LIBRARY

Michigan State

University

 

 

PLACENRE‘I'URN BOXtomnmmbMMflomyum
TO AVOID FINES Mum on orbdoro duo duo.

 

L

 

Cw
. & \UQ:)~ .} (

l-x,

Wr’ ’ e 312’ if

‘13:!
_

 

 

DATE DUE DATE DUE DATE DUE :I

 

 

 

 

 

:1

 

 

 

 

::J

:‘T—‘i

 

 

 

MSU IcAn Afflnnm Awm/Eqnl Opponmny lea-tum
m

m1

 

 

 

 

 

AN INVESTIGATION OF FLUID FLOW DURING INDUCTION STROKE OF A
WATER ANALOG MODEL OF AN IC ENGINE USING AN INNOVATIVE
OPTICAL VELOCIIVIETRY CONCEPT - LIPA

By

Bernd Stier

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Mechanical Engineering

1 994

ABSTRACT

AN INVESTIGATION OF FLUID FLOW DURING INDUCTION STROKE OF A
WATER ANALOG MODEL OF AN IC ENGINE USING AN INNOVATIVE
OPTICAL VELOCIMETRY CONCEPT .- LIPA
By

Bernd Stier

Optical measurements on an axisymmetrical quartz component engine research model
were made to evaluate the flow field encountered during induction. The measurement
technique is LIPA (Laser Induced Photochemical Anemometry), a non-intrusive
velocimetry concept that provides an investigator of fluid flow with a tool to attain planar
information about three-dimensional velocity and vorticity vectors in a single measurement
step. The goal of this investigation is to further develop this measurement technique and
apply it to study the induction stroke of a water analog model of a four-stroke internal
combustion engine.

The research conducted in the water analog model is a fundamental scientific inquiry
into the flow fields that develop in the induction stroke of an engine at idling engine speeds.
As this is the first investigation of its kind using LIPA technique, our goal has been to
quantify, in a preliminary manner, the flow field features that develop during the intake
stroke. In the process a more comprehensive understanding of the flow field features was
developed, and tied to the quantifications. The study evaluated the flow field of the intake
stroke by estimating fields of velocity and vorticity. On the basis of these data, information
about fluid dynamics during induction at engine speeds of 10, 20, and 30 RPM
(corresponding to 170, 340, and 510 RPM respectively, when air is the flowing medium)
for three different valve lifts was obtained. The overall development of the flow field, its
energy content (kinetic, fluctuation) for the different settings of the engine parameters,
vorticity information, and cyclic variations have been quantified. These have been discussed

in terms of mixing performance.

ACKNOWLEDGMENTS

I would like to take this opportunity to thank the many people whose help and
assistance were invaluable for the realization of this work. The first 'thank you' goes to my
advisor, Prof. Robert E. Falco, for providing me with the initial impulse and the guidance
for this doctoral program. I thank the members of my committee, Profs. Daniel N ocera,
Manoochehr M. Koochesfahani, and Harold Schock, for their fruitful scientific comments
and procurement of insight throughout the project. Thomas Stuecken, Michael
Hetherington, John Brandon, and Colin McKinnon assisted me in many questions of
mechanical, electronical, and computational problems; I am intensely grateful for their
efforts and input.

I thank Michael for his stimulating visits and my fiancee Amy for her warm
encouragement via telephone as support during my solitude in a 'human-deserted'

laboratory.

Finally I would like to express my gratitude for the financial support endowed by the
Ford Motor Company in Dearborn, MI, by NASA Langley Research Center, and by the
Center of Fundamental Material Research (CFMR) at the Michigan State University

without whose financial aid this project would not have been possible.

iii

TABLE of CONTENTS

List of Tables ............................................................................................................... viii

List of Figures .............................................................................................................. ix

List of Abbreviations, Nomenclature, and Symbols ....................................................... xv

I Introduction ............................................................................................................. 1
1.1 History of Research performed in IC Engines and pertinent free

Jet Experiments, and of Applications of LIPA ........................................... 1

1.2 Outline of Investigations performed for this Work ..................................... 10

H Description of LIPA and LIPA'S Implementation in experimental

Envirpnment ............................................................................................................ 12
11.1 LIPA as Velocimetry Technique ................................................................ 13
11.1.1 LIPA's concept ............................................................................. 13
11.1.1.1 DemandtowardsehemistryasgodfatherofLIPA ........................... 14
11.1.1.2 Principle of evaluation of array of velocity and vorticity vectors ....... 14
11.1.2 Phosphorescent characteristics of employed LIPA - chemical ........ 17
11.2 Experimental Apparatus and Method of Data Acquisition .......................... 20
11.2.1 Water analog model of four-stroke 1C engine ............................... 20
11.2.1.1 Refractive index matching of experimental environment .................. 22
11.2.1.2 Recipe for engendering of LIPA - chemical .................................. 24
11.2.2 Aspects of data acquisition facility and its facilitation .................... 25
11.2.2.1 Generation of laser grid - optical assemblage ............................... 25
11.2.2.2 Onhowtoacquiregridsoflaserlinesthatconstitute LIPA-data ...... 28
11.2.2.3 Results from preliminary preparation-activities for series of

LIPA experiments .................................................................... 29
11.2.2.4 Matrix of conducted LIPA experiments ........................................ 30

11.2.3 Estimates for parameters relevant for description of quality of
LIPA - data .................................................................................. 32
11.2.3.1 Delay time ............................................................................. 32
11.2.3.2 Exposm’e time ........................................................................ 34
11.2.3.3 Quality ofLIPA - data .............................................................. 35

iv

111 On Methodplpgy and Accurapy of Data Reduction .................................................. 37
111.1 Illustration of two Approaches to deduce Coordinates of Intersection

Points from LIPA Images, and Pre-Conditioning of these Images .............. 39
111. l . 1 Reconstruction of Intersection Area (RIA) ................................... 4O
111.1.2 Search for Highest Intensity Spot in IA (HIS) .............................. 42
111. 1 .3 Issues concerning the two IP detection approaches and
the decision for one over the other ................................................ 44
111.1.4 Computational manipulation of LIPA images ............................... 47
111.2 Mathematical Formulation of Fluid Dynamic Quantities ............................. 49
111.2.1 Reynolds’ decomposition as basis for a model to separate
sources of velocity fluctuations ..................................................... 49
111.2. 1 . l Formulations employed for ensemble averaging ............................. 50
111.3 Critical Assessment of Accuracy Issues for LIPA ...................................... 52
111.3.1 Overview of encountered sources of error in connection
with the LIPA technique ............................................................... 53
III.3 .1. 1 Movement of intersection point within intersection area
as main source of error ............................................................. 53
111.3.1.21ntersection angle, resolution of LIPA images, and dynamic
rangeasissuesinthediscussionofaccuracy ................................. 56
111.3. 1.3 Miscellaneous sources of error ................................................... 60
111.3.2Actual quality ofLIPA - data ....................................................... 61
[11.3.3 Estimation of errors in plots of velocity vectors ............................ 65

IV The Me Flow Field in the axisymmetric Engine examined by Flow

 

I on ............................................................................................................ 69
IV] Results fiom Application of Photochromic Dye ......................................... 71
1V.l.1Experimental procedure ............................................................... 71

1V. 1.2Flow visualization results .............................................................. 73

IV .2 Results from Application of Fluorescent Dye ............................................ 78
IV.2. lFlow visualization Mode 1 ............................................................ 79
IV.2.2Flow visualization Mode 11 ........................................................... 81

1V .3 Comparison of Flow Visualization Results at two different Engine Speeds. 84

V 'en ' Fin ' in R ch of En 'nes obtained fr m LIPA - Data ..................... 87
VI Description of Induction Stroke Flow Field in Terms of Maps of
Ensemble Averaged Velocity Vectors ........................................................ 89
V. 1.1 Genesis and fluid dynamical behaviour of features of the
flow field in water analog model ................................................... 89
V. 1.2 Profiles of induction angle ............................................................ 94
V2 Discrimination of Flow Fields as enforced by different Parameter
Settings of the Engine ............................................................................... 97
V.2.1 Definition of parameters ............................................................... 97

V.2.2 Information about difierent flow fields in terms of defined
parameters ................................................................................... 99
V23 Discussion of the correlation between intake jet flow and
the fluid dynamics of the intake stroke with special reference

to the area of scrutiny ................................................................. 103
V.2.3.1 Review of background of annulus free jet flow ............................ 104
V23 .2 Application of knowledge about annular fi'ee jet flow to

fluid dynamical aspects in engine ............................................... 105
V.2.3.3 Flow visualization results revisited ........................................... 107
V3 Data from Evaluations of Instantaneous Velocities .................................. 110
V.3.l Maps of instantaneous velocity vectors ....................................... 1 10
V3. 1 .1 Results fi'om individual engine cycle for 20 RPM and
3 mm valve lift ...................................................................... l 1 1
V3. 1.2 Variation of flow field with engine cycle and change in
engine parameters .................................................................. l 14
V32 Time history of a velocity vector - an approach towards

cycle resolved data analysis ......................................................... 116
V3 .2.1 Presentation of velocity profile ................................................. 1 18
V.3.2.2 Notes on cycle resolved analyses .............................................. 120
V.3.2.3 Procedure and results of cycle resolved analysis .......................... 123

V4 Vorticity Contour Maps -- Indicator for Mixing Performance ................... 129
V.4.1 First and second derivative of vorticity distribution in area

of scrutiny ................................................................................... 132
V.4.l.1 Introduction ofmeasures for thefirstandsecondderivative ............ 133
V.4.1.2 Application of measures .......................................................... 135

V.4.2 Energy contained in vortical motion ............................................ 136
V.4.3 Miscellaneous topics involving vorticity distributions .................. 139
V.4.3.1 Axisymmetry ofthe flow field .................................................. 139
V.4.3.2 Vorticity distribution in the EA domain ...................................... 141
V.4.3.3 Influences of crank-angle and engine cycle on vorticity
distributions .......................................................................... 142
V.5 Annulus Intake Jet ................................................................................... 146
V.5.1 EA domain ................................................................................. 147
V.5.l.l 20 RPMenginespeed,3mmvalvelift ....................................... 147
V.5.l.2 30 RPM enginespeed,3mmvalvelift ....................................... 154
V.5.l.3 20 RPM enginespeed,9mmvalvelift ....................................... 155
V5. 1.4 Comparison between investigated cases ..................................... 157
V.5.1.5 Theintake flowinterms ofnon-linear chaoticsystemdynamics ...... 158
V.5.2 Analysis of individual engine cycles ............................................. 160
V.5.2.l 20and30 RPM enginespeed,3mmvalvelift ............................. 161
V.5.2.2 20 RPM engine speed, 9 mm valve lifl ....................................... 164
V.5.2.3 Momentum of the free annular intake jet .................................... 165
V.5.3 Velocity fluctuation within the intake jet ..................................... 166

VI Wk ......................................................................................... 170

Appendix A. ................................................................................................................ 17 9
Generation of Piston Motion ............................................................................ 179
Sealing and Lubrication between Piston Head and Cylinder Wall ...................... 180

Appendix B ................................................................................................................. 182
Issues on refractive Index Matching ................................................................. 182
Optical Issues .................................................................................................. 183

Appmdppg ................................................................................................................. 185
Synchronization Aspects, its Dimculties, and related Uncertainties ................... 185

Appendix D ................................................................................................................. 190
Specification of Hardware ................................................................................ 190

Imager System. ....................................................................................... 190
Image digitizing facility ........................................................................... 191

Appendix E ................................................................................................................. 193
FORTRAN Program ........................................................................................ 193

Bibliggrgaphy ................................................................................................................ 201

Iiigtges ........................................................................................................................ 207

Number Page
11.2.1 209
11.2.2 23
11.2.3 31
111.3.1 66
V3.1 125
V3.2 126
V4.1 143
V5.1 148
V.5.2 158
V5.3 166
V5.4 166

LIST of TABLES

Caption

Engine geometry {Incorporated in Figure 11.2.3}
Indices of refraction for diverse materials

Matrix of experiments performed for the investigation of (a) whole
engine cylinder and (b) intake jet.

Uncertainties in the maps of velocity vectors

RMS fluctuation and intensities in connection with turbulent motion
and cycle-to-cycle variability at a location in the vicinity of the axis
of the intake jet.

RMS fluctuation and intensities in connection with turbulent motion
and cycle-to-cycle variability at a location in the area of scrutiny.

Comparison of S1, SG, and VE between two difi‘erent crank-angle
positions for individual engine cycle as well as averaged over 42 ECS.

Assignment of individual velocity vectors to flow structures

Occurrence of specific events in profile of induction angle; the
numbers indicate the degree of crank-angle.

Normalized momentum as a function of engine parameter setting.

Comparison of ratios as encountered for momentum and for kinetic
energy.

viii

Number

11.1.1

11.1.2

11.1.3

11.2.1

11.2.2

11.2.3

11.2.4

11.2.5

Page
207

208

208

209

210

211

212

213

LIST of FIGURES

Caption

Evaluation of velocity and vorticity vectors from convection of
investigation grid with the fluid flow.

Degradation of intensity of radiation emitted fiom investigation grid
with time; influence of bleaching and shelf-life.

Results of attempt for refinement of used LIPA - chemical by
adding supplements of the alcohol component t-BuOH.

(a) Schematic of the water analog model showing the essential
quartz pieces [included in (a) are the measures of the engine
geometry]. (b) The quartz cylinder, head, piston, and valve along
with diverse other components of the engine configuration. Also
shown are the four metal 'riders' for the four sides of the outer tank
- used for the calibration of axisymmetry -, that hold the fluid
reservoir at the right distances fiom the outer tank after the collar
established axisymmetry.

Apparatus used to create the grid of laser lines in the quartz engine.
(a) Schematic showing the paths of light and grid location.

(b) Photograph of the setup, with the image acquisition system in
the background.

Grids of laser lines for investigation of (a) most of the cylinder
volume and (b) intake jet. For both grids a typical intersection area
is shown magnified and scrutinized.

(a) Schematic of the timing system; (b) Photo of the image
intensified CCD camera in data taking position; (c) Schematic of
the overall data acquisition system.

Black and white images holding LIPA grids generated by the
optical system. (a) The reference grid (light enters the cylinder wall
on the right, and can be seen coming out of the other side of the
quartz valve - near the center of the picture).

(b - e) Four independent photos showing distortions in the laser
grid larger than useful for data acquisition (long delay time).
Engine speed is 20 RPM.

Number Page

11.2.6

11.2.7

111.1.1
111.1.2

111.1.3

111.1.4

111.2.1
111.3.1

111.3.2

111.3.3

214

215

216

216

217

217

218

219
220

220

221

222

Caption

Color images of LIPA grids showing a sequence of four data
photos and the calibration grid used with them from preliminary
tests. (a) Calibration grid. (b - e) Four successive data frames; these
frames also reveal the flow pattern of intake vortices because

the LIPA - chemical irradiated in the previous laser pulses is still
glowing after it has been warped around the vortices (20 RPM).

Examples of the undistorted (a) and distorted (b, c) LIPA grids
acquired with the IIL/camcorder combination. To optimize the
CCD format, the camcorder was turned 90° so that the valve is on
the right side of the image, and the cylinder wall on the bottom.

Comparison between results of inference of coordinates of
illuminated 'particles' from LIPA images deduced by hand (cross,
square, and diamond) and by machine (dotted lines).

