3;: w

's» ”why

 

 

. a.
' if».

 

= 2;?
i211

VHE‘BB

/ .
9081/
J" ‘7 .N' 77<H

This is to certify that the
thesis entitled

MEASUREMENTS OF CYCLE-TO-CYCLE VARIABILITY OF
FUEL INJECTORS

presented by
Joshua C. Bedford

has been accepted towards fulfillment
of the requirements for the

Masters of Science degree in Mechanical Engineering

 

Major Professor’s Signature
7 /2o /0 ‘7’

Date

 

 

MSU is an Affirmative Action/Equal Opportunity Institution

 

LIBRARY

Michigan State
University

 

 

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJClRC/DateDue.p65-p.15

—_——E *7 —' —--

MEASUREMENTS OF CYCLE-TO-CYCLE
VARIABILITY OF FUEL INJECTORS

By

Joshua C. Bedford

A THESIS

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

MASTER OF SCIENCE

Department of Mechanical Engineering

2004

MEASCR

Ti
cycle var
design.

emission

Engine
\‘anabil
nltIOge]
velocitj
injecto
6H5, dc
this pr
fluid n
measu
i”J'E‘cte

flex-131

ABSTRACT
MEASUREMENTS OF CYCLE-TO-CYCLE VARIABILITY OF FUEL INJECTORS
By

Joshua C. Bedford

The goal of this project was to develop a technique for measuring the cycle-to-
cycle variability of fuel injectors. This method can then be used to improve injector
design. More consistent and precise fuel injectors have the potential to improve
emissions, fuel efficiency, and engine performance.

The experiments for this study were conducted at the Michigan State University
Engine Research Laboratory on a test setup specifically designed to evaluate the
variability of fuel injectors. The setup consists of a vessel pressurized by compressed
nitrogen, a Dantec laser Doppler anemometry (LDA) system that measures the centerline
velocity of fuel through a quartz tube, and a Cosworth IC 5460, which controls the
injector. The detector on the LDA system is capable of resolving Doppler bursts at up to
6us, depending on the level of seeding, thus giving a detailed time/velocity profile. From
this profile, the mass injected in each injection event was calculated using appropriate
fluid mechanics equations. These calculated values were compared with cycle-averaged
measurements to validate the accuracy of this technique. Finally, profiles of the mass
injected per cycle have been generated and the variability calculated in terms of standard

deviation, coefficient of variation, etc.

 

1 311‘
research. ‘
Schock for
his continu;
the unstead
would like
Stuecken 1
experiment
assistance;
Sasak for
members h
Yuan Shen
engines frc
l‘d like to
Prayers. su

ACKNOWLEDGMENTS

I am truly grateful to the people who have helped me through my schooling and
research. I would like to extend my sincere thanks and appreciation to Dr. Harold
Schock for giving me the opportunity to work at the Engine Research Laboratory, and for
his continual guidance and support. Thanks to Dr. Giles Brereton for his assistance with
the unsteady flow calculations and to Dr. Farhad Jaberi for being on my committee. I
would like to thank Ed Timm for teaching me how to use the LDA system, Tom
Stuecken for his creative input and for fabricating the mounts required for the
experimental setup, and Andy Fedewa for help at the lab. I genuinely appreciate the
assistance and fi'iendship they have provided throughout my time here. I also thank Andy
Sasak for his help in writing the calculation program. Thanks to all the remaining
members here at the lab including Jan Chappell, Mulyanto Poorte for his editing skills,
Yuan Shen, and Boon-Keat Chui. Thanks to Jack Tennis for encouraging my interest in
engines from a young age and continually answering my automotive questions. Finally,
I’d like to thank my entire family, particularly my parents Jon and Angie, for your
prayers, support, and encouragement throughout the years. I would have never made it to

where I am today without you.

iii

 

List of Table
List of Figu
List of 831T
CHAPTER
lntroductic
ll

ii

If

in

l.
CHAPTL
Experimi

b.) Ex) I -J

CHAP
Expen

TABLE OF CONTENTS

List of Tables .......................................................................... vi

List of Figures ......................................................................... vii

List of Symbols and Abbreviations ................................................. ix

CHAPTER 1

Introduction ........................................................................... 1
l . 1 Motivation ........................................................... 1
1.2 Fuel Injection History ............................................. 2
1.3 Gasoline Port Fuel Injector Design .............................. 3
1.4 Previous Work ...................................................... 5
1.5 Calculating the Mass Flow Rate ................................. 11

CHAPTER 2

Experimental Equipment ............................................................ 14
2.1 Dantec Fiber LDA Measurement System ....................... 14
2.2 Injectors .............................................................. 16
2.3 Injector Control System ........................................... 16
2.4 Fuel Delivery System .............................................. 17
2.5 Mounting Hardware and Alignment ............................. 18
2.6 Oscilloscope ......................................................... 19
2.7 Seed and Dispersal Devices ....................................... 20
2.8 Mass Balance ......................................................... 20
2.9 Delay Box ........................................................... 21
2.10 Vacuum Pump ...................................................... 21
2.11 Software .............................................................. 21

CHAPTER 3

Experimental Procedure .............................................................. 23
3.1 Complete Test Rig Setup and General Procedure .............. 23
3.2 LDA Technique .................................................... 25
3.3 Post-processing ....................................................... 28

iv

 

CHAPTER .
Results and
4.1

CHAPTER
Conclusions

C HAPTER
Recommem

APPENDIC

App
Deta

App
Calc

App
EVol

REFEREN(

CHAPTER 4

Results and Discussion ......................................... _ ..................... 30
4.1 Results ................................................................ 30
4.1.1 Injector #1 Results ......................................... 30
4.1.2 Injector #2 Results ......................................... 33
4.2 Discussion ........................................................... 36
4.2.1 Comparison of Injectors #1 and #2 ...................... 36
4.2.2 Sources of Variability ..................................... 39
4.2.2.1 Cosworth Variability ........................... 39
4.2.2.2 Sources of Error ............................... 41
CHAPTER 5
Conclusions ............................................................................ 44
CHAPTER 6
Recommendations .................................................................... 47
APPENDICES ........................................................................ 49
Appendix A
Details of the Mass Flow Equation ........................................ 50
Appendix B
Calculation of the LDA Probe Volume ................................... 52
Appendix C
Evolution of Experimental Technique .................................... 56
REFERENCES ........................................................................ 66

 

Table 1: Pot
Table 2: Fm
Table 3: lnj
Table 4: C c
Table 5.: Pr.

Table 6: Rt

LIST OF TABLES

Table 1: Potential gains for individual cylinder fuel control .................... 7

Table 2: Fuel/Air ratio range ........................................................ 39
Table 3: Injection control system variability statistics .......................... 40
Table 4: Comparison between mass balance and LDA measurements ....... 43
Table 5: Probe volume variables .................................................... 54
Table 6: Results of probe volume calculations ................................... 55

Images in this thesis are presented in color.

vi

 

   
   

Figure 1: Th
Figure 2: Eu
Figure 3: Is
Figure 4: Is
Figure 5: Pre
Figure 6: Th

Figure 7: Sc

Figure 8: Re
Figure 9: Gr;
Figure 10: D
Figure 11: D
Figure 12: n
FiLlure 13: Cr
Flsure 14; FL
Ftame 15; M
Figure 16: Hi
Figure 17: A.‘
Figure 18:31‘

FlgUre 19 E\|

 

“gm" 20: Ph

l

LIST OF FIGURES

Figure l: The Rochester fuel injection system for the ’57 Corvette ........... 2
Figure 2: Fuel injector diagram ..................................................... 4
Figure 3: Ismailov’s test setup ...................................................... 6
Figure 4: Ismailov’s velocity and mass flow rate profiles ...................... 6
Figure 5: Presence Probability Image (PPI) for a set of 30 injections. . . . .....9
Figure 6: Three different spray patterns overlaid on the PPI ................... 9
Figure 7: Schematic of ambient test performed by Delphi for

determining concentration and droplet sizing ....................... 10
Figure 8: Results of measuring concentration and droplet sizing .............. 10
Figure 9: Graph of the unsteadiness of a fuel injection event .................. 12
Figure 10: Dantec ion laser .......................................................... 14
Figure 11: Dantec transmitting (top) and receiving .............................. 15
Figure 12: Fuel injectors used ....................................................... 16
Figure 13: Cosworth injector control module .................................... 17
Figure 14: Fuel delivery system .................................................... 18
Figure 15: Mounting setup .......................................................... 19
Figure 16: Hewlett-Packard Infinium oscilloscope .............................. 19
Figure 17: AND GX-4000 mass balance .......................................... 20
Figure 18: Stanford Research Systems delay box ................................ 21
Figure 19: Experimental setup ...................................................... 23
Figure 20: Photograph of experimental setup .................................... 25
Figure 21: .LDA schematic (shown in back-scatter configuration) ............ 26

vii

 

Figure 22: (j
\

Figure 23: Tl
Figure 24: .\1

Figure 25: C
1

Figure 26: Tl
Figure 27: M
Figure 28: A
Figure 29: Pl
Figure 30: Pl
Figure 31: V
Figure 32: V
I:igure 33: P;
Figure 34: p
Fame 35; 3.
Figure 36: A
Figure 37; V
Figure 38: W
Figure 39: BL
Figufe 40: V‘

Flgul-e 4]: \.K
81

Figure 42: Vt

Figure 43: Ch I

 

Figure 22:

Figure 23:
Figure 24:

Figure 25:

Figure 26:
Figure 27:
Figure 28:
Figure 29:
Figure 30:
Figure 31:
Figure 32:
Figure 33:
Figure 34:
Figure 35:
Figure 36:
Figure 37:
Figure 38:
Figure 39:
Figure 40:

Figure 41:

Figure 42:

Figure 43:

Cycle-averaged centerline velocity plot

with injector voltage ................................................... 31
Three consecutive cycles from Injector #1 .......................... 32
Mass injected chart for Injector #1 .................................... 33

Cycle averaged centerline velocity plot with

injector voltage ......................................................... 34
Three consecutive cycles from Injector #2 ........................... 35
Mass injected chart for Injector #2 .................................... 36
Average centerline velocities along with applied voltage ......... 37
PDF of mass injected for Injector #1 ................................. 38
PDF of mass injected by Injector #2 ................................. 38
Voltage plot of the injection control system ........................ 4O
Velocity plots for Injector #1 with and without air ................. 45
Probe volume diagram and intensity distribution ................... 52

