensional model was used for detailed scavenging and combustion
simulations and to provide estimates of emissions levels. The combustion
simulations were used to adjust zero-dimensional combustion correlations when
experimental data was not available. Scavenging simulations were performed with
shrouded and unshrouded intake valves.

7.2 Conclusions
Several conclusions can be drawn from the simulations performed in this study:

a) Experimental investigations of novel engine systems are greatly enhanced by
prior use of thermodynamic and multidimensional modelling to guide the
design of prototypes.

213
b) The use of thermodynamic and multidimensional modelling in parallel with
experimental investigations should greatly reduce the time and cost of
prototype development compared with experimentation alone.

c) Thermodynamic and multidimensional modelling of novel engine systems
without validation by experiments provides useful predictions of behaviour
throughout the system with the exception of emissions predictions.

d) A novel two-zone scavenging model has been developed and validated by
multidimensional modelling. It provides very accurate estimates of
scavenging efficiencies over a very broad range of engine speeds and boost
pressures.

e) Multidimensional scavenging simulations confirm previous studies that
indicate scavenging performance approaching that of a uniflow-scavenged
cylinder is possible. To achieve this, an intake shroud or other means of
reducing short-circuit flow is necessary. Intake shrouds larger than
approximately 105 reduce both the delivery ratio and scavenging efficiency.
This supports previous studies that concluded shrouds of approximately 90
were optimum.

f) Two-stroke poppet-valved engine systems can have a lower indicated specific
fuel consumption than equivalent four-stroke engines. This is because the
turbulence generated by high flow rates past the shrouded intake valve
increases the rate of combustion of fuel. This in turn means that a greater
fraction of heat release occurs at higher pressures, and the piston can extract
more work.

g) Parasitic losses from the external air pump were significant at low loads,
leading to increased brake specific fuel consumption relative to the four-
stroke cycle. Air pump losses were mitigated by the use of turbocharging
followed by charge air cooling, which reduced the specific volume of the
fresh charge. This in turn allowed the air pump size to be reduced by
approximately 36% compared to the un-turbocharged engine system.
214

Two-stroke poppet-valved engine systems are likely to achieve substantially
the same fuel efficiency as equivalent four-stroke engine systems.

h) Two-stroke poppet-valved engine parameters can not accurately be optimised
for both fuel consumption and emissions in the absence of experimental data
from a similar two-stroke poppet-valved engine. If such optimisation is
required, then it entails the construction and instrumentation of one or more
prototypes.

i) Two-stroke poppet-valved engine systems will have much greater demands
on the valve train than four-stroke engine systems.

7.3 Further work
7.3.1 Improvements to the thermodynamic model
The use of one-dimensional elements in the thermodynamic model should be
investigated. The overall air consumption and pressure traces showed satisfactory
agreement with experimental data and multidimensional modelling, but KIVA-ERC
modelling showed that significant variations do occur in intake and exhaust ports. A
trade-off analysis would have to be performed to determine whether any
improvements in system behaviour were worth the reduction in program speed.

Presently, the thermodynamic model is capable of automatically surveying the user-
defined parameter space. The implementation of an efficient automatic optimisation
scheme capable of optimising engine system parameters for multi-mode engine
cycles would be a great improvement.

One significant improvement to the thermodynamic model would be parallelisation
on a machine such as the Silicon Graphics Origin 3000.

A review of submodel behaviour would be desirable to identify ways to reduce the
computational effort of the ordinary differential equation integrator and speed
convergence to a periodic solution without affecting the overall model accuracy.
215

Numerous small improvements to the user interface and program source code can be
made.

7.3.2 Improvements to multidimensional modelling
Attempts to calibrate KIVA-ERC with experimental data indicated that trends were
reliably predicted, but that model constants required adjusting at different engine
operating conditions. The emissions estimates available from the multidimensional
modelling of the two-stroke poppet-valved engine systems must be considered
tentative until experimental data becomes available.

The effects of initial residual gas distribution on ignition delay, combustion and
emissions should be investigated further to determine whether the current approach
assuming a uniform distribution based on thermodynamic modelling is sufficiently
accurate. One alternative is to do full cycle simulations with a 360 mesh which may
entail a prohibitive amount of computational effort for some applications.