Nomenclature for intersection area

Weight distribution used for the evaluation of intensity distribution
in an IA according to which 'HIS' chooses the highest intensity Spot.

Comparison between the results of applications of 'IAR' and ’HIS'
to four intersection area.

Depiction of an intersection area in terms of pixel intensity;

(a) shows the pixel intensity distribution for an 1A as it is delivered
at the output of the CCD camera, (b) exhibits the same distribution
after the image was conditioned.

Methodology of ensemble averaging

Efl‘ect of change in number of pixel per image and extent of object
inimageinx-andz-direction.

Velocity fluctuations exhibited by reference laser grids employed in
investigation of (a) whole engine cylinder and (b) the intake jet as
close up.

Uncertainty about determination of initial positions in reference
grids for 20 RPM engine speed and 3 mm valve lift as encountered
for the close-up investigation grid; (a) in z - direction,

(b) in x - direction.

Flow visualization of the intake stroke using a photochromic dye
(color change upon irradiation). The piston head is visible near the
bottom of each photo, and the cylinder head and the intake valve
are visible nera the top. It is interesting to note that a spiral up-flow
of fluid that has reached the piston head exists throughout the
sequence.

Number Page
IV.2.1 223
IV.2.2 224
V. 1.1 226
V. 1.2 228
V. 1 .3 229
V2.1 230
V2.2 231
V2.3 231
V2.4 232
V2.5 232
V3.1 233
V3.2 236

Caption

Flow visualization of the intake stroke using a fluorescent dye in the
reservoir above the valve (MODE 1). A laser sheet of 308 nm light
is slicing the cylinder both above and below the valve (the
impurities in the quartz of the piston head make it clearly visible).
Results are at 20 RPM and 3 mm valve lift. The numbers are in
degrees crank-angle. By 40° CA the cylinder is filled with dye and
the experiment is stopped. Some air bubbles have formed from air
in the water and partly obscure the view.

Flow visualization of the induction stroke, where the inducted fluid
is dye-free and the fluid in the cylinder has been dyed (MODE 11).
Engine speed is 20 RPM and valve lift is 3 mm. The LIPA results
for CAS 32°, 52°, and 92° (section V5. 1) are superimposed on the
visualizations.

Maps of ensemble averaged velocity vectors (over 42 ECS) at
selected crank-angle positions for 20 RPM engine speed and 3 mm
valve lift illustrating the development of intake flow.

Ensemble averaged velocity vectors at 132° crank-angle for all
nine settings of the engine parameters.

Profiles of induction angle in dependence of engine speed and valve lift.

Specific kinetic energy inherent in area of scrutiny at 132° crank-
angle as a function of valve lift and engine speed;
(a) KE versus valve lift, (b) KE versus engine speed.

Energy content in velocity fluctuations for z - direction inherent in
area of scrutiny at 132° crank-angle in dependence of valve lifi and
engine speed.

Comparison of angular momentum in area of scrutiny for difl‘erent
settings of valve lift and engine speed at 132° crank-angle.

Mechanism of vortex ingestion

Schematic decomposition of an axisyrnmetric jet flow field into two
characteristic regions of flow. [Sourcez Yule 1978]

Maps of instantaneous velocity vectors for one engine cycle at
selected crank-angle positions for 20 RPM engine speed and 3 mm
valve lift illustrating the development of intake flow.

Maps of instantaneous velocity vectors for three consecutive engine
cycles at 20 RPM and 3 mm VL. Crank-angle 132°,
engine cycles #27, 28, and 29. (Time delay: 0.64 ms)

Number Page

V.3.3

V.3.4

V.3.5

V.3.6

V.3.7

V3.8

V.4.]

V.4.2 -
V.4.10

V.4.l l

V.4.12

V.4.13

V.4. 14

237

238

239

239

240

241

242

243
25 l

252

252

253

254

Caption

Maps of instantaneous velocity vectors for three consecutive engine
cycles at 30 RPM and 3 mm VL. Crank-angle 132°,
engine cycles #38, 39, and 40. (Time delay: 0.32 ms)

Maps of instantaneous velocity vectors for three consecutive engine
cycles at 10 RPM and 9 mm VL. Crank-angle 132°,
engine cycles #15, 16, and 17. (Time delay: 0.96 ms)

Velocity profile for 10 RPM engine speed and 3 mm valve lifi;
location is in the vicinity of the axis of the intake jet.

Velocity profile for 10 RPM engine speed and 3 mm valve lifi;
location is in the area of scrutiny.

Profiles of the z - component of the instantaneous velocity vector
located in the region governed by the intake jet for eight different
engine cycles; the plots also contain linear curve fits for the profiles.

Profiles of the z - component of the instantaneous velocity vector
located in the area of scrutiny for eight difi‘erent engine cycles; the
plots also contain linear curve fits for the profiles.

Investigation scheme for vorticity distributions. Scheme includes
hexagonal pattern formed by vorticity vectors from area of scrutiny.
The 116 vorticity vectors are derived fiom the 74 intersections in
the grid of laser lines exhibited in Figure 11.2.3a.

Difl‘erent perspectives showing the distribution of the circumferential
component of the vorticity vector. The area covered is about
50x50 mm. {For nine data sets}

Evaluation of slope in circumferential vorticity distribution in area
of scrutiny as a firnction of valve lift and engine speed at

132° crank-angle. Error bars are evaluated in the same manner as
for Figures V.2. la and V.2.2.

Sign switches of slope in vorticity distribution in area of scrutiny as
a function of valve lift and engine speed at 132° crank-angle. Error
bars are evaluated in the same manner as for

Figures V.2.1a and V.2.2. Error bars for 20 RPM are representative
for other two speeds.

Comparison of kinetic and vortical energy stored in area of scrutiny
as a function of valve lift and engine speed at 132° CA; (a) denoting
magnitudes, (b) after normalization by values for 9 mm VL.

Illustration of boundary efl'ect on the fluid dynamics of the main
toroidal recirculation pattern.

Number

V.4.15

V.4.16

V.4.17

V.4.18

V.4. 19

V51

V.5.2

V.5.3

V.5.4

V.5.5

V.5.6

V.5.7

Page
255

256

257

258

259

260

261

261

262 _

264

264

265

Caption

Comparison of vorticity distributions for 20 RPM engine speed and
3 mm VL between vorticity results from ensemble averaged and
from instantaneous velocity field, and also comparison between
vorticity distribution fi‘om CAS 132° and 180° (end of induction
stroke) within one engine cycle (EC #27). Investigated area is
about 50x50 mm.

Sequence of vorticity data from three consecutive engine cycles for
20 RPM engine speed and 3 mm valve lift at 132° crank-angle.
Investigated area is about 50x50 mm.

Slope determined from the distribution of vorticity in the area of
scrutiny at crank-angle 132° for individual engine cycles for 20 RPM
engine speed & 3 mm VL.

Sequence of vorticity distributions for one engine cycle (EC #17)
for 10 RPM engine speed and 9 mm VL around midstroke at crank-
angles 88°, 90°, and 92°. Investigated area is about 50x50 mm.

Sequence of vorticity distributions for one engine cycle (BC #17)
for 10 RPM engine speed and 9 mm VL at crank-angles 150°, 152°,
and 154°. Investigated area is about 50x50 mm.

EA velocity vectors depicting the intake jet as close-up for 20 RPM
and 3 mm VL; time delay: 0.34 ms.

Fields of ensemble averaged velocity vectors depicting the intake jet
as close-up for 30 RPM engine speed and 3 mm valve lift at
selected crank-angle positions; time delay: 0.17 ms.

Fields of ensemble averaged velocity vectors depicting the intake jet
as close-up for 20 RPM engine speed and 9 mm valve lift at
selected crank-angle positions; time delay: 0.51 ms.

Maps of instantaneous velocity vectors from one individual engine
cycle illustrating the dynamics of the intake jet at selected CAS for
20 RPM engine Speed and 3 mm valve lift; time delay: 0.34 ms.

Maps of instantaneous velocity vectors from one individual engine
cycle illustrating the dynamics of the intake jet at selected CAS for
30 RPM engine speed and 3 mm valve lift; time delay: 0.17 ms.

Maps of instantaneous velocity vectors fi'om one individual engine
cycle illustrating the dynamics of the intake jet at selected CAS for
20 RPM engine speed and 9 mm valve lift; time delay: 0.51 ms.

Velocity fluctuations for (a) RMS fluctuations and (b) fluctuation
intensities at 132° CA for 20 and 30 RPM engine speed and 3 and
9 mm valve lift.

Number Page Caption

A1 266 (a) The difi‘erences between the piston motion of a normal four-stroke
engine and the water analog model used in this work. The differences,
required because of incompressibility of water, are also shown in phase
space. (b) A picture of the cam-shalt used to create the piston motion.

Cl 267 Drawing of circuit of 'electronic signal valve'

C2 268 Virtual synchronization between timing on engine side (optical
encoder) and continuously running camcorder on camera Side.

xiv

LIST of ABBREVIATIONS and NOMENCLATURE

BDC bottom dead center

Ber Bromonaphthalene

CA crank-angle

CCD charge coupled device

CD Glucosyl-fl-cyclodextrin

CFD computational fluid dynamics

CtC cycle-to—cycle (variability)

DHN Decahydronaphthalene

EA abbreviation referring to data in the context of ensemble averaged results
EC engine cycle

HIS Highest Intensity Spot; algorithm for deduction of positions of illuminated particles
IA intersection area

IAR Reconstruction of IA; algorithm for deduction of positions of illuminated particles
1C internal combustion

11L image intensifier lens

1P intersection point

KSCN Potassium Thiocyanate

LDA laser Doppler anemometry

LIPA Laser Induced Photochemical Anemometry

M mol per liter

RMS root mean square

RPM revolutions per minute

t-BuOH tert-butyl alcohol

TDC top dead center

TRP toroidal recirculation pattern

UV ultra violet

VL valve lift

2D two-dimensional

3D three-dimensional

bufl‘er zone An area on top of the piston head in which no consistent flow pattern

exists.

induction angle Angle under which the intake jet enters the engine cylinder.
interface zone A region in which the intake jet and the main recirculation pattern

interact.

recirculativity A measure to evaluate how much structure a flow pattern contains and

by how much fluid flow is recirculated within this flow pattern.

XV

E

:meeevga

5' E1“;
PT

Hows 2.9.5» rt

gggcifma

§O< <.<l<l

LIST of SYMBOLS

angular momentum over certain area

width of nozzle

measure for velocity fluctuation in connection with cyclic variations
difl‘usion coeficient

spatial distance between two circumferential components of the vorticity vector
infinitesimal area

infinitesimal line segment

specific energy in conjunction with velocity fluctuations

running indices

momentum flux

specific kinetic energy

characteristic length

characteristic length of the order of laser line distortion

length of intake jet axis between valve and cylinder wall
symbol for number of events

index of refiaction

normal (perpendicular) vector to area

quality factor

radius

Reynolds number

anhydrous solute weight per cent

Schmidt number

average of slope encountered in vorticity distribution

Sign changes in first derivative applied to vorticity distribution
time between the acquisition of a distorted grid and the application of the laser grid
time for which scene of investigation is exposed to camera
measure for velocity fluctuation in connection with turbulence
local velocity vector

velocity component in z - direction

x - and z - component of a velocity vector

characteristic velocity

velocity at centerline of a fi'eestream jet

velocity of inner jet in coaxial jet

velocity of outer jet in coaxial jet

velocity vector

mean velocity

velocity fluctuation

velocity at annular intake jet centerline

specific vortical energy

xvi

x,y.z

e.a>gw‘01:<.:s*.:s'-IQ<1
n

8
0

mean velocity within an engine cycle evaluated by curve fit

reference coordinate system of engine cylinder

derivative operator

induction angle; angle under which the annulus intake jet enters the engine cylinder
circulation

self-preservation parameter for free jets

normalized parameter of distance C divided by the width of the free jet
kinematic viscosity

angle

density

coordinates in streamwise and lateral directions of the intake jet
circumferential direction of engine reference frame

vorticity vector

vorticity component in circumferential direction

xvii

Chapter I
INTRODUCTION

The underlying scientific objective of this work is essentially two-fold. The two
'projects' consist of the application of Laser Induced Photochemical Anemometry (LIPA)
- an innovative optical velocimetry concept - to a relevant and important fluid dynamics
problem (induction fluid flow within an engine cylinder) in order to demonstrate the
technique's potential and usefirlness in the field of fluid dynamics in general, and the
attempt of advances in engine research - especially for four-stroke internal combustion
(1C) engines - by describing phenomenons of engine performance on fluid dynamical
grounds based on the data acquired by LIPA in particular. The four-stroke IC engine will
be represented by a motored axisymmetric water analog model. Water is the required

medium for the employment of LIPA in this application.

1.1 History of Research performed in IC Eng;r_r° es and pertinent free Jet
Experiments, and of Applications of LIPA

The engine research community is constantly seeking further optimization of an engine's
performance. The latest round of discussions about improvements in engine performance
on the verge of the completion of the millennium was partly encouraged by the US Vice
president's ecological imperatives to produce 21St century automotive vehicles that achieve
higher engine eficiency accompanied by a lower level of pollution than their ancestors. To
accomplish this honorable goal by pushing for advances in engine performance the engine
itself cannot be regarded a black box with respect to fluid flow generated within an engine

cycle anymore; instead, the physics of fluid flow within the engine cylinder have to be
investigated.

An approach favored by the engine research community presently in the investigation of
fluid dynamics experienced in an engine is to apply a hybrid experiment/calculation scheme
where computational fluid dynamics (CFD) are employed to the compression process --
where the flow is relatively less Complex and CFD can yield higher accuracy -- to evaluate
fluid dynamical quantities (Trigui et.al. 1994a), and seek initial conditions for the
compression stroke fiom experimental applications -- such as PTV/PIV (Trigui et.al.
1994b) or LDA - to the induction stroke which is computationally more diflicult to grasp.

The current study introduces a useful new optical velocimetry tool (LIPA) which is
based on the phosphorescence of the flowing medium to examine the flow field in an
engine -- particularly the induction stroke - and to derive important parameters to
describe, e.g., mixing of air/fire] phases during induction. The knowledge about these

parameters enables engine designers to firmish improvements of engine performance.

There exist investigations that examined fluid flow in axisymmetrical assemblies - the
most simplified and abstract realization of an engine - of four-stroke engines. They
performed flow visualization (Bkchian and Hoult 1979, Narnazian et.al. 1980, Regan eta].
1987, Eaton and Reynolds 1989, and Stier and Falco 1993) as well as quantitative Studies
using Hot-wire and laser Doppler anemometry (LDA) techniques. From these works it
becomes clear that the setup of the flow field consists essentially of the incoming annulus
jet and two toroidal recirculation patterns one of which occupies most of the engine
cylinder volume; the second rotates in the upper corner of the engine cylinder between
cylinder wall and chamber head. At the Imperial College of Sciences and Technology in
London an axisymmetrical engine configuration - which is also a water analog model as in
this study - has been the object of thorough and intense study. Arcoumanis et.al. (1982a),
Arcoumanis et.al. (1982b), Bicen et.al. (1985), Vafidis and Whitelaw (1986),

Arcoumanis eta]. (1987), and Arcoumanis and Whitelaw (1987) evaluated fluid flow in
the model cylinder employing LDA at difi‘erent locations in the engine chamber, and flow
visualization, determining velocity and turbulence intensity distributions in selected planes

as well as streamline plots.

