: 3;: g E, 4::

 

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the
thesis entitled

1-D SIMULATION OF HCCI ENGINE PERFORMANCE USING
KNOCK-INTEGRAL IGNITION PREDICTION WITH WIEBE
FUNCTION COMBUSTION MODELING, AND COMPARISON TO

M.S.

ADVANCED SI ENGINE PERFORMANCE

presented by

Andrew Michael Huisjen

has been accepted towards fulfillment
of the requirements for the

degree in Mechanical Enfleering

 

 

#Vv’étfi/ O/Y‘f/flw’yy/

Major ProfeSBo'r’s Signature
,37/ 3/21/76)

Date

MSU is an Affirmative Action/Equal Opportunity Employer

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:/Proj/Acc&Pres/ClRC/DateDue.indd

l-D SIMULATION OF HCCI ENGINE PERFORMANCE USING KNOCK-
INTEGRAL IGNITION PREDICTION WITH WIEBE FUNCTION COMBUSTION
MODELING, AND COMPARISON TO ADVANCED SI ENGINE PERFORMANCE
By

Andrew Michael Huisjen

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
Mechanical Engineering

2010

ABSTRACT
l-D SIMULATION OF HCCI ENGINE PERFORMANCE USING KNOCK-
INTEGRAL IGNITION PREDICTION WITH WIEBE FUNCTION COMBUSTION
MODELING, AND COMPARISON To ADVANCED SI ENGINE PERFORMANCE
By

Andrew Michael Huisjen

A study of the fuel consumption benefits of Homogeneous Charge Compression
Ignition (HCCI) engine operation has been done using GT-POWER l-D engine
Simulation software. A port fuel injection spark ignition engine model was used as a
starting point for making modifications to create an HCCI model by implementing a
negative valve overlap breathing scheme and a simple knock-integral ignition timing and
Wiebe combustion model based on the in-cylinder conditions at a 2000 RPM and 2 bar
BMEP test point. Simulations run over a wide range of valve timing schemes produced a
range in which brake specific fuel consumption from this HCCI engine are expected to
fall. Further variations of the base PFI SI model were also developed to explore the fuel
consumption benefits of other competing current or near production-ready technologies
aimed at improving fuel consumption, and the simulation results of these models at the
same load and speed test point were compared with those of the HCCI model. These
models included variations on variable cam phasing, direct injection, intake valve
throttling, and lean combustion. In the SI simulations, results showed expected decreases
in BSFC as model complexity was increased. The HCCI simulations showed further
reductions in BSFC, and the improvements fell very much in line with published
experimental HCCI data. It was shown that simple model modifications can enable the

use of GT—POWER for effective approximate results for HCCI operation.

ACKNOWLEDGEMENTS

I would like to thank Dr. Harold Schock for giving me the opportunity to work on
this project and for his support and encouragement throughout the life of it. I would also
like to thank Dr. Gouming Zhu for his help in getting me up and running with GT-
POWER, John Opra and Lurun Zhong from Chrysler for their help with getting the GT-
POWER model calibrated and giving me the tools needed to do this project, and my
office mates Cody Squibb and Mulyanto Poort for their help in making work at the

engines lab interesting on a daily basis.

iii

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................ v

LIST OF FIGURES ...................................................................................................... vi

LIST OF SYMBOLS AND ABBREVIATIONS ......................................................... ix

CHAPTER 1

INTRODUCTION AND BACKGROUND ................................................................. 1
1.1 Overview of SI Operation ........................................................................... 2
1.2 Overview of HCCI Operation ..................................................................... 3

CHAPTER 2

l-D ENGINE MODELING .......................................................................................... 10
2.1 Base Model ................................................................................................. 11
2.2 Cam Phasing Model .................................................................................... 13
2.3 Direct Injection ........................................................................................... 18
2.4 PFI Lean-Bum Operation ........................................................................... 22
2.5 Intake Valve Throttling ............................................................................... 23
2.6 Burn Curve Duration and Phasing Sweeps ................................................. 27

CHAPTER 3

1-D HCCI ENGINE MODELING ............................................................................... 29
3.1 NVO Valve Timing ................................................................................... 30
3.2 Burn Curves ................................................................................................ 33
3.3 Lambda Control for Load ........................................................................... 38
3.4 Heat Transfer Model ................................................................................... 38
3.5 Knock Integral Calculation ......................................................................... 41

CHAPTER 4

RESULTS AND DISCUSSION ................................................................................... 44
4.1 SI Results .................................................................................................... 44
4.2 HCCI Results .............................................................................................. 53
4.3 Heat Transfer Sensitivity ............................................................................ 58
4.4 Higher Compression Ratio HCCI ............................................................... 60
4.5 Overall Comparison .................................................................................... 64

CHAPTER 5

SUMMARY AND CONCLUSIONS ........................................................................... 67

REFERENCES ............................................................................................................. 70

iv

LIST OF TABLES

Table 1: Specifications of base case GT-POWER model ................................................. 11
Table 2: Base case valve profile specifications ................................................................. 13
Table 3: Specifications used for direct injection system. .................................................. 19
Table 4: BSFC results for injection timing Sweeps of D1 models. .................................... 21

Table 5: Valve lift specifications obtained from intake valve throttling model
optimization ....................................................................................................................... 26

Table 6: Valve timing specifications for all NVO-multiplier combinations tested in HCCI

simulations ......................................................................................................................... 33
Table 7: Specifications of HCCI burn curves used in GT-POWER simulations .............. 37
Table 8: BSFC results for burn duration sweep of Atkinson cam phasing model ............ 45

Table 9: BSFC results for burn duration sweeps of D1 models with base case cam phasing

........................................................................................................................................... 48
Table 10: BSFC results for burn duration sweeps of DI models with Atkinson cam
phasing ............................................................................................................................... 49
Table 11: BSFC results for burn duration sweeps of all PFI lean models ......................... 51
Table 12: Comparison of air flow through the engine for stoichiometric and lean air-fuel
ratio, base case cam phasing and Atkinson cam phasing .................................................. 52
Table 13: BSFC results for burn duration sweeps of intake valve throttling model ......... 53
Table 14: Interpolated results of all successful HCCI simulations .................................... 58

Table 15: Best case BSFC results and percent improvement over base case for all GT-
POWER models created .................................................................................................... 65

LIST OF FIGURES

Figure 1: Comparison of negative valve overlap scheme with standard base case valve

timing ................................................................................................................................... 5
Figure 2: Base case valve timing ....................................................................................... 12
Figure 3: Base case mass fraction burned curve ................................................................ 14
Figure 4: Cam phasing BSFC map with overlays of Residual Gas Fraction levels .......... 15
Figure 5: BSFC map with overlay of RGF levels from refined cam phasing sweep ........ 15
Figure 6: Optimized cam phasing valve lift profiles shown with base case profiles ........ 16
Figure 7: BSFC vs. injection timing for various DI setups ............................................... 21

Figure 8: BSFC map with RGF overlay from initial valve timing sweep for intake

throttling model with burn duration of 31.1 CAD ............................................................. 25
Figure 9: Refined BSFC map with RGF overlay for intake valve throttling model with

burn duration of 50 CAD ................................................................................................... 26
Figure 10: Valve Profiles obtained from intake valve throttling model optimization ....... 27

Figure 11: Valve lift profiles using six different valve lift multipliers for the 180 NVO
setup ................................................................................................................................... 32

Figure 12: Comparison of standard scaled GT-POWER burn curve, modified GT-
POWER burn curve, and calculated Wiebe function burn curve with burn duration of
13.3 CAD ........................................................................................................................... 35

Figure 13: Sample of four MFB curve comparisons between GT-POWER modified
function and calculated Wiebe function ............................................................................ 35

Figure 14: Points used for determining the set of burn curves used in HCCI simulations...
........................................................................................................................................... 37

Figure 15: GT—POWER convection heat transfer multiplier for base case model, overlaid
with net mass flow rate through valves and valve lift profiles .......................................... 39

Figure 16: GT—POWER convection heat transfer multiplier modified for NVO valve
timing used in the HCCI model ......................................................................................... 40

vi

Figure 17: Sample temperature and pressure traces used for knock ignition integral
calculation .......................................................................................................................... 43

Figure 18: Examples of knock integral calculations shown graphically vs. CAD along

compression stroke ............................................................................................................ 43
Figure 19: BSFC results for burn duration sweep of Atkinson cam phasing model ......... 45
Figure 20: BSFC results for burn duration sweeps of all DI models ................................. 50
Figure 21: BSFC results for burn duration sweeps of all PFI lean models ....................... 51
Figure 22: BSFC results for burn duration sweeps of intake valve throttling model ........ 53
Figure 23: BSFC verses intake valve lift duration for different NVO setups .................... 55
Figure 24: Residual Gas Fraction verses intake valve lift duration for different NVO

setups ................................................................................................................................. 55
Figure 25: Lambda verses intake valve lift duration for different NVO setups ................ 56
Figure 26: Lambda verses residual gas fraction for different N VO setups ....................... 56

Figure 27: BSFC of HCCI simulations in regions in which autoignition is predicted to

occur for each NVO setup ................................................................................................. 57
Figure 28: Effect of NVO convection heat transfer multiplier on burn duration in HCCI
simulations with 180 NVO ................................................................................................ 59
Figure 29: Effect of NVO convection heat transfer multiplier on BSFC in HCCI
simulations with 180 NVO ................................................................................................ 60
Figure 30: Effect of compression ratio on BSFC in HCCI simulations with 180 NVO .......
........................................................................................................................................... 62
Figure 31: Effect of compression ratio on burn duration in HCCI simulations with 180
NVO ................................................................................................................................... 62
Figure 32: Effect of compression ratio and intake valve duration on amount of trapped
mass per cylinder in HCCI Simulations with 180 NVO .................................................... 63
Figure 33: Effect of compression ratio on lambda in HCCI simulations with 180 NVO .....
........................................................................................................................................... 63
Figure 34: BSFC results of all GT-POWER models ........................................................ 65

vii

Figure 35: Percent improvement in BSFC over the base case for all GT-POWER models.
........................................................................................................................................... 66

Images in this Thesis are presented in color.

viii

BMEP

BSFC

ISFC

DI

GDI

PFI

SOI

CAD

TDC

TDCF

TDC GE

ATDC

ATDC GE

BTDC

NVO

EVO

EVC

IVO

IVC

ECL

ICL

LIST OF SYMBOLS AND ABBREVIATIONS

Brake Mean Effective Pressure
Brake Specific Fuel Consumption
Indicated Specific Fuel Consumption
Direct Injection

Gasoline Direct Injection

Port Fuel Injection

Start of Injection

Crank Angle Degrees

Top Dead Center

Top Dead Center Firing

Top Dead Center Gas Exchange
After Top Dead Center

After Top Dead Center Gas Exchange
Before Top Dead Center

Negative Valve Overlap

Exhaust Valve Opening

Exhaust Valve Closing

Intake Valve Opening

Intake Valve Closing

Exhaust Center Line

Intake Center Line

ix

SOC

MFB

CAO

CASO

HCCI

IC

RGF

SI

CR

RPM

zgn

A6

Start of Combustion

Mass Fraction Burned

Crank Angle of 0% Mass Fraction Burned
Crank Angle of 50% Mass Fraction Burned
Homogenous Charge Compression Ignition
Internal Combustion

Residual Gas Fraction

Spark Ignition

Compression Ratio

Revolutions Per Minute

Temperature

Pressure

Equivalence Ratio

Ignition Delay

Characteristic Burn Duration
Stoichiometry—based Air—Fuel Ratio

Wiebe Function Exponent

CHAPTER 1

INTRODUCTION AND BACKGROUND

A study of the fuel consumption benefits of Homogeneous Charge Compression
Ignition (HCCI) engine operation has been done using GT-POWER l-D engine
simulation software. A 4-cylinder gasoline spark ignition engine model was used to
compare the benefits of HCCI relative to standard spark ignition (SI) operation.
Variations of the base SI model were also developed to explore the fuel consumption
benefits of other current or near production-ready technologies aimed at improving fuel
consumption. The goal of these studies was to verify the effectiveness of HCCI as an
implementable technique for efficiency gains by modifying an existing production—grade
SI engine model to simulate HCCI operation and demonstrate improvements similar to
what have been seen in experimental research. While some modeling of HCCI has been
done which involves complicated chemical kinetics systems for predicting combustion,
the goal in this study was to Show that simpler methods of modeling combustion based on
existing experimental correlations can be used effectively to produce expected results

which much less computational effort.