7.3.3 Two-zone scavenging model
Two-way mass transfer between the two zones should be investigated. This has the
potential to improve the exhaust gas purity curve. If this can be achieved without
compromising the presently excellent agreement with multidimensional scavenging
simulations, it would be a substantial improvement.

The physical interpretations of the scavenging parameters x and k should be
investigated further.

The applicability of the two-zone scavenging model to other scavenging
arrangements (e.g. cross-flow, loop, uniflow) could be investigated.

7.3.4 Further system studies
The systems reported in this study are applicable to the retrofitting of a diesel engine
to two-stroke operation and re-use at the same operating conditions. The two-stroke
poppet-valved engine system is likely to have substantially different operating
216
characteristics to a four-stroke engine, so different load-speed conditions may yield
much better results. Much further work is required to evaluate the overall potential of
two-stroke poppet-valved engine systems. In particular, the potential of this type of
engine for increases in power density should be investigated.
217
References

Abthoff, J., Duvinage, F., et al. (1998). "The 2-Stroke DI Diesel Engine with
Common Rail Injection for Passenger Car Application." SAE Paper 981032.
Amsden, A. A. (1993). KIVA-3: A KIVA Program with Block-Structured Mesh for
Complex Geometries. LA-12503-MS (March 1993), Los Alamos National
Laboratory.
Amsden, A. A. (1997). KIVA-3V: A Block-Structured KIVA Program for Engines
with Vertical or Canted Valves. LA-13313-MS, July 1997, Los Alamos National
Laboratory.
Amsden, A. A. (1999). KIVA-3V, Release 2, improvements to KIVA-3V. LA-13608-
MS (May 1999), Los Alamos National Laboratory.
Amsden, A. A., O'Rourke, P. J., et al. (1989). KIVA-II: A Computer Program for
Chemically Reactive Flows with Sprays. LA-11560-MS, May 1989, Los Alamos
National Laboratory.
Assanis, D. N. and Heywood, J. B. (1986). "Development and Use of a Computer
Simulation of the Turbocompounded Diesel System for Engine Performance and
Component Heat Transfer Studies." SAE Paper 860329.
Avallone, E. A. and Baumeister, T. (1996). Mark's Standard Handbook for
Mechanical Engineers, 10th ed. New York, McGraw-Hill.
Bazari, Z. (1992). "A DI Diesel Combustion and Emission Predictive
Capability for Use in Cycle Simulation." SAE Paper 920462.
Beatrice, C., Belardini, P., et al. (1996). "Assessment of predictivity of CFD
computations of combustion and pollutants formation in D.I. diesel engines."
Modeling and Diagnostics in Diesel Engines SAE Special Publications: 81-92.
Benson, R. S. and Brandham, P. T. (1969). "A Method for Obtaining a Quantitative
Assessment of the Influence of Charging Efficiency on Two-Stroke Engine
Performance." Int. J. Mech. Sci. 11: 303-312.
218
Borman, G. L. and Johnson, J. H. (1962). Unsteady Vaporization Histories and
Trajectories of Fuel Drops Injected into Swirling Air. SAE Paper 598C. National
Powerplant Meeting, Philadelphia, PA.
Bosch (1986). Automotive Handbook, 2nd ed. Warrendale, PA, Society of
Automotive Engineers.
Bowman, C. T. (1975). "Kinetics of Pollution Formation and Destruction in
Combustion." Prog. Energy Combust. Sci. 1: 33-45.
Brown, P. N., Byrne, G. D., et al. (1988). VODE, A Variable-Coefficient Ode Solver.
UCRL-98412 Preprint, June 1988, Lawrence Livermore National Laboratory.
Byrne, G. D. and Hindmarsh, A. C. (1975). "A Polyalgorithm for the Numerical
Solution of Ordinary Differential Equations." ACM Transactions on Mathematical