Hot-wire and LDA technologies have been widely and exclusively employed in engine
science up to recently for the quantification of flow fields (Lancaster 1976, Witze 1977,
Rask 1979, Lee et.al. 1993, and Yoo 1994). These technologies are single probe
measurement concepts, thus, their application provides velocity information at only one
location in the flow field at a time. In order to extract planar information of the flow field,
such as vorticity vectors and length scales, fi'om velocity information one has to assume
homogeneity and isotropy of the flow field. This assumption has been shown not to be
true. Several investigators proved that the intake stroke and the beginning of the
compression stroke are clearly inhomogeneous and anisotropic (Semenov 1963, Winsor
and Patterson 1973, and “Witze 1977). Recently, velocimetry approaches have become
multi-point in nature with the implementation of techniques such as PIV and PTV (particle
imaging and tracking ideas). LIPA, also a planar velocimetry approach, promises to be a
very useful and important tool to investigate the physics of flowing fluid, as, for instance,

in the case of this study during the intake stroke of a reciprocating engine.

Since water is the flowing medium in this application and water is incompressible, fluid
dynamical issues can be explored only during the induction stroke. Nonetheless, previous
results Show that the induction stroke is an important part during an engine's cycle. The
induction stroke sets up the flow field for the subsequent compression stroke which in turn
provides the initial flow situation for the combustion process of the cycle. The induction
stroke has, concerning mixing of air and fuel for example, a very direct influence on
combustion performance since most of the mixing process takes place during intake.

Furthermore, there has been evidence for a profound correlation between the

establishment of fluid dynamic quantities during intake and compression, and combustion
performance. In particular, Witze et.al. (1984) showed that an increased level of
turbulence intensity exhibited by the flow field is favorable for the quality of combustion.
Optimization of the combustion process is the ultimate goal and crucial step in the attempt

for improvement of engine performance.

The complexity of the flow field experienced during induction requires the flow field to
be dissected into fluid dynamically meaningful elements, e. g. vortex structures and intake
jet, which have to be identified and then analyzed separately. Investigations of singled out
flow structures and their effect on engine performance have been conducted in numerous
investigations examining flow structures like swirl - a rotational flow about the cylinder
axis - and tumble - rotational motion about an axis perpendicular to the cylinder axis --
which are generated during induction by specially designed boundary conditions. Both
experimental as well as computational studies have been performed. AS it turns out,
existence of these flow patterns improved engine combustion. The effects of swirl have
been investigated by Kent et.al. (1989), Henriot et.al. (1989), Shack and Reynolds (1991),
Moriyoshi et.al. (1991), and A'r‘ta et.al. (1991). The effects of tumble on engine
performance have been studied by Kent et.al. (1989), Henriot et.al. (1989), Omori et.al.
(1991), and Hadded and Denbratt (1991).

Complex flow structures such as swirl and tumble, whose characteristic lengths scale
with the radius of the engine cylinder, provide high stratification of the flow field and store
and preserve (angular) momentum. Also, transport is mostly done through these large
scales. By stratifying the flow, swirl and tumble inhibit mixing mechanisms that would
intermingle the difi‘erent phases inherent in the flow field -- firel/air -- thoroughly. In
breaking up, these structures provide the flow field with high turbulence intensities which
are preferable for enhanced mixing and subsequently for improved combustion

performance. A compromise must be found in connection with the timing of the break-up

of these large scale structures because the sooner break-up occurs the more
homogeneously the flow field will be mixed at the time of combustion. The later swirl or
tumble breaks up, the higher will be the degree of turbulent intensity by the time of
combustion; then again, the later they break up the larger are the length scales of flow

structures prior to combustion.

When employing Reynolds' decomposition method to an instantaneous velocity vector
(see Tennekes and Lumley 1972 for a modern account) to take into consideration the
turbulent character of the flow field encountered during engine performance the mean
velocity in the decomposition is usually evaluated by an ensemble averaging technique.
The velocity fluctuation part of the decomposition cannot, for engines, be attributed
directly to turbulence inherent in the flow field. The engine process being periodic, more
precisely reciprocating, introduces velocity fluctuations that are due to variations in bulk
flow velocity fi'om engine cycle to engine cycle which is termed in engine science
temrinology 'cycle-to-cycle variability'. Separation of velocity fluctuations due to cycle-to-
cycle variability fi'om those due to turbulent flow within an engine cycle has been
attempted in many studies. A review on the most important techniques has been provided
by Catania and Mittica (1989). A summary of the review is furnished in section V.3.2.2 of
this work. A simplistic model to separate the two sources responsible for velocity
fluctuations was assumed in this investigation and its implementation as well as results on

the basis of the procedure are presented in section V.3.2.3.

In the age of exploding computer power it becomes very attractive to attempt the
prediction of flow fields in the field of fluid dynamics using computational abilities
(computational fluid dynamics, CFD). Naturally, CFD has been employed in engine
research Some interesting papers on numerically simulated flows in engine science have
been published. Henriot et.al. (1989), Moriyoshi et.al. (1991), and A'I'ta et.al. (1991), for

instance, undertook an efl‘ort to model swirling motion in an engine. Gosman et.al. (1984)

and Ai’ta et.al. (1990) calculated three-dimensional air motion and port-valve-cylinder
flow, respectively. Recall fi'om the first paragraph in section 1.1 that Trigui et.al. (1994a)
most recently applied CFD to the combustion stroke of an engine.

Unreliable and inaccurate initial as well as boundary conditions are partially responsible
for difi‘erences between calculated and experimental results. Furthermore, calculated
results diverge fi'om experimental ones with time or crank-angle as the in-cylinder flow
develops. An attempt has to be made to provide accurate initial and boundary conditions.

This investigation will yield some for an axisymmetrical engine.

The understanding of fluid dynamics of the annulus jet that enters the engine cylinder
through the valve is crucial for the comprehension of the fluid dynamics exhibited by the
entire flow field in an engine cylinder during induction. After studying the flow field
development during the intake phase of the water analog model at three difl‘erent valve
lifts and constant engine speed on the basis of fields of velocity vectors provided by LIPA

the importance of the intake jet became obvious.

A fi'ee jet in general, alter leaving the geometrical constraint and entering the ambient
fluid, incorporates basically two regions which can be described as transitional flow' and
'turbulent flow' (Yule 1978, see Figure V.2.5). At the beginning of the transitional flow,
first order instabilities lead to waviness of the jet flow immediately after it leaves the
geometrical constraint (this is not valid for very low Reynolds numbers, however, valid for
the range of Reynolds numbers covered in this study). These first order instabilities are
followed spatially by a roll-up of the Shear layer of the jet forming a coherent vortex ring
with simultaneous occurrence of entrainment of ambient fluid. At the end of the
transitional flow the coherent vortex ring breaks up into small scale vortices which are
embedded in large scale eddies. The core of the jet which exhibits potential flow in the
transitional jet region breaks up as well. From then on the flow is turbulent.

Yule (1978) examined an axisymmetric jet. He found that the length of the transitional
flow is Reynolds number dependent and that for a Reynolds number of about 5000 --
based on the diameter of the nozzle and the velocity of the flow in the nozzle - the break-
up of vortex rings occurs after four diameters of the nozzle. Dahm et.al. (1992) performed
flow visualization on coaxial flows. For one parameter set examined, the velocity of the
outer jet was significantly higher than the velocity of the inner jet and the velocity jump
across the inner mixing layer (between inner and outer jet) was of roughly the same
magnitude as the jump across the outer layer (between outer jet and quiescent ambient
fluid). This is similar to what is encountered in the case of the incoming jet of an engine
and, indeed, the street of vortex rings created in the transitional phase, as visible in the
visualization results by Dahm et.al., exhibit strong similarity to vortex streets visualized by
experiments within this work. Dahm et.al. found that the potential core of the transitional
flow ends abruptly (for the parameter setting given above) at about 7 widths of the annular
orifice - (do-di)/2 - downstrearn. The variation of the length of transitional flow with
annular orifice turns out to be an important feature in the overall development of the flow
field encountered in engine research during induction (section V2.3). Champagne and
Wygnanski (1971) also investigated coaxial jets employing Hot-wire technology; they
found the length of the external (outer) potential core of the jet to be approximately 8
times the width of the annular orifice and assert that the length appears to be independent
of the velocity ratio between external and internal jet.

Ko and Chan (1978) analyzed an annular flow, i.e. the velocity of the internal jet of a
coaxial jet is zero, using Hot-wire probes. They named the transitional jet region 'initial
merging zone'. In their study the position at which the annular potential core ends was
found at about 6 widths of the annular orifice downstream and the investigators infer fiom
their data that the length is independent of the conditions in the center of the annular jet.
Furthermore, they cite several earlier works in which annular and two—dimensional dual jet

flows had been studied. Ko and Chan extract fi’om information given in these publications
the positions at which reattachment of the outer jets occur, i.e. the merging of two
maxima in the velocity profile to only one maximum, and plot the results in a diagram of
'Distance to Reattachment' vs. 'Diameter Ratio' of orifice. A conclusion from this plot can
be drawn stating that reattachment length - and thus the length of the potential core
which is arguably proportional to the reattachment length - Scales with the width of the
annular orifice. In another study by Chan and KO (1978) on annular jets the investigators
observed that the annular-jet results agree in connection with overall pressure intensity
very well with the results for a single jet and coaxial jets within the first four-and-a-half
widths of the annular nozzle downstream fiom the orifice. They suggest that the jet
vortices are not so easily affected during their formation and during their initial
propagation downstream.

Rockwell and Niccolls (1972) examined planar, two-dimensional jets by applying flow
visualization. Although Rockwell and Niccolls do not explicitly state the length for which
a potential core exists in the jet flow in their application, their pictures depicting the planar
jet indicate that the length of the potential core - and thus the length of the transitional
flow - is longer than 6 widths of the annular orifice. They outline, however, that any
study of the 'breaking length' is relatively meaningless without careful specification of the
upstream spectral history of the confined flow. Furthermore, they found that the region
where vortices are defined moves downstream with decreasing values of the Reynolds
number. Also for two-dimensional jets, Miller and Comings (1957) distinguished a
'transition flow region' fiom the 'fully turbulent region' where the transition flow region

extends from the nozzle to a location 7 times the width of the nozzle downstream.

Experimental information on fi'ee jets (White 1991) suggests that the mean-velocity
profile for the velocity component in axial direction is self-preserving beyond about twenty

jet diameters (x/d>20). The mean-velocity U of the profile in a cross-sectional plane can

be, after establishment of self-preservation, evaluated at any point by U=Usf(n) where Us
is the velocity at the centerline and f is a function of n according to fisechzm) with
n=7.67y/x. It should be noted that the beginning of self-preservation of the mean-velocity
profile coincides with the disappearance of the core of irrotational flow, and that it can be
observed fi'om experimental results (flow visualization) that the existence of the potential
core is terminated by the break-up of vortex structures. The observation of the termination
of the potential core establishes a bridge between analytical and flow visualization results.

The fact that the transitional length scales with the width of the orifice for fixed
Reynolds numbers leads to the assumption that the transitional length of the incoming jet
in the model engine of this study difl‘ers with valve lift. In section V.2.3 an attempt is made
to link fluid dynamical behaviour exhibited by the flow field during induction to the length
of the transitional flow experienced by the intake jet.

The optical velocimetry technique employed in this investigation (LIPA) was
introduced and developed at the 'Turbulence Structure Laboratory‘ at Michigan State
University by Professor RE. Falco. Development of the method and implementation of
LIPA to study a model of a four-stroke engine will be discussed in detail in chapter 11. So
far, LIPA has been applied in several research projects at the 'Turbulence Structure
Laboratory'. Studies involving LIPA as a measurement technique to evaluate flow fields in
terms of fields of velocity vectors were performed for measurements in a Stokes' layer and
of the circulation of a vortex ring (F alco and Chu 1987), evaluations of vortex structures
in connection with an airfoil (F alco et.al. 1988), determination of the skin fiiction on a
compliant surface (Falco and Chu 1988), a nozzle flow at supersonic velocities
(F alco et.al. 1989), quantification of the flow field of a two-stroke engine (Hilbert 1991,
Hilbert and Falco 1991), examinations of a subsonic jet (Gieselmann 1991), studies of
flow properties of wet solids (F alco and Nocera 1991), and evaluation of the flow of a

two-phase fluid in a vertical pipe (Buchner 1992). For all these investigations, maps of

10

velocity and vorticity information were derived fiom the flow field of the flowing medium.
Experiments were conducted using three difi‘erent media as basis for the working fluid
which served as vehicle for the phosphorescent dye: Kerosene, Nitrogen, and water. LIPA

was extensively used as photochromic visualization technique by Chu (1987).

1.2 Outline of Investigations performed for this Work

One purpose of this work was to apply LIPA to the model of a four-stroke 1C engine in
order to gain new qualitative and quantitative understanding of the fluid dynamics of the
engine's intake flow. This application of LIPA was to provide fields of velocity and
vorticity vectors which will, among other goals, answer questions about mixing
performance. We also hoped to establish an alternative to conventional processes of
examination of an induction flow, e.g. LDA, PIV techniques. Preliminary LIPA results
have been published by Stier and Falco (1994).

Investigations were performed in an effort to study in single measurement steps the
flow field of the induction stroke in the entire engine cylinder as well as the annular free
intake jet. For the former project, the influence of the variation of valve lift are examined
in the water analog model at three difi‘erent engine speeds in a range of 10 to 30 RPM
which correspond to 170 to 510 RPM, respectively, for air. For the second project, results
will be presented for two valve lifts (3 and 9 mm) at two engine speeds (20 and 30 RPM;
corresponding to 340 and 510 RPM, respectively, in air). Fields of instantaneous as well
as ensemble averaged velocity vectors will illustrate the development of the flow field
during induction. Vorticity distributions are generated fiom the velocity fields which will
be used to examine mixing performances. Several parameters are defined in this work
which attempt to analyze flow fields stemming fi'om different settings of the engine's

parameters valve lift and engine speed.

11

As mentioned in section 1.1 a simple approach offers a separation of velocity
fluctuations due to cycle-to—cycle variations and turbulent motion inherent in the flow
field. RMS fluctuations as well as turbulent intensities are presented.

To grasp the fluid dynamics in a qualitative approach flow visualization experiments
were performed. Results of the unfolding of the flow field during intake will be given for
1.5 and 20 RPM engine speed at 3 mm valve lift.

As for the velocimetry technique LIPA itself, several requirements for this technique
had to be established prior to its employment. These included finding optimal quantities
for the constituents of the phosphorescent dye, writing a grid of laser lines into the flow
field, and implementing algorithms to reduce data automatically. Only vague sketch-book

ideas existed on the latter issue prior to this application of LIPA.

The strategy for the generation of fluid dynamical results on engine flow within this
work was to, within the available timefrarne, collect as much information as possible on
the flow field encountered during intake of the axisymmetric model, to draw conclusions
on the basis of this information, and to participate in the overall attempt of painting a
picture that exposes fluid dynamics encountered during induction.

The study will, however, barely go into depth at any point with respect to aspects of
fluid dynamics. In general, the author had to be satisfied with scientifically scratching the
surface of topics relevant within engine science (e.g. mixing and cycle resolved analysis).
However, going into depth was not the prime goal of this work. The goal was to establish
the general mode showing the viability and promise of engine research using LIPA With
this work in hand as basis, a researcher can concentrate on special issues in the study of
the induction stroke of this water analog or any geometrically more challenging model and
apply suitable LIPA configurations for designed tasks.