HCCI engines operate by indUcing autoignition of the air—fuel mixture due to the
temperature and pressure conditions in the cylinder as the piston nears the end of the
compression stroke—there is no spark used to initiate combustion as in a standard SI
gasoline engine. Thus, the operation of an HCCI engine is dependent on controlling the
temperature and pressure in the cylinder, as well as the air-fuel ratio, the residual gas

fraction (RGF) trapped in the cylinder, and the compression ratio; all of these parameters

together determine the combustion characteristics in the engine. The goal in HCCI
operation is to control these variables in a way that will in turn control the phasing and

duration of combustion consistently in each cycle.

1.1 Overview of SI Operation

Conventional gasoline-fueled IC engines operate by igniting the air-fuel mixture
in the cylinder with a spark from the spark plug. The amount of air allowed into the
engine is controlled by a throttle plate upstream of the intake valves, and an appropriate
amount of fuel is mixed into this air by the fuel injector to form a stoichiometric mixture.
While this design is very robust in its simplicity and ease of control, it does have fuel
consumption limitations; these can be partially overcome by the use of additional
technology, though in the future, additional improvements to the standard SI engine may

not be enough to meet fuel consumption needs and regulations.

Cam phasing can be used to help an engine breathe more efficiently and also to
trap high levels of exhaust gas from the previous cycle to create a higher residual gas
fraction (RGF) in the cylinder. A cam phasing scheme that utilizes Atkinson cycle style
valve timing is an example of an extreme cam phasing strategy that can increase
efficiency. Furthermore, a fully variable intake valve (both lift and duration control) can
be used to eliminate the throttle plate and further reduce pumping losses. Direct injection
(DI) can be used to cool the air-fuel mixture in the cylinder and allow for higher

compression ratios. In some cases, operating lean at lambdas greater than 1 can reduce

fuel consumption, partially due to reduced pumping losses due to a more open throttle

position.

1.2 Overview of HCCI Operation

HCCI operation allows for an improvement in fuel efficiency over normal SI
engines because of the combined effects high RGF dilution levels, high lambda (lean)
conditions, high compression ratio, and the characteristic short combustion duration.
While the burn duration (the 10-90% burn period) time in an SI engine can be on the
order of 20 to 60 CAD depending on load, speed, and set-up conditions, ideal HCCI
operation can occur with burn durations as Short as 5 to 10 CAD. HCCI is achieved
using a throttle-less setup—metering of the intake air is performed entirely by the valves.
The reduction of pumping losses from this and the theoretical increase in cycle efficiency
from the Short burn duration give HCCI much potential for part load fuel consumption
improvements [1]. Dahl et al. saw a 15% to 22% BSFC improvement for HCCI operation
[2]; Urata et al. achieved a similar 14% to 22% improvement [3]. Zhao et al. achieved up
to a 30% improvement at low loads and a 10%-15% improvement near the 2 bar BMEP
and 2000 RPM load point used in the test here [4]. Persson et al. achieved a greater than

15% improvement in ISFC [5].

The key to successful HCCI operation is controlling the temperature and pressure
conditions in the cylinder in a way that combustion is initiated at the appropriate moment

Shortly before top dead center (TDC). If conditions are too hot or at too high a pressure,

autoignition can be induced too early and cause harmful knock; if conditions are too cool
or at too low a pressure, autoignition may not happen at all, causing a misfire, or it may
happen so late that efficiency gains are compromised because of the lengthened burn
duration that results [6]. Much of the control of these conditions can be performed using

advanced valve timing techniques.

1.2.1 Valve Timing for HCCI Operation

Two common valve timing techniques used to achieve HCCI operation are
Negative Valve Overlap (NVO) and exhaust rebreathing. Exhaust rebreathing can be
done in two different ways; the exhaust valve can either be left open long into the intake
stroke as seen in [l], or the valve can be closed at its conventional time and then quickly
re—opened for a brief period during the intake stroke as performed in [6] and [7]. Both of
these techniques allows for hot residual gases to be drawn back into the cylinder. The
NVO technique has been used in many studies such as ones by Adomeit et al who refer to
it as CCR for Controlled Chamber Recirculation [l], Dahl et al. [2], Babajimopoulos et
al. who refer to it as early exhaust valve closing [7], Urata et al. who refer to it as NOL
[3], and Law et al. [8]. NVO involves closing the exhaust valve early and opening the
intake valve late. The early exhaust valve closure traps some of the residual gases in the
cylinder before they can be fully discharged during the exhaust stroke. The NVO period
refers to the number of crank angle degrees near TDC during which neither the intake or
exhaust valves are open. NVO is frequently done symmetrically—that is, the number of

CAD between exhaust valve closing (EVC) and TDC is often set to be the same as the

number of CAD between TDC and intake valve opening (NO) as in [3]; furthermore, the
lift and duration period of the exhaust and intake valves are set to be equal in a
symmetrical NVO scheme. This creates a valve lift scheme that is symmetrical about
TDC (Figure 1), though this is not always the way it is implemented—Dahl, Denbratt,
and Koopmans used a non-symmetrical NV 0 setup in their HCCI experiments [2]. The
NVO period (the number of CAD between EVC and NO) can be varied as a way to help
control RGF levels; thus, NVO is an important key to understanding behavior of an

HCCI engine, and an easy way to parameterize experiments and results ([5], [9]).

 

 

   
  
  
     
  

 

 

 

 

 
 

 

 

 

 

 

 

 
   

Valve Lift
l
— Exhaust NVO
— Exhaust Base
3 I ‘ “ ‘ — Intake NVO
— Intake Base
6 . L -
Lift
[mm]
4 ._, _ L
2 _
lI \\
0 iiiiiiiiiii “H;- uuhnn ....... ““41““.
0 360 720
TDCF BDC TDC GE TDCF

Crank Angle [degrees]

Figure 1: Comparison of negative valve overlap scheme with standard base case valve timing.

Using NVO to control the RGF level in the cylinder in turn helps to determine the
temperature conditions that will be achieved during compression. Since the trapped RGF
is hot, high RGF levels (50% and above) are desirable for creating hot enough
temperature conditions for well-timed autoignition. However, the total effects of RGF
dilution are complicated, because while high temperatures can advance the start of
combustion, RGF levels which are too high can also slow the start of combustion.
Having too much RGF dilution can lead to very low exhaust temperatures, and while this
can be desirable from an emissions standpoint (low temperatures inhibit the formation of
NOx), too low a temperature hurts the ability to achieve auto-ignition through use of the

heat of the residual gases.

Another aspect of NVO schemes is that it uses only small amounts of valve lift.
While normal SI engine operation might use valve lifts in the neighborhood of 8 to
10mm, NVO schemes often use much Shorter valve lifts, as shown in Figure 1; while the
base case profiles shown both have lifts of approximately 9 mm, the HCCI NVO valve

lift profiles Show only about 5 mm of lift.

1.2.2 HCCI Combustion

As previously noted, controlling the temperature and pressure conditions in the
cylinder is vital to achieving proper HCCI combustion. For modeling purposes, it is
necessary to have a way of using the model-simulated in—cylinder conditions to predict

the start and duration of the combustion event.

To predict the start of combustion from in-cylinder conditions, a knock integral
calculation approach can be used. For any given time in the cylinder, an ignition delay,
or induction time, t, can be calculated based on instantaneous conditions. Expressions
for calculating this induction time vary from source to source. One expression which
comes from iso-octane rapid compression research by He, et al [10] and has been

successfully implemented in previous HCCI modeling work by Chang et al. [6] is

_1.41 * 33700
ign 02 CXPI (1)

R*T

where P is the pressure in atmospheres, T is the temperature in Kelvin, ¢ is equivalence

ratio, X 02 is the mass percentage of oxygen, and R is the gas constant in cal/mol-K.
Once this Tign term is calculated for each crank angle increment along the compression

stroke, the actual autoignition timing is determined by integrating its inverse along crank

angle until the integral reaches a value of l:

CAO

[(I/r. }1CA=1.0 (2)
Ign

IVC

The limits of integration range from intake valve closing (IVC), determined by the valve
timing, to the start of combustion (or CAO for crank angle at 0% burn), which is the

variable being determined from this expression.

Once equation (1) and (2) have been used to determine CAO, the combustion
duration has already effectively been determined. Experimental data in [6] has shown
that burn characteristics are affected primarily by the timing of the start of combustion
and are relatively independent of the mixture composition. In general, earlier phasing of
CAO results in a shorter burn duration; trends from experimental data can be curve fit so
thata database of possible realistic HCCI combustion curves can be created from the

Wiebe function

_ w+l
CA CAO) ] (3)

MFB=C *1— -
efi‘ CXPII A0

where the mass fraction burned (MFB) is calculated as a function of crank angle (CA),

start of combustion (CAO), combustion efficiency (C ), characteristic burn duration

615‘

(A6 ), and Wiebe exponent (w). Cejf

, A6 , and w are determined from experimental

data curve fitting found in [6].

1.2.3 Compression Ratio

Increasing compression ratio is one technique that can be used to aid in achieving
auto-ignition in a HCCI engine. Clearly, an increase can lead to higher pressure and
temperature conditions in the cylinder, and these could potentially allow for more RGF
diluted or leaner mixtures to become feasible for use. However, from a practical
standpoint, Since most HCCI work is done under the assumption that an actual production
engine would need to be able to run in both HCCI mode and standard spark ignition
mode for a full range of operation, as outlined in [4], not all HCCI work is done at high
compression ratios. Compression ratios are seen across the board, ranging from [0.3:]
in work found in [4] and [5] and 10.521 in [8] up to 17:1 in [7] and 18:1 in [11]; the area

in many tests are performed is between 1 1:1 and 15:1, as in [6], [l], [2] ,[3] ,and [9].

CHAPTER 2

1-D ENGINE MODELING

Modeling was performed using the l-D engine simulation software GT-POWER.
A working model of a 2.4L 4-cylinder gasoline PFI engine was provided by Chrysler as
the starting point for the simulations. This model was set up to be run over a range of
load and speed points, and all specifications were based on an actual production engine—
the model included complete intake and exhaust systems (i.e. manifolds, runners, throttle,
air cleaner, catalytic converter, muffler, etc) as well as cylinder, head, valve and port
geometries. Operational parameters such as cam phasing, combustion phasing and
duration, and air-fuel ratio were determined by built-in look-up table maps based on load
and speed of the simulation case. Key specifications are shown in Table l. The chosen
test point of interest for all simulations run in this study was 2 bar BMEP and 2000 RPM,

used to simulate a typical cruising Speed and low load operation point.

The modeling process was broken up into two main parts. The first was to make
numerous modifications to the base case model to create a group of SI models that
demonstrated various modern techniques aimed at improving fuel efficiency. The second
was to modify the base model to run in HCCI mode. The goal behind all modeling was
to create expected approximate brake specific fuel consumption (BSFC) figures for each
explored technique, and then compare those to the expected BSFC figures from HCCI

Simulations.

Table 1: Specifications of base case GT-POWER model

 

 

 

 

 

 

 

 

 

 

 

 

 

Enflie Laput Type Inline 4
Bore 88 mm
Stroke 97 mm
Displacement 2.4 L
Compression Ratio 10.5
Intake Valve Diameter 32.1 mm
Exhaust Valve Diameter 25 mm
Valve Train 4 valve/cyl
Fuel Infiction Type PFI

Fuel Type lndolene

 

Development of the SI models involved modifying specifications and setups and

performing parameter optimization sweeps aimed at minimizing BSFC.

2.1 Base Model

The base model was run at the desired 2000 RPM speed and 2 bar BMEP load
with the use of a throttle controller to target the load. No modifications were made to this
model—it was run to create a baseline set of operational parameters and results from

which to develop and compare other models.