Probe volume dimensions .............................................. 53
Beam refraction sketch and values .................................... 54
Average plot of early experiments .................................... 56
Velocity profiles for a 15-inch copper fuel delivery line .......... 58
Velocity profiles for a 20-inch rubber fuel delivery line ........... 58
Bubble sizes used to seed the flow .................................... 59
Velocity profiles taken through time with seed and water. . . . . 60
Velocity profiles taken through time with water

and bubbles ............................................................. 61
Velocity plots for Injector #1 with and without air .................. 63
Cut-away of Injector #2 ................................................ 64

viii

 

deg

DI
ECU

HC
HCCI

b

Hz

1C FC
IMEP
kHz

LDA

mg
mm
ms
mV
MHZ
mW

NOX
0C
PDA
Pdf
PP]
psi

SAE
SO]

Sy1_nbol

Avg

CCD
cm
CO

deg

DI
ECU

HC
HCCI

Hz

ICFC
IMEP
kHz
LDA

mg

ms
mV
MHz
mW

NOx
OC
PDA
Pdf
PPI

psi
ROSA

SAE
SOI

LIST OF SYMBOLS AND ABBREVIATIONS

Description

cross-sectional area

average

wave speed

charged-coupled device

centimeter

carbon dioxide

decibel

elastic modulus, expander ratio

degree

diameter of laser

direct injection

engine control unit

focal Length

wall thickness

hydrocarbons

homogeneous charge compression ignition

mercury

hertz

incidence

individual cylinder fuel control

indicated mean effective pressure

kilohertz

laser Doppler anemometry

mass flow rate

milligrams

millimeters

milliseconds

millivolts

megahertz

megawatts

nitrogen

nitrous oxides

optical concentration

particle dynamics analyzer

probability density function

presence probability image

pounds per square inch

radius

rapidly operating electromagnetic secondary
actuator

Society of Automotive Engineers

start of injection

ix

 

212“”

V ellipse

N‘DQDNAJQo

 

 

<: 52mm"
,2...
G

Vellipse

CEQ‘DQNNQ:

time, transmitted
second

velocity

average velocity
centerline velocity
weighting fimction
volts

probe volume

incremental value
wavelength

3. 141 59

angle

density
non-dimensional time
micrometer

kinematic viscosity

 

1.1 M
As
engines. \
importanc
fuel flow.
focuses 0
system. I
Olliput V0
measured.
to test inj.
and will 1
Potential

cOI'III'OI ()y

1.2 Fu

Fuc
its dEVElop
engines. 7
Pressure. 8”

CHAPTER 1

INTRODUCTION

1.1 Motivation

As engineers strive to improve the performance of modern internal combustion
engines, well-designed engine control systems and their components become of greater
importance. The more accurately the engine control system measures air flow and meters
fuel flow, the more control engineers have over the output of the engine. This project
focuses on measuring the cycle-to-cycle variability of a fuel injector and its control
system. If a stable injection control system is developed, and there is no variation in the
output voltage, current, and duration, the variation of the fuel injector alone can then be
measured. The technique developed during this study will allow injector manufacturers
to test injectors thoroughly, viewing real-time performance on an individual cycle basis,
and will assist in improving injector designs. Improved injection consistency has the
potential to reduce cycle-to-cycle variability in combustion quality, thus improving

control over fuel economy and emissions.

1.2 Fuel Injection History

Fuel injection technology has made great advances since Robert Bosch pioneered
its development in the 1920’s. Bosch originally developed fiiel injectors for use in diesel
engines. These early fuel injection systems were entirely mechanical, in that timing,
pressure, and droplet distribution were all controlled by mechanical means. The only

inputs to this early control system were throttle position and engine speed. During World

War 11, Bosch developed a fuel injection system for German airplanes. The injectors
were spring-loaded open full-time and oscillated rapidly to maximize atomization. With
such crude parts and primitive controls, early fuel injection systems were essentially
“controlled leaks” [1].

After the war, most aircraft industries did not continue fuel injection research, but
instead concentrated their efforts on the development of turbine engines. Automotive
manufacturers were content to make minor developments to the inexpensive carburetor,
so advances in fire] injection were temporarily on hold [2]. In 1949, a young American
hot-rodder named Stuart Hilbom re-ignited interest in this field when he developed a fuel
injection system for his race car. Shortly after, Mercedes sold fuel-injected models in the
early 1950’s, but with little success. Chevrolet worked with carburetor manufacturer

Rochester to develop a fuel injection system called the Rochester Ramjet (pictured below

 

 

 

 

Enrichment
Coasting Shutofl‘ Fuel Control Diaphragn Diaphragm
D‘T’Km mam“ PW“ “Tum” Fast Idle Cam Electric Choke
Maximum Idle Adjustment g
\ Overrun Vacuum Line ’ U r Enrichment Vacuum
\. I \ Air
l . .. -/~
/ “a: 4.1-writ: ~- «9‘ ~
iIII g r” E; . i. ‘ u
i W ~ 4. Li] _ J; ”in I A— :‘y
Fuel ‘EL‘J‘ y Air Cleaner
Supply rffi - . I a . -
i9) ‘ lll'lk
Vi. .
Fuel Bowl Fuel Lines

      
  
   
 

  

High Pressure Fuel Pump

Intake Manifold \I‘U/

\

Throttle

Fuel valve control valve

   

 

Figure 1: The Rochester fuel injection system for the ’57 Corvette [3].

from article by Woron [3]) for the ’57 Corvette and the Pontiac Bonneville. Even with
these advances, fuel injection was not widely accepted until the 1970’s when emissions
and fuel economy concerns coincided with advances in electronics technology to make
fuel injection more desirable and affordable.

Early systems employed a throttle-body design in which one or two injectors
replaced the carburetor to meter fuel. With the development and advancement of
microprocessors, multi-port and sequential multi-port fuel injection systems were
developed. Currently, all new cars and trucks produced and sold in the US. are fire]-
injected.

Modem injection systems are highly developed feedback control systems that take
into account several parameters, including engine speed, load, throttle position, mass air
flow, oxygen concentration in exhaust, coolant temperature, manifold pressure, etc. Fuel
injection performance is better today than ever before, but there is clearly room for

improvement in the area of cycle-to-cycle variability in the fuel delivery systems.

1.3 Gasoline Port Fuel Injector Design

In order to understand possible sources of fuel injection variability, it is important
to have an understanding of the basic components of a typical fuel injector and how it
fiinctions. In this study, gasoline port fiiel injectors were used to perform the
experiments. Figure 2 on the following page shows a cutaway diagram of the
components that make up a standard gasoline electronic fuel injector for a port injection
configuration. A fuel injector is essentially an electronically controlled valve. When the

injector is energized by the engine control unit (ECU), an electromagnet (solenoid)

 

moves 3 p111
to reveal a
injector. the
opened [4].
combustion
and fuel inj
little as a ft
high speed.

stroke in 0.2

 

Flier

pressure 0

 

flUCIUanOnS

 

moves a plunger that is connected to the pintle. This plunger only moves about 0.15 mm
to reveal a calibrated annular passage. Because a pressurized fuel line supplies the
injector, the fuel travels through this passage and sprays out the nozzle when the pintle is
opened [4]. The nozzle is designed to atomize the fuel for improved mixing and
combustion. When the injector is no longer energized, the return spring closes the pintle
and fuel injection ceases. For solenoid-actuated injectors, this entire process may take as
little as a few milliseconds for multiple injection systems or for engines running at very
high speed. Piezoelectric injectors, on the other hand are capable of completing one

stroke in 0.2 ms [5].

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. . [[3 Pressurized fiael supply
3 .u
Plug- Supply from EC_U Filter Screen I
Solenoid windings
Return spring_
_ Valve needle
Pintle

 

 

 

Figure 2: Fuel injector diagram [3].

There are several potential sources of variability in a fuel injection system. The
pressure of the fuel supplied should be constant to reduce variability. Pressure

fluctuations can be caused by a fuel pump or by pressure waves propagating in the fuel

 

lines as a
cause wil
this prOJ
Voltage :
pulse wit
not to re

keep the:

1.4 P
R
variabilit
Complete
Laborato;
mfiaSUre‘n
research i
Six injecn'
System_
EvaluarjOI

I’LIECtion g

lines as a result of injection events [6]. Air in the fuel lines or in the injector itself will
cause wild oscillations that contribute to variability, as was discovered over the course of
this project. Dirty or partially clogged injectors certainly contribute to variability.
Voltage and current supplied to the solenoid must be consistent as they directly control
pulse width, or the duration of the injection event. Although the focus of this project is
not to redesign fuel injectors, but to measure and quantify variability, it is important to

keep these sources of variability in mind.

1.4 Previous Work

Relatively little research has been reported in the area of measuring cycle-to-cycle
variability of the total mass injected by fuel injectors. Previous work in this field was
completed by Dr. Murad Ismailov at the Michigan State University Engine Research
Laboratory in 2003. He studied high-pressure (up to 30,000 psi) diesel injectors using
measurement techniques like those discussed in this report. The focus of Ismailov’s
research was to develop an injector control system that was capable of delivering up to
six injections per cycle. Ismailov also calculated the mass injected by this multiple-burst
system. His work is published in two SAE papers. In his paper titled “Performance
Evaluation of a Multi—Burst Rapidly Operating Secondary Actuator applied to Diesel
Injection System” [7], Ismailov mentions that cycle-to-cycle variation is observed. He
attributes this largely to the cyclic pressure deviation in the common rail fuel delivery
system.

In his next paper, “Quantification of Instantaneous Diesel Flow Rates in Flow

Generated By 3 Stable and Controllable Multiple Injection System (ROSA)” [8],

Ismailov discusses the measurement technique in greater detail and includes centerline
velocity plots. A diagram of the setup is shown in Figure 3, and graphs of centerline

velocity and mass flow rates are shown in Figure 4.

 
   
  

Waveform ”Signal
Generator ocesscr
Injector Dauicfitor
Driver
Photo receiving
_ optics 7.
Transparent tube in i . 4
intersection ‘ we"

-.4
fl. -
If

Transnnttmg t. w
Optics

Figure 3: Ismailov’s test setup [8].