Chapter 11
DESCRIPTION OF LIPA AND LIPA'S IMPLEMENTATION
1N EXPERIMENTAL ENVIRONMENT

Velocimetry techniques based on the photoluminescence of the working fluid have been
experimented with more intensively after laser power has been made available as a strong
source of energy. Popovich and Hummel (1967) first suggested the idea of
photoluminescent flow visualization and quantification. At the 'Turbulence Structure
Laboratory' of Michigan State University various studies have been released employing
LIPA (Laser Induced Photochemical Anemometry), a velocimetry concept based on a
phosphorescent working fluid. These studies investigated, for example, vortices in
connection with an airfoil (F alco et.al. 1988), the flow field of a two-stroke engine
(Hilbert 1991, Hilbert and F alco 1991), and the flow of a two-phase fluid in a vertical pipe
(Buchner 1992), to repeat three of the projects that were outlined in the 'Introduction' of
chapter I. Just recently, several investigations involving LIPA and LIPA - related
velocimetry approaches at other universities (Chu et.al. 1991, Yurechko and Ryazantsev
1991, Lempert et.al. 1993) have documented the usefirlness of this measurement concept
in laboratories around the world. The techniques taking advantage of the
photoluminescence of the convecting fluid provide planar - and it is possible to extent the
technique to the third dimension - instantaneous velocity and vorticity information about

a flow field.

In this work LIPA has been applied to the induction stroke of a water analog model of
a four-stroke internal combustion engine in order to yield two-dimensional flow fields in
terms of maps of velocity vectors. From these, vorticity distributions will be inferred

immediately. The purpose of this research in the field of engine science employing LIPA is
12

13

- as outlined in the introduction - to demonstrate the potential of and promote LIPA in a
practical application of LIPA in a discipline of fluid dynamics, and to, among other issues,
illustrate cycle-to-cycle variability of the engine's reciprocating process and to undertake
an attempt to attain information on mixing performance during induction in order to

achieve advances in engine research.

II.1 LIPA as Velocimer Technigpe

11,1,1 LIPA'S conc_ept

LIPA is a fairly new velocimetry concept which takes advantage of phosphorescent
properties of the working fluid. It is optical, and thus non-intrusive, in that it is based on
the excitation of molecules - which exhibit photoluminescence, or photochromism (color
change) upon stimulation by an energy input - in the fluid whose flow field is to be
examined. LIPA'S underlying idea for the evaluation of a velocity vector is to simply place
an energy input at any location of the flow field, thus illuminating a fluid 'particle', and
visually following the illuminated and radiating fluid particle as it convects with the flow
field. There is no Slip between the traced fluid particle and the flow field as can be
encountered in most common measurement techniques such as LDA or PIV. In fact, not
even temperature gradients (if neglecting an insignificant increase of temperature in the
flow field due to the energy input) or concentration gradients (since the phosphorescent
additive dissolves thoroughly within the working fluid) exist which could signify a source
of error. Furthermore, after an energy input to the fluid has been released, the fluid
exhibits only negligible changes in its properties (e. g. bleaching of phosphorescent matter)
and can therefore readily be used for next excitation. The fluid dynamical approach in
connection with LIPA is Lagrangian as opposed to LDA and Hot-wire anemometry.

14

11.1.1.1 Demand towards chemistry as godfather of LIPA

Chemistry has to provide the dye that enables the marking of fluid particles in the flow
field. The dye has to incorporate photoluminescent properties, has to dissolve thoroughly
in the flowing medium, and ought to be safe to use and reasonably priced.

The phosphorescent dye used for LIPA which serves as an additive to append
photoluminescence to the list of properties of the working fluid was designed and
produced by the Department of Chemistry at Michigan State University under the
direction of Professor D. Nocera. The 'LIPA dye' consists of three constituents; the
vehicle for the phosphorescent compound is water (1-120). The actual phosphorescent
matter (providing excitable molecules) is Bromonaphthalene (Ber). Ber would, if
exposed directly to water molecules, experience a quenching efi‘ect (lose its
phosphorescent property) necessitating a protection (coat) for the Ber molecules. This
coat is facilitated by the other two components in the compound of the LIPA dye, namely
a sugar, Glucosyl-B-cyclodextrin (GB-CD or for short CD), and an alcohol, terr-butyl
alcohol (t-BuOH). These two components represent, in terms of a model, a cup and a lid
for this cup, respectively, in which the Ber molecules are embedded. Water as flowing
medium together with the composition of the three constituents being dissolved in water is
referred to as 'LIPA - chemical'. An optimum for the ratio of amount of alcohol, t-BuOH,
necessary for a given amount of sugar, CD, had been evaluated by Buchner (1992). A
recipe for the manual engendering of the LIPA chemical is given in section 11.2.1.2. More
information about the LIPA dye can be obtained fi'om Ponce et.al. (1993).

11.1.1.2 Principle of evaluation of array of velocity and vorticity vectors

LIPA'S purpose is to deduce from a three-dimensional flow field planar two-
dirnensional (2D) velocity vectors which are projections on the imaging plane. Intrinsically

15

LIPA provides 2D velocity vectors because data are inferred from 2D images captured by
an imager that contain two-dimensional convection information of the illuminated
particles. Since it is very dificult, if not impossible, to provide point energy inputs at
different locations in the flow field at the same time to achieve planar velocity information,
the excitation and illumination of fluid particles is accomplished in this work by imposing a
grid of laser lines onto the flow field where the laser is the source of energy. Now,
intersections of lines define the fluid particles which ought to be traced.

In this work, planar fields of velocity vectors are evaluated by recording the imposed
laser grid when the working fluid is quiescent (here, engine is not running) to obtain the
initial origin locations fiom where fluid particles convect (A', B', C', and D' in
Figure 11.1.1b), and capturing LIPA grids during engine performance a certain time interval
after a laser pulse illuminated fluid particles to obtain the locations where fluid particles
convect to (A", B", C", and D" in Figure 11.1.1b). Velocity vectors can then be evaluated in
the flow field by determining in the recorded images the distances fluid particles convected
during the chosen time interval (also referred to as time delay). The evaluation is a simple
division of traveled distance by the time interval. The time delay should not be too long
since LIPA'S intrinsic assumption of linearity of the fluid particle's path between its initial
and convected position should not be violated. A scheme of evaluations of velocity vectors
at the four points of a quadrolateral can be seen in Figure 11.1.1b. Note that the origin of

the evaluated velocity vector is located half-way between A' and A".

Vorticity vectors can easily be deduced fi'om the distribution of velocity vectors. If one
considers a quadrolateral spanned by the origins of four velocity vectors (Figure 11.1.1c)
one can consider the area of this quadrolateral a fluid particle associated with a
characteristic length scale of the order of the resolution of the laser grid. Under
consideration of distortion and rotation of this fluid particle set forth by the convection

mechanisms at the four corners one is able by applying fluid dynamic identities which

l6

incorporate information on the velocity vectors at the comers to infer a vorticity vector for
this fluid particle where the origin of the vector is situated at the center of the area of the

quadrolateral.
In general, vorticity is calculated fi'om a velocity field by
E) = V x V (11.1.3)
which can be reduced for two dimensions (x, z - plane) to

Bu aw
=___ 11.1.b
0’ 3 a ( )

where for this notation to is a 'point' quantity.

An alternative approach for vorticity evaluation incorporates the determination of
circulation. Along the contours of an assigned area, e. g. the quadrolateral in Figure 11.1.1c,
the circulation about that area can be evaluated by the following line integral,

r=§iflfi (11.1.c)

the line being a closed curve in the velocity field. The circulation for the abovementioned
quadrolateral can be determined by a piecewise summation of averaged velocity
projections onto the considered quadrolateral (Figure 11.1.1c) multiplied by the distances
between the corresponding velocity origins. The following equation then links vorticity

and circulation:

r = I(&‘J-fi)d4 or (11.1.d)

'03:.Lw dA (11.1.6)

A

17
In this form an 'area averaged' vorticity is evaluated.

The evaluation scheme for 'area averaged' vorticity vectors presented by
equations 11.1.c and 11.1 .e is more suited for the form in which velocity information is
available in this investigation (namely discrete values within the flow field) and thus
adopted for this study.

A thorough analysis of error inflicted by the employment of LEA on the basis of
double differencing the LEA grid - which is the actual practical procedure to obtain

vorticity -— can be found in F alco and Nocera (1993).

11.1.2 Phosphorescent characteristics of employed LEA - chemical

From empirical experience it can be asserted that the five-component mixture [water as
the convecting medium, KSCN to reflactive index match water to quartz -- see
section 11.2.1.1 - and three components flom the LEA dye] used as working fluid,
possesses, with respect to the ability of phosphorescence, hardly any shelf-life. An attempt
to reactivate a four week old compound by adding alcohol - which is the component that
goes out of solution at the highest rate through evaporation during the LEA experiments --
was unsuccessfirl. Even within 24 hours it could be evidenced that the fluid lost a

considerable amount of its phosphorescent potential.

To quantify the above stated observations, an investigation was carried out to
determine the comportment of the phosphorescence of the LEA - chemical with time
when not being in use - shelf-life - as well as its behaviour through the conduction of sets
of experiments - problems with evaporation of alcohol component, exposure to laser
energy, etc. The objective was to find out (a) the efl‘ect of intermolecular action on the
phosphorescent material while the LEA - chemical is in storage and (b) for what time
flame it will be scientifically meaningful to employ the phosphorescent chemical in LEA

18

experiments before the quality of the laser grid in terms of level of illumination decreases
to a non-acceptable level. A third chemically interesting issue is whether the
phosphorescent matter can be reactivated after depletion of phosphorescent activity -- as

suggested by the designer of the LEA dye - by supplements of the alcohol component.

In order to obtain information on the time-dependence of the LEA - chemical - being
in storage and in use - a scheme of experiments had been devised. Eight point data series
that determine the time interval of detectability of the phosphorescence after excitation by
an energy input were to be performed. These data series were taken before and after a
body of actual LEA experiments as well as shortly before the chemical is put into storage
and immediately after it is filled into the model engine. For every point in the data series a
grid of laser lines was recorded with the fluid quiescent for a fixed exposure time and
f-stop of the imager for eight different time delays (2, 4, 6, 8, 10, 13, 16, and 20 ms) -- a
time delay being the time between the pulsing of the laser and the recording of the
illuminated scene. In order to establish a curve where the degree of light intensity of the
radiation emitted by laser grids is plotted versus time delay, the mean pixel intensity in the
area covered by the laser grid was evaluated.

The results flom the above outlined experimental operations are summarized in
Figures 11.1.2 and II. 1.3. In Figure 11.1.2 curves of pixel intensity versus time delay were

evaluated on

CL 111: the day after engendering of LEA - chemical; before body of
experimental procedure (LEA)

CL IV: the same day; cyter body of experimental procedure
CL V: day 2 after engendering of LEA - chemical; no LEA experiments
CL VI: day 4 after engendering of LEA - chemical; no LEA experiments

Note that no LEA experiments were performed between the evaluations of curves IV
and VI.

19

It would have been very interesting to observe, qualitatively, the phosphorescence of
the LEA - chemical before & alter a set of experiments on the very same day it was
generated and the degradation of it between day zero and day one of its engendering.
Unfortunately, the decrease of laser power during the experimental phase on day zero of
the generation of the LEA - chemical and the recharging of the laser prior to the
experiments on day 1 (CL 111 and IV) turned the two data series, CL 1 and 11, into
ambiguous information. For curves 111 through X laser power was constant.

Curves III and IV illustrate the degradation of the LEA - chemical because it was in
use. Apart flom some waviness in curve 111 at the lower time delays - which are
presumably due to limitations in accuracy of the applied detectability evaluation
procedure - curve IV experiences a substantial shift down to a lower intensity level. This
decrease in intensity level is due to bleaching of the phosphorescent dye.

The shift down in intensity flom curve 1V to V - which is conceivably smaller than the
shift flom curve 111 to curve IV - is caused by intermolecular activity within the LEA -
chemical while the chemical is not in use (Shelf-life). The same argument applies for the
decrease in intensity flom curve V to V1. The decrease flom the intensity level as
encountered 2 days after the LEA - chemical had been mixed (curve V) to the level as
encountered 4 days after the LEA - chemical had been generated (curve V1) is smaller
than the decrease in intensity level flom curve IV to curve V although only one day passed
between the evaluations of the phosphorescence—detectability curves in the latter case.
This suggests a non-linear behaviour of the degradation of the phosphorescent properties
of the LEA - chemical when in storage.

In Figure 11. 1.3 the efl‘ect of adding successively four complements of t-BuOH, 5 ml
each, to the LEA - chemical after the evaluation of the data series leading to curve V1, is
demonstrated. It can be asserted that adding amounts of the alcohol component to the

LEA - chemical does not yield the desired refinement. For the most part of the range of

20

time delays examined the addition of alcohol does not change the level of light intensity
radiated flom the grid of laser lines. At small time delays a decrease of intensity is even

encountered.

11.2 Experimental Apparatus and Method of DxtgAcguisition

 

11.2.1 Water galog model of four-stroke IC engine

The experiments for which LEA was employed were performed in a motored water
analog model of a four-stroke IC engine with flat cylinder and piston heads using an
axisymmetrically seated, stationary valve. This most basic and simplified configuration as
apparatus for investigation in engine research was chosen to enable the investigator to
concentrate on the extraction of fluid dynamic quantities and their influence on mixing of
air and firel with residual fluid in the engine cylinder - stemming flom the previous cycle -
during induction, without being compelled to include considerations about the influence of
complex boundary conditions. And also to demonstrate the applicability of LEA without

casting expensive engine environments.

Since LEA is an optical measurement technique and the energy source used for this
study is a UV laser (see section 11.2.2. 1) all components of the engine model in the vicinity
of the area of investigation (piston head, engine cylinder wall, valve, and bottom of fluid
reservoir or cylinder head) are made of quartz - for a complete depiction of the apparatus
of the engine model in form of a drawing as well as of an anatomization of its elements see
Figure 11.2.1 - to ultimately accomplish thorough optical and visual access. Ultimately,
because the difl‘erence between the reflactive indices of quartz and water, and also
between quartz and ambient air which causes electro-magnetic waves to bend at
water/quartdair interfaces, requires an efi‘ort to reflactive index match the three phases. If

the index of reflection of, for instance, the working fluid was difl‘erent flom that of the

21

quartz cylinder the curved cylinder wall would fimction as a lens and the mapping of
intersection points onto the plane of the imager would not reflect their real locations in the
physical space.

Reflactive index matching of the entire system was attained by embedding the engine
model in an aqueous medium that was reflactive index matched with respect to quartz and
by using index matched water as the working fluid in the cylinder. A Plexiglas tank holds
the aqueous index matched fluid (see Figure 11.2.1a). The tank has a quartz window to
allow the UV light energy source to enter the test facility. A more detailed description of
reflactive index matching procedures can be found in section 11.2.1.1 ’Refiactive index
matching of experimental environment '.

Figure 11.2.1b which displays the dissected components of the model engine shows a
Rulon O-ring. Rulon was found to be a material (plastic) that could be used as O-ring for
a water/quartz environment (for more information on the O-ring see Appendix A).