The cam phasing specifications in the base model use Intake Center Line (ICL)
and Exhaust Center Line (ECL) as the anchor points in the lift profiles, which, for the
chosen test point, are shown in Figure 2. These specifications in the base case came from
the load-and-Speed-based lookup tables that were built in to the provided GT-POWER
engine model—the model was equipped with specifications that simulate variable cam

phasing on both the intake and the exhaust cams.

ll

Valve Lift—Base Case

 

 

 

 

 

 

 

 

 

 

 

 

 

   

 

10
— Intake
— Exhaust
8 a _ _a_ - -_ __L.
6
Lift \
[mm]
4. - - _-- _- -s
2 ., - L _._- - _ - -___1
o .............. L.) . ..
360 540 720

 

TDCF BDC TDC GE BDC TDCF
Crank Angle [degrees]

Figure 2: Base Case valve timing.

The specifications for cam phasing can either be based on TDC Firing (TDCF) or
on TDC Gas Exchange (TDC GE). For most engine specifications (spark, injection
timing, etc.), TDCF is used, and is generally just referred to as TDC. If TDCF is
specified as 0 CAD (as Shown in Figure 2), then the ECL and ICL for the base case are at
265.6 CAD ATDC and 469.9 CAD ATDC, respectively. However, since valve timing
events take place around TDC GE and are part of the gas exchange process, it is often
times more convenient to specify valve timing and cam phasing specs using TDC GE.
This makes it possible to specify timings as a negative value for the exhaust valve and a
positive value for the intake valve, and this makes it easy to immediately visualize how

the valves are phased relative to each other and to TDC GE. Under this convention, ECL

l2

and ICL for the base case are -94.4 CAD ATDC GE and 109.9 CAD ATDC GE,
respectively (Table 2). This TDC GE convention will be used to refer to valve timing
and cam phasing from this point on unless otherwise noted, though in figures showing

valve profiles, the TDCF convention is still used for the scale on the timing axis.

Table 2: Base case valve profile specifications

 

 

 

 

 

 

 

 

 

Intake Valve Lift 9.1 mm

Exhaust Valve Lift 8.6 mm

Intake Valve Duration 262 CAD

Exhaust Valve Duration 317.4 CAD

Intake Cam Timing (ICL) 109.9 CAD ATDC GE
Exhaust Cam Timing (ECL) -94.4 CAD ATDC GE

 

The air-fuel ratio the stoichiometric 9:] condition. The base combustion model
was a Simple Wiebe function using a crank angle of 50% burn (CASO) anchor point of
7.5 CAD ATDC and a 10-90% burn duration of 31.] CAD (Figure 3). The simulations

for the base case resulted in a baseline BSFC of 324.3 g/kW-h and an RGF of 27.4%.

2.2 Cam Phasing Model

The goal of the cam phasing model was to optimize the base model for best BSFC
using extreme cam phasing. This was done by setting up sweeps of ECL and ICL to
create operation maps of BSFC over ranges of these cam phasings, using optimizing
software to find the region of minimum BSFC on these maps, and then confirming that
point by running further follow-up simulations at that point. The same throttle controller
from the base case model was used to control the 2-bar BMEP load as variations of ICL

and ECL were performed.

l3

 

0 Burned Fuel Fraction (Normalized by Fuel Mass)

 

0.8 4 '1

0.6 ,__

0.2 //

 

 

 

 

 

 

 

 

 

 

 

 

0.40 -2o 0 20 4o 60 so 100

[CAD ATDC]

Figure 3: Base case mass fraction burned curve.

2.2.1 BSFC Map Optimization

Initially, ICL was swept from 125 to 170 CAD ATDC GE while ECL was swept
from —85 to -65 CAD ATDC GB. The resulting BSFC and RGF plots were overlaid in
order to target areas of minimum BSFC at specific target RGF levels (Figure 4). With
the base case RGF level of 27.4% in consideration, an RGF limit of 30% was chosen for
this cam phasing optimization, in agreement with Heywood’s approximation of the
maximum RGF a typical SI engine can handle at part load [12]. The map shows that the

area of interest (circled in white) is near ICL=160 and ECL=-74. Thus a second sweep

l4

-65

    

-70 5 40 0
ECL BSFC
[CAD] [g/kW-hr]
300
'75 303
305
308
.80 . 311
r 314
316
‘8‘IZ5 130 135 140 145 150 155 160 165 170 319
ICL [CAD]
Figure 4: Cam Phasing BSFC map with overlays of Residual Gas Fraction levels.
-74.0
-74.5
ECL
[CAD]
-75.0
-75.5
'76'960 164 166

 

ICL [CAD]
Figure 5: BSFC map with overlay of RGF levels from refined cam phasing sweep.

15

was run which focused just on an area surrounding this target zone, sweeping ICL from
160 to 170 and ECL from -76 to -74 (Figure 5). The results from this model were run on
GT-POWER’S post-processor optimizing software DOE-Post to create a mathematical
model to target cam phasing that would theoretically yield the lowest BSFC when held to
a given RGF constraint. The result of this optimization process was an optimized cam
phasing of ECI;-75 and ICL=165; that is, the exhaust cam was retarded by 19.4 CAD,
and the intake cam was retarded by 55.1 CAD. Figure 6 shows the optimized profiles as
solid lines next to the base case profiles, which are shown as dashed lines. Since
optimization only involved phasing, the Shapes and specifications of the valve lift profiles

(duration and lift) were left unchanged from the base case.

o-Valve Profiles-Cam Phasing Optimization

 

 

  
 
  
 

 

 

 

 

 

 

 

 

 

 

 

 

    

1
- - Intake Base
— Intake Phased
8 -- ~ - - Exhaust Base
— Exhaust Phased
6 I
Lift ,’
two] i
4 - f
I
I
2» .. .:
I
l
I
[I
Oouuuuu “186 1111111111111
TDCF BDC TDC GE

Crank Angle [degrees]
Figure 6: Optimized cam phasing valve lift profiles shown with base case profiles.

16

The model was re-run with this optimized valve timing to confirm that the
mathematical model did indeed find a point near the optimum. This optimized cam
phasing model resulted in a BSFC of 307.8 g/kW-hr, a 5.1% improvement over the base

C386.

2.2.3 Atkinson Valve Timing Technique

The cam phasing used for this model essentially creates an Atkinson cycle type
valve timing scheme—the intake valve is left open far into the compressions stroke,
causing some fresh charge to be forced back into the intake port. This reduces pumping
losses on part load operation, but also reduces effective compression ratio; the region
where the tradeoff between the efficiency gains from pumping loss reduction and losses
from low compression is evident in Figure 4 and Figure 5 where the BSFC curves reach a
minimum near ICL=164 CAD and then begin increasing with further retarding of the

intake valve timing.

An important limit to check with this Atkinson cycle style cam phasing is to make
sure IVC occurs prior to Spark. In this case, using the base case combustion curve, the
IVC of 61 CAD BTDC is clearly early enough for the start of combustion timing of 25
CAD BTDC. However, if the lower effective compression ratio caused by the late IVC
were to cause a lengthened combustion duration, a much earlier spark timing would be

necessary to preserve the original or optimal CA50 phasing; in this case the spark timing

17

could potentially become a limiting factor for how far the intake timing could be

retarded.

2.3 Direct Injection

Creating a direct injection (DI) model involved an injection system swap—the
port fuel injectors were removed and replaced with injectors on the cylinders. A benefit
of a D1 system is that the evaporation of the fuel in the cylinder after injection has a
cooling effect on the charge in the cylinder. In a practical sense, a major benefit of this
cooling is that it allows an engine to be run at higher compression ratios than in normal
PFI operation. Thus, along with the use of a new injection system, compression ratio
effects were introduced to the model—it was run at compression ratios of 10.5 (same as

the base case) and 12.5.

Direct Injection model simulations were run using both the base case cam timing

and the optimized cam timing from Section 2.2. Rather than re-optimize the cam phasing

again for the DI model, it was decided that a better comparison of PFI vs. DI performance

could be made if the cam phasing was left the same for the two models.

2.3.1 Injector Specifications

Since no actual injector Specifications for a D1 setup existed for this particular

engine model, necessary specifications (injection rate, fuel temperature, etc.) were based

18

on those found in a built—in gasoline direct injection (GDD GT -POWER model (Table 3).
An evaporation model was also added to the cylinder that defined a 50% evaporation

crank angle specification to determine how quickly the injected liquid fuel evaporated in
the cylinder. This was set to 25 CAD, also borrowed from the generic GT-POWER GDI

model.

Table 3: Specifications used for direct injection system.

 

 

 

 

 

 

Injection Rate 6.5 g/s

Start of Injection Dependent on valve timing
Vapor Fraction of Fuel 0

Fuel Temp 300 K

50% Evaporation Time 25 CAD

 

 

 

2.3.2 Injection Timing

The start of injection ($01) in the GT-POWER GDI example mode was 368 CAD
ATDC. However, since the valve timings of this example model and the actual base case
model used for simulations differed, it was unclear what specifications to use for 801 in
the Chrysler-based DI model. Thus, a sensitivity study was done by varying injection
timing over a range of crank angle degrees and examining the impact on BSFC; this was
performed on DI models using both the base case cam phasing and the optimized
Atkinson-style cam phasing. $01 was varied from 360 to 420 CAD ATDC for the base
case and 360 to 450 CAD ATDC for the Atkinson case. These sweeps were performed

for compression ratios of 10.5 and 12.5.

19

The plotted results of the sweeps (Figure 7) Show that injection timing does play a
role in fuel consumption. For base case valve timing (shown in blue and green), the
difference between the maximum and minimum BSFCs over the swept region was only
about 1.9 g/kW-hr, less than a 1% difference (Table 4). Within the sweep, there is a
slight local minimum at 390 ATDC, and also a minimum at the lower limit of 360 CAD
ATDC; however, injecting near the 368 CAD ATDC timing used in the GDI example or
before 360 ATDC (which is TDC GE) is not physically practical since this would mean
injecting into only the clearance volume of the cylinder. Interestingly, the region model
appears to be the worst region for efficiency on the simulation model. It was hoped that
looking deeper into the SOI timing relative to IVO would provide some more insight as
to when the timing should be, but this proved to be unpractical. The IVO timing in the
GDI example was 356.9 CAD ATDC, meaning that SOI occurs 11.1 CAD after IVO.
Using that same 1 1.1 CAD gap from IVO to injection for the Chrysler model, SOI would
be set at 353 CAD ATDC, which is an impractical injection time Since it is before TDC
GE. Thus, the 368 CAD ATDC timing of the GDI model came to be viewed as a limit of
how early the injection timing can be advanced, but not necessarily a guideline for
optimum timing in all setups; the optimum for this case was determined to be at 390
CAD ATDC, where the local minimum appears in Figure 7. The apparent local

minimum at 390 was thus used for the DI model with base case cam phasing.

20

 

BSFC [gIkW-hr]

 

BSFC vs. Injection Timing at Varying Compression Ratios

330

 

 

 

325 w

 

L L A A 7 ~ +CFI=10.5
W +cn=12.5

320 . L WW w «r— - - —' ‘ ' ‘ ‘ “' —x—cn=10.5, Atkinson

+ CR=12.5, Atkinson

 

 

 

O)
_L
U"
{

 

 

 

 

 

 

 

 

 

310
305 W X x X X
300
295 H ,7
290 I I I I T
350 370 390 41 0 430 450 470

Start of Injection [CAD ATDC]

 

Figure 7: BSFC vs. injection timing for various DI setups.

Table 4: BSFC results for injection timing sweeps of DI models.