Software 1(1): 71-96.
Caterpillar (1998). Press Release, October 23, Caterpillar Inc.
Caterpillar (2002). On-Highway Engines Info/Specs: Non-current 3406E 14.6L,
www.cat.com/industry_solutions/shared/truck_engines/truck_engines.html,
Accessed:October 2002.
Chen, C., Veshagh, A., et al. (1992). "A Comparison Between Alternative Methods
of Prediction of Performance and Gas Flow in Internal Combustion Engines." SAE
Paper 921734.
Das, S. and Dent, J. C. (1993). "A CFD Study of a 4-Valved, Fuel Injected Two-
Stroke Spark Ignition Engine." SAE Paper 930070.
Dickey, D. W., Ryan, T. W., et al. (1998). "NOx Control in Heavy-Duty Diesel
Engines - What is the Limit?" SAE Paper 980174.
Fleck, R., Cartwright, A., et al. (1997). "Mathematical Modelling of Reed Valve
Behaviour in High Speed Two-Stroke Engines." SAE Paper 972738.
Freudenberger, R. (1995). Two Stroke Progress,
www.autosite.com/GARAGE/encyclop/ency19k.asp, Accessed:July 2000.
219
Fuchs, T. R. and Rutland, C. J. (1998). "Intake Flow Effects on Combustion and
Emissions in a Diesel Engine." SAE Paper 980508.
Gilmore, D. B. (1998). Torque Improvement (extract),
www.rotecdesign.com/Accreditation/Torque.htm, Accessed:October 2001.
Gordon, S. (1982). Thermodynamic and Transport Combustion Properties of
Hydrocarbons With Air. Part IProperties in SI Units. NASA TP-1906.
Gordon, S. and McBride, B. J. (1994). Computer Program for Calculation of
Complex Chemical Equilibrium Compositions and Applications I. Analysis. NASA
RP-1311, October 1994.
Halstead, M., Kirsh, L., et al. (1977). "The Autoignition of Hydrocarbon Fuels at
High Temperatures and Pressures - Fitting of a Mathematical Model." Combust.
Flame 30: 45-60.
Hampson, G. J. and Reitz, R. D. (1995). "Development of NO sub x and soot models
for multidimensional diesel combustion." Environmental Control Fuels and
Combustion Technologies American Society of Mechanical Engineers,
Environmental Control Division Publication, EC 1: 187-198.
Hampson, G. J., Xin, J., et al. (1996). "Modeling of NOx emissions with comparison
to exhaust measurements for a gas fuel converted heavy-duty diesel engine."
Diagnostics and Modeling in SI Engines SAE Special Publications: 191-205.
Han, Z. and Reitz, R. D. (1995). "Turbulence Modeling of Internal Combustion
Engines Using RNG k- Models." Combust. Sci. and Tech. 106: 267-295.
Han, Z. and Reitz, R. D. (1997). "A Temperature Wall Function Formulation for
Variable Density Turbulent Flows with Application to Engine Convective Heat
Transfer Modeling." Int. J. H&M Transfer 40: 613-625.
Han, Z., Uludogan, A., et al. (1996). "Mechanism of Soot and NOx Emission
Reduction using Multiple-Injection in a Diesel Engine." SAE Paper 960633.
Hardenberg, H. and Hase, F. (1979). SAE Paper 790493.
220
Hessel, R. (2003). Guidelines for Matching KIVA to Experimental Data,
www.erc.wisc.edu/~hessel/research/faq/guidelines_for_matching_kiva_to_experime
ntal_data.htm, Accessed:April 2003.
Hessel, R., Kong, S.-C., et al. (2003). Itape Inputs,
www.erc.wisc.edu/modeling/ThreeD/KIVA-II_input_file_presentation/ppframe.htm,
Accessed:April 2003.
Heywood, J. B. (1988). Internal Combustion Engine Fundamentals. New York,
McGraw-Hill.
Heywood, J. B. (1989). Internal Combustion Engine Fundamentals. New York,
McGraw-Hill.
Heywood, J. B. and Sher, E. (1999). The Two-Stroke Cycle Engine: Its Development
Operation, and Design. New York, McGraw-Hill.
Hiroyasu, H. and Nishida, K. (1989). "Simplified Three-Dimensional Modeling of