The experimental rig has been provided by the Ford Motor Company, Dearbom, MI.

An advantage associated with the use of water as the base for the working fluid is that
the kinematic viscosity of water is 17 times lower with respect to the kinematic viscosity

of air - the standard flowing medium for engines. Thus, if equating Reynolds numbers,

Re = —, (11.2.a)

where L is a characteristic length, U is a characteristic velocity, and v is the kinematic
viscosity, piston speeds can be 17 times lower in the water analog model in order to
produce equivalent flow field dynamics. Typical Reynolds numbers - based on the width
of the annulus (orifice) and the velocity of the flow in the orifice - range in this
application between low Re and Re=6300. For this engine configuration, the Reynolds
number is - to within a negligibly small margin - independent of the valve lift; it is

 

 

22

however, dependent on the crank-angle. The main influence of the Reynolds number on
the fluid flow in an engine is manifested by the dependency of the creation and destruction
of toroidal vortex structures - in the shear layer of the intake jet - with Re.

A disadvantage associated with water as the base for the LEA - chemical is its
property of incompressibility. Hence, an effort had to be made to avoid the compression as
well as the expansion stroke during engine performance (for the layout of the cam-shaft

employed to omit compression & expansion see Appendix A).

11.2.1.1 Reflactive index matching of experimental environment

Considerations of reflactive index matching is always worthwhile contemplation when
phases of different reflactive indices are involved in an optical system. Every medium has
its specific index of reflaction n. The index of reflaction of a medium can be manipulated
by changing the composition of the medium's chemical synthesis. The most convenient
medium to be manipulated is a liquid phase. The index of reflaction of water (1-120), for
instance, which is smaller than that of quartz (see Table 11.2.2), can simply be matched to
that of quartz by adding an agent with an index of reflaction that is greater than that of
quartz and that dissolves in water thoroughly.

The point in a parameter space where a medium, a liquid for instance, is reflactive
index matched - e. g. with respect to the quartz used for the model engine nquam -- is a
function of the wavelength of the light for which index matching is to be achieved as well
as of the ambient temperature (see Table 11.2.2). For the experiments performed for this
study the index matching agent for water was Potassium Thiocyanate (KSCN). KSCN has
an index of reflection greater than quartz, dissolves in water, and is transparent with
respect to energy flom a UV laser beam. Analytically, the quantity of KSCN needed to

reflactive index match its solution with water to quartz can be calculated by the following

23

formulation derived by Buchner (1992) whose derivation was based on equations in a

review about reflactive index matching considerations by Mohr and Ertel (1986):

S[%] = "W “"1 (11.2.b)
n. -nr

 

where mom, has to be equal to nquam in order to index match for quartz, n1 is the index of
reflaction for water and n; the one for KSCN. S represents the anhydrous solute weight
per cent.

Apart flom the theoretical prescription of the quantity needed for KSCN to index
match water to quartz, scientific research incorporating trial and error led to the following
'rule of thump' recipe flom empirical observation: 2000 ml H20 + 2000 g KSCN +
270 ml H20 index matches quartz at about 15° C for 308 nm wavelength. When the
temperature of the solution increases more KSCN has to be added in order to reflactive
index match. The mixing of water and Potassium Thiocyanate is strongly endothermic

(i.e. the solution becomes very cold).

Table II. 2.2: Indices of refiactz'on for diverse materials

 

 

Conditions for lndecee of Refraction
Temperature and
Wavelength Water KSCN Quartz | DHN Hexane
20 degree C, 505 nm 1.455
20 degree C, 559.5 nm 1.555 1.552 1.455
25 degree C, 559.5 nm 1.474 1.572

For flow visualization experiments involving a photochromic chemical by Kodak -
Pfocedure and results will be described in section IV .1 - a chemical that belongs to the
may of oils is preferable for being the basis of the working fluid. Decahydronaphthalene

«3101“ 1 3) turned out to work very well in connection with the Kodak chemical.

24

Decahydronaphthalene (DHN) absorbs almost no energy at 308 nm wavelength and has an
index of reflaction of 1.474 - at 25° C and 589.3 nm wavelength, Table 11.2.2 -- which is
fairly close to the one of quartz. To completely match the index of reflaction of DHN to
that of the quartz engine another organic component with a low index of refraction had to
be found that furthermore exhibits high transparency in the UV range of the spectrum of
electro-magnetic waves. During the phase in which this suited partner for DHN was
sought it was experienced that the higher the ratio n/m in the formula CmHn for organic
chemicals, the higher was the transparency of the chemical at 308 um. At the end of the
testing phase, Hexane (C6H14) was chosen as the index matching agent for DHN whose
index ofreflaction is 1.372 - also at 25° C and 589.3 nm wavelength, Table 11.2.2.
Contributions of 74.44 Vol-% Decahydronaphthalene and 25.56 Vol-% Hexane constitute
a working fluid that is reflactive index matched to the quartz used for the model engine at
operating conditions. Note that the evaporation rate for Hexane is considerably high when
the DHN/Hexane solution is exposed to air.

Due to the fact that oils are not aggressive towards the metal components in the engine
apparatus - in the contrary, they rather serve as lubrication - the DHN/Hexane mixture
was used as the index matched fluid in the outer tank embedding the quartz cylinder. This

constitutes a rather prophylactic measure for the possibility of leaking.

For firrther empirical observations on reflactive index matching refer to Appendix B.

11.2.1.2 Recipe for engendering of LEA - chemical

After the flowing medium for the LEA experiments (water/KSCN solution) was
properly index matched for quartz the fluid had to be 'loaded' with the phosphorescent
compound (Ber molecules protected by sugar/alcohol shield; see section 11.1.1.1). The
loading process was accomplished by the following steps:

25

1. Add necessary amount of CD (for 4.5E-03 M CD in 800 ml H20/1(SCN, an amount
of 4.62 g CD is needed); stir vigorously.

2. Add necessary amount of t-BuOH (for 4.5E-03 M CD in 800 ml H20/1(SCN, an
amount of 42.3 ml t-BuOH is needed; this is 5.78E-01 M); stir.

3. The current mixture is, with respect to the desired matching of refractive indices to
quartz, low on KSCN; add about 14 g KSCN to index match again.

4. Filter fluid through regular cofl‘ee filter.

5. Add one drop of Ber as phosphorescent material; stir vigorously.

11,2.2 Aspects pf dag amuisition and its facilitation

11.2.2.1 Generation of laser grid - optical assemblage

The optical accessories necessary for the generation of the LEA grid of laser lines
which is superimposed onto the flow field and their configuration are exhibited in
Figure 11.2.2. The energy source used (providing monochromatic, coherent light) is the
excimer laser LPX2 flom Lambda Physik Lasertechnik. It was operated at 308 nm
(UV waves in electro-magnetic spectrum) by using a Xenon Chloride (XeCl) gas mixture,
the buffer gas being Helium (He). The exiting laser beam is of rectangular shape with
dimensions 20 by 65 mnr; the flequency range of laser pulses comprises 1 to 200 Hz with
a pulse length of 17 ns; the maximum output attainable is 250 m1 per pulse; the laser can
be triggered externally.

The implementation of the grid of laser lines as employed for the majority of LEA
experiments -- examination of fluid dynamics in a plane at one circumferential position 0
that covers most of the engine cylinder -- is accomplished as follows (for a sketch of the
setup refer to Figure 11.2.2a): The exiting laser beam - after diverting the beam by two
mirrors - is splitted by a half-transmitting mirror into two sub-beams - under 45° the

half-transmitting mirror splits the beam into two sublets of equal intensity. For each sub-

26

beam a convex quartz lens (500 mm focal length) focuses the beam at the measurement
site. The long focal length assures the thinness of the laser grid lines at the region of
interest. After passing through the lenses each sub-beam is divided into several parallel
lines by means of a grating which is incorporated by an array of aluminum coated, parallel
slim mirrors glued onto a base body. For a scrutinizing view of the beam divider see
Figure 11.2.2a. The laser lines flom the two coplanar sets of parallel lines intersect at the
area of investigation under a certain angle. This angle constitutes a compromise: On the
one hand, the angle was to be as close to 90° as possible for reasons that will become
clear in section 111.3. However, since both sets of laser lines enter through the cylinder
wall and since the Plexiglas reservoir as well as the metal piston are not transparent with
respect to UV - light the two sets of parallel laser lines had to be adjusted in such a way to
avoid blockage of energy by these mechanical parts. [A solution where one set of laser
lines enters vertically through the quartz piston head in order to achieve intersection
angles of 90° was considered and implemented, but had to be rejected, since flom a
certain crank-angle during the induction stroke on, the penetration of the laser lines was
insuflicient to sustain a grid in the upper part of the cylinder (attenuation, too much dye);
information could only be obtained during early intake] The plane spanned by the
generated grid of laser lines intersects the cylinder axis of the engine. Local assignment of
intersections of the laser grid, thus the definition of locations where the velocity vectors
were to be evaluated, occurred rather randomly in the implementation phase of the grid.
The reader is referred to Appendix B 'Optz'cal issues' for more information on the
generation of the LEA laser grid.

The laser grid implemented by the setup in Figure 11.2.2a is depicted in Figure 11.2.3a; it
covers for quadrant I of the x, z - plane at the investigated circumferential position 0 most
of the cylinder. 74 locations were considered for the examination to be distinct enough and
therefore qualified for data reduction. By these 74 locations a spacing between points of

about 6 mm could be achieved. For one intersection of two laser lines the major axes of

27

the formed typical ridge-shaped core are, in the digitized image, about 11 x 30 pixel in
length; thus, the area of the core is represented by 330 pixel (Figure 11.2.3a). Laser lines
intersect in general under an angle of about 50°. The longer major axis of an intersection

coincides roughly with the x - axis of the reference flame.

In section VS a more thorough investigation of the annular intake jet is furnished by
employment of an investigation grid ofl‘ering a closer spatial spacing of laser line
intersections than the one achieved by the optical setup in Figure 11.2.2a. In order to attain
the more dense spacing, 150 mm focal length lenses replacing the 500 mm lenses were
placed behind the grating (the setup is shown in the photograph of Figure 11.2.2b).

At 15 locations velocity information could be obtained. The locations are depicted in
Figure 11.2.3b along with the geometry of a typical intersection of two laser lines. The
number of pixel within the ridge-shaped core of a typical intersection is here significantly

higher than for an intersection of two laser lines for the investigation grid in Figure 11.2.3a.

Since the process under investigation is a reciprocating one where one boundary moves
vertically, throughout parts of the intake stroke laser lines were blocked by the piston and
only a certain number of intersection areas - which increases with crank-angle -- could be
evaluated for the first part of the induction stroke. Consequently, for the laser grid in
Figure 11.2.3a, not until shortly alter midstroke could all 74 locations be studied. Thus, in
the data reduction process, a map of locations had to be assigned to each crank-angle

before midstroke to notify the data reduction process where information is available.

Optical systems generating the grids of laser lines were implemented using a mechanical
support construction consisting of simple iron rods and aluminum elements connecting the
rods where mirror and lens components were held in place by clamps that were screwed to
the iron rods (Figure 11.2.2b). Naturally, the degree of fine tuning of Optical elements

achievable with such an optical system is rather low; however, suficient precision was

28

accomplished for the generation of laser grids employed for the experiments conducted
within this study. A system based on an optical bench or table with micrometer adjustable
enclosures for optical elements will enable the experimentalist to realize grids with close

spacing of intersections of laser lines where laser lines are very thin.

11.2.2.2 On how to acquire grids of laser lines that constitute LEA - data

Data acquisition was performed for the bulk of the experiments according to the
scheme in Figure 11.2.4a. Figure 11.2.4c contains a more physical description of the
abstract quotation of devices in Figure 11.2.4a; Figure 11.2.4b displays the engine apparatus
and the imager in a photograph.

According to Figure 11.2.4a, a motor puts out a constant angular torque that drives the
cam-shalt. (The unique characteristics of the cam are described in Appendix A) The carn-
Shafi generates for the induction stroke a sinusoidal movement of the piston - the
induction stroke being the mere subject of study in this work - as input to the 'Volume of
Investigation'. The measure for data acquisition of LEA images is provided by the running
engine model (for more detailed information on timing of the data acquisition process the
reader be referred to Appendix C). An optical encoder which is mounted to the shaft
between motor and cam and which serves as mechanical-electronic interface puts out a
signal at every degree of rotation of the (cam-) shaft. This signal is transferred to an
'electronic signal valve' which serves as an 'AND' filter letting through information only
when a signal fulfills two requisites: It is every 13‘, or 20°, ..., or 10th (choosable) signal
AND it belongs to the crank-angle range of between 10° prior to the beginning of the
intake stroke and 10° alter the end of the intake stroke. (More on the 'electronic valve' can
be obtained flom Appendix C.) The output flom the 'valve' is forwarded to the excimer
laser and triggers it to pulse (to build up the grid of laser lines in the engine cylinder),

contributing to the 'Volume of Investigation' as another input, Figure 11.2.4a. The signal

29

from the ‘valve' is also conveyed to a flequency generator that can manipulate incoming
electronic information (e.g. delay transmission of incoming signal). This 'delay' generator
schedules data acquisition for the imager. The imager receives input flom the 'Volume of
Investigation' in optical form. Because of the timing procedures enforced by
synchronization requirements (Appendix C), for 10 RPM engine speed, data were
obtained for every other degree crank-angle; for 20 RPM the flow field was examined at
every 4th, and for 30 RPM at every 6th degree crank-angle.

To capture LEA images an assembly of a regular Video camera by Canon and an
image intensifier lens (11L) mounted in float of the camera has been used (for more
information about the imager see Appendix D). The optical axis of the imager system is
perpendicular to the plane spanned by the grid of laser lines. Information on the timing
procedure for the imager system can be gathered flom Appendix C. The Video camera
saves the received information onto tape (Figure 11.2.4c). This information is digitized by
an imaging processing facility [incorporated by MegaVision and working in conjunction
with a Sun 3/260 Workstation; for a thorough description of the MegaVision system read
Appendix D] and the images containing LEA grids are then stored in the memory of a
Sparc2 SunWorkstation. Implemented software, then, extracts coordinates of illuminated
fluid particles flom the LEA images with subsequent calculation of fluid dynamic
quantities such as velocity, vorticity, and fluctuation intensities. Implemented algorithms
will be discussed in chapter 111 'On the Methodology of Data Reduction'.

11.2.2.3 Results flom preliminary preparation-activities for series of LEA experiments

After an extensive experimental phase preceding the actual LEA experimental series
where the LEA method and chemical were tested thoroughly and results were recorded
using the 35mm-4ML movie camera (series 2000) by Photo-Sonics, Inc., onto

T-max 3200 black&white film flom Kodak, the acquisition of an image intensifying system

30

enabled the application of the above mentioned Video camcorder flom Canon to capture
LEA - data on video tape. In the data acquisition setup using the movie camera
implemented for the preliminary testing of the LEA concept, the movie camera triggered
the laser on a 60 fps (flames per second) basis to build up the laser grid in the flow; no
synchronization with the running engine was accomplished. Some results flom the ground
setting quest with the movie camera are presented in Figures 11.2.5 and 11.2.6.

Figure 11.2.5 shows a laser grid depicting the origin locations of intersection points
prior to convection, and grids exhibiting distorted laser lines for 20 RPM engine speed and
6 mm valve lift in black&white for the beginning of an engine cycle. Note that one set of
laser lines enters through the piston head. The pictures in Figure 11.2.5 hold valuable LEA
information.