 

 

 

 

 

 

 

_S_Q_I BSFC
|CAD ATDC] lgflgW-hrl
Base Cam Phasing Atkinson Cam Phasing
CR=10.5 CR=12.5 CR=10.5 CR=12.5
360 321.9 313.8 302.1 294.0
370 323.2 315.3 302.4 294.1
380 322.6 314.6 304.9 296.5
390 322.1 314.0 306.6 298.4
400 322.5 314.2 306.5 298.3
410 323.2 314.9 306.1 297.9
420 323.8 315.7 305.8 297.6
430 - - 305.9 297.7
440 - - 306.1 297.8
450 - - 306.2 297.8
Range: 1.9 1.8 4.5 4.4
%diff over range: 0.59% 0.58% 1.50% 1.49%

 

 

 

 

 

 

21

 

For the Atkinson DI model, the trends are more complicated. The BSFC range
over the swept SOI region was 4.5 g/kW-hr, about a 1.5% difference. However, the
lowest BSFCS are found at the extreme advanced side of the range, at the 360 and 370
timings, as seen in the lower red and yellow curves in Figure 7. With the Atkinson-style
cam phasing, IVO does not occur until 397 CAD ATDC, and this leads to the question of
whether injecting before IVO is a realistic strategy to pursue. Though the intake valve is
not yet open, the exhaust valve, which closes late, is still open during the start of the
intake stroke; thus, the fuel is injected into hot burned gases that are being rebreathed
through the open exhaust valve. If this is a poor strategy, the next best option is to inject
at the 420 CAD ATDC timing where the slight local minimum for BSFC occurs; this
would also avoid potential problems of unburned hydrocarbon emissions and piston
wetting from injecting too early while the exhaust valve is open and the piston is near top

dead center.

2.4 PFI Lean-Burn Operation

Lean-burn versions of the PFI models, both base case cam phasing and Atkinson-
style cam phasing, were created by running them at a 1:1.4 condition. While this limit is
at the extreme of general guidelines outlined by Heywood [12], which says practical
normal engine limits are around 2:1.1 with high CR fast—buming engines around A=1.25,
this limit, chosen under the guidance of Chrysler, would give the theoretical best
efficiency gain if an engine were able to run at this level. Injection timings and valve

timings were the same for all the Lean-bum models. DI models were not considered in

22

this lean analysis, because a major part of lean—bum DI operation involves charge

stratification. This effect is hard to model in a l-D environment such as GT-POWER.

2.5 Intake Valve Throttling

The goal of intake valve throttling is to improve fuel consumption by eliminating
the throttle plate through the use of variable lift, duration, and timing of the intake valve.
This reduces flow restriction caused by a nearly closed throttle plate at low or part load.
This concept has been utilized in production engines by BMW with their Valvetronic

system [13] and more recently by Fiat with their Multi-air system.

2.5.1 Model Modifications

For Simplicity of simulation setup, variable lift was only applied to the intake
valve—the exhaust valve profiles were unchanged. ICL and ECL were still swept over a
range of timings, but instead of controlling the 2 bar load with a throttle controller, the
throttle plate was now set wide open and load was controlled through use of valve profile
multipliers, which were controlled by PID controller tuned to respond Similarly to the
response of the throttle controller in the base case model. These multipliers were applied
equally to both the valve lift and to the duration; for example, a multiplier of 0.5 would

halve both the max lift and the lift duration of the base profiles.

23

For combustion modeling, two possible approaches were investigated. While the
standard base case burn duration of 31.1 CAD was used in the optimizations of other
models such as Atkinson cam phasing and DI, experimental data provided by Chrysler
pointed to the possibility that this combustion duration was not realistic for an intake
valve throttling set-up—their data showed a much longer burn duration of between 50
and 60 CAD. Thus, BSFC optimization maps were created for each of these cases in

order to see the effect that burn duration had on the optimized valve timing and lift.

2.5.2 BSCF Map Optimization

Initial timing sweeps were run by varying ICL from 25 to 75 CAD TDC GE and
ECL from -.95 to -55 CAD TDC GE (Figure 8). From this map, the area of interest for
further refinement was determined to be in the region of ICL from 45 to 60 CAD TDC
GE and ECL from -80 to -65 CAD TDC GE . The map created using this more refined
range was used for in the DOE-Post optimizing package in the same manor described in
Section 2.2.1 for the Atkinson-style cam phasing. This process was done twice, using the
base case burn duration of 31.] CAD first, and then using a 50 CAD burn duration in the
second. Results were compared to see what effect the burn duration had on the
optimization; the differences were small—only about 1 and 3 CAD for the intake and
exhaust timing, respectively (Table 5). It was decided that the specifications for the
longer burn duration were more sensible to use considering the experimental data

available; an optimization done using unrealistic assumptions is meaningless. The Sweeps

24

for the refined 50 CAD burn duration are found in Figure 9, and the final valve timing

scheme used is Shown in Figure 10.

-55

-65

ECL
[CAD]

-75

-85

-95
25 35 45 55

ICL [CAD]

 

BSFC
[glkW-hr]
298

303
307
311
31 5
320
324
328

Figure 8: BSFC map with RGF overlay from initial valve timing sweep for intake throttling model

with burn duration of 31.1 CAD.

25

 

45 48 51 54 57 60
ICL [CAD]
Figure 9: Refined BSFC map with RGF overlay for intake valve throttling model with burn duration
of 50 CAD.

Table 5: Valve lift specifications obtained from intake valve throttling model optimization.

 

 

 

50 CAD Burn Duration 31.] CAD Burn Duration
ECL -70.5 CAD ATDC GE -71.5 CAD ATDC GE
ICL 54.7 CAD ATDC GE 51.8 CAD ATDC GE
Intake Duration 61.3 CAD 60.7 CAD
Intake Lift 2.45 mm 2.43 mm

 

 

 

 

26

Valve Profiles—Intake Valve Throttling

 

 

 

 

 

 

 

 

 

 

 

 

1O .
— Exhaust
— Intake
8
6 _.------_-M-_--a
Lift
[mm]
4 LL__~ L L L
2 t- LL- _.L___L_,-.L.L.L
0 ”1.114111“. “L141“...HILLLULHIL unit“ IIILILLLIIIIU...
0 180 360 540 720
TDCF BDC TDC GE BDC TDCF

Crank Angle [degrees]
Figure 10: Valve Profiles obtained from intake valve throttling model optimization.

2.6 Burn Curve Duration and Phasing Sweeps

It is not known how to model the burn characteristics accurately for the many
different models created. AS noted in Section 2.5, a change in engine setup, whether for a
different injection system, a different breathing system, or a different cam phasing, can
drastically affect the burn duration of the fuel-air mixture in the cylinder. However,
without a physical engine on which to test these setups, it is not possible to know exactly
how each model should be expected to burn. Thus, for each model created, a burn

duration sweep was executed. By specifying 10-90% burn durations ranging from 20

27

CAD to 70 CAD in the Wiebe function specifications in GT-POWER for all models,

approximate comparisons were made, and results were also Shown graphically.

Under normal conditions, small changes in burn duration do not have a
particularly large effect on BSFC [12], especially if durations are fairly short, around 30
CAD. It is only when the duration increases Significantly (as noted in the intake
throttling case) that it causes a major reduction in BSFC when compared to Shorter burn

durations.

Sweeps of combustion phasing were also investigated on some models in order to
see trends of how late combustion could affect fuel consumption. The CA50 anchor
timing of the Wiebe model was swept from 7 CAD ATDC (its approximate value in the
base case PFI model) to 22 CAD ATDC, and expectedly, BSFC increased. Data suggest
that 7 CAD ATDC is ideal for best efficiency, so the CA50 phasing was left alone for all

burn duration sweeps.

28

CHAPTER 3

1-D HCCI ENGINE MODELING

Developing a model for HCCI operation in GT-POWER required a somewhat
different approach to that of the development of previous spark ignition models. For
HCCI operation, the characteristics of combustion are based solely on the conditions
found in the cylinder as the air-fuel mixture undergoes compression. Thus, it is not
possible to make simple sweeps of valve timing and combustion duration to gather
trends, because there would be no way of knowing if the prescribed combustion curve
was a realistic model for that set of operation parameters. Instead, many valve timings
were investigated, each with full sweeps of possible HCCI combustion curves (Section
1.2.2); then in-cylinder conditions from each of these simulations were checked in order
to see which Simulation conditions matched with the conditions that theoretically would
induce autoignition at the appropriate time. In short, many sweeps of combustion curves
were done, and the results were then back-checked to see which ones would realistically

occur according to theoretical models.

The basic process of running the HCCI simulations consisted of four steps. First,
an NVO was Specified. Second, for each specified NVO, a Sweep of different valve lift
multipliers was set up. Third, for each NVO—multiplier combination, a set of many
combustion curves (consisting of predetermined burn duration and phasing) was inserted
into the model. Finally, to control the load for each of these NVO-multiplier-combustion
cases, the lambda ratio was adjusted via a PID controller on the fuel injector—the

leanness of the air-fuel ratio was adjusted until the load reached the 2 bar BMEP test

29

point. After these sweeps were simulated, the in-cylinder conditions were checked to see
if they matched the necessary conditions that would induce autoignition at the CAO that

had been specified for that particular case.

3.1 NVO Valve Timing

At the heart of the HCCI GT-POWER model is the NVO valve timing technique
(Section 1.2.1). To effectively model NVO, the exhaust valve specifications (lift, lash,
etc.) in the base case model were replaced with copies of the intake valve specs to create
identical profiles. The anchor points for the profiles were changed to allow Specification
of an NVO value—for the intake valve, the anchor point was changed from ICL to IVO,
while for the exhaust valve, it was changed from ECL to EVC. EVC was defined to be
equal to IVO, thus creating symmetry about TDC GB for the two profiles. NVOS used in
the simulations were 160 CAD, 170 CAD, and 180 CAD. Much experimental
investigation has been done around the 180 NVO level. Some work has been done at
levels larger than this by Urata et al. who used NVO ranges up to 230 CAD [3]; Persson
et al. studied effects from NVOS ranging from 140 to 200 CAD [5]. For this study, 160
was chosen to be the lower limit after early NVO sweeps pointed to increasingly poor

performance at shorter NVOS.

Multipliers were used on the valve lift and duration in much the same manner as

in the intake valve throttling model—a Single multiplier was applied to both lift and

duration equally to shrink the profile; these were also applied equally to the intake and

30

exhaust valves. By using these multipliers with the modified valve timing anchors
discussed above, lift and duration could be varied while keeping the specified NVO
constant. For each NVO, lift multipliers were varied in increments of 0.025, and ranged
from roughly 0.4 to 0.55; Figure 11 shows these multipliers applied to the 180 NVO
setup, and specifications for all valve timings used are found in Table 6. In this figure, it
can be seen that EVC and IV 0 stay the same for each multiplier while EVO and IVC are
varied. Unlike the intake throttling model, a PID controller was not used to adjust valve
lift duration, because the valves were no longer variables being used for load control in
the HCCI setup. Outside these ranges of NVOS and multipliers, HCCI combustion was
either not practical or could not provide good efficiency gains; at Shorter multipliers, fuel
levels were too lean to induce combustion, while at longer multipliers, intake valve
timing was such that fresh charge was lost at the start of the compression stroke, thus

causing temperature and pressure conditions to be insufficient for auto-ignition.

31

6.25

 

1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

— Exhaust
- Intake
5.00 L\
3.75 - - - L __ -
Lift
[mm]
2.50 ~~ L
1.25
0.0 111111111111 11_L1L111 1111111111111L11 111111L 111111111111
00 180 360 540 720
TDCF BDC TDC GE BDC TDCF

Crank Angle [degrees]

Figure 11: Valve lift profiles using six different valve lift multipliers for the 180 NVO setup.