 

 

 

 

 

 

 

 

 

 

 

80 r r I
_ A A p=62 bar, unseeded
E E 40 V!
E E
S) E": 0 ‘ 'r r. , AA
7 bar, seeded p= 7 bar. seed:
_40 l l
0 60 120 180 240 300 360 0 60 120 180 240 300 360
Angular Phase [deg] Angular Phase [deg]

Figure 4: Ismailov’s velocity and mass flow rate profiles [8].
From the paper, it is unclear what equations were used to accomplish this, but it is

clear that the centerline velocity trace was involved in the calculations. Nevertheless,

Ismailov’s work was instrumental in that it provided the basis of the experimental
procedure used in this study.

While no other published reports were found where velocity was measured before
the injector, other papers measured variability in fuel injectors. Delphi Corporation has
shown interest in this topic for several years. A 1999 SAE paper written by Kainz and
Smith titled “Individual Cylinder Fuel Control with a Switching Oxygen Sensor” [10]
discusses controlling the fuel/air ratio by closely monitoring the oxygen content in the
exhaust for each cylinder. They also developed an adaptive control algorithm called
Individual Cylinder Fuel Control (ICF C) that precisely controls the fuel delivered to each
cylinder based on current inputs such as load, throttle position, etc. and also the oxygen
sensor history for that particular cylinder. By modeling this control system, Delphi was
able to predict several significant engine performance gains. These gains are listed in the

table on the next page. The emissions were reported to be lower, more consistent, and

Table 1: Potential gains for individual cylinder fuel control [10].

 

 

 

 

 

 

 

 

 

 

 

 

HO CO Max Torque IMEP
(”'4")
Cylinder
35:! max 6.0 49 70 60 5.6%
m
Batmcod
"m 5.5 29 14 61.21 2.5%
Cyli (+20%)
Emissions (ppm) Fuel Drive ability
Economy

 

mor
eco
cyl'
ope

inc

lnj
Ti

an

Cy
PF
1h:
C (
ell

fo‘.

more repeatable. Variability in IMEP was also lowered using this technique. Fuel
economy increased because the lean limit was extended. Typically, the leanest-running
cylinder determines the lean drivability limit. For an engine with ICFC, each cylinder is
operated closer to the lean limit. This resulted in improved idle quality as well as
increased torque and improved engine efficiency. While this study did not focus on
improving the performance of the fuel injector itself, it shows the benefit of a fuel
injection system operating with greater control and precision.

Another interesting study is described in an SAE paper titled “Application of an
Imaging—based Diagnostic Technique to Quantify the Fuel Spray Variations in a Direct-
Injection Spark-Ignition Engine,” written by Hung, Chmiel, and Markle (Delphi) [1]].
This paper focuses on capturing the magnitude of pulse-to-pulse variability in penetration
and spray geometry. Using a high-resolution grayscale CCD digital camera and triggering
it by pulsing a laser, they were able to generate pictures of the spray distribution for each
cycle. The images were then post-processed with OptimasR image analysis software to
produce a Presence Probability Image (PPI) like the one shown in Figure 5. Images like
this give important insight into how consistently the fuel and air are likely to mix.
Consistent mixing is vital if one is concerned with the cycle-to-cycle performance of an
engine, particularly for homogenously charged compression ignition (HCCI). The

following figures illustrate the results of this research effort.

100% 50% 0%

Probability Scale

   

 

Total of 30 Images

Figure 5: Presence Probability Image (PPI) for a set of 30 injections [11].

 

100% 50% 0%
Probability Scale
Image #1 Image #2 Image #3

Figure 6: Three different spray patterns (black outline) overlaid on the PPI (color) [1 1].
There are several publications in the field of droplet sizing of the spray from fuel
injectors. Lefebvre discusses several measurement techniques (mechanical, electrical,
and optical methods) in his book Atomization and Sprays [12]. Another noteworthy
publication is an SAE paper by Hung et al., “A Novel Transient Drop Sizing Technique
for Investigating the Role of Gasoline Injector Sprays in Fuel Mixture Preparation” [13].

This method measures the concentration and droplet sizing at various locations from the

injector. The setup and results are shown in the figures that follow. In the second figure,

SOI refers to the start of injection.

Fuel Pressure Regulator

 

   

Incandescent ligiting for

Me scattering
120 deg. w.r.t camera

 

 

Laser Dififi'actioni
........ - Receiver t :".‘f.'

 

 

 

 

 

 

 

 

 

 

Injector control
aid logic

tinting unit ”91's?“ cm“

 

 

 

Figure 7: Schematic of ambient test performed by Delphi for determining concentration and
droplet sizing. [l3]

 

 

 

 

70 + 00 :7 Dv90 200

3‘3 60 ’8‘
g 50 150 g
40 2

8 1
a so 0° .3
'3 m
.3 20 8'
a 10 5° .5

0 0

0 20 40 60 80 100

Time after 301 (ms)

 

Figure 8: Results of measuring concentration and droplet sizing [13]

10

These results are interesting because they show a very large range of droplet sizes.
The size of the droplet directly affects how quickly the fuel mixes with the air. In port
injection, the hot intake valve assists in vaporizing the fuel; but in a direct injection

engine, droplet size could have a significant effect on performance and emissions.

1.5 Calculating the Mass Flow Rate

Determining the mass injected per cycle is at the core of this measurement
technique. Some of the key parameters available for calculation include fluid density,
viscosity, the areas and lengths of the fuel delivery and quartz tubes, and centerline
velocity plotted as a function of time. The problem can be viewed as an internal flow in a
circular pipe of problem. The mass flow rate was first calculated. This was then plotted
as a function of time. From this, the mass injected could be calculated by simply
integrating over the correct portion of time. Initially, a plug flow profile was assumed

n'z=p-uc,-A (1)

(Eqn. 1) to simplify calculations. In this equation, p is the density, ad is the centerline
velocity, and A is the cross-sectional area of the tube through which the fluid is flowing.
Later on, a laminar parabolic velocity profile was assumed (Eqn. 2). The plug flow
assumption over-predicted the mass injected

. I
m=E-p-ud-A (2)

(based on an average measured value over 50 cycles) by roughly 60% because it did not
account for the viscous interaction between the fluid and the wall. The steady state,
laminar, parabolic profile assumption under-predicted the mass flow by about 45%. It

was reasonable to assume laminar flow because the Reynold’s number typically ranged

ll

from 0-2000 with only a very small portion occasionally spiking to 2100. After further
analysis, however, it was determined that a local, or quasi-steady assumption was not
valid.

Equation (2) is an equation for measuring the level of unsteadiness in a flow [14].

This equation is a ratio of the unsteady to viscous term in the

ldu R2

udt V

21 (3)

 

 

streamwise momentum equation. A flow is considered to be unsteady for absolute values
greater than or equal to 1. Figure 9 is a graph of the unsteadiness of a typical injection

event. Clearly this shows a great deal of unsteadiness. Therefore, a laminar unsteady

pipe flow solutions must be implemented.

 

Injector #1 Unsteadiness

 

 

 

 

 

 

 

 

Unsteadiness Factor

 

 

 

 

 

Time [ms]

 

 

 

Figure 9: Graph of the unsteadiness of a fuel injection event.

A solution to this highly unsteady flow pipe problem was developed previously

by Brereton [14]. This solution is an exact unsteady solution of the laminar Navier-

12

Stokes equations for arbitrary, unsteady duct flow and essentially consists of a quasi-
steady solution plus an unsteady correction term. The solution takes into account the
flow’s “history”. The influence of the velocity profile history is weighted by using an

inverse convolution integral. The general form of the solution is given below where
n'1(t)=10-(7t'1'?2)'U(t) (4)

2 C, I C,

In Equation (4), the first term is the quasi-steady velocity term and the second term is the

unsteady correction term. Also, ud is the centerline velocity and W(t) is the inverse

convolution weighting function. The details for this weighting function are given in

Appendix A.

13

CHAPTER 2

EXPERIMENTAL EQUIPMENT

2.1 Dantec Fiber PDA Measurement System
The Particle Dynamics Analysis (PDA) system was used to measure the
centerline velocity profile during each injection event. The Dantec Fiber PDA system

consists of the following major components:

0 Laser: one 120-mW ion laser

 

Figure 10: Dantec ion laser [15].

0 Transmitting optics: Fiber Flow optical system consists of a beam splitter, Bragg
cell, fiber optic cables, and a transmitting probe with a 310 mm focal length

0 Receiver: 57x40 Fiber PDA receiving optics

 

Figure l 1: Dantec transmitting (top) and receiving (bottom) [16].

o 58N70 Fiber PDA detector unit

0 Signal processor: 58N80 PDA enhanced Particle Dynamics Analyzer

0 Computer: for post-processing and storage of data

0 Sofiware: PDA Flow and Particle Sofiware

0 Mounting equipment (breadboard, C-clamps, traverses, etc.)

Because the flow was measured in a very small quartz tube, the setup needed to be
capable of fine adjustments to ensure that the probe volume is located in the center of the
flow. The system was noted to be sensitive to vibrations sent through the floor. For this
reason, all other large machines that produce these vibrations were shut off during

experimentation.

2.2 Injectors

Two gasoline port fuel injectors were used for analysis. The first was an injector
from the 2004 Toyota Prius Hybrid (1.5 liter four-cylinder engine). This injector had 12
tiny holes for increased fiiel atomization. The second injector was a port injector made
by Siemens (111084) for the 2004 Daimler-Chrysler Hemi 5.7 liter V8 engine. Pictures

of these injectors along with close-up views of their nozzles are shown in Figure 12. For

 

Figure 12: Fuel injectors used (left— Toyota Prius injector, right- Chrysler Hemi injector).

the remainder of this report, the Toyota injector will be referred to as Injector #1 and the

Chrysler injector as Injector #2.

2.3 Injector Control System
The fuel injector was controlled by the Cosworth IC5460 Engine Control System.

Using the Flowbench interface sofiware, the injector was set to fire every 200 ms with a

m

t: $7

m in a
.. _ ‘V'g'mmzkom-on avru *
mm» wavy

 

Figure 13: Cosworth injector control module.
9 ms duration. In order for the results to be accurate concerning the fuel injector’s
variability, it is vital that the injector controller introduce as little variability as possible.