In Figure 11.2.6 color (ASA 1000, E6 processing) pictures taken with the movie camera
of the early induction stroke of the model engine are shown where, again, one set of lines
enters through the piston head. Figure 11.2.6 also exhibits images of one laser grid
showing origin locations as well as grids containing convected intersections of laser lines.
In the background of the laser grid showing distortion and thus carrying the current LEA
information - Figure 11.2.6b through e -- one can perceive, nebuously, the highly distorted
remains of the laser grid that was generated 16.67 ms before the grid that represents the
current LEA information. A detectability of the LEA - chemical well beyond 17 ms
enabled the acquisition of this picture. The recording of this image constitutes a genuine

success for performing quantification and qualification of the flow field in one step.

11.2.2.4 Matrix of conducted LEA experiments

Experiments involving the application of LEA to quantify the flow field of the
induction stroke of a model of an IC engine comprise a variation of two engine

parameters, namely engine speed and valve lift (VL; the valve lift is defined as the vertical

31

length of the orifice between valve and head of engine chamber). For the experiments in
the investigation of the whole engine cylinder (74 locations, Figure 11.2.3a) the matrix in
Table 11.2.3a depicts the settings of engine parameters examined. (A double-'x' indicates
that there is flow visualization information available for that particular setting to which the
LEA results can be compared.) Table 11.2.3b denotes the experimental series that have
been conducted in order to investigate the intake jet with the close-up look at the valve
(15 locations, Figure 11.2.3b). Note that by conversion of the engine speed as encountered
for the water analog model in this work to an equivalent speed for an engine using air as
working fluid - by equating Reynolds numbers - investigated engine speeds are 170, 340,
and 510 RPM.

It should be noted at this point that due to memory limitations of the SunWorkstations
fluid dynamical quantities could be averaged only over a confined number of engine
cycles. For 10 RPM engine speed (data available at 90 crank-angle events within one
realization of an intake stroke, see section 11.2.2.2) the number of engine cycles over
which averaging was performed is 21; for 20 RPM (45 crank-angle events) the number
is 42, and for 30 RPM (30 crank-angle events) it is 63 ECS. The higher the resolution in
crank-angle the lesser is the number of engine cycles due to constant memory of
computer", however, on the other hand, the smaller the engine speed the lower are the

cyclic variations (see section V2.2) and the less important is thus the need for sample size.

Table II. 2. 3: Matrix of experiments performed for the investigation of
(a) whole engine cylinder and (b) intake jet.

. '3 Speed Valve ~'

10

3

x
20 xx
30 x

 

32

 

11.2.3 Estimates for parameters relevant for description of quality of LEA - data

11.2.3.1 Delay time

In general, a map of velocity vectors is evaluated by taking two images of the grid of
laser lines during convection of the flow field and determining the distances that
corresponding intersection points traveled during the time interval between the acquisition
of the two images.

For this work, the LEA - data representing the 'first' image are, as mentioned above,
taken with the working fluid quiescent (i.e. the piston is not moving). These images
contain grids which depict the origins of intersection locations of laser lines prior to
convection - they will be referred to as 'undistorted' or 'reference’ grids (Figures 11.2.5a &
11.2.6a) - and were recorded always right before a set of data incorporating one setting of
the engine parameters was acquired. Uncertainties about accurate inference of the origins
flom the reference grids will be discussed in section 111.3.

In order to resolve the range of velocity magnitudes encountered in the investigated
model engine (0 - 3.5 m/s), and with the prerequisite of linearity of particle paths in mind,
time delays for acquisition of the LEA information consisting of laser grids that present
illuminated particles alter their convection with the flow, were in the range between 0.3
and 1.1 ms. No imager device was available to take images at a rate of 1000 Hz or faster

which would enable the experimentalist to record successively the reference image and its

33

distorted relative. This lack of equipment forced the investigator of this study to record
the undistorted grids with the working fluid quiescent. Undistorted grids were acquired a
certain time delay after the laser pulses (in fact the same time delay as was used for the
corresponding data set was chosen) in order to avoid fluorescent background noise in the
images.

The images containing the convection information were taken during engine
performance a certain time delay (between 0.3 & 1.1 ms) after the laser pulsed and the
energy input illuminated the fluid particles. These images hold data that are referred to as
'distorted' grids (Figures 11.2.5b-e & 11.2.6b-e). Now, a field of velocity vectors can be
inferred by calculating the quotient of the distance each fluid particle convected to the time
delay between the application of the laser grid and the acquisition of the distorted grid.
This time delay is a crucial parameter for the conduction of LEA experiments. It ought to
be estimated by considering the highest occurring velocity magnitude in the flow field that
the investigator is interested in to evaluate. LEA is, for the current study, based on the
assumption of linearity between a fluid particle's original location and the location it
convected to, as mentioned earlier. Linearity means that only the first term in an expansion
is used and that higher order terms are negligible. Under this assumption velocities are
deemed to be constant and locally uniform between the acquisition of the two images
which contain the reference and the distorted investigation grid. Nonlinearity of a particle's
path is usually associated with the region of the highest velocities and the highest spatial
velocity gradients, it is the region where one generally finds highest strain. In this
application of LEA to the water analog model of an IC engine the highest velocities and,
thus, the highest nonlinearities occur in the region governed by the incoming annulus jet.
Hence, it is meaningful to estimate a reasonable delay time by the reciprocal value of the

quotient of the velocity at the jet centerline vc and a characteristic length 2,,“ which is of

the order of distortion allowed under the requirement of linearity.

34

ATM = (11.2.c)

(dist

For 20 RPM engine speed and 6 mm valve lift a typical velocity magnitude in the seat
area of the valve is calculated to be about 2 m/s. Linearity is assured, according to visual
observations during the distortion of laser grids, at convection distances smaller than
1.3 mm (corresponding to 10 pixel in a LEA image). Plugging these numbers into
equation 11.2.c yields a delay time of 0.65 ms. This first estimate was used in a preliminary
test phase. Eventually, the delay time for 20 RPM and 6 mm VL was corrected to 0.7 ms.

11.2.3.2 Exposure time

The exposure time expresses the time interval for which the CCD array of the
camcorder is exposed to the illuminated scene in the model engine. The exposure time
ought to be as small as possible to avoid blurring in the image. Variation of illumination
intensity within the investigation grid that is spanned by the laser lines in the measurement
area -— in x - and z - direction of the reference flame -- inflicts a manual adjustment of the
exposure time.

Variation of intensity within the laser grid occurs in the x - direction because the laser
line entering the cylinder through the wall experiences attenuation so that the intersection
points further in towards the cylinder axis have lower maximum intensity. Then, the level
of energy intensity inherent in each laser line difl‘ers flom line to line before the lines'
entrance into the cylinder - basically for two reasons: the initial laser beam that leaves the
laser is inhomogeneous over its rectangular extent, and, due to the fact that the two
splitted laser beams go through the focusing lens before they are divided by the grating so
that the mirrors situated on the beam divider further away flom the lens reflect higher
intensity light (Appendix B). Furthermore, there are two laser lines that pass through the

35

quartz bottom of the reservoir which are for this reason weaker than the other lines
involved in the generation of the grid.

The exposure time was adjusted so that the intersections of laser lines close to the
lower cylinder wall exhibiting highest light intensity would not saturate the phosphor in the
image intensifier [saturation of the phosphor is equivalent to assuming pixel values of 255
after digitization into the 8-bit grey scale scheme provided by the MegaVision system].
The exposure time was thus found empirically by trial and error. For the main body of

experiments it was between 1 and 10% of the delay time.

By leveling out the difl‘erences in intensity distribution exhibited by the laser lines the

determination of exposure times can be made easier.

11.2.3.3 Quality of LEA - data

In order to create a quantitative and objective measure to compare the grade of a
particular realization of a laser grid within the flow field, a quality factor is defined. This
quality factor will also enable the investigator to decide what kind of data reduction
process can be employed (section 111.1). The suggested quality factor Q incorporates
information about intensity of illumination (which depends on laser power, concentration
of LEA dye and age of chemical, and so forth), the sharpness and thinness of the laser
lines, the grade of index matching, the distance of penetration of laser lines into the flow,
and the degree of image intensification (photon magnification) and resolution provided by
the imager. Q is proportional to these parameters. Q is reciprocal to the intensity level of
the background, the exposure time (fuzziness of lines), and the time delay between the
application of the laser mid to the flow and the recording of the LEA image (diflirsion of
energy in lines). In a very simple, just-plug-in formulation of the above mentioned
influences on the quality of LEA mids a mathematical proportionality relation reads:

36

INT* SHARP t THIN* INDMA TCH .. PENEII MAGN .. RES
(BACK +1): ATM .. ATM,

 

Q (11.2.d)

Ill

where INT/(BACK+1) can be replaced by information about contrast of the laser lines in the
test facility achieved in the image. BACK is given as pixel intensity. This is why BA CK is
incremented to avoid a Q of infinity when BA CK is zero. Since no data are available to compare

an evaluation of this quality factor to, no quantification of Q will be firrnished in this work.

To get a qualitative impression of the grade of LEA - data that were acquired for this
work three images bearing information on one undistorted - (a) -- and two distorted grids
-— (b) and (c) - recorded for 20 RPM engine speed and 3 mm valve lifi as data for the
investigation of the whole engine cylinder are displayed in Figure 11.2.7. The images were
recorded with the image-intensifier/Video combo. As mentioned in Appendix C, the
camcorder was turned by 90° in order to make most use of the CCD array of the Video
camcorder, thus, the valve is on the right side of the picture, the cylinder wall at the
bottom, and the piston head on the left side. In Figure 11.2.7b and c one can hardly
perceive distortion of laser lines. Highest distortion is, during engine performance,
encountered at the axis of the intake jet. This axis is in the lower right corner of the
pictures. With a scrutinizing look one can discern slight waviness of the laser lines. Low
distortion assures the assumed linearity of particle paths. Furthermore, the pictures
demonstrate that fairly deep penetration -- beyond the axis of the cylinder -- could be
accomplished in the experiments.

In Appendix D it is noted that for calibration purposes a fixed reference point was
added to the LEA images. This reference point can be seen in the pictures of Figure 11.2.7
in the upper left corner. It is situated within a crescent-shaped area formed by the mirror
that reflects the light flom an LED onto the CCD array of the camera.

Chapter 111
ON METHODOLOGY AND ACCURACY OF DATA REDUCTION

LEA can be eflicient and survive profound criticism flom the velocimetry community
only if the data reduction process of raw data provided by the acquisition of LEA
information (distorted mids of laser lines) flom LEA's application to a flow field, provides
a high level of accuracy along with quickness in the deduction of results. Consequently,
the data reduction process ought to be thoroughly automated. Quickness of the data
reduction process, then, in terms of automation, is ultimately limited by the speed and
communication abilities of computer systems; quickness is, thus, basically a matter of
hardware and only to a limited extent a matter of software. Accuracy is, as far as data
reduction is concerned, considerably affected by the method and architecture of the
software that infers fluid dynamical information flom the raw data. The focus of research
in the field of computer science, as far as LEA is concerned, has to be on the
implementation of software that is suited to reduce LEA - data with the highest degree of
accuracy attainable, while keeping a squinting eye on the progress of computer hardware.

For this work, all essential hardware/software requisites for data reduction of LEA
information were available with the exception of an algorithm that extracts coordinates of
illuminated fluid particles flom an image, and algorithms for the calculation of fluid
dynamical quantities as well as their statistics. On the administrative side, a demand
towards the data reduction process incorporates the process being able to manage the vast
amount of data necessary for statistical independence of flow quantities that describe
turbulence inherent in an encountered flow field. Data administration utilities and routines
desimred for this purpose were implemented on a UNIX platform. Limitations in computer

memory forced the process of data reduction to seek a compromise between

37

38

administrative necessities and sample size for the generation of fluid dynamic quantities.
These limitations confine, in this work, the total of flames over which ensemble averaged
results can be obtained to a certain number.

The generation of algorithms that calculate the actual fluid dynamical quantities
associated with a flow field was performed on a FORTRAN platform; the algorithms do
not bear any significant or firndarnental scientific innovations and will, thus, not be
deliberated in this work. The same holds true for the utilities which are written in a UND(
environment and serve to administer the LEA - data. Therefore, only the software that

detects illuminated fluid particles in the LEA images will be discussed.

In order to implement the software inferring coordinates of illuminated fluid particles
flom LEA images, the illuminated particle itselfas it is represented in an image has to be
examined thoroughly. The particular realization of a mid of laser lines as it was achieved
in this study, in connection with the resolution provided by the imager system and
digitizing facility, yielded at intersections of laser lines illuminated areas that consist - in
the digitized images -- of about three hundred pixel (Figure 11.2.3a); hence a fluid particle
is represented in an image by an area of illumination. Within these areas which are
surrounded by a background of constant intensity, the distribution of pixel intensities (on a
range of 256 grey scale levels) resembles -- speaking in metaphorical terms - that of a bell
or an igloo sitting on flat mound; see Figure 111.1.4b. The intensity values within the
intersection area (IA) are significantly greater than those of the backm'ound. In order to
calculate a velocity by considering the spatial distance between two temporally separated
occurrences of the same fluid particle, this intersection area - spanned by the crossing
laser lines - ought to be represented by one intersection point (E) which comprises the
center of a pixel. It is an imperative to any algorithm that determines intersection points

flom LEA images to find the representative E within this area consistently. AS long as

39

this representation is chosen consistently the intersection point can be any characteristic

point within the IA, e. g. the spot with the highest pixel intensity.

Motivation for the generation of software that was to automatically infer the
coordinates of intersection points (illuminated fluid particles) flom LEA images was
derived flom the fact that deduction by hand is enormously time consuming and greatly
imprecise. In fact, a suficient amount of data necessary for statistical independence of
velocity fluctuation information would not have been assured if data reduction had been
performed manually.

Fimrre III.1 gives an idea of the uncertainty about hand-sampled coordinates of
illuminated particles. For this plot three different persons deduced intersection points flom
a distorted laser mid in a LEA image independently, each intersection point five times.

- The dotted lines are results flom the application of an algorithm that reduces extensive
laser lines to one-pixel, skeleton representations of a laser line; crossings of these Skeleton
lines denote intersection points of the laser mid. - Not only is there considerable
fluctuation within one cluster of the five points for one intersection and one person, but
the centers of the three clusters neither coincide with each other for one intersection nor

do they coincide with the intersection point suggested by automatic data reduction.

111.1 Illustration of two Approaches to dedpce Coordingtes of Intersection Poin1s_
flom LEA Images, and Pre-Conditioning of these Images

 

For the inference of the coordinates of illuminated particles from LEA images --
namely the coordinates of their intersection point representations - two basic approaches
had been considered. One approach incorporates the reconstruction of two laser lines in an
intersection area that lose the information about their edges within the IA due to their

crossing. After the recovery of the lost information the laser lines can be shrunk to

40

'skeleton' lines or 'centerlines' (with a width of one pixel) whose crossing denotes the
intersection point for the particular intersection area. The other approach constitutes a
more simplistic model where the algorithm looks essentially for the area of highest
occurring pixel intensities within an intersection area and chooses its peak as intersection
point. For both algorithms FORTRAN is the computer language they are composed in and
FORTRAN7 7 the utilized compiler.