32

Table 6: Valve timing specifications for all NVO-multiplier combinations tested in HCCI simulations

 

 

 

NVO Multiplier EVO EVC IVO IVC Duration
[CAD] [ATDC GE] [ATDC GE] [ATDC GE] [ATDC GE] [CAD]
180.0 0.550 -237.1 90.0 90.0 237.1 147.1
180.0 0.525 -229.1 -90.0 90.0 229.1 139.1
180.0 0.500 -221.0 -90.0 90.0 221.0 131.0
180.0 0.475 -213.0 -90.0 90.0 213.0 123.0
180.0 0.450 -204.9 -90.0 90.0 204.9 114.9
180.0 0.425 -196.9 -90.0 90.0 196.9 106.9
170.0 0.550 -232.1 -85.0 85.0 232.1 147.1
170.0 0.525 -224.1 -85.0 85.0 224.1 139.1
170.0 0.500 -216.0 -85.0 85.0 216.0 131.0
170.0 0.475 -208.0 -85.0 85.0 208.0 123.0
170.0 0.450 -199.9 -85.0 85.0 199.9 114.9
170.0 0.425 -191.9 -85.0 85.0 191.9 106.9
160.0 0.550 -227.1 -80.0 80.0 227.1 147.1
160.0 0.525 -219.1 -80.0 80.0 219.1 139.1
160.0 0.500 -211.0 -80.0 80.0 211.0 131.0
160.0 0.475 -203.0 —80.0 80.0 203.0 123.0
160.0 0.450 -194.9 -80.0 80.0 194.9 114.9

 

 

3.2 Burn Curves

The burn characteristics used in the HCCI modeling were derived from the
equations, relationships, and experimental data discussed in Section 1.2.2. A database of
combustion curves, parameterized by CAO timing was created from these relationships,
and these curves were input as Wiebe functions into GT-POWER. The Wiebe function
used in the HCCI modeling differed from the standard GT-POWER Wiebe function in
the base case model—simply specifying the same burn duration and phasing in both did
not result in identical curves because of a few key differences. The HCCI curves were
based on a burn efficiency of 95.5%; in [6], this was shown to be fairly constant with

early CAO times and variably dropping off as CAO approached TDC, but for Simulation

ease, it was assumed to be constant across the full range. Experiments in [2] showed
approximately this same burn efficiency of just over 95% at the same 2 bar and 2000
RPM test point used in these simulations. With this assumption, the CA50 timing had to
be modified in the GT—POWER specification—rather than using the 50% burn location,
the 47.75% burn was used (which comes from 50%*0.955), because GT-POWER
normalizes its Wiebe function inputs by the amount of fuel burned, not the total amount
of fuel present in the cylinder. The Wiebe exponent, used to modify the length of the end
of the burn curve, was modified in GT-POWER to match that of the correlated data found
in [6]; in GT-POWER’S standard model, this is a constant set equal to 2, but for the
experimental HCCI curves, it is varied from 2 down to 1.1 as burn CAO timing
approaches TDC and burn duration is increased. Figure 12 shows an example of the
changes made by showing a calculated HCCI Wiebe function curve, a standard GT-
POWER Wiebe function curve scaled down to 95.5% bum efficiency, and the updated
GT-POWER HCCI Wiebe function curve for the same duration and phasing
specifications; there is very close agreement between the modified GT-POWER curve
and the calculated curve, while the original GT-POWER curve shows a faster burn
initially and a slower burn toward the end. This technique worked consistently for all
burn durations tested—Figure 13 shows four sample curves used in the simulations with
very close agreement between the GT-POWER curve and the calculated Wiebe function

CUI‘VC.

34

 

HCCI Wiebe Model Modifications

 

 

   
  

 

0.9
0.8 I
I
E 0.7.
3
m 0.6 «
S
.8 0.5 <
a
L: 0.4 « . L , LL
g —Calculated Wiebe
2

e GT-POWEFI Modified
:GT-POWER Base

 

 

 

 

T I T—"7 ‘1

-5 0 5 10 15 20 25 30
CAD ATDC

 

 

 

Figure 12: Comparison of standard scaled GT-POWER burn curve, modified GT-POWER burn
curve, and calculated Wiebe function burn curve with burn duration of 13.3 CAD.

 

 

 

I-IOCI Wiebe Model Comparison

 

0.9 1
0.8 ~
0.7 ~
0.6 -‘
0.5 ~

      

 

[L—

g . GT-POWEFI10.SBDUR
l e GT-POWER13.SBDUR
. GT-POWER15.7BDUR ,
i . GT-POWEFI18.1BDUFI
— Wiebe Models of same duration»

.4

Mass Fraction Burned

 

 

 

 

 

 

-5 0 5 10 15 20 25 30
CAD ATDC

T " T fi— ’1

 

L*_ _L___—¥__J

 

 

Figure 13: Sample of four MF B curve comparisons between GT-POWER modified function and
calculated Wiebe function.

 

35

Using equation (3) and data for A6 and w from [6], a database of curves was
created. This database was parameterized by CAO timing, and gave specifications of
CA50 and 10—90 burn duration for each CAO chosen (Table 7). Figure 14 shows the
CA50 timings, the black points, and the burn durations, red points, as functions of the
CAO timing. A sweep of the burn curves constructed from these specifications was

applied to each combination of NVO and valve lift multiplier.

The general trend is that the later the start of combustion, the longer the
combustion duration. Longer duration has a negative impact on fuel consumption, so the
goal in HCCI operation is to phase CAO early enough that combustion happens quickly
and so that CA50 timing stay close to just a few CAD ATDC without being too early,

which would cause a severe knock.

36

Table 7: Specifications of HCCI burn curves used in GT-POWER simulations.

HCCI Combustion Curve Tabulated Data

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

CAO 10—90°/o Burn
[CAD CA50 Duration A6 Wiebe
ATDC] [CAD] [CAD] [CAD] exponent
-0.07 16.5 22.5 19.8 1.12
-0.30 15.6 20.9 18.9 1.19
-0.54 14.8 19.5 18.1 1.26
-0.77 14.0 18.1 17.3 1.33
-1.01 13.1 16.9 16.5 1.40
-1.24 12.2 15.7 15.7 1.47
-1.47 11.3 14.4 14.8 1.54
-1.71 10.4 13.3 14.0 1.61
-2.17 8.9 11.6 12.8 1.75
-2.63 7.9 10.5 12.0 1.89
-3.10 6.9 9.7 11.4 2.00
-3.55 5.8 9.1 10.7 2.00
95.5% Burn Characteristics
25
C
.9 I
5 I CA50 [CAD ATDC] .
g 20 I B1090 [CAD] I
e ' .
g H 15 -_ .._I _ . 1
.5 D I
E 5 I I
a "" 10 ,LL LLLl I - J
E I
‘2'
a 5 ~——— -i — ~ — ~— —— ——
E
8
0 r Y I T
-3.00 -2.50 -2.00 -1.50 -0.50 0.00
CAO [CAD ATDC]

 

Figure 14: Points used for determining the set of burn curves used in HCCI simulations.

37

 

3.3 Lambda Control for Lead

With throttle—less operation and fixed valve timing and burn characteristics, the
logical way to control the load of the engine was with a Lambda controller on the fuel
injection system; a similar experimental setup has been used in [2]. A PID controller was
developed to find the appropriate lean air-fuel ratio that would produce the target 2 bar

BMEP load.

3.4 Heat Transfer Model

The in-cylinder heat transfer model for the base case model consists of a Woschni
model that can be customized with a convection heat transfer multiplier that varies
throughout one cycle of the engine. In the base case model, the multiplier (Figure 15)
was set to higher valves during periods of high flow through the valves—there is a
moderate spike in the multiplier during the exhaust stroke and a larger plateau during
most of the intake stroke. Essentially, this is saying that convection heat transfer is
enhanced under moments of high flow, and it is enhanced even further when that flow
involves a temperature difference (such as taking in cool intake air into a hot cylinder).
Figure 15 shows a plot of multiplier (black curve) overlaid with a plot of net mass flow
rate through all valves (green curve) and the valve lift profiles (red and blue for exhaust

and intake, respectively).

38

 

0.050

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5 1 I I
-I-HT Multiplier ; i j I
—~ Intake ,‘ l| I I
4 1~ Exhaust '3 ‘, f - I 0.040
5 Net Flow i I, ’i I.
A; l j I g A I '5'
E; 3 ~ ;- r——————+‘~ ,'~~ T —4P 0.030 2
*-' ._ I I l
E ‘5 I I I '3
9 E 2 - a ' ‘ ‘ I - 0020 E
To |— I V ' I ' ' I"
> *5 l \ 2
I
L‘s-h“ i I
1 l *1— 0.010
0 . . . /. \s . f . 0.000
0 90 180 270 360 450 540 630 720
CAD

 

 

Figure 15: GT-POWER convection heat transfer multiplier for base case model, overlaid with net
mass flow rate through valves and valve lift profiles.

Due to the unique NVO valve timing setup, using this same heat transfer model
did not seem reasonable for the HCCI Simulations. In the base case model, the heat
transfer (HT) multiplier made a sudden jump at IVO from approximately 1 to a value of
2. It stayed at this value of 2 for much of the intake stroke until mass flow through the
valve tapered off. For the NVO operation, IVO is drastically later, so this spike in the
heat transfer coefficient was retarded to coincide better with the IVO timing and thus the
beginning of intake flow into the cylinder. The period during NVO (between EVC and
IVO) was set to a value of l, the default value. Figure 16 Shows these changes to the

multiplier along with the valve lifts and flows for HCCI operation.

39

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5 0.1
+Base HT
+HCCI HT

.—. A ——~ Intake
E 4 , \ L ~ 0.08
E. I/ I. ,K ——-»- Exhaust
, I I H
E g. {I I — Net Flow %
g 3 ', ; I I- 0.06 E-
; I / I 2
t: ‘ I u"
5 I I I a
E 2 « - ' 4» 0.04 g
5 a
o z
D
E 1< -— 7' - - 0.02
O I I I I T T T 0
0 90 180 270 360 450 540 630 720
CAD

 

 

 

 

Figure 16: GT-POWER convection heat transfer multiplier modified for NVO valve timing used in
the HCCI model.

The modified heat transfer coefficient was not a precisely developed model. It
was especially unclear what exactly the heat transfer multiplier should be set at during the
NVO period. During compression and combustion (while both valves are closed), the
base case model has this multiplier set to a value of 0.9; during the period just before
IVO, while the exhaust valve was still open, the base case value dropped to 1.16. It was
decided that the NVO period simulates a condition somewhat similar to both of these

conditions, so the value of 1, between those two values, was chosen as a reasonable

estimate.

This heat transfer multiplier has a significant effect on the operation of the HCCI
model. Because of the high dependence on temperature and pressure conditions in the

cylinder, an over-estimate of the convection heat transfer multiplier causes too much heat

40

to be lost from the hot residual gas left in the cylinder during NVO; likewise,
underestimating it slows heat transfer down significantly enough that the in-cylinder
conditions become hot enough to cause early knock. Furthermore, the heat lost from the
trapped gases during NVO has effects not only on autoignition timing, but also on overall
cycle efficiency—a loss of heat causes a Slight pumping loop to appear during this

recompression that can have a direct impact on fuel consumption [4].

3.5 Knock Integral Calculation

With a Simulation result database consisting of a sweep of burn curves for each
NVO-multiplier combination, it was necessary to sift through to find the cases that
theoretically could happen for each setup. This was done by using the in-cylinder
pressure and temperature traces from GT-POWER to numerically integrate the knock
ignition integral (see Equations 1 & 2). Figure 17 shows samples of such cases for a
simulation using a 180 CAD NVO with a lift multiplier of 0.475 and thus an IVC of 573
CAD ATDC, which is where the traces start. Equations 1 and 2 were adapted for
numerical application by modifying them to become equations 4 and 5 below. Equation

4 shows a simple midpoint rule used to numerically integrate the inverse of Tign using

along each crank angle degree increment. ACA was set to 1 CAD, so Pi and Tisimply

became the instantaneous temperature and pressure at each 1 degree increment of CAD

after IVC.

41

L+_1_

 

 

CAO 1 n 2'. r. 1
I —~dCAz Z—l 1+ (AC/1):] (4)
2'. . 2
IVC lgn l=1
Tl. 1.310 Pl. a x02 exp ——R*T. (5)
l

Plotting this integration process along CAD provides an easy way to see when the
integral reaches the value of 1 that signifies autoignition. Figure 18 shows examples of
two different cases of this calculation which predict different autoignition times. The
blue curve Shows the integral as calculated from the pressure and temperature data found
in Figure 17 from a 180 CAD NVO case, while the red curve shows the same calculation
using a 170 CAD NVO and a shorter 0.425 lift multiplier; in this case, it is clear that the
first ease predicts an earlier CAO timing since it crosses a value of 1 at an earlier CAD.
The cases that were considered realistic were the ones in which this predicted
autoignition time from integration matched closely with the imposed CAO of the
prescribed Wiebe function defined in GT-POWER. Generally, this meant that for each
NVO-multiplier pair, there were two imposed combustion curves that bracketed a
theoretical realistic curve—for one imposed curve, the integration calculation would
predict autoignition occurring too late, and for the next, the integration would predict it
occurring too early. The results of interest (such as BSFC, CAO, Burn Duration, CA50,
lambda, RGF, etc) were interpolated from these two bracketing curves; the interpolation
was done by calculating when the CAO trend of the GT-POWER data would cross the

CAO trend of the knock integral data.