In order for this to occur, the frequency and pulse width must be extremely consistent.

2.4 Fuel Delivery System

Compressed nitrogen was used to pressurize the fire] delivery system. This
method was chosen over a fiiel pump because it provides constant, even pressure. A
beaker containing the fuel is placed in a pressure vessel connected to the compressed
nitrogen and the fuel injector. The system was then raised to 50 psi. A diagram of the
fuel delivery system is shown in Figure 14. The material and length of the fuel delivery
tube greatly affected the flow dynamics due to the propagation of pressure waves.
Several materials and lengths were tested, and a nylon tube was eventually selected. It
was also determined that the presence of air bubbles or air pockets located either in the
fire] lines or in the fuel injector itself greatly affects the centerline velocity profile. The
details of these tests will be discussed in a later section. The final setup consisted of a
nylon fiiel delivery tube with a diameter of 6 mm and a length of 60 cm, a quartz tube

with a diameter of 2.97 mm and a length of 10.5 cm, a fuel injector, and a drain tube.

 

 

 

 

Fuel beaker Pressurized fuel
container

 

 

Injector

 

 

 

Drain tube

 

Figure 14: Fuel delivery system.

2.5 Mounting Hardware and Alignment

When designing the test rig for this experiment, an effort was made to insure that
the laser transmitting and receiving probes could be positioned precisely. It was also
important that these components remain fixed in place. The laser transmitter, receiver,
and fuel injector were all mounted on a large breadboard. The injector was mounted on a
custom-made bracket that was attached to a NeWport optical mounting rod (Model 75).
This rod allowed for vertical adjustment. The laser-transmitting probe was mounted on a
dual rod system because of its size and weight. This dual rod setup was mounted on an
x-y translation stage (Newport Model 400). The laser-transmitting probe had a built-in
swivel base that allowed for rotation in the x-y plane. Thus, the probe could be adjusted
in the x, y, and z axes and was free to rotate in the x-y plane. The receiving optic was
mounted in a similar fashion but used a single optical mounting rod. When these rigs

were complete, they were capable of making the fine adjustments necessary to properly

18

align the optics. As a result, good data rates were achieved. Pictures of the mounting

hardware are shown below.

   

Figure 15: Mounting setup.
2.6 Oscilloscope
The Hewlett-Packard Infinium oscilloscope (Model # 54810 A) was used to
monitor Doppler bursts as well as to measure the variability of the Cosworth injector
controller. In order to obtain valid results, it was important to get a good strong Doppler

signal from the detector. The oscilloscope was particularly usefirl for making fine

 

Figure 16: Hewlett-Packard Infinium Oscilloscope.
adjustments in the detector alignment. It was also used to measure the variability in the
frequency and voltage output of the Cosworth controller. Statistics such as standard

deviation of voltage, frequency, and rise time were recorded.

2.7 Seed and Dispersal Devices

Two types of seed were used in the experiments for this project. Early on,
polyamid seeding particles from Dantec were used. These particles have a mean particle
diameter of 5 pm and are recommended for use in fluids with densities similar to water.
In order to distribute the seed throughout the working liquid, a Nuova II magnetic stirrer
was used. Later, a microbubble seeding technique was developed and implemented. A

hi gh-speed blender was used to entrain and distribute these bubbles throughout the liquid.

2.8 Mass Balance

An AND GX-4000 mass balance was used to measure the mass of the fuel
injected over a series of 50 injections. The value was then divided to calculate the cycle-
averaged mass injected per cycle. This value was compared with the calculated values

for validation purposes.

 

Figure 17: AND GX-4000 mass balance.

20

2.9 Delay Box
A Stanford Research Systems delay box (Model DG535) was used to center the
injection pulse in the time plot. Without the delay box, the injection pulse was located at

the endpoints of the graph making it more difficult to analyze.

, .‘rgs'xilfiflr’l‘ei'fiiflrifi' 4. 4.1" ..

 

Figure 18: Stanford Research Systems delay box.

2.10 Vacuum Pump
A Dayton Electric Speedaire vacuum pump (Model 78866) was used to evacuate
air from the fuel delivery system as well as from the fuel injector. This pump is capable

of pulling up to 23 inches of mercury (in Hg).

2.11 Software

Microsoft Excel and C++ were used extensively for performing analysis and
calculations. The program developed in C++ requires a .txt file that is exported from
Dantec’s PDA Flow and Particle Software. It has inputs for the tube radius, fluid density,
kinematic viscosity, and approximate beginning and ending of injection event. For
simplicity, the program defaults to the test conditions used in the lab (water at standard
atmospheric conditions). The program then sorts the data so that it is arranged in

consecutive injections, performs the necessary unsteady, laminar, pipe flow mass

21

calculations as previously discussed (Appendix A), and outputs the mass injected per
cycle along with a sparse value that describes the density of the data over the cycle. This
sparse value is important for understanding whether the output is reasonable. If the time
resolution of the data is too small, interpolation error becomes a problem. The program

also outputs statistics such as average, standard deviation, and coefficient of variation.

22

CHAPTER 3

EXPERIMENTAL PROCEDURE

3.1 Complete Test Rig Setup and General Procedure

Vacuum
pump

ybn trite (ID - 6 mm)
/.r';?3" f ‘1‘, Compressed nitrogen

 

QuartztlbeaD-Bmm) {5" “Wk.“ Sflpdtouymm

    
  

 

 

 

Laser -transmitting optic \ i

Photo detector

   

7,,

27.22:.

 

 

 

 

 

 

 

 

Coeworthfml injector Dantec LDA data analyst
control system processor and software

Figure 19: Experimental setup.

General Procedure:

1.) Attach injector to quartz tube using rubber hose and hose clamps. Fasten the
drain tube to the nozzle end of the injector. Hook up the wires from the injector
control unit to the connector.

2.) Turn on the Cosworth control system, laser, computers, oscilloscope, and delay

box.

23

3.) Load the PDA Flow and Particle software to control the LDA system and load
the Flowbench software to control the Cosworth injection control device.

4.) Seed the working fluid (water for this case) and fill the beaker located in the
pressure vessel. Seal the pressure vessel completely. Any leak will result in a
rapid evacuation of the liquid when pressure applied.

5.) Connect the pressure vessel to a tank of compressed nitrogen. Slowly bring up
the pressure and look for leaks. If no leaks exist, raise the pressure up to 50 psi.

6.) Evacuate any air from the fuel lines and injector using the vacuum pump. Make
sure the valve to the container vessel is shut off so that only air is pulled out of the
fuel supply line and injector and not all of the fluid from the beaker exits.

7.) When laser comes on, align the laser transmitter and detector.

8.) Fire the injector and observe the Doppler bursts on the oscilloscope. Make fine
adjustments to the LDA detector until the bursts are at their maximum.

9.) With the injector still running, trigger the PDA system to collect data.

10.) If the time resolution is not good enough (Velocity Data Rate less than 2 kHz),

adjust the High Voltage and S/N Validation parameters and collect data again.

11.) If time resolution is good, collect as much data as is needed, then depressurize

system, and power down electronic devices.

12.) Post-process data using Excel or C++ program.

This procedure was developed through months of testing. The details of these various

tests and some of the important lessons learned can be viewed in Appendix C.

24

 

Figure 20: Photograph of experimental setup.

3.2 LDA Technique

Laser Doppler Anemometry (LDA) was chosen to measure the centerline velocity
of the fuel entering the fuel injector. , LDA was desirable because of it is non-intrusive,
directionally sensitive, and has high spatial and temporal resolution. In general, LDA
works by processing data from laser light that is reflected by particles in the flow field
(Doppler bursts). Two lasers intersect in the flow, creating a probe volume. When
particles suspended in the working fluid intersect this probe volume, light is scattered
with frequencies that are mathematically related to the velocity of the fluid in the probe
volume. The drawing that follows shows the general setup and flow of data. The system

is capable of detecting negative velocities because

25

& Bragg cell
Laser 4
Photo detector
Transmitting] »~-

{8081171118 optics

      

Flow with
seeding particles

/ I
/ Measurement
Volume

 

Sigial Processing

Figure 21: LDA schematic (shown in back-scatter configuration) [16].
a Bragg cell is used to shift the frequency of one of the laser beams. The forward-scatter
configuration was chosen to achieve higher data rates. In forward-scatter, the detector is
separate from the transmitting probe as opposed to back-scatter, where the detector is
integrated in the probe.

The setup used in this experiment consisted of transmitting and receiving probes
with focal lengths of 310 mm and 400 mm respectively. The dimensions of the resulting
probe volume were 77 x 77 x 945 pm. These dimensions result in a probe volume of
0.004198 mm3 in a fused quartz tube. The details of this calculation are presented in
Appendix B. The cross-section of the probe volume was 0.081407 mm2 as compared to
the quartz tube cross-sectional area of 6.9279 mmz. More importantly, the length of the

probe volume is approximately one third of the quartz tube’s inner diameter. This

26

comparison was important to determine if the velocity readings could be approximated as

centerline velocities. It was determined from previous experiments that the optimal
angle for the receiving optics was 39° off-axis angle [9]. The fringe spacing was 3.15

pm, the frequency shift was 40 MHz, and the cyclic length was set to 360°.
Several of the previously mentioned parameters are entered into the Dantec
program called PDA Flow and Particle Software. This program controls the detector
settings, initiates and terminates data collection, and has several useful post-processing
features. The software also requires a bandwidth. For this experiment, 1.20 MHz was
chosen because it corresponds to a velocity range of -l.890 m/s to H890 m/s.
Preliminary tests indicated that the peak velocities would fall within this range. Two
parameters that can be modified to improve data collection rates are the High Voltage
parameter and the Data Validation number. High Voltage corresponds to the voltage
supplied to the photomultipliers. In effect, it controls the detector sensitivity. The High
Voltage used for data collection ranged from 700 to 800 Volts. Too low of a voltage
results in a low data collection rate, and too high results in excessive noise. The Data
Validation value corresponds to the way the data is filtered. The signal/noise validation
value typically varied from -1dB to ldB.
Once these values were set, the experiment could then be run and data collected.
The program was set up so that it outputted arrival time, cycle time (0-200 ms), the
corresponding crank angle (0-360°), and velocity. These values were tabulated and
could be exported for further analysis in Microsoft Excel or C++. The Dantec program
also outputted a cycle-averaged plot, and the number of data points collected at each

crank angle was tabulated. This was particularly useful for determining if the resolution

27

was good throughout the injection event. The cycle-averaged plot was helpful for getting

an idea of the injection profile immediately. In another window, a histogram showed the

total distribution of data points collected over the 200 ms event. There was also a table
that showed the validation data rate, velocity validation, elapsed time, and reset pulses.
The velocity validation rate roughly indicated the time resolution of the run. In order to
get a velocity reading for nearly every millisecond, a validation data rate of greater than
or equal to 2 kHz was required. The velocity validation feature was enabled to achieve
more accurate results. Velocity validation values of 95-100% were achieved during these

experiments. The elapsed time simply recorded the length of the data collection and the

reset pulses indicated the number of injections collected.