As has been experienced during data reduction, the point of highest intensity in an
intersection area moves with respect to the intersection of the skeleton representations of
the two intersecting laser lines. The problems in connection with these occurrences will be

discussed later in this chapter.

1111.1 Reconstruction of Intersection Area (MA)

This approach is the more complex and precarious one. In order to apply it to LEA
images to infer intersection points flom the mid of laser lines, a considerable contrast
between laser lines and background is a necessary prerequisite. The process of E
deduction, which involves the recovery of information about the edges of the two
intersecting lines which make up an intersection area and thus the reconstruction of the IA
(RIA), consists in its most general form of the following computational procedure (for
nomenclature refer to Figure [1111):

E] First, an edge detection routine extracts within the intersection areas those dots flom
the two laser lines involved in an intersection that depict edges of the lines. Edge detection
was based on a threshold criterion: An edge is defined at a location within a LEA image
with the property that a transition occurs of pixel intensity flom intensities lower (higher)
than a certain threshold to higher (lower) than this threshold. The applied threshold is

determined based on information of local pixel intensity distribution by employing a

41

statistical analysis on a set of pixel values that consists of randomly chosen pixel intensities
drawn flom an IA and its vicinity.

El Next, each single edge element in the detected cloud of dots is assigned to its
respective line it lies on by clustering the elements according to the nearest-neighbor rule
(Jain and Dubes 1988). Thus two crossing lines are portrayed by the 'chaining' of edge
elements. During the chaining procedure four simrificant points -- 'crucial points' - are
marked: these points are localized where the edges of the two laser lines cross. They are
detected when during the chaining procedure a simrificant change in direction of the I
congregating chain is experienced.

[I] The next step is to consider two of these crucial points (any pair except a pair of two
opposing points). The goal is to interpolate the edge of an interrupted laser line between
these two considered points. In order to do so the ten closest edge elements of the two
corresponding edges (which are chained away flom the IA) at each of the two crucial
points are considered. With these twenty data points one can perform any interpolation
procedure thinkable. For this study a first order least sqriare fit had been applied. Now, the
recovery of information about laser line edges which was lost in the intersection area due
to the crossing of the laser lines has been accomplished.

D The ultimate objective of the RIA algorithm is, however, to deduce the
representative intersection point from the intersection area. For this, the recovered laser
lines are shrunk to skeleton lines using the edges of the laser lines found by interpolation
and the pixel intensity distribution within the intersection area. By intersecting these
skeleton centerline representations of the laser lines it is possible to find the E

representation of the fluid particle for this approach.

The results achieved with the RIA approach were satisfactory when, in a preliminary
test phase, applied to a set of undistorted LEA mids originating from a different LEA

project.

42

The outlined procedure is established to find the E for any IA; no a priori knowledge
about shape, size, intersection angle, or orientation of the examined IA is necessary.
Concern about deformed intersection areas after distortion of the laser grid which would
cause problems when applying versions of RIA that require specific information about an

IA was motivation to seek an algorithm of generalized form.

Preliminary procedures incorporating the shrinking of grid lines to a width of one pixel
with subsequent intersection of the shrunk lines in order to find a one-pixel representation
of the IA by using components of an existing software package provided by MegaVision
(refer to Appendix D for more information on MegaVision) constituted the first step of
and were the basis for the RIA approach. It turned out that the accuracy obtained by the
MegaVision software was not sufficient so that software had to be implemented that was

better suited for the task.

Edge detection software flom the computer science literature such as 'Sobel' and 'LoG'
(see Ballard and Brown 1982) had been implemented to serve as step 1 in the
computational procedure outlined above. This software has been applied to LEA images
with minor success. All edge detection algorithms employed require by definition an
abrupt change in pixel intensity as one examines a cross-section in an image in order to
acknowledge an edge. The distribution of pixel intensity within a laser line is, however, of

rather Gaussian nature condemning the edge detection algorithms to fail.

1111.2S hfrHi estIntensi SotinIA S

The second approach was originally implemented to support the first approach outlined
in the preceding section. It was meant to serve as a 'second Opinion' corroborating the
result of the application of the RIA approach to a LEA image. For the approach discussed

in this section, the area of highest pixel intensity is found within an intersection area and its

43

center or peak - the highest intensity spot (HIS) - is defined as the intersection point of
an IA. The ofllset between the Es detected by the RIA option and the HIS option and its
variation for one IA with several LEA images of information was to be used as a measure
for accuracy. The HIS approach represents a more simplistic model; its mere presumption
is that the laser lines intersect properly enough in order to generate a bell-shaped intensity

distribution within an 1A.

The computational procedure for HIS comprises a scanning routine which scrutinizes
an intersection area to determine its subregion of highest pixel intensity and to appoint its
center as the representative intersection point for an intersection of two laser lines.

The process of finding the highest intensity spot begins with a coarse assignment of a
starting point in the vicinity of the area of highest intensity within an IA. Information
about starting points is readily obtained flom the locations of the Es in the reference
mids. Around this starting point an area of 10 x 10 pixel is examined. In order to find the
highest intensity spot in the square area consisting of these hundred pixel around the
reference E the computer algorithm integrates subsequently for every one of the hundred
pixel the intensity distribution over a square matrix of 7 x 7 pixel where the pixel under
investigation is the center of this applied matrix. The algorithm, thus, takes into account
information that is contained in the vicinity of each examined pixel in the 10 x 10 pixel
array. There is a preassigned weight for each of the 49 members of this matrix. The
weights in the matrix were chosen on the basis of an optimization procedure that
incorporated empirical observations; the three-dimensional distribution of the weights
resembles that of a 2D Gaussian distribution and is depicted in Figure 111.1.2. The center
pixel of that matrix that exhibits the highest integrated intensity within the 10 x 10 pixel
array when the moving average is applied to an IA is selected the E for an intersection

area. \Vrth this method the coordinates of an E can be evaluated to within a half pixel.

44

This is one contributing issue in connection with spatial resolution achievable with LEA.

The program implementing the HIS approach is listed in Appendix E.

 

111.1.3 Igpes concerning the two E detection approaches afl the decision for one over
the other

The advantage of the HIS over the RIA approach is that it is more stable and less
sensitive to the 'landscape' of pixel distribution in a LEA image. The disadvantage
associated with HIS is that HIS's detection algorithm is confined to the digital pixel
pattern (resolution) established during the digitization of the images. Furthermore, HIS is
more sensitive to noise in the measurement procedure (for the occurrence of noise and its
impact on accuracy read section 111.3); a consequence of this sensitivity is that the moving
of the highest intensity spot within the intersection area during convection of the IA as a
whole is detected. A necessary criterion for RIA is the presence of a certain contrast in the
images to discriminate between background and IA In terms of the quality factor defined
in section 11.2.3.3 by equation 11.2.d, RIA requires a higher value of this factor than HIS.
RIA also requires a certain extension of the intersection area. An advantage in conjunction
with RIA is that it overcomes the limitation prescribed by the chessboard-like pixel pattern
by Shrinking laser lines to their skeleton representations. All things considered, the RIA
approach yields more accurate results.

As a matter of fact - as experienced during manual data reduction - the human
cognition system (i.e. eye and brain) uses both forms of information: it looks for the
highest intensity occurring within an IA and at the same time incorporates information
about the shape and orientation of the laser line edges that form the IA A phase II

algorithm should try to simulate and adopt human nature behaviour.

Only in an ideal case do the two intersection points as deduced by RIA and HIS fall
together. That is when the two crossing laser lines exhibit the theoretically exponential

45

cross-sectional distribution of intensity and the phosphorescent chemical is distributed
homogeneously within the intersection area. The application of RIA and HIS to a LEA
image generated within a previous project that employed LEA as velocimetry technique
yielded coordinates for Es that did not coincide, however, lay very closely together (see
Figure 111.1.3). It is not necessary that RIA and HIS find their respective E at the same
coordinate pair as long as the axis connecting the two detected ES assumes a consistent
length and position in 2D space for one IA for the flames containing the undistorted and
the distorted LEA information. As soon as this assumption is violated, however, models
have to account for the variation of this axis.

In the mid of laser lines employed in the experiments conducted for this study the
intersection point evaluated by HIS was generally off in x - direction (longer major axis of
IA) when compared to the E detected by RIA It was ofi‘ in the direction towards the

cylinder wall, i.e. towards the origin of the laser lines.

The quality of the LEA images that contain the distorted and undistorted mids which
were acquired within this study was insuflicient to apply the RIA approach, i.e. Q was too
low. In particular, the condition that the contrast between the illuminated areas and the
background being sufficient was violated. This was because of high background noise in
the regions between the laser lines caused primarily by lower order laser light that is
difl’racted at the grating (read Appendix B 'Optical issues' on influence of diffraction).
Thus, for the data reduction of all LEA information, HIS was employed. This implies that
one has to be concerned and conscious about relocation of the highest intensity spot
within the 1A It was possible to perceive this relocation for some intersection points by
eye, essentially for the close-up mids in the investigation of the intake jet; the E as
deduced by HIS moved considerably within the intersection area when observed over
several reference mids. The question emerges of what specifically are the causes for the

movement of the intersection point within an intersection area. Two sources for the

46

movement are thinkable as most prevailing: Vibration of the experimental rig and small
scale convection mechanisms in the flow field where only vibration in the apparatus yields
uncertainties in connection with reference mids. The efiects of convection mechanisms
forcing ES to roam within intersection areas for distorted LEA mids will be discussed
next.

In order to resolve the fluid motion associated with a certain length scale the
intersection area of two laser lines has to be much smaller than this length scale. Fluid
dynamics on length scales that are of the order of the major axes of an IA distort the
intersection area. This causes IAS to not only translate but to also rotate and twist, thus
deform. In this context, incorporating convection of the highest intensity spot within the
intersection area into the translation of the IA as a whole by an application of the HIS
approach, assures LEA'S strict assumption to merely consider translation of a fluid
particle. The movement of the highest intensity spot within the borders of IAS during
convection contains also valid convection information which through HIS is used in the
data reduction process. Deformation of the IA was encountered mainly in the experiments
in connection with the study of the intake jet -- as mentioned above - and there
particularly at the edges of the jet. This data should be analyzed in the awareness of the
problem of relocation of the highest intensity Spot within the 1A AS soon as it is possible
to quantify the mechanisms within the flow field that cause the highest intensity spot to

move within the IAs more precisely, it is legitimate to utilize this data unrestrictedly.

To accommodate partly for the disadvantage in connection with HIS, namely the
restriction to the resolution-imposed pixel pattern, was one reason why coordinates of
intersection points in the reference grid were averaged over several reference mids
yielding a non-integer number for the origin of a fluid particle before convection. Through
the application of the averaging procedure considerable fluctuation of E coordinates for

one IA over several 'undistorted' grids was encountered. This 'phenomenon' is due to the

47

movement of the E within an intersection area; it will be dealt with in more detail in
section 111.3. By averaging pixel intensity over a bell shaped area of 7 x 7 pixel in order to
infer the highest intensity spot in an intersection area (as described above) a firrther
increase in certainty about the actually evaluated location of the E, and thus accuracy, is
achieved. The application of the HIS algorithm enables the investigator to reduce data so
ambiguous -- for which intersection areas exhibit rather flat regions of highest intensity --
that it cannot be reduced properly by hand.

111.1.4 Computational manipulation of LEA images

The LEA images as they are ofl'ered at the output of the video camera, stored on tape,
and then digitized by MegaVision are, as they present themselves in digitized form, of
minor quality concerning pixel intensity distribution. The images require comprehensive

conditioning.

First of all, every other horizontal line contains no intensity information because there
was information available for only every other line due to the manner of data acquisition
(sampling in R8170 mode) as discussed in Appendix C. This problem was taking care of
by just filling in every line that is without information with the content of the neighboring
line. It is likely that better results are obtained by interpolating between preceding and
proceeding lines instead of just copying the neighboring line.

Next, as can be seen in Figure 111. 1.4a the distribution of pixel intensities is very rough.
High frequency fluctuation of pixel intensity with moderately high mamritudes is
encountered in the intensity distribution within a LEA image. The tribute had to be paid
for using a non-scientific video camera purchased flom the commercial video market.
Scientific cameras would offer an additional electronic feature that intelligently prepares
the image for scientific use. The problem was being taken care of by applying a simple
3 x 3 averaging procedure on the image which yields an astonishingly smooth surface of

48

the intensity distribution in the 3D plot; see Figure 111.1.4b. By this averaging process the
lines in a LEA image which carried no information and were replaced by a Simple copying
process (as discussed in the preceding paragraph) are, as a positive side effect of the
averaging process, adjusted to their neighboring lines so that one arrives now at a
smoothed image with 512 x 512 pixel. The disadvantage with averaging procedures is, on
the other hand, that since averaging is of an integrating nature there is a loss of contrast
experienced in the images. Transitions are diminished or at least weakened. This loss of
contrast contributes to the inhibition of the application of the RIA approach for the
deduction of intersection points flom LEA images.

In order to increase the dynamic range and thus the accuracy of the data reduction
process every pixel unit was cut into four sub-pixels named 'quarts'. After the liberation of
the quarts flom their constraint to assume the value of the pixel 'host' a 3 x 3 averaging
procedure is applied on the 1024 x 1024 quarts (stemming flom the 512 x 512 original
array of pixel) in which, as initial condition prior to the averaging process, the quarts keep
the values of their pixel hosts. The coordinates of Es can be evaluated now to within a

quarter of a pixel of the original image.

All three elements that condition the LEA image are incorporated in the FORTRAN
promam that implements the HIS approach as a prestage preparation prior to the
application of HIS itself. For a listing of the software see Appendix E.

49

111.2 Mathematical Formulation of Fluid Mammal Mfifies

111.2.1 Remolds' decomposition as basis for a model to separate sources of velocig
fluctuations

According to Reynolds' decomposition theorem for turbulent flow (Tennekes and
Lumley 1972, Fox and McDonald 1973, Potter and Foss 1982) an instantaneous local
velocity vector can be decomposed into an invariable mean part and a part constituting
velocity fluctuations. The mean velocity can be obtained by different methods, for example

ensemble averaging procedures. Reynolds' decomposition reads:
v = V+v' (III.2.a)

There are three components responsible for velocity fluctuations v' as encountered in
the reduced data of this work, two of which are due to fluid dynamics of the flow field
— cycle-to-cycle variability and turbulent motion of the flow field - whereas the third
component is related to 'noise' in the velocimetry system. Noise is defined to consist of all
errors that are encountered in the data reduction process (e.g. vibration of experimental

apparatus). Thus v' can be decomposed as follows:
v'= v'ac +v'm +v'Nm (III.2.b)

In this study the mean velocity is calculated using ensemble averaging (EA) techniques.
Considering this and plugging 111.2.b into 111.2.a, 111.2.a yields for an instantaneous local

velocity vector as encountered in sections V3 and V.5.2
v = vu + v'ac +v'm +v'Nm (III.2.c)

An efl‘ort has been undertaken in section 111.3 to eliminate the 'noise' component flom
equation 111.2.c. In section V.3.2 an attempt is made to quantify the individual fluctuation

sources v'ac and v'mb which are in this work functions of crank-angle, location within flow

50

field, and engine cycle. Summation of VEA + V'CrC gives a local mean velocity for a

particular crank-angle and engine cycle.