42

 

 

Temperature and Pressure Traces after IVC
NVO=180, Lift Multiplier=0.475, BDUR=11.6

 

 

 

 

 

 

 

 

 

 

 

 

 

1600 7 - Temperature 40
1400 #—~ 0 Pressure —- —~ We: —~ 35
._. 1200 « w 30
E- 'E
.. 1000 -. .. 25 .5
5 ....
E: 800 - a— 20 g
8 8
g 600 ~ .. 15 9
,2 n.
400 .1 A» 10
200 < L 5
0 O
570 600 630 660 690 720

CAD

 

* L _ ___ 7 —_ __LL

 

 

Figure 17: Sample temperature and pressure traces used for knock ignition integral calculation.

 

Knock Integral Calculation as CAD approaches TDC

-I— 180 NVO, 0.475 Lift
+ 170 NVO, 0.425 Lift )1

,1 if
‘
(ii-LL
,O/
I"!

 

 

 

 

Integral of 1/Tau

 

 

 

CAD

 

 

 

Figure 18: Examples of knock integral calculations shown graphically vs. CAD along compression
stroke.

43

CHAPTER 4

RESULTS AND DISCUSSION

4.1 SI Results

The goal of formulating the different spark ignition models was to find
approximate percentage improvements in BSFC relative to the base case model that could

potentially arise from implementing each technology.

4.1.1 Base Case Results

The base case for this engine at this operating condition provided a BSFC figure
of 324.2 g/kW-hr with an RGF level of 27.4% using a 10-90% burn duration of 31.1
CAD. All forthcoming BSFC improvements are referenced relative to this base case

BSFC.

4.1.2 Cam Phasing Results

The Atkinson valve timing that resulted from the cam phasing optimization of
Section 2.2 resulted in a BSFC that ranged from 307.1 to 329.8 g/kW-hr over a 30 to 60
CAD 10-90% burn duration sweep (Table 8). If burn duration is assumed to stay the
same as in the base case, at around 30 CAD, this model represents about a 5%

improvement; if burn duration were to lengthen 40 CAD, the improvement would drop to

44

3.5%, and if to 50 CAD, then only 1.2%. Also worth noting is the small gap between
IV C and SOC as burn duration increases. If combustion phasing is to stay the same, an
increased burn duration requires a much earlier spark. In the case of the 60 CAD burn
duration, start of combustion was -55 CAD ATDC while IVC occurred only a few CAD
earlier at

-61 CAD ATDC.

Table 8: BSFC results for burn duration sweep of Atkinson cam phasing model.

 

 

Burn Duration RGF BSFC
[CAD] [°/o] [g/kW-hr] % Improvement
30 29.6 307.1 5.3
40 29.1 312.8 3.5
50 28.5 320.5 1 .2
60 27.8 329.8 -1 .7

 

 

 

 

BSFC vs Burn Duration, Atkinson Cam Phasing

/

 

 

+Atkinson Carn Phasing
3300 LL 4* Base Case

 

 

 

 

 

325.0 4.

 

BSFC [glkw-hr]
(13
1'0
0
O

 

310.0 /
/

305.0 e i T r r . . .
20 25 30 35 40 45 50 55 60 65

Burn Duration [CAD]
I L4

Figure 19: BSF C results for burn duration sweep of Atkinson cam phasing model.

 

 

 

 

 

 

45

It is of note in the results for the cam phasing model that as burn duration is
increased, RGF decreases. If the actual burn duration were known, the valve timing
could be re-optimized to increase the RGF towards the 30% limit that was chosen;
however, Since this limit was chosen somewhat arbitrarily, and since the base case model
had only a 27.4% RGF, the re-optimization was not performed for either this cam phasing

model nor for any of the other SI models.

4.1.3 Direct Injection Results

The results for the direct injection model consist of tests at two different
compression ratios, 10.5 and 12.5; furthermore, for each compression ratio, two injection
timings were investigated. The DI tests were also run using both the base case valve

timing and the Atkinson timing from the cam phasing model.

For the DI cases using base case cam phasing, adding DI alone yielded an
approximately 0.8% BSFC improvement when burn duration was assumed the same as in
the base case model. Increasing compression ratio to 12.5 increased this improvement to
3.3%, which corresponds to a BSFC of 313.4 g/kW-hr. Furthermore, with the possibility
that a higher compression ratio could lead to a faster burn rate, this improvement could
increase even further—with a 20 CAD burn duration, the high compression model
yielded a BSFC of 309.5 g/kW-hr, a 4.5% improvement. These figures are all based on
the optimized 390 injection timing (Section 2.3.2); with the GT-POWER GDI model 368

CAD ATDC injection timing, all BSFC improvements decreased by approximately 0.3%.

46

Full results Showing two injection timings and two compression ratios are shown in Table

9.

Implementing the DI system on the Atkinson cam phasing model yielded further
improvements. On the base 10.5 CR model, BSFC with a 30 CAD burn duration
(similar to the base case) was 301.6 g/kW-hr (for the 368 CAD ATDC injection
timing), a 7.0% improvement; even if burn duration were to increase to 50 CAD, this
model still predicted a 2.9% BSFC improvement (compared to only 1.2% without DI for
the same burn duration). Increasing CR to 12.5 improved these BSFCS to 293.7 g/kW- hr
for a 9.4% improvement with a 30 CAD burn duration and 307.6 g/kW-hr for a 5.1%
improvement with a 50 CAD burn duration. In all these calculations, it is assumed that
the injections timing is 368 CAD ATDC injection timing; the 420 CAD ATDC timing
from Section 2.3.2 increased fuel consumption by about 1% for all cases. Full results of
these sweeps are Shown in Table 10. Figure 20 Shows the fuel consumption trend of all

four DI models.

47

Table 9: BSF C results for burn duration sweeps of D1 models with base case cam phasing.

 

 

 

 

 

 

CR=10.5
SOI Burn RGF BSFC

[CAD ATDC] Duration [CAD] [°/o] [g/kW-hr] °/olmprovement
368 20 27.9 319.0 1.6
368 30 27.6 322.6 0.5
368 40 27.1 328.7 -1.4
368 50 26.6 336.6 -3.8
368 60 26.1 346.2 -6.8
368 70 25.4 357.2 -10.2
390 20 27.8 317.9 2.0
390 30 27.5 321.5 0.8
390 40 27.1 327.5 -1.0
390 50 26.6 335.4 -3.5
390 60 26.0 345.0 -6.4
390 70 25.3 355.9 -9.8

CR=12.5
SOI Burn RGF BSFC

[CAD ATDC] Duration [CAD] [°/o] [g/kW-hr] °/olmprovement
368 20 27.8 310.7 4.2
368 30 27.4 314.6 3.0
368 40 26.8 321.1 1.0
368 50 26.3 329.4 -1.6
368 60 25.7 339.3 -4.7
368 70 25.0 350.6 -8.1
390 20 27.8 309.5 4.5
390 30 27.3 313.4 3.3
390 40 26.8 319.8 1.4
390 50 26.2 328.1 -1.2
390 60 25.6 337.9 -4.2
390 70 24.9 349.2 -7.7

 

 

48

 

Table 10: BSF C results for burn duration sweeps of DI models with Atkinson cam phasing.

 

 

 

 

 

 

 

CR=10.5
SOI Burn RGF BSFC
[CAD ATDC] Duration [CAD] [°/o] [ng-hr] °/o|mprovement
368 20 33.0 298.0 8.1
368 30 32.7 301.6 7.0
368 40 32.2 307.4 5.2
368 50 31.7 314.8 2.9
368 60 31.0 323.9 0.1
368 70 30.2 334.3 -3.1
420 20 29.9 301.9 6.9
420 30 29.6 305.3 5.8
420 40 29.2 310.9 4.1
420 50 28.5 318.5 1.7
420 60 27.8 327.8 -1.1
420 70 27.0 338.5 -4.4
CR=12.5
SOI Burn RGF BSFC
[CAD ATDC] Duration [CAD] [°/o] [g/kW-hr] °/olmprovement

368 20 32.9 289.6 10.7
368 30 32.5 293.7 9.4
368 40 31.9 299.6 7.6
368 50 31 .2 307.6 5.1
368 60 30.6 317.0 2.2
368 70 29.8 327.8 -1.1
420 20 29.9 293.3 9.5
420 30 29.4 297.0 8.4
420 40 28.9 303.1 6.5
420 50 28.3 311.0 4.1
420 60 27.6 320.5 1 .1
420 70 26.8 331.4 -2.2

 

 

49

 

BSFC vs. Burn Duration for DI Models

 

+CR=10.5, 390 801
+CR=12.5, 390 801

 

 

 

 

 

 

 

 

 

 

 

 

 

 

360.0 ~~
+ CR=10.5, 420 801, Atkinson /
350.0 «4 -—X——CFI=12.5, 420 801, Atkinson
_ 0 Base Case /0//.
340.0 ”/4
37? 330.0 X
g / /://
253,. 320.0 K .
1.1.
3 310.0 -I~— ~
300.0 I /
290.0 ~~
280.0 1 i i i i .
1O 20 30 4O 50 60 70 80
Burn Duration [CAD]

 

 

 

Figure 20: BSFC results for burn duration sweeps of all DI models.

4.1.4 PFI Lean Results

The PFI lean Simulations, at lambda=l.4, were performed for both the base case
cam phasing and the Atkinson cam phasing. With burn duration assumed to stay near 30
CAD, operation at this lambda yielded an 8.1% improvement in BSFC, reducing it to
298.0 g/kW-hr; under the Atkinson phased cam timing, this dropped to 284.4 g/kW-r, a
12.3% improvement. However, running lean permits the possibility of a lengthened burn
duration. When running these models with burn duration increased to 60 CAD, the

improvements dropped to 319.8 g-kW-hr, a 1.3% improvement, for the base cam

50

phasing, and 305.7 g/kW-hr, a 5.7% improvement, for the Atkinson lean model. Full

results are seen in Table 1 l and Figure 21.

ech [gikw-hr]

 

Table 11: BSFC results for burn duration sweeps of all PFI lean models.

Base Case Cam Phasing

 

 

 

Burn RGF BSFC
Duration [CAD] 4%] [g/kW-hr] %Improvement
30 25.1 298.0 8.1
40 24.6 303.7 6.3
50 24.1 311.0 4.1
60 23.5 319.8 1.3
70 22.9 329.9 -1.7

 

 

Atkinson Cam Phasing

 

 

 

Burn RGF BSFC
Duration [CAD] [°/o] [g/kW-hr] %Improvement
30 27.1 284.4 12.3
40 26.6 290.0 10.6
50 26.1 297.1 8.3
60 25.5 305.7 5.7
70 24.8 315.5 2.7

 

 

 

BSFC vs Burn Duration, Lean Operation

 

 

 

 

 

 

 

 

—e—Lean, Base Cam Phasing
340-0 —— +Lean, Atkinson Cam Phasing
9 Base Case
330.0 4 ,
0

320.0 1
310.0 A
300.0 «
290.0 4
280.0 44 —

20 30 40 50

Burn Duration [CAD]

60 7O

 

Figure 21: BSF C results for burn duration sweeps of all PFI lean models.

51

 

 

A significant benefit of running lean is the reduction in pumping loss that comes
from a more open throttle plate—for all bum durations, the throttle plate was opened
approximately 1 degree further under lean operation than for stoichiometric operation.
Though a small difference in plate angle, at low loads and small throttle angles, this small
change has a large effect on airflow. The lean cases all showed an approximately 30%
higher rate of air flow into the engine than when base case model and Atkinson model

were run over the full sweep of burn durations, shown in Table 12.

Table 12: Comparison of air flow through the engine for stoichiometric and lean air-fuel ratio, base
case cam phasing and Atkinson cam phasing.