3.3 Post-processing

Initially, post-processing was done in Excel. The data was sorted and the mass
injected per cycle was calculated using a plug flow assumption. When it was determined
that this should be done on a large scale and that the unsteady equations should be used
for analysis, a program was developed in C++ to handle this task and was described in
Chapter 2. In order to validate the accuracy of this program, a simpler test data set was

constructed and evaluated using Excel. Brereton’s program in FORTRAN was also run

for comparison purposes. The C++ program showed good agreement with these

programs, so it was accepted and used for analysis. To verify that the equations were
reasonable, cycle-averaged results from a mass balance were compared to the predicted
values outputted by the program. To begin this procedure, the mass balance was first

zeroed. Fifty injections were collected in a beaker on the mass balance, and the resulting

28

mass was divided by fifty to determine the average mass injected per injection. When
this number was compared to the output data of the C++ program, it was found to be

within 6% for Injector #1 and 8% for Injector #2. This agreement will be discussed in

greater detail in the next section.

29

CHAPTER 4

RESULTS AND DISCUSSION

4.] Results

After perfecting the test rig and experimental procedure, tests were run using the two
injectors previously introduced. The results of each injector will be discussed
individually, and then a comparison of injector variability will be made at the end of this

chapter.

4.1.1 Injector #1 Results

The following plots are samples of the data collected for the Toyota Prius injector.
The first plot shows an average of centerline velocity along with the voltage applied to
the injector. The following plots are three individual, consecutive injection events.

Lastly, a bar graph shows the variability in the mass injected for each event.

30

 

 

 

 

 

 

Centerline
Velocity [mls]

Injector #1: Centerline Velocity and Voltage Plots

4-8-04
0.6 ( )

o N
Injector Voltage

.09
#01

.0
no:

fire #‘flw Ma»

9.0
0A

 

ON-bODCD—i-e

.33
A

0 50 100 150 200

Time [ms]

 

—- Injector #1 velocity

 

 

—-—- Voltage

 

 

Figure 22: Cycle-averaged centerline velocity plot with injector voltage.

31

 

 

 

 

 

 

 

 

 

 

 

 

 

Injector #1 Individual Cycle (4-8-04) 7
E
E.
t
g Pacle 1]
.E
t
8
c .
8

0 50 100 150 200
Time [ms]

Injector #1 Individual Cycle (4-8-04)
E
E.
3'
'8
2
g {—o— CYCle 2
.E
'C
i

o 50 100 150 200
Time [ms]

 

 

 

 

 

 

Centerline Velocity [mls]

Injector #1 Individual Cycle (4-8-04)

0.7
0.6
0.5
0.4
0.3
0.2
0.1

L; Cycle 3]

l

-0.1
-0.2

 

0 50 100 1 50 200

Time [ms]

 

 

 

Figure 23: Three consecutive cycles from Injector #1.

32

J

 

 

EPA Injector- Calculated mass injected
(4-8-04)

meow
01001

..x
01

Mass Injected [mg]
8 8

01

 

v—COI-Ome—O'JLDNC)
‘— ‘—

‘_-‘_‘_.

(OI-ON
NNN

‘— O)
N N
Injection
Average: 27.0378mg, Standard Deviation: 1.0664mg.
Coefficient of Variation: 3.94499,

Measured Mass (cycle avg): 28.84mg

 

 

 

Figure 24: Mass injected chart for Injector #1.
The individual cycles show the dense distribution of data at the point of injection.
The time resolution (averaging 4 data points per millisecond) gives some insight into the
flow dynamics. Significant variability can be observed between the three graphs shown
in Figure 23. The bar graph above shows that there is, in fact, a large amount of
variability between consecutive injection events. This variation will have an effect on the

fuel/air ratio, particularly for direct injection engines.

4.1.2 Injector #2 Results

The following plots are for the new Chrysler Hemi fuel injector. As before, the

first plot shows an average of centerline velocity along with the voltage applied to the

33

 

injector. This is followed by plots of three individual consecutive injection events and

finally a bar graph showing the variability in the mass injected for each event.

Injector #2: Centerline Velocity and Voltage Plots

 

 

 

 

 

 

 

 

 

 

 

 

(5-14-04)

0.8 14

. . ~— 12
'5' 0.6 Iv 3:
o ‘ ~~ 10 9
5 £- 04 3
t . ~— 8 >
a E e
r: g 0.2 - ~~ 6 o
8 '6 Ma‘fli-‘VWi‘Qg-fiefi’fiufl‘fig-J‘Ifitk“ 4 g
> o - 'E
2 .—

‘O.2 r r r 0

O 50 100 150 200

Time [ms] —— Injector #1 velocity

 

 

——-——.—- Voltage

 

 

 

 

 

 

Figure 25: Cycle-averaged centerline velocity plot with injector voltage.

34

 

Injector #2 Individual Cycle (5-14-04)

0.8
0.6
0.4 (I

0.2 [+ Cycle 1 l

-o.2 -
-o.4

 

 

 

 

 

 

 

 

Centerline Velocity [mls]

 

 

 

0 50 100 1 50 200

Time [ms]

 

 

 

Injector #2 Individual Cycle (5-14-04)

0.8

 

0.6

 

 

 

 

0.4
i L—o— Cycle 2
0 2

 

 

 

 

Centerline Velocity [mls]

 

 

-0.2 . i
0 50 100 150 200

Time [ms]

 

 

 

 

Injector #2 Individual Cycle (5-14-04)

0.8
0.6

 

 

 

 

 

 

0-2 b-CYEE‘ET

 

 

-o.2
I
43.4

Centerline Velocity [mls]

 

 

 

0 50 100 1 50 200
Time [m s]

Figure 26: Three consecutive cycles from Injector #2.

35

 

 

 

 

 

Injector 2- Calculated mass injected
(5-14-04)

Mass Injected [mg]

 

mekms-OOLONO) MID
n—r-v-V-P NN

NO)
NN

‘5.
Injection

Average: 30.6147 mg, Standard deviation: 1.44552 mg
Coefficient of variation: 3.94499. Measured mass (cycle avg): 33.24 mg

 

 

Figure 27: Mass injected chart for Injector #2.

Again, significant variability can be observed in the individual velocity traces.
Initially, the bar graph shows great variability. The injector was allowed to run for
several cycles before data was collected, so this variability at the beginning should not be
attributed to some kind of startup condition. It is believed that this is simply a display of
the varied nature of the injector. Note that the standard deviation is greater for Injector

#2 than it is for the #1.

4.2 Discussion
4.2.1 Comparison of Injectors #1 and #2
Figure 28 shows the average centerline velocity plots from Injectors #1 and #2 as

well as the voltage applied. It very interesting to note how similar the pressure wave

36

 

Injector Comparison (5-14-04)

14

.3
C

.0 .O .O
N co
Injector Voltage

A

it ”747’ “*Ts’t-fifiwv‘t—awe ii:

Centerline Velocity
[mls]
O

c'>
'_.

 

omemoo

.5
N

0 50 1 00 1 50 200

Time [ms] _"- InjeCtor 1
—— Injector 2

———— Voltage

 

 

 

 

 

 

Figure 28: Average centerline velocities and applied voltage.

oscillation is. The most notable difference is that Injector #2 flows considerably more
fuel. This is expected, as the cylinder displacement of a Chrysler Hemi is about twice
that of the Toyota Prius.

The standard deviation of mass injected for Injector #1 was 1.0664 milligrams as
compared to 1.4552 milligrams for Injector #2. It appears as though the Prius injector
was designed with more precise fuel control in mind. This is not surprising because the
Prius is a car whose designers were concerned with fuel economy. Also plotted are plots

of the probability density functions.

37

 

Injector #1- PDF Plot of Mass Injected

 

 

 

 

 

 

 

 

I. 115 Injections

 

 

 

 

 

Number of Injectlons
o a B 8 B 8 8

 

 

 

24 25 26 27 28 29 30 31
less Injected [my

 

 

 

Figure 29: PDF of mass injected for Injector #1.

 

Injector #2- PDF Plot of Mass Injected

 

[.115 Injections]

 

Number of Injection:
0 m a a B at 8 81 e a

 

171819 20 2122 23 24 25 26 2728 29 30 3132 33

less Injected [my

 

 

 

Figure 30: PDF of mass injected by Injector #2.
A useful way of viewing the injector precision is to assume that this injector
variability could translate into fuel/air ratio variability. If the average were taken to be
the amount of fuel required for a stoichiometric mixture, then the maximum and

minimum fuel/air ratios would be as follows. This is a large range and would be highly

38

Table 2: Fuel/Air ratio range.

 

jlnjectorlrnjector
1 2
Max 15.54 15.85
[Min 13.27 13.48

 

 

 

 

 

 

 

undesirable from an emissions and fuel economy standpoint. For port injection systems,
however, the fuel/air ratio may not fluctuate quite as much. There is a phenomenon
whereby the fuel film on the intake runner, port, and intake valve serves as a cycle-to-
cycle filter. In other words, the extreme variations in fuel injected may not result in sharp
variations in the in-cylinder fuel/air ratios for port injection systems [7],[10]. This is not
the case for direct injection gasoline engines, however. These engines require precise
fuel metering for smooth operation. Variation in the amount of fuel injected can also
have dramatic effects on the emissions produced, particularly oxides of nitrogen (NOx)

and hydrocarbons (HC).