Mixture Formation and Combustion in a DI Diesel Engine." SAE Paper 890269.
Hohenberg, G. F. (1979). "Advanced Approaches for Heat Transfer Calculations."
SAE Paper 790825.
Horrocks, R. W. (1994). "Light Duty Diesels - An Update on the Emissions
Challenge." Proc. Instn Mech. Engrs 208: 289-298.
Huh, K. Y., Kim, K. K., et al. (1993). "Scavenging Flow Simulation of a Four-
Poppet-Valved Two-Stroke Engine." SAE Paper 930500.
Hundleby, G. E. (1990). "Development of a poppet-valved two-stroke engine. The
flagship concept." SAE Trans 99: 1657-1663.
Joyce, A. (1999). "Thermodynamic Mathematical Modelling," in Diesel Engine
Reference Book. B. Challen and R. Baranescu (ed.). Woburn, MA; Oxford, England,
Butterworth-Heinemann.
221
Kang, K. Y., Lee, J. W., et al. (1996). "The Characteristics of Scavenging Flow in a
Poppet-Valve Type 2-Stroke Diesel Engine by Using RSSV System." SAE Paper
960368.
Kittelson, D. B. (1998). "Engines and Nanoparticles: A Review." J. Aerosol Sci.
29(5/6): 575-588.
Knoll, R. (1998). "AVL Two-Stroke Diesel Engine." SAE Paper 981038.
Kong, S.-C. (2002). Personal communication, 10th February.
Kong, S.-C. (2003). Email dated 14th March 2003.
Kong, S.-C., Han, Z., et al. (1995). "The Development and Application of a Diesel
Ignition and Combustion Model for Multidimensional Engine Simulation." SAE
Paper 950278.
Krieger, R. B. and Borman, G. L. (1966). The Computation of Apparent Heat
Release for Internal Combustion Engines. 66-WA/DGP-4, New York, ASME.
Ladommatos, N., Abdelhalim, S. M., et al. (1996). "The dilution, chemical and
thermal effects of exhaust gas recirculation on diesel engine emissions Part 1: Effect
of reducing inlet charge oxygen." SAE Paper 961165.
Ladommatos, N., Abdelhalim, S. M., et al. (1996). "The dilution, chemical and
thermal effects of exhaust gas recirculation on diesel engine emissions Part 2: Effects
of carbon dioxide." SAE Paper 961167.
Ladommatos, N., Abdelhalim, S. M., et al. (1997). "The dilution, chemical and
thermal effects of exhaust gas recirculation on diesel engine emissions Part 3: Effects
of water vapour." SAE Paper 971659.
Ladommatos, N., Abdelhalim, S. M., et al. (1997). "The dilution, chemical and
thermal effects of exhaust gas recirculation on diesel engine emissions Part 4: Effects
of carbon dioxide and water vapour." SAE Paper 971660.
222
Ladommatos, N., Abdelhalim, S. M., et al. (1998). "The effects on Diesel
Combustion and Emissions of Reducing Inlet Charge Mass Due to Thermal
Throttling with Hot EGR." SAE Paper 980185.
Liu, A. B. and Reitz, R. D. (1993). "Mechanism of Air-Assisted Liquid
Atomization." Atomization and Sprays 3: 55-75.
McBride, B. J. and Gordon, S. (1996). Computer Program for Calculation of
Complex Chemical Equilibrium Compositions and Applications II. Users Manual
and Program Description. NASA RP-1311, June 1996.
Middlemass, C. (2002). Facsimile message dated 20th August 2002.
Montgomery, D. T. (2000). An Investigation into Optimization of Heavy-Duty Diesel
Engine Operating Parameters when Using Multiple Injections and EGR, PhD
Thesis, Mechanical Engineering, University of Wisconsin-Madison. Madison, WI.
Montgomery, D. T. and Reitz, R. D. (1996). "Six-Mode Cycle Evaluation of the
Effect of EGR and MultipleInjections on Particulate and NOx Emissions from a D.I.
Diesel Engine." SAE Paper 960316.
Morita, Y. and Inoue, I. (1996). "Development of a poppet-valved two-stroke
engine." Suzuki Technical Review 22: 1-7.
Nagle, J. and Strickland-Constable, R. F. (1962). Oxidation of Carbon between 1000-