 

 

 

Airflow Into Engine [kg/hr]
Burn Duration Atkinson Atkinson

[CAD] Base Lean Base Lean
20 36.3 -- 34.5 --
30 36.8 47.6 34.9 45.4
40 37.4 48.5 35.5 46.3
50 38.3 49.7 36.4 47.4
60 39.4 51.1 37.4 48.8
70 40.7 52.7 38.7 50.4

 

 

 

 

4.1.5 Intake Throttling Results

The intake valve throttling model, which was optimized for running with a 50
CAD burn duration, yielded a 6.0% BSFC improvement of 304.6 g/kW-hr. If the burn
duration were longer, at 60 CAD, this would fall to a 3.2% improvement at 313.8 g-kW-

hr. Full results are seen in Table 13 and Figure 22.

52

Table 13: BSFC results for burn duration sweeps of intake valve throttling model.

 

 

 

 

 

 

Burn RGF BSFC
DurationjCAD] [°/e] [g/kW-hr] %Immvement
30 30.9 291.8 10.0
40 30.5 297.2 8.3
50 30.0 304.6 6.0
60 29.3 313.8 3.2
70 28.6 324.5 -0.1
BSFC vs. Burn Duration, Intake Valve Throttling [I

330.0

C
320.0 /

/

300.0 / + intake Valve Throttling —
/ 9 Base Case

290.0 1 i i i i

20.0 30.0 40.0 50.0 60.0 70.0 80.0
Burn Duration [CAD]

 

 

 

 

BSFC [glkW-hr]
22
O
O

 

 

 

 

 

 

 

 

 

 

 

Figure 22: BSFC results for burn duration sweeps of intake valve throttling model.

4.2 HCCI Results

The BSFC of all successful cases of the HCCI model fell between 258.5 g/kW-hr
and 268.9 g/kW-hr. The best efficiency, which represents a 19.8% improvement over
base case, was achieved with an NVO of 180 CAD and a valve lift multiplier of 0.45,
which created lift durations of 114.9 CAD. For this case, in-cylinder conditions predicted

a calculated CAO timing of 1.8 CAD BTDC, with a burn duration of 13.1 CAD (these

53

numbers are interpolated from the two cases in which the calculated knock-ignition

integral times bracketed the GT-POWER CAO times).

Figures 23 through 25 show results of BSFC, RGF, and lambda for each NVO
tested as function of the intake valve duration. The lift duration comes from the specified
valve multiplier for each case. The results from these figures are the interpolated values
that come from calculating where the crossover point falls between the imposed GT-

POWER combustion curves and the calculated knock-integral combustion curves.

By comparing the three tested NVO levels of 180, 170, and 160 CAD, it may be
seen that a longer NVO period allows for shorter valve durations and lower fuel
consumption (Figure 23). The 180 NVO tests yielded the best performance with valve
lift durations of 115 CAD, while the 170 NVO and 160 NVO cases saw their bests with
durations of 123 CAD and 131 CAD, respectively. At longer valve openings, the results
for each NVO seemed to trend toward a similar fuel consumption region, but for shorter

openings, the 180 NVO cases showed clear gains.

In looking at Figures 24 through 26, it can been seen that the longer NVO period
leads to significantly higher RGF, which in turn leads to lower lambda ratios, not just
because the mixture needs more fuel to operate, but also because the higher RGF leads to
less fresh air being present in the cylinder; thus, for the same amount of fuel, lambda falls

because of lower amounts fresh air. The combination of hot RGF and lower lambda

54

 

helps keep in-cylinder temperatures high enough for auto-ignition to occur earlier than it

can with the shorter NVO periods.

 

 

 

-hr]

BSFC [

BSFC vs Intake Valve Duration

269

 

268 1
267 <
266 <
265 I
264 <
263 «
262 1» —— —
261 ~
260
259 . . e . 4

 

 

 

 

--180 NVO
-e—170Nvo.‘
-—160Nvo?‘

   
   

I

  

 

 

100 105 110 115 120 125 130 135 140 145
Intake Valve Lift Duration [CAD]

150

 

 

Figure 23: BSFC verses intake valve lift duration for different NVO setups.

 

 

RG F°/o

RGF vs Intake Valve Duration

 

i-I-lfilwo—
6‘ -I-170 NVO
59 +160 NVO

 
 
 

 

 

 

 

 

 

 

 

57-
55-
531

51,4 _ rt/

49*

 

 

47 I T I I T fl I T I
100 105 110 115 120 125 130 135 140 145

Intake Valve Lift Duration [CAD]

 

Figure 24: Residual gas fraction verses intake valve lift duration for different NVO setups.

55

 

150

 

 

 

 

 

 

 

  
 
  

 

 

 

 

 

 

 

 

 

 

 

Lambda vs Intake Valve Duration

2.2

2.0 M

1.8 . ~ ~- ’—"—
1 \
E 1.6 4» ~—- ————- —— . —fi
.5

1.4 +180 NVO _

+170 NVO
1-2 «» +1160 N119 _
1 .0 1 r i r r r r r r
100 105 110 115 120 125 130 135 140 145 150
Intake Valve Lift Duration [CAD]

 

 

 

Figure 25: Lambda verses intake valve lift duration for different NVO setups.

 

 

  
 

 

 

 

 

 

 

 

 

 

 

 

 

Lambda vs Residual Gas Fraction
2.2
2.0 «
1.8 4
a
'u
g 1.6 —-—180Nvo
-| 1 4 -I—170 NVO
' —-— 160 NVO
1.2 .
1.0 , . r . . . .
48 50 52 54 56 58 60 62
Residual Gas Fraction [%]

 

 

 

Figure 26: Lambda verses residual gas fraction for different NVO setups.

Figure 27Figure 27 shows this trend of shorter burn durations for the 180 NVO
cases. The gray dots in this figure represent all of the HCCI cases that were simulated—

every combination of NVO, valve lift multiplier, and burn duration is shown here. The

56

red, green, and blue dots are the cases that were found to be likely to actually happen
according to the knock-ignition integral calculation—they are the points that were closest
to matching the GT-POWER imposed CAO timing with the knock-integral prediction
CAO timing. Together, these points form regions in which HCCI is achievable with this
particular engine model setup. The final interpolated results from all of these simulations

are found in Table 14.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

BSFC Theoretical BDUR Sweeps
285.0 - w— ~—«—— ~- —- «H ..._____ _- — _._.___._ - ..
i . All Simulated Points
280 0 + 180 NVO Predicted Range °
' —l— 170 NVO Predicted Range ’ ’
~I~ 160 NVO Predicted Range ’ e ’
275.0 ' :
'3 ; 1
E 2700 : ;
8 265.0 .
m
m 0 o 3
260.0 % a $
0 x o
t t 2
255.0 3 1, .
250.0 I I I
5 10 15 20 25
BDUR [CAD]

 

 

 

Figure 27: BSFC of HCCI simulations in regions in which autoignition is predicted to occur for each
NVO setup.

57

Table 14: lnterpolated results of all successful HCCI simulations.

 

 

 

 

Intake CAO Percent
NVO Duration Intake Lift [CAD BDUR BSFC Improvement
[CAD] [CAD] [mm] ATDC] [CAD] Lambda RGF [°/e] [g/kW-hr] [°/e]
180.0 147.1 5.0 -1.9 12.8 1.2 62 268.3 17.2
180.0 135.1 4.8 -2.0 12.3 1.4 61 265.1 18.2
180.0 131.0 4.5 -2.2 11.7 1.5 60 262.5 19.0
180.0 123.0 4.3 -2.2 11.7 1.6 58 260.6 19.6
180.0 114.9 4.1 -1.8 13.1 1.8 57 259.9 19.8
180.0 106.9 3.8 -1.3 15.7 1.9 55 261.5 19.4
170.0 147.1 5.0 -1.3 15.9 1.4 59 267.4 17.5
170.0 135.1 4.8 -1.4 15.1 1.5 58 264.4 18.4
170.0 131.0 4.5 -1.5 14.5 1.7 56 262.1 19.2
170.0 123.0 4.3 -1.4 15.4 1.8 55 261.6 19.3
170.0 114.9 4.1 -1.0 17.1 1.9 54 262.5 19.0
170.0 106.9 3.8 -0.6 19.0 2.0 52 264.5 18.4
160.0 147.1 5.0 -0.7 18.9 1.6 55 268.1 17.3
160.0 135.1 4.8 -0.7 18.1 1.7 54 265.4 18.1
160.0 131.0 4.5 -0.8 18.0 1.8 53 263.9 18.6
160.0 123.0 4.3 -0.6 19.1 1.9 52 264.3 18.5
160.0 114.9 4.1 -0.4 21.0 2.0 51 266.1 17.9
4.3 Heat Transfer Sensitivity

Certain cases of the HCCI model were run using different convection heat transfer

multipliers during the NVO period. The main HCCI simulations were run using a value

of 1.0 for this multiplier during the NVO period (Figure 16), but since this value directly

affects the in-cylinder conditions that lead to autoignition, a small sensitivity test was

performed. The l80 NVO setup was used with valve lift multipliers of 0.425, 0.45, and

0.475, which created valve opening durations of 106.9 CAD, l 14.90 CAD, and 123

CAD, respectively. These valve timings were run with the multiplier set to 1.1 and to

0.9, and the results were compared to the original data from this set of valve timings that

had the multiplier set at 1.0 (Figure 28, Figure 29). _

58

 

 

Burn duration with different H'I' multipliers

 

 

 
 
 
 

 

18.0 1
—0— Convection Miitiplier=0.9
—I-— Convection MJitipIier=1.0

16'0 ‘ ‘ + Convection MJltiplier=1.1

 

 

~~§\\“
12.0 \\\’.
10.0

8.0

Burn Duration [CA 3

 

 

 

106 108 110 112 114 116 118 120 122 124
IV duration

 

 

 

Figure 28: Effect of NVO convection heat transfer multiplier on burn duration in HCCI simulations
with 180 NVO.

Increasing the convection multiplier 1.1 causes the trapped residual gases to cool
more during the NVO period. The effect of this is that the cooler temperatures during
compression retard autoignition, thus increasing burn duration and increasing fuel
consumption. Lowering the multiplier to 0.9 has the opposite effect; the gases do not

lose as much of their heat, and thus autoignition is advanced and fuel consumption

reduced.

59

 

 

BSFC with varying HT multiplier

270.0

 

 

  
  

:Convection Multiplier=0.9
+Convection Multiplier=1.0
+ Convection Multiplier=1.1

 

 

 

 

 

 

 

 

 

 

 

:1 265.0 1— ~ 4 --
‘F
E ‘
<5 k
m 2600 + 4" l 44
a
H
+—
255-0 l I Y f T T I Y
106 108 110 112 114 116 118 120 122 124

IV Duration [CAD]

 

 

 

Figure 29: Effect of NVO convection heat transfer multiplier on BSF C in HCCI simulations with 180
NVO.

4.4 Higher Compression Ratio HCCI

Much work has been done on HCCI running at high compressions ratios, reaching
17: I [7] and 18:] [I I]. Higher compression ratios have potential to create autoignition
conditions at even leaner mixtures and with less overall charge in the cylinder. To
explore this potential, cases of the 180 NVO setup were run at a compression ratio of
13.5. While for the HCCI tests run with a CR of 12.5, the smallest valve multiplier used
was 0.425, the 13.5 CR allowed the valve lift and duration to be decreased more before

BSFC gains began to diminish (Figure 30).

60

The best interpolated BSFC achieved with this higher compression ratio was
255.2 g/kW-hr with an intake valve duration of 106.9 CAD (0.425 valve multiplier).
This is a 1.8% improvement over the best case 12.5 CR HCCI BSFC, which was 259.9
g/kW—hr with an intake valve duration of 114.9 CAD (0.45 valve multiplier).
Interestingly, longer valve lift durations for the higher CR model led to shorter burn
durations (Figure 31), but did not lead to better BSFC. When the intake valve duration
reached 123 CAD (0.50 multiplier), the knock integral calculations predicted autoignition
occurring earlier than any of the HCCI Wiebe function burn curves were set up for;
estimates were made by extrapolating data from the shortest duration combustion curve
simulations that were run, and these extrapolation estimates are shown in the figures as
light dashed—line extensions of the main data curves. The predicted CA50 timing of these
simulations was earlier than 5.8 CAD ATDC, and phasing combustion this early
worsened efficiency. Furthermore, the effects of EVO and IVC under longer valve
durations worsened efficiency. With the earlier EVO of the longer valve duration cases,
there is not as much work extracted during the expansion stroke, and with IVC happening
later, some mass is forced out of the cylinder before IVC at the start of the compression
stroke, leading to less total mass to be compressed (Figure 32). These two factors offset
efficiency gains and force a less lean lambda ratio, as lambda can be seen to reach a peak
around the 100 CAD intake valve duration case in Figure 33; however, this figure does
not show any obvious trend of compression ratio effects on lambda . Similar trends of
shortened burn duration, smaller amounts of trapped mass, and lower lambda ratios

leading to increased BSFC are seen when valve lift durations are decreased too much.