4.2.2 Sources of Variability
4.2.2.1 Cosworth Variability

A great deal of effort was taken to isolate the injector so that only the variability
of the injector would be measured. Nevertheless, there is still the possibility of
variability from the Cosworth control module. To investigate this, the control module
was monitored using an oscilloscope to measure and plot the voltage. Three injection
events are shown in the following figure. One can observe slight variations in the peak

voltage. To better quantify this variability, a table was constructed from the statistics

39

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Cosworth Voltage Plots
14
12 ——«
r— 11 r1
2 10 «I -
o 3 j
°’ I
g 6 l
O i . ‘ 1 . . ,. - 1
> 4 tier-“+4; . , ...4 f 1 125% t c #14" 4‘54»;‘14“‘1-5'1421-3‘13’1
2
0 T F T T
0 100 200 300 400
Time [ms]

500

 

 

Figure 31: Voltage plot of the injection control system.

recorded by the oscilloscope. These values were recorded over a series of about 200

injection events. It would be interesting to find out if this variability had any effect on

the motion of the needle inside the injector. It is possible that this variability

Table 3: Injection control system variability statistics.

 

 

 

 

 

 

 

I tandard "J
Mean Deviation lMinimum Maximu
[Volts peak-to-peakl7.965 V 111 mV 7.87 V 8.92 V
[Period I199.998 m 32 us 199.61 m 200.03ms
IFrequency I5.0004 Hz 794.459 sz 4.9994 Hz'5.0098 Hz
[like Time I442 us 24 us 320 us 2.15 ms

 

 

 

had an affect the performance of the injector. For that reason, the variability presented

should be considered the variability of the injector and the control module.

40

4.2.2.2 Sources of Error

As with any experiment, there is always the potential of errors affecting the end
result. The uncertainty errors and systematic errors associated with this experiment will
be discussed at this time.

Uncertainty errors in this experiment arose from the measurements of the quartz
tube diameter, centerline velocity, as well as the temperature measurements needed to
determine the fluid density. Recall that the general form of the equation used to solve
for the mass flow rate was

m(t)=,0°(fl°R2)'U(t) (4)

where

1 d *
U(t)—§uc,(t)+E(uc,(t)) W(t) . (5)

The uncertainty associated with the quasi-steady portion (gut, (1)) is given by taking the

partial differential of (4) using the following equation
dn'z=-;—[(p-u~dA)+(p-A-du)+(A-u-dp)]. (7)

The uncertainty from the radius measurements is i: 0.005 mm, from the velocity
measurements (LDA precision) i 0.0005 m/s, and from the density i 0.05 kg/m3. The
resulting uncertainty from this quasi-steady calculation is i 0.001839 mg/ms. To put that
into perspective, the quasi-steady mass flow rates varied from 0 to 2.07416 mg/ms. Thus,
the uncertainty contributions from measurements and tabulated values were quite low.
Systematic errors are likely present, though their direct contributions may be more

difficult to quantify. One key assumption that was made was that the quartz tube was

41

perfectly circular and that it had a uniform cross-sectional area throughout its length.
Obviously the calculations are quite sensitive to the diameter measurement as it is
squared in the area calculation.
There are also some errors associated with measuring the centerline velocity. As
previously stated, the recorded velocities were assumed to be centerline velocities. In
reality, these are the probe volume-averaged velocities. While the probe volume area is
small compared to the flow area, the length of the probe volume is approximately 1/3 of
the diameter. This could contribute to some error as a result of this averaging. With the
assumed parabolic profile, it is reasonable to assume that the error in the measured
velocity introduced as a result of the probe volume is 0-5%. A key parameter that affects
the accuracy of the velocity is the measurement of the angle between the laser beams.
Based on previous experiments, the velocity is likely affected by this parameter by about
i- 2%. Finally, processing error could also come into effect. The documentation on the
LDA system indicated that the processor accuracy was i 0.5% full spectrum. Since the
velocity range was from —1 .2-1.2 m/s, this translates to 0.006 m/s. It is likely that these
errors would average out since several tens of thousands of samples were taken. While
these certainly are not all the potential sources of error, they were thought to be the major
contributors.
Having discussed the potential sources of error, however, it should be understood
that for the experiments run, the agreement was very good with the cycle-averaged
measurements. The table on the following page shows this agreement. The LDA
measurements of the mass injected by Injector #1 was typically within 6% and Injector

#2 within 8% of the cycle-averaged measured values from the mass balance. Moreover,

42

 

the relative cycle-to-cycle variability shown by the previous bar graphs is unaffected by
this uncertainty. The only change is in the magnitude of the mass injected. The reason
for this is that the calculation procedure for the mass injected for each cycle is the same
and the LDA equipment was not moved between measurements. Any error present

should be consistent.

Table 4: Comparison between mass balance and LDA measurements

 

 

 

 

 

 

 

ln'ector £1 Injector 52
Mass Balance LDA Injection Percent Mass Balance LDA Injection Percent
Measurement (mg) Average (mg) Difference Measurement (mg) Average (mg) Difference
I 28.84 27.343 5.191 33.24 30.7676 W
l 28.84 27.2857 5.389 33.24 30.4943 8.260]
I 28.84 27.0378 6.249 33.24 30.6147 7.89%

 

 

 

 

 

 

 

43

CHAPTER 5

CONCLUSIONS

In this study, the real time cycle-to-cycle variability of a fuel injector and its
control system was quantified. The approach was quite different from traditional
measurement techniques as it involved measuring the centerline velocity before the
injector using LDA as opposed to making measurements after the injector. It was also

found that good time resolution is necessary, and the seed used to scatter the laser light

must not plug the injector if the results are to be trusted.

The results of these experiments have shown that there is a significant amount of
cycle-to-cycle variation for the two injectors used. The Toyota Prius injector (Inj. #1)
had an observed standard deviation of 1.0664 mg while that for the Chrysler Hemi (Inj.
#2) injector was 1.4552 mg. If the variability in mass injected was directly related to the
fuel/air ratio, this would result in a large amount of variation and would be highly
undesirable for clean, efficient combustion. For port injection systems, this may not be
the case because of fuel film that is a result of wetting on the valves and/or intake ports.
This causes a filtered or damped response so that the actual fuel/air ratio seen in the
cylinder may not show as much variability as the fuel injector [7],[10]. In the case of
directly injected engines, however, this variability is much more closely related to the in-

cylinder fuel/air ratio. There have also been some recent developments in fuel-injected

two-stroke engines. More precise electronic fuel control could make these engines

cleaner and more acceptable in the near future [I 7]. The measurement technique outlined

44

 

 

in this report could prove to be a useful tool for companies striving to design more
precise and consistent fuel injector performance for such applications.

While developing and perfecting this measurement method, several important
discoveries were made. These discoveries are discussed in detail in Appendix C and will
be summarized at this time.

First, the microbubble seeding technique used in this experiment proved to be
extremely effective. By introducing thousands of these tiny bubbles into the working
fluid, excellent data rates were achieved without plugging the fuel injector. Because of
the small size of these bubbles, they did little or nothing to affect the velocity profile and
bulk fluid density (proved experimentally).

The second useful discovery was the presence of air in the fuel injector and the
effects it has on the centerline velocity. Figure 30 shows the difference in the centerline

velocity profile with and without air in the injector. These oscillations affect the

 

Injector #1: Centerline Velocity and Voltage Plots

 

 

(4-8-04)
1 14

. ,g‘ 1 12.
g 05 7 7 7 777 777 7 7m7 7 77 105
i E o A 4— ' .I 3
g e 1 V 5
g -o.5 g.
0 _
o -1

 

 

 

 

Time [ms] l—— purged injectorj

I
i ~77 unpurged injectorl

_ git/9113,99 , ,

l

 

 

Figure 32: Velocity plots for Injector #1 with and without air.

45

 

consistency of the injector. In some cases, the oscillation did not die out before the next

injection event, adding to injector variability. Even though it was initially thought that

these bubbles would dissolve or “work themselves out” of the system, this was not
observed over the course of several days of intermittent testing.

Lastly, it was determined that a fuel injector is only as good as the control system

that actuates it. If there are fluctuations in the voltage sent to the injector, increased

variability is inevitable. Stable control systems must be developed to minimize this

problem.

46

 

CHAPTER 6

RECOMMENDATIONS

After completing this project, it was clear that there were several other areas that

would be worth to investigating.

0 Use gasoline in test rig and automotive injection control unit as well as stock
fuel delivery setup to simulate more realistic conditions. Is there any
variability introduced by the addition of a fuel pump? How much?

0 Find a way to quantify the variability of the injector with air inside. How
much of an improvement is there when the air is evacuated?

o Evacuate air from injectors in an engine on a dynamometer/emissions test cell
to see how performance and emissions changes as a result.

0 Measure variability in high-pressure diesel injectors. It is difficult to design a
quartz window that can withstand these high pressures. Such a window has
already been developed here at the Michigan State University Engine
Research Laboratory, however. This window has been tested up to about
30,000 psi [8].

- Streamline software for sale to companies which develop fuel injectors for
commercial use.

0 Compare similar injectors made by the same company to measure variability
from one injector to another. Is the manufacturing process consistent enough?

0 Measure variability in a piezoelectric injector (Siemens). Are they more

consistent? What is the consistency for multiple injection systems?

47

0 Test injectors for D1, HCCI engines. Since there is no wall-wetting filtering

effect, precise fuel metering is more important.

Great strides have been made in fuel injection technology since the 1920’s, but
there is still much more that can be explored in this field. More precisefuel injectors and
control systems will enable automotive manufacturers to achieve greater fuel efficiency

and cleaner emissions than ever before.