2000 degrees Celsius. Proceedings of the Fifth Conference on Carbon, London,
Pergamon Press.
Nakano, M., Sato, K., et al. (1990). "Two-stroke cycle gasoline engine with poppet
valves on the cylinder head." SAE Trans 99: 1972-1984.
NIST (2003). Chemistry WebBook,
http://webbook.nist.gov/cgi/cbook.cgi?ID=C629594&Units=SI, Accessed:January
2003.
Nomura, K. and Nakamura, N. (1993). Development of a New Two-Stroke Engine
with Poppet-Valves: Toyota S-2 Engine. A New Generation of Two-Stroke Engines
223
for the Future? Proceedings of the International Seminar, Rueil-Malmaison, France,
Editions-Technip, Paris.
Reitz, R. D. and Diwakar, R. (1987). "Structure of High-Pressure Fuel Sprays." SAE
Paper 870598.
Rutherford, R. and Dunn, P. (2001). FREEDOMAIR for Diesel Engines: Low
Engine-Out Emissions of NOx and PM with High Power Density. Rotec Design Ltd,
Brisbane, QLD, Australia.
Sato, K., Takahashi, S., et al. (1981). "On the Scavenging Action of a New Type
Two Stroke Cycle Engine with Scavenging and Exhaust Valves on the Cylinder
Head." Journal of the Marine Engineering Society in Japan 16(11): 31-40.
Sato, K., Ukawa, H., et al. (1992). "Two-stroke cycle gasoline engine with poppet
valves in the cylinder head. Part II." Two Stroke Engine Diagnostics and Design SAE
Special Publications. n 901: 197-205.
Schindler, K. P. (1997). "Why Do We Need The Diesel?" SAE Paper 972684.
Senecal, P. K., Montgomery, D. T., et al. (2000). "A Methodology for Engine Design
Using Multi-Dimensional Modelling and Genetic Algorithms with Validation
Through Experiments." Int. J. Engine Research 1(3): 229-248.
Shampine, L. F. (1994). Numerical Solution of Ordinary Differential Equations.
London, Chapman and Hall.
Stokes, J., Hundleby, G. E., et al. (1992). "Development experience of a poppet-
valved two-stroke flagship engine." Two Stroke Engine Diagnostics and Design SAE
Special Publications. n 901: 251-261.
Stone, R. (1992). Introduction to Internal Combustion Engines. London, MacMillan.
Streeter, V. L. and Wylie, E. B. (1983). Fluid Mechanics. Singapore, McGraw-Hill.
Streit, E. E. and Borman, G. L. (1971). "Mathematical Simulation of a Large
Turbocharged Two-Stroke Diesel Engine." SAE Paper 710176.
224
Timoney, D., Brophy, B., et al. (1997). "Heat Release and Emissions Results from a
D.I. Diesel with Special Shrouded Intake Valves." SAE Paper 970900.
Tschoeke, H. (1999). The Fuel Injection System. Ypsilanti, MI, Society of
Automotive Engineers International-Diesel Engine Design Academy.
Van Wylen, G. J. and Sonntag, R. E. (1985). Fundamentals of Classical
Thermodynamics, 3rd ed. New York, John Wiley & Sons.
Watson, N., Pilley, A. D., et al. (1980). "A Combustion Correlation for Diesel
Engine Simulation." SAE Paper 800029.
Wright, C. C. (2001). Development and Application of a 1-Dimensional Multi-
Cylinder Turbocharged Engine Cycle Simulator, Master of Engineering Science
Thesis, Mechanical Engineering, University of Wisconsin-Madison. Madison, WI.
Yang, X., Ishima, T., et al. (1997). "High speed video recording of fog-marked
scavenging flow in a motored poppet-valved two-stroke engine." Design and
Performance of Two and Four Stroke Engines SAE Special Publications SAE
972736: 23-36.
Yang, X., Okajima, A., et al. (1999). "Numerical Study of Scavenging Flow in
Poppet-Valved Two-Stroke Engines." SAE Paper 1999-01-1250.
Zhu, Y. and Reitz, R. D. (1999). "A 1-D Gas Dynamics Code for Subsonic and
Supersonic Flows Applied to Predict EGR Levels in a Heavy-Duty Diesel Engine."
Int. J. of Vehicle Design 22(3/4): 227-252.

Você também pode gostar