111.2. 1.1 Formulations employed for ensemble averaging

Figure [11.2.1 helps to explain the procedure of ensemble averaging which was
performed to obtain V in equation 111.2.a as mentioned in the preceding section. Sketches
of LEA images are exhibited for one engine cycle (EC; #5) at difi‘erent crank-angles -
horizontal mode - and at one crank-angle (28°) for eight difl‘erent engine cycles -- vertical
mode. The averaging software ensemble averages at one crank-angle and one location
within the flow field over all investigated engine cycles. This yields the local ensemble
averaged velocity vector at that particular crank-angle. EA velocity vectors are functions
of location within the examined flow field and of crank-angle vm' a. 1n mathematical
terms this reads:

_ l N
vloc.C4 = Verb,“ = fizvbcmacj (111.2.d)
j=1

Velocity fluctuations are expressed in terms of a 'Root Mean Square' (RMS) value
evaluated over all v' encountered during the application of an ensemble averaging

procedure. VRMS is calculated according to the following formula:

 

1 N I2
Vans“ = JEZV loc.Gd.EC}
i=1

= J ‘1;— 2”: (Vrocmec, ’ Vega )2

j=1

(III.2.c)

 

51

N in equations 111.2.d and 111.2.e is the number of engine cycles available for the
averaging procedure. AS mentioned in section 11.2.2.4 this number had to be limited due to
confined workspace of the employed computer system.

To find out about the legitimacy of the ensemble averaged numbers the convergence of the
ensemble averaging procedure was checked during the calculations performed by the
software, by monitoring the change of the RMS value of the velocity fluctuations as data
were being added by the ensemble averaging process. From this, satisfactory convergence
was observed during the ensemble averaging procedure where towards the end of the
incorporation of all data available for a certain set, the change of the RMS value
describing the fluctuation of events about the mean was only between 2 and 3% between

the incorporation of two succeeding events.

52

111.3 Critical Assessment of Accuracy Issues for LEA

At the end of the two chapters that introduced LEA as a velocimetry technique and
that described the general data acquisition and reduction procedure, a section discussing
critical issues in connection with the application of LEA in general, and for this study in
particular, is mandatory to follow. Some topics concerning accuracy have already been
discussed in depth elsewhere in this work so that in this chapter, for reasons of
completeness, a mere repetition has been furnished of these matters. - In fact, throughout
the entire text of this work one can find issues concerning the accuracy of LEA that are
being assessed critically. -

LEA is based on the underlying assumption that the influence by convection processes
on the development of the flow field is much larger than the influence by difl‘usion
processes. The parameter that describes the relationship between the flux of momentum
and the rate of mass transport between flow structures is Schmidts number which is
defined by

Sc = (III.3.a)

e|<

where v is the kinematic viscosity and D the mass difiirsion coeficient. Schmidts number
for water assumes a value of about 600. This number indicates that the influence on fluid
flow associated with convection processes with water as the base for the working fluid is
three orders of mamritude higher than that associated with diffusion mechanisms which is
suficient to consider diffusion processes as negligible in this work. For air, Sc is about 1.
Thus, in air LEA has to be careful about the effects enforced on the flow field by
difl‘usion.

53

1113.1 Overview of encountered sources of error in connection with the LEA technique

11131.1 Movement of intersection point within intersection area as main source of error

Artificial or virtual velocity fluctuation enforced by the method of evaluation of an
instantaneous local velocity is intrinsic - to various degrees - to almost all anemometry
techniques; for every technique one has to seek a minimization of this effect. The primary
source of error as it presents itself in this work is established by the fact that the
intersection area (1A) -- marking of fluid particle - does not consist of just one pixel
which would be considered the representative intersection point for the IA, but has
extension and thus a certain intensity distribution within the area. As mentioned earlier a
relocation of the spot representing highest intensity - which is detected by HIS - within
the intensity distribution of an intersection area can be observed for one IA and several
reference mids. This relocation of the E is undesirable and constitute the main error
source. The relocation due to convection as deliberated upon in section 111.1.3 is captured
by the HIS approach and included in the calculation of velocities. In order to separate the
relocation mechanism due to noise in the system flom that due to convection and study
these issues individually, undistorted reference mids of laser lines where only influences
by, e.g., vibration in the apparatus are encountered since the images are recorded with the
engine quiescent, should be considered for the study of the influences of noise on the

relocation of the highest intensity spot. This has been done in the proceeding paragraphs.

It has been observed during the interrogation of reference grids that the intensity
distribution within an intersection area varies flom laser pulse to laser pulse with respect
to highest intensity regions due to a non-homogeneous, vacillating distribution of the
LEA - chemical in the Water/KSCN solution. The alteration of the non-homogeneous
distribution in between laser pulses, specifically within the intersection areas, is caused by

'noise' where supposedly vibration of the experimental rig constitutes the main influence.

54

The redistribution of the LEA - chemical on the molecular level enables the highest
intensity spot to move within the IA Since the HIS approach employed in this study finds
the spot representing the highest intensity within an [A the algorithm finds a different
intersection point with every laser pulse. A quantitative measure for the movement can be
found in Table 111.3.1.

The uncertainty in connection with the relocation of the highest intensity region within
an intersection area accounts for the strongest contribution to the third component of the
threefold velocity fluctuation model, the third component being the contribution of system
"noise" V'Noise to velocity fluctuations in connection with the application of the
velocimetry technique; see section 111.2.], equation 111.2.b.

Statistical law implies that the uncertainty in conjunction with the undesired moving of
the highest intensity region within an intersection area due to 'noise' cancels out for the
reference mids when an average is taken over an infinite number of instantaneous
velocities. Hence, the effect comprising the relocation of the E within an IA vanishes
hypothetically in the EA domain. The mean becomes unafl‘ected by the movement of the
highest intensity region within an intersection area, not the fluctuations. One has to, thus,
pay enhanced and more carefirl attention to the plots that contain instantaneous velocity
data. The uncertainty in connection with noise-related movement of the E is removed as
well in the evaluation of mean velocity profiles in cycle resolved analyses when curve fit

routines are employed to instantaneous velocity profiles as will be done in section V3.2.

In this study reference mids are taken prior to a set of experiments with the working
fluid quiescent. This approach offers some advantages as outlined in chapter H. What the
acquisition of reference data with the engine standing still also simrifies, however, is that
one does not really follow one specific fluid particle during data acquisition of distorted
LEA mids in the experimental phase, as is the idea of the Lagrangian approach assumed

by LEA, since acquisition of initial origin location of fluid particle and location particle

55

convected to are decoupled. This constitutes a source of error because due to 'noise', that
relocates an intersection point within an intersection area, the location of an E within an
IA for a particular intersection of laser lines in an undistorted mid recorded with the
working fluid quiescent is statistically difi‘erent flom the one in the reference mid that
contains the initial positions for an acquired distorted mid and that cannot be captured
because the sampling rate of the imager is not high enough; section 11.2.3. 1. In this work
sets of initial fluid particle locations flom undistorted grids are therefore generated by
statistically averaging over twenty reference mids.

The fact that the acquisition of reference information is decoupled flom the acquisition
of the distorted mids together with the fact that averaging is performed over 20 reference
mid realizations to deduce the initial positions of fluid particles implies, with respect to the
Lagrangian approach employed by LEA, that one follows a fluid particle that presumably
leaves flom a location where one assumes it to be based on statistical probability. Since
one cannot average over an infinite number of LEA mids in the ensemble averaging
process, i.e. remove uncertainty due to 'noise' in the system completely, the uncertainty in
connection with the movement of the highest intensity area within an IA because of 'noise'
remains in the data discussed in this work. This uncertainty is a function of the degree of
fluctuation of the intensity distribution, and therefore movement of the intersection point,

within the 1A as well as of the extent of the intersection area.

There are ways to avoid the above discussed problems: 1) The extension of the
intersection area should be only a flaction of the distance the average fluid particle travels.
(An enhanced efl‘ort on the optical side of the LEA technique should be undertaken.)

2) For every distorted mid its own original reference laser mid at the moment when it is
imposed onto the flow field should be recorded during engine performance so that one
considers the intrinsic origin of the convected fluid particle. (A camera ofl‘ering an

appropriate sampling rate has to be available for this.)

56

111.312 Intersection angle, resolution of LEA image, and dynamic range as issues in the

discussion of accuracy

The fact that for the investigation grids employed in this study two crossing laser lines
do not intersect under an angle of 90° poses another unfortunate condition upon the data
reduction process. Due to the thus established inequality in length of the two major axes in
each intersection area, unequal uncertainty in conjunction with the relocation of an
intersection point within an 1A is forced upon the detection of the intersection point
coordinates for the x - and z - direction. As mentioned in section 11.2.2.1 the laser lines of
the mid that covers most of the engine cylinder exhibit intersection angles of about 50°
yielding a ratio of the major axes of about 3. Thus, because of the ridge-shaped pattern of
the 1A with the longer extension being in x - direction the uncertainty in the deduction of
an E in connection with the x - direction is potentially three times higher. In the x -
direction there is a higher sensitivity to a change in the intensity distribution due to system
noise within the IA

An interesting issue to be discussed can be formulated by the question of how the
number of pixel in LEA images (not necessarily resolution) effects the accuracy of the
data reduction process in conjunction with the uncertainty impelled by the movement of
highest intensity regions within an intersection area. The influence of the number of pixel
in a LEA image on the extraction of accurate positions of intersection points has been
investigated in the following by detecting ES in ten undistorted mids employing the 1118
approach, evaluating the fluctuation in the detection of these E locations for a particular
intersection area, and comparing these results for different LEA images having different
numbers of pixel displaying the mid information. The values representing a measure for
the fluctuation of the detected E locations that will be compared for difl‘erent LEA

images with difl‘erent numbers of pixel in the images, are obtained according to the

 

57

following scheme: For one IA, the variation of the coordinates of detected intersection
points over several (10) reference mids is evaluated in an RMS approach. This variation is
determined for each intersection area in the set of the ten images with the same number of
pixel with subsequent evaluation of the mean of these variations over all intersection areas.
Two representative numbers for the x - and z - direction are obtained as measure for the
fluctuation encountered. The results for four sets of LEA images where undistorted grids
are represented by a different number of pixel for each set are presented in Figure 111.3.1.

First of all, there is fluctuation, indicating the existence of error due to system noise
(ideally the fluctuation for the reference mids should be zero, of course). Secondly, the
eminently higher uncertainty for proper evaluation of Es in the x - direction for cases I, H,
and III is due to the abovementioned fact that the extension of the intersection area of
laser line crossings is greater in the x - direction than in the z - direction (non-
perpendicular intersection of laser lines).

In Figure [11.3.1 the four cases I - 1V represent the following stages of manipulation
(conditioning) of the reference mids: Number 1 depicts the uncertainties about accurate
detection of initial positions in the images containing undistorted mids after all
conditioning steps as outlined in section 111.1.4, namely the filling in of missing
information with subsequent application of an averaging procedure and quartering of pixel
units, have been conducted (1024 x 1024 pixel in image). Number 11 denotes the
uncertainty after filling in lines of information and averaging over the image; no quartering
of pixel units is performed (512 x 512). For number HI only the missing information has
been added to the image (256 x 512). Number IV reflects the uncertainty about the
detection of Es in a reference mid after those horizontal lines that do not contain any
infOl‘rnation have been taken out in the images (256 x 512). This move corresponds to a
Virtual 50°/o compression of the -- real physical world -- engine cylinder in x - direction by
the mapping of the cylinder into an image. Along with this compression in the x - direction

 

58

the extension of the longer axis in the intersection areas is cut in halfand, thus, reduced to
be fairly close to the width of the IA in z - direction.

A very obvious feature in the plot of Figure 111.3.1 is the dramatic decrease of the RMS
fluctuation in connection with the detection of intersection points in x - direction for
case 1V to almost the level of uncertainty experienced in the z - direction. This can be
attributed to the fact that for case IV the length of the intersection areas is only slightly
longer than the width of the IAS and that therefore the accuracy is of the same order for
the two directions. The remaining difference is arguably not due to the difference in the
number of pixel used to display the mid information for the x - and z - direction but is due
to the fact that the axis of the IA in x - direction is still longer than the one in z - direction;
thus indicating the minor influence of the number of pixel in an image on accuracy in
connection with the movement of an E in an IA, since the number of pixel for the z -
direction is twice as high when compared to the one for the x - direction. The increase of
the variations about detected E locations in x - direction as encountered in a transition
flom case IV to case 111 illustrates the disadvantage in conjunction with the elongation of
one major axis of the IA and thus the demand for laser lines to intersect perpendicularly in
order to avoid non-isotropy in the data reduction process. The fluctuation in the
unmanipulated z - direction when comparing 111 and IV is considerably smaller for case 1V
- although for 111 and IV the same number of pixel is used to display undistorted mids
(256 x 512) - since the effect of the modifications performed on the LEA images in the
x - direction for case IV enters the calculations of the fluctuation of E coordinates for the
Z - direction through the use of the 'Gaussian' shaped bell that was utilized to deduce Es
(seetion 111.1.2). The bell couples results obtained for quantities in x and z.

The RMS fluctuations in connection with the detection of Es in intersection areas as
Presented in Figure 111.3.1 appears to be efl‘ected only marginally by the increase in the
mufiber of pixel within a LEA image (see sequence 111 to I in Figure 111.3.1). However,

there is a weak but perceivable trend for the fluctuations that suggests that with increasing

 

59

number of pixel in a LEA image the variations about the detection of coordinates of

intersection points inherent in a sequence of images containing reference mids are
decreasing. - For validity considerations of the numbers in Figure 111.3.1 bear in mind that

a sample size of ten is not exceedingly representative. - With an increase in the number of
pixel in a LEA image two efi‘ects concerning the sought after decrease in fluctuation
intensity and thus increase in certainty are working against each other: Since the centers of
neighboring categories (pixel units) into which highest intensity spots can fall are closer
together in the case where a larger number of pixel are used in an image, fluctuation
decreases with an increase of pixel units. However, on the other hand when using a
Smaller number of pixel in an image, the probability that events fall into the same pixel unit
is more likely. A more detailed study including considerations flom this phenomenon
would be beyond the scope of this study.

In summarizing the above analysis one can say that an increasing number of pixel in a
LIPA image does not effectively improve the accuracy in the detection process of
reference data in conjunction with the attempt of elimination of errors due to system noise
that forces the region of highest intensity to move within an intersection area. Of course,
an increase of the number of pixel representing distorted and undistorted mids improves

the accuracy of the data acquisition process as a whole (e.g. dynamic range) and that is

desired.

The accuracy of LEA in general is strongly dependent on the resolution that can be
aOllieved during digitization of the flow field. For the ‘whole cylinder' approach 7 pixel
re11>r¢iesent about one millimeter in a digitized LEA image. A maximum distortion of
1 - 3 mm in real space was allowed in one direction in order to ascertain linearity between

two Points. This 1.3 mm is approximately represented by 10 pixel alter digitization. The
Macy in connection with the detection of intersection points in the digitized images is
h. with respect to the pixel scheme in the images - a quarter of a pixel after quartering

¥

60

pixel units (as was outlined in section III.1.4). This leads to a hypothetical dynamic range
of 1:40 - based on velocity magnitudes - if one considers 10 pixel of maximum
convection for a fluid particle in each direction. The actual effective dynamic range is

smaller as will be delineated in section 111.3 .3 when uncertainties