48

Appendices

49

APPENDIX A
Details of the Mass Flow Equation
The exact unsteady solution of the laminar Navier-Stokes equations is discussed in this
section as developed by Brereton [14]. As previously mentioned, the general expression

for calculating the mass flow rate in uniform density flow is
mow-(8129479) (4)

2 (‘I d, (.I

where W (t) is a known weighting function and * the convolution operator. This exact

solution applies to laminar, fully developed, constant property duct flow undergoing
arbitrary unsteadiness from an initially steady state. Recall that the é—ud (t) term is

simply the momentary velocity term for a steady laminar parabolic velocity profile. The
g—(uc,(t))* W (t) term is an unsteady correction term. By definition of the convolution
t

operator *,

d :1: _ Ifi _ ' . ' .
5W”) mn— 0j dt (t t) W(t) du (7)

For simplicity of evaluation, the following non-dimensional term is now introduced:

V't
72?. (8)

Here, v is viscosity, t is time, and R is the measurement tube radius. W (2’) is an inverse

convolution integral term which can be described as follows:

For 1 < 0.01, W(r) = 0.5 — 2.2567J1+ 1.1251 (9)

(gt-73'7” (0.253393 cos(26. 1 2881) + 0.499595 sin(26. 12881))

+ (””05“ (0.0816947 cos(58.5689z') + 0.289019sin(58.56891))
For 1 2 0.01, W(1) = + e7377»58'5°’((0.0402422 cos(94.02701) + 0.200663 sin(94.02701))

+ 1376629727” ((0.0240313cos(13 1 .5101) + 0.153211sin(131.5101))

+ 61°27'47“)”((0.0160209cos(l 70.5211) + 0.123770 sin(1 705211))

as given in Brereton, Schock, Rahi, and Bedford [15].

In this solution, the angles are in radians. Once this “modified” area-averaged
velocity in (5) is developed from the measured centerline velocity history, the mass
injected can then be calculated by carrying out the convolution integral, at each instant in

the time series, and multiplying 17(1) by the fluid density and the cross-sectional area of

the tube. It was determined that the unsteady correction term contributes nearly as much
(~ 45%) to the calculation of the total mass injected as the quasi-steady laminar parabolic

profile portion.

51

 

APPENDIX B
Calculation of the LDA Probe Volume
In an LDA system, two laser beams cross in a fluid flow. The drawing below
shows the general layout of an LDA system and shows the ellipsoidal shape of the probe

volume. The intersecting lasers produce a series of fringes. As particles cross this

 

 

 

 

 

 

 

 

 

 

Transmitting
W system
w...»-
fmfifiz
D
a... . ,,
f
I
X .
/1\ IntenSIty
lie 2) \X distribution
52 A
5,, Z

   
  

..... ,------- Measurement
‘ volume
Y
Figure 33: Probe volume diagram and intensity distribution [16].
probe volume, light is scattered. The fluid velocity can then be determined based on the
Doppler shifi of the light reflected from the moving particle.

When calculating the probe volume, the following diagram and equations are very

useful.

52

 

Length:

Width:

Height:

Fringe separation:

Number of fringes:

X

 

Figure 34: Probe volume dimensions [16].

4F}. (10)

 

6, — (11)

6 =——— (12)

5] =_— (13)

(14)

53

 

Table 5: Probe volume variables.

mm
514.5 nm

beam 4.684
nder ratio 1.95
diameter of laser 1.350 mm

 

When the test fluid is air, these equations can be used as they are. For the case of fluid
flowing through a quartz tube, refraction must be considered. The refractive index of
quartz and the working fluid, water in this case, must be known and the effective beam
angle must be modified according to the following equation:

nl sin, = 712 sin 6, (15)
In this equation, 11 is the refractive index of the medium, i stands for incidence, and t
stands for transmitted. The following diagram shows how the angle changes as it passes

through each medium. Now that all the variables are known, the dimensions can be

 

0002926
.3315?
.45843

 

 

 

 

 

Figure 35: Beam refraction sketch and values.

calculated and finally the probe volume, using the equation for the volume of an ellipsis

shown below.

V . = _7[ ..... _
ellipse 3 2 2 2 ( l 6)

54

The results are shown for calculating the probe volume in air as well as inside a quartz
tube with water.

Table 6: Results of probe volume calculations.

 

 

 

 

 

 

 

 

 

 

[Dimension Value in air Value in quartz tube/water
By (mm) 0.077217 0.077217

I8x 0.077475 0.077345

I62 0.945584 1 .34233

Volume (mm"3) 0.002926 0.004198

IFringe Separation 0.1m) 3.15 3.15

Number of Fringes 24 24

 

 

55

APPENDIX C

Evolution of Experimental Technique

Over the course of this project, several discoveries were made that were quite
significant. This section discusses these discoveries and the lessons learned from them.
Fuel Delivery Tube

Early in the experimental phase, the test rig was designed with a rubber fuel
injection hose that connected to the quartz tube and supplied the injector with fuel. The
working fluid was mineral spirits because it possessed a density and viscosity that was
similar to gasoline but is safer to work with. The mineral spirits was then seeded with the
previously mentioned Dantec polyamid seed. The initial plots looked like the one shown

on the following page. After overlaying the voltage plot on top of the velocity plot, it

 

 

 

 

 

__—_.——.___.—————~—

Cosworth/E PA Injector measurements (1 -1 5-04)

 

 

 

 

 

 

Volts

 

Velocity [mls]

 

 

0 50 100

 

 

 

Figure 36: Average plot of early experiments.

56

 

was proposed early on that the first spike corresponded to the injector opening and that
the remaining oscillations were due to pressure waves. This was later found to be
partially correct. At the time, it was thought that the pressure waves reflected back to the
surface of the beaker containing the fuel. However, when wave speeds were estimated

using the following equation and, it was determined that the speed oscillations was far to

c: E_h (17)

slow. In this equation E is the Young’s modulus, h is the wall thickness, p is the density
of the liquid, and R is the inner radius of the tube. At that point in time, the cause of this
oscillation was unclear. In order to observe the effects of pressure waves, the fuel
supply line was modified. Two tests were designed: the first tube was a 20-inch rubber
tube, the second a 15-inch copper tube. The resulting velocity plots are shown in Figures
35 and 36. From these graphs, it was evident that the type of fire] delivery line affects the
rate and duration of oscillation. The copper tube shows the greatest amplitude and
longest duration of oscillation, while the rubber tube has the smallest amplitude and
shortest duration. The reason for this is that the walls of the rubber tube appear to absorb
pressure fluctuations while the rigid walls of the copper tube do little to absorb them.
Because the purpose of this investigation is to isolate the variability of the fuel injector,
the long rubber tube is preferred. This was later changed to a clear nylon tube for

viewing purposes.

57

 

Injector Centerline Velocity using a 15" copper
fuel delivery tube

11

 

d

 

.0
0'1

 

 

 

Centedine Velocity [rer]
O

 

. .6
.. 01
1:1 1——
l

50 100 150 ‘ 200
Timelms]

 

 

 

Figure 37: Velocity profiles for a 15-inch copper fuel delivery line.

 

Injector Centerline Velocity using a 20"
rubber fuel delivery tube

 

 

 

 

 

1:11:01:

 

 

 

 

 

 

 

 

 

 

 

.0151:

 

 

 

 

manomumcn-n
1 1

Centerline Velocity [mls]

 

 

 

1
.0
CD

I

I

1

I

|

I

I

0 ’ 50 100 ‘ 150200

 

 

 

Figure 38: Velocity profile for a 20-inch rubber fuel delivery line.

Seeding
Another interesting observation was that the shape of the velocity profile had a

tendency to change throughout time. Several causes were proposed. One thought was

58

that as the fuel beaker emptied, the distance traveled by the pressure waves was
shortened. Another possible explanation was that the seed was building up in the injector
and plugging it. This theory was developed after observing a thin film of seed that
remained on the walls of the beaker that supplied fuel to the system. To test this theory,
STP Super Concentrated Fuel Injector Cleaner was run through the injector undiluted.
The cleaner was allowed to soak in the injector and then purged the following day. After
cleaning the injector, data was again collected. These velocity profiles resembled earlier
profiles, so it was determined that the injector was in fact plugging due to seeding the
flow.

To remedy this problem, a microbubble seeding technique was devised. To
accomplish this, water was placed in a high-speed blender with a small amount of
concentrated liquid soap. When the blender was turned on, air was entrained and finely
distributed into the water. The soap coated these bubbles and slowed down diffusion
allowing the bubbles to remain suspended in the water for several minutes. This method
resulted in greater data rates and much more consistent velocity profiles. The size

distribution of these bubbles can be seen in the PDA bar graph that follows.

Counts
200 7

 

 

 

0.000 200.000 400.000
PDA 0 [pm]

Figure 39: Bubble sizes used to seed the flow.

59

The following profiles show how seed eventually plugs the injector, alters the

velocity profile, and how the use of bubbles remedies this problem.

1

 

 

 

 

 

 

     

 

 

 

 

 

  

 

 

 

 

 

 

 

 

       

 

 

 

  

 

  

  

 

 

 

   

 

 

   

 

 

PDA U1-Meon[m/s] PDA U1 -Meon [m/s]
1.0‘ .7 ; ID] 1 : I
0.87 777777777 t 7777777777777777 7: 77777 0.8 ---------- ------- 7 -------- -----
0.6:} 77777777 i 7777777 t 777777 i 77777 0.6:: -------- E --------------- 7: -----
1..., -------- 1 ------- - ------ a ----- 0.4:: ........ 1 ................ a .....
02.. ........ g ....... .. 021: ........ 1 ............... 1: .....
00; E ‘ ‘é 0.0? l . .
41277 77777777 f 77777777 7 771: 77777 41.2: -------- i -------- 1: -------- 1: -----
---------- ------ -
100.000 200.000 300.000 . 100000 . 200000 . 3001000 -
Angle Bin [deg] Angle Bin [deg]
PDA UI-Meon [mls] 3 PDA U1-Meon[m/s] 4
100 j I I 1-0 I i if
0.8:: -------- j 7777777 ' 77777777 7: 77777 0.8:: -------- :1 ------- I -------- i -----
0.5; 77777777 777777777777777 77777 0.63> } ------- ' -------- 7: -----
0.47 777777777 t 777777777777777 1 77777 0.4“ -------- f --------------- 7, -----
0.2:; -------- j ------------ i ----- 0.2: -------- j -------------- a: -----
00 i . 0.0? :
-0.2< --------- 1 -------- . ------ 1' ----- 02:1 -------- l -------- 1' ------- 1: -----
4.41 ----- . 3.4:: -------- é -------- -----
100.000 200.000 300.000 - 100000 . 200000 - 300000
Angle Bin [deg] Angle Bin [deg]
PDA U1-Meon[m/s] 5
1.00 f T f
0.8;: 77777777 :r 7777777 i 77777777 i 77777
0.5: --------- ,1 --------------- 7: -----
0.4;L i“ """"""""" ‘E “““
021 --------- 1 ------------ J. -----
0.0? i -ll :
1.2:: ........ 1 ........ ll ....... .= .....
-----
100.000 200.000 300.000
Angle Bin [deg]

Figure 40: Velocity profiles taken through time with seed and water.