61

 

 

BSFC vs Intake Valve Duration, Compression Ratio Comparison

265.0

 

 

esrc [glkW-hr]
N
0')
O
O

 

-—I-—CR=12.5
—¢—CFI=13.5
— e CR=13.5 est.

 

 

 

 

 

255.0 1 . . r
80.0 90.0 100.0 110.0 120.0 130.0 140.0

Intake Valve Duration [CAD]

 

j

 

Figure 30: Effect of compression ratio on BSFC in HCCI simulations with 180 NVO.

 

 

Burn Duration vs Intake Valve Duration, Compression Ratio Comparison

 

 

 

 
    

 

 

 

 

 

 

18.0
—-l-—CR=12.5
16.0 —e-—CR=13.5 —1
'9: t - CR=13.5 est.
2.
c 14.04
.2
E
P 12.0—
E
3
m
10.0«
A
A
8.0 . . . . .
80.0 90.0 100.0 110.0 120.0 130.0 140.0

Intake Valve Duration [CAD]

 

Figure 31: Effect of compression ratio on burn duration in HCCI simulations with 180 NVO.

 

62

 

 

 

Total Trapped Mass in Cylinder [mg]

560.00

540.00 «~

520.00

500.00 .

480.00

Trapped Mass vs Intake Valve Duration, Compression mtio Comparison

 

—I—CR=12.5
—O— CR=13.5
a CR=13.5 est

 

 

 

 

 

 

 

 

 

80.0

7

90.0 100.0 110.0 120.0 130.0 140.0
Intake Valve Duration [CAD]

 

Figure 32: Effect of compression ratio and intake valve duration on amount of trapped mass per
cylinder in HCCI simulations with 180 NVO.

 

 

Lambda

~77A‘~—»» A

Lambda vs Intake Valve Duration, Compression Ratio Comparison

 

 

 

 

 

 

 

2.5
2.0 ~ m
1.5
1.0 .
—a—CR=12.5_
0.5 » —o— CR=13.5
A CR=13.5 est.

0.0 V Y 1 V j

80.0 90.0 100.0 110.0 120.0 130.0 140.0

Intake Valve Duration [CAD]

 

Figure 33: Effect of compression ratio on lambda in HCCI simulations with 180 NVO.

63

 

 

4.5 Overall Comparison

HCCI showed potential for a significantly larger fuel consumption improvement
than any of the other simulated technologies. In Figure 34, the BSFC results of all
simulated models are shown. In this figure, the solid bar represents the best BSFC
estimate produced by each model based. For all of the 81 cases, it is assumed that burn
durations were similar to that of the base case; the shaded region above the solid bar
represents the range that BSFC could fall into depending on how the changes to the
model could affect (lengthen) the burn duration. For the HCCI model, the shaded region
is the full range of all of the NVO-multiplier combinations simulated; the results of these
simulations show that HCCI has potential to reach a BSFC in the neighborhood of 260 to
270 g/kW-hr, far less than the SI models. This is an improvement of between 15% and
20% (Figure 35), which is in agreement with expected results found in other studies of
HCCI (see Section 1.2). Figure 35 shows the percent improvement results for all
simulated cases. In this figure, the solid bars represent the worst case minimum
improvement using lengthened burn duration assumptions, while the shaded regions
show how far this improvement could be stretched with shorter burn durations more in

line with the base case model. All of these results are summarized in Table 15.

64

Overall BSFC Comparison

 

 

 

 

 

 

 

 

Figure 34: BSFC results of all GT-POWER models.

 

 

 

 

 

 

360 T8 EstimatedRange
340 «H '-:1:-:=:=:1:3. —-:- _ —i — _—J I BeSt case
I; 3 2 O A ...............
E1 280 - I
. E 260 I
‘ g 240 ‘1‘ I
200 ,
o ‘o ‘0 0 <0 0 ‘o o \
or? s a a" s of” a“) 5» . 90 0°
\ \ ‘3‘ V 0 xi. .90 ‘2‘
a?” Q 0 Y5 V5 «‘0‘ ‘5“ 6*
<55 0\ o O\ 0
a a
e V

 

 

 

 

Table 15: Best case BSFC results and percent improvement over base case for all GT-POWER
models created.

 

 

 

 

Best BSFC Best °/o
[g/kW-hr] Improvement

Base Case 323.7

DI, CR=10.5 321.5 0.8

DI, CR=12.5 309.5 4.5
Atkinson Cam Phasing 307.1 5.3
DI Atkinson, CR=10.5 301.6 7.0
Lean 298.0 8.1

Intake Throttling 297.2 8.3
DI Atkinson, CR=12.5 289.6 10.7
Lean Atkinson 284.4 12.3
HCCI 259.9 19.8

 

 

65

 

Improvement [%]

25

Overall Percent Improvement Over Base Case

 

201

_ -—,_ _ __ 7.. v—‘

I a Estimated Range
r—LIMinimum

__fi ___ r__——__—4

 

 

 

 

 

 

 

 

 

 

Figure 35: Percent improvement in BSFC over the base case for all GT-POWER models.

66

 

CHAPTER 5

SUMMARY AND CONCLUSIONS

Simulations have been performed to compare the potential fuel economy gains of
HCCI operation relative to the expected performances of numerous SI operating
techniques. Starting with a base case production-spec PFI 2.4L 4-cylinder gasoline
engine model on GT-POWER, modifications were made to create models utilizing
Atkinson cam phasing, direct injection, lean combustion, and intake valve throttling.
Further modifications were made to create an HCCI model by implementing a negative
valve overlap scheme and using in-cylinder pressure and temperature conditions to
predict autoignition and combustion characteristics using a knock integral ignition delay
calculation and existing experimental correlations. All simulations were run at the same

2-bar BMEP load at 2000 RPM.

For the SI models, the major unknown factor in modeling was the burn duration
of the combustion event. Thus, for each model variant, a sweep of burn duration was
performed in order to create an expected range of BSFC. For the HCCI model, a test grid
of valve timings was set up using three different levels of NVO (160, 170, and 180
CAD), each with a sweep of valve lift multipliers that created different valve lifts and
durations. For each of these valve timings, simulations were run using a sweep of
imposed combustion curves. The data generated from these sweeps was analyzed using a

knock integral calculation in order to match the combustion characteristics that would

67

realistically take place. The results of these HCCI simulations was a range of expected

BSFC that could be compared to the results of the SI simulations.

In the SI simulations, results showed decreases in BSFC as model complexity was
increased. Adding a simple DI system improved BSFC by a small amount (0.8%);
combining DI with a higher compression ratio yielded a further improvement (4.5%);
combining DI and a high CR with an Atkinson-cycle cam phasing scheme, which by
itself yielded moderate improvement (5.3%), yielded further improvements (10.7%).
Eliminating the throttle plate and using intake valve throttling yielded significant
improvements (8.3%) even when used by itself as the only modification. Lean Operation
alone showed a similar improvement (8.1%), while combining lean operation with the
Atkinson-cycle cam phasing yielded a SI model best improvement of 12.3%. All of these
improvements are based on a best case burn duration, and would be expected to drop if

burn duration for a particular model were to increase.

The HCCI simulations showed reductions in BSFC or between 15% and 20%.
This was better than the improvements shown by all of the other SI models, and also was

in agreement with results of numerous experimental HCCI studies.

A conclusion that can be drawn from this study is that a simple Wiebe combustion
model can be effectively implemented in a l-D code to show feasibility of HCCI engine
operation. The HCCI simulations used modeling techniques that are computationally

inexpensive compared to other HCCI modeling attempts that use complicated chemical

68

kinetics to predict and model combustion characteristics. While more complex models
such as these can give more detailed results about what is going on during an HCCI
cycle, approximate results from simpler prescribed combustion models can still provide
useful information for showing HCCI feasibility, and they can provide a method of
performing quick and effective parameter studies in order to target HCCI operation

schemes that hold the greatest potential for improving fuel consumption.

69

REFERENCES

1.) Adomeit, P., Sehr, A., Weinowski, R., Stapf, K., Seebach, D., Pischinger, S.,
Hoffmann, K., Abel, D., Fricke, F., Kleeberg, H., Tomazic, D., “Operation Strategies for
Controlled Auto Ignition Gasoline Engines”, 2009-01-0300, Society of Automotive
Engineers Technical Paper Series, 2009.

2.) Dahl, D., Denbratt, I., Koopmans, L., “An Evaluation of Different Combustion
Strategies for SI Engines in a Multi—Mode Combustion Engine”, 2008-01-0426, Society
of Automotive Engineers Technical Paper Series, 2008.

3.) Urata, Y., Awasaka, M., Takanashi, J., Kakinuma, T., Hakozaki, T., Umemoto,
A., “A Study of Gasoline-Fuelled HCCI Engine Equipped with an Electromagnetic Valve

Train”, 2004-01-1898, Society of Automotive Engineers Technical Paper Series, 2004.

4.) Zhao, H., Li, J., Ma, T., Ladommatos, N., “Performance and Analysis of a 4-
Stroke Multi-Cylinder Gasoline Engine with CAI Combustion”, 2002-01-0420, Society
of Automotive Engineers Technical Paper Series, 2002.

5.) Persson, H., Agrell, M., Olsson, J., Johansson, B., Strom, H., “The Effect of
Intake Temperature on HCCI Operation Using Negative Valve Overlap”, 2004-01-0944,
Society of Automotive Engineers Technical Paper Series, 2004.

6.) Chang, K., Babajimopoulos, A., Lavoie, G., Filipi, Z., Assanis, D., “Analysis of
Load and Speed Transitions in an HCCI Engine Using l-D Cycle Simulation and
Thermal Networks”, 2006-01-1087, Society of Automotive Engineers Technical Paper
Series, 2006.

7.) Babajimopoulis, A., Lavoie, G., Assanis, D., “Modeling HCCI Combustion With
High Levels of Residual Gas Fraction—A Comparison of Two VVA Strategies”, 2003-
01-3220, Society of Automotive Engineers Technical Paper Series, 2003.

8.) Law, D., Kemp, D., Allen, J., Kirkpatrick, G., Copland, T., “Controlled
Combustion in an IC-Engine with a Fully Variable Valve Train”, 2000-01-0251, Society
of Automotive Engineers Technical Paper Series, 2000.

9.) Koopmans, L., Denbratt, I., “A Four Stroke Camless Engine, Operated in
Homogeneous Charge Compression Ignition Mode with Commercial Gasoline”, 2001-
01-3610, Society of Automotive Engineers Technical Paper Series, 2001.

10.) He, X., Donovan, M., Zigler, B., Palmer, T., Walton, S., Wooldridge, M., Atreya,
A., “An experimental and modeling study of iso-octane ignition delay times under

homogeneous charge compression ignition conditions”, Combustion and Flame,Vol. 142,
pp. 266-275, 2005.

70

I l.) Olsson, J., Tunestal, P., Johansson, B., “Closed-Loop Control of an HCCI
Engine”, 2001-01-1031, Society of Automotive Engineers Technical Paper Series, 2001.

12.) Heywood, J., Internal Combustion Engine Fundamentals, McGraw-Hill, 1988.

13.) Flierl, R., Kluting, M., “The Third Generation of Valvetrains—New Fully
Variable Valvetrains for Throttle-Free Load Control’, 2000-01-1227, Society of
Automotive Engineers Technical Paper Series, 2000.

71

l
l
I
l
l
l
l
l
5
I
Ill
l
l
l
l
l

IIIIIIIIII

718