60

 

 

 

 

 

 

 

 

 

 

 

    

 

 

 

 

    

 

 

 

 

 

 

 

 

   
   

 

 

 

 

 

 

PDA Ul-M
eon [mls] PDA U1-Meon[m/s]
1.0 .
7’ : 1.0 . v r
0.87 --------- r -------- " I
‘* l 0.87 777777777 r
0.5‘ """"" ‘ """"" 0 I
7L I 0.67 --------- f
0.4‘ ““““ ‘ """" ‘ 1
1 I 0.47 --------- f
._ ........ i. ........ o .
"72.. : | 0.24 t
0.0' ' I ' " ‘ e
.. . 0.07
-0.27 --------- ' -------- 17 ;
[ r l -0.27 --------- r
-0.47 ---------------- " :
e i e i e . ‘ -0.47---------r . A - A .
. 20 . 300.000 ' ‘ ' ' ' ' ‘
100 0T" l B' 0 :00 100.000 200.000 300.000
9 e '"l 99] Angie Bin [deg]
PDA U1-Meon[m/s] 4
1 0 PDA U1-Meon[m/s]
' 7’ : 1.0 f . .
0.87 --------- 7 4 ; ' g
" . 087 --------- e ---------------- 1 -----
0.6‘ """"" t 0 : :
0 i 06' --------- t --------------- 7" -----
0.4? -------- f 'i I I
'i I 04‘ --------- t -------------- Jl -----
0.2" """" If ‘1 r I
‘f I 0.24 --------- t ----------- J. .....
0-0‘ 1 1+ .
" ; 0.0 v‘:~‘w *1 W74 7 .
-0.2‘ """"" r 1* 1 :
‘* : -0.277 77777777 i 7777777777 7i 77777
—0_471 -------- r 1f : j
5 C 5 3 5 ‘ .0‘411. -------- f ........ 1 ..... 1 .....
100.000 200.000 300.000 100:000 7 2005000 7 3005000 7
Angle Bin [deg] ' _ ' ‘
Angle Bin [deg]
PDA UI-Meon [m/s] 5
1.0 I T r
0.87 --------- f
0.67 --------- :~
0.4 --------- i
0.27 777777777 f
0.07 '
70.27 777777777 '1
1t :
0471-7777777: _

 

 

 

     

 

100.000 200.000

300.000
Angle Bin [deg]

Figure 41: Velocity profiles taken through time with water and bubbles.

6]

 

Air Pockets

Up until this point, mass flow calculations were performed assuming plug flow.
As mentioned earlier, this assumption proved to be inaccurate and did not agree well with
cycle-averaged measurements made with the mass balance (over-predicted by 60 %).
Assuming the flow was quasi-steady and assuming a laminar parabolic profile was also
inaccurate, under-predicting by 45%. At this point, it was determined that the flow could
not be assumed as quasi-steady. The equations developed by Brereton [14] and outlined
in Section 1.5 and Appendix A of this report were then employed. After the program
written to perform these calculations was completed and debugged, it was run using
centerline velocity data to find the mass injected per cycle. These results over-predicted
the mass injected by about 30%. It was then proposed that the system might have air in
the injector. The equations used determine the mass flow at the location of the centerline
velocity measurement. In order to find the mass injected by the injector, it was assumed
that what went into the injector must exit. Since the calculations were performed during
the duration of the voltage applied to the injector, air in the injector could allow more fuel
to exit than entered due to compression of the air pocket.

To test this theory, a vacuum pump was connected to the fuel line. The rubber
fiiel line was replaced with a clear nylon line so that air bubbles could be observed. The
water was also dyed for the same reason. The system was also equipped with extra
valves to facilitate the removal of air without removing excessive amounts of liquid.
Once the setup was modified, the pump was turned on. It was immediately obvious that
there was in fact air in the injector, as bubbles came out of the injector. This was viewed

through the quartz tube. It was also clear that air tended to remain in the fuel line. After

62

all the air was removed from the system, new data was collected. The resulting velocity
trace did not look anything like previous traces. It did, however, have a very close
resemblance to the trace of the voltage applied to the injector. A plot of velocity traces
with and without air in the lines is shown on the following page along with the voltage
trace. The data from this run was run through the program to find the mass injected per
cycle. The average outputted by the program was within 6% of the cycle—averaged
measurement from the mass balance. Also, the sharp oscillation at the peak and just after
the velocity returns to zero on the purged injector plot have wave speeds similar to the

pressure wave speeds calculated earlier.

 

 

Injector #1: Centerline Velocity and Voltage Plots
(4-8-04)

.0
or

Centerline Velocity
[mls]
O

 

 

 

 

l
-1 i

‘ 0 50 100 150 200 l
1 Time [ms] :— purged injector l
—~—- unpurged injector !

L [ -——- voltage j

 

 

Figure 42: Velocity plots for Injector #1 with and without air.
After seeing these results, one obvious question arose: How does this air remain

in the fuel injector? It does not appear that this air is simply flushed out of the injector or

diffused into the water, because weeks of testing yielded the same oscillating velocity
profile. Only when a vacuum was attached to the fuel line did the air come out of the
injector. In order to investigate this, a spare Chrysler Hemi injector was milled to reveal
the internal components and passages of the injector. The picture of this injector can be
seen below. Looking at this picture, there appears to be a fairly large crevice. Further
inspection reveals that this is the location where the lower nozzle portion of the injector
joins with the upper portion. This could very well be the location where air pockets are

trapped inside the injector.

 

 

 

 

/ Supply from ECU

 

 

 

 

 

[ Solenoid windings

 

 

 

 

 

 

 

Valve needle

 

 

 

 

 

 

“I

 

 

Pintle 7H?

 

 

Figure 43: Cutaway of Injector #2. Note crevice zoom.

64

It would be very interesting to see how the removal of this air influences the
variability of the fuel injector. This technique does not allow for that comparison
directly. In order to measure the mass injected using centerline velocity measurements, it
was assumed that what goes into the injector comes out. If air is present, this is not the
case as the bubble is capable of compressing. Thus it is impossible to determine the
precise mass injected for comparison with air in the injector using this technique.

In conclusion, air located in the fuel system and/or the injector itself drastically
changes centerline velocity. Since most previous studies measure the fuel spray, these
bubble dynamics have gone largely unnoticed. Removing the bubbles produces an
accurate centerline velocity profile that can then be used as an input to solve more

precisely for the mass injected for each injection event.

65

References

66

REFERENCES

[l] Nixon, M., littp;//www.motorcycleproiect.com/motorcycle/tcxt/iniect.html .
[2] Sparks, L., http:(Velma/thunder.com/fuel‘ib201njccti0119112011istoryhtm .
[3] Woron, W. “Road Test: 1957 Chevrolet Corvette,” Motor Trend, April 2003.

[4] HowStuffVVorkscom educational website, http:/'/auto.liowstuffworks.com/fuel-
iniectionhtm .

[5] Heywood, J. 3., Internal Combustion Engine Fundamentals, MCGraw-Hill
Publishing Company, 1988.

[6] Winterhagen, Johannes, “Siemens VDO Automotive‘s Piezo Technology Helps
Improve Gasoline Engine Fuel Efficiency”, Siemens web article,
http://www.Siemenscom/indexJ511?de rli=null&sdc flags=null&sdc_sectionid7—70
&sdc_secnavid=0&sdc_3dnvlstid-7—&sdctcountryid=08zsdcjmpid7-=O&sdc‘unitid=9
99&sdc_conttvpe==2&sdc contentidrtl l74392&sdc langid=l& 2004.

[7] Stone, R., Introduction to Internal Combustion Engines, SAE International, 2000.

[8] Ismailov, M. M and Schock, H. 1., “Performance Evaluation of a Multi-Burst Rapidly
Operating Secondary Actuator applied to Diesel Injection System,” SAE Technical
Paper 2004-01-0022, 2004.

[9] Ismailov, M. M. and Schock, H. J ., “Quantification of Instantaneous Diesel Flow
Rates in Flow Generated By a Stable and Controllable Multiple Injection System
(ROSA),” SAE Technical Paper 2004-01-0028, 2004.

[10] Kainz, J. L. and Smith, J. C., “Individual Cylinder Fuel Control With 8 Switching
Oxygen Sensor,” SAE Technical Paper 1999-01-0546, 1999.

[l I] Hung, D. L., Dhmiel, D. M., and Markle, L. E., “Application of an Image-based
Diagnostic Technique to Quantify the Fuel Spray Variations in a Direct-Injection
Spark-Ignition Engine,” SAE Technical Paper 2003-01-0062, 2003.

[12] Lefebvre, A., Atomization and Sprays, Taylor and Francis, 1989

[l 3] Hung, D. L., Humphrey, W. A. et al., “A Novel Transient Drop Sizing Technique

for Investigating the Role of Gasoline Injector Sprays in Fuel Mixture
Preparation,” SAE Technical Paper 2004—01-1349, 2004.

67

[l4] Brereton, G. B., “The Interdependence of Friction, Pressure Gradient, and Flow
Rate in Unsteady Laminar Parallel Flows,” Physics of Fluids, Vol. 12:3, March
2000.

[15] Brereton, G.B., Schock, H.J., Rahi, M.A., and Bedford, J.C., “Some Indirect
Techniques for Measuring Instantaneous Flow Rates in Unsteady Duct Flows,”
submitted to Experiments in Fluids, 2004.

[16] Dantec Dynamics Website, http://www.dantccmtcont/"Lsz-T’riiicip/lndcx.lilml .

[17] Ribbens, W. B., Mansour, N.P. et al.,Understanding Automotive Electronics, 6”
Edition, Newnes, 2003

68

   

l11111111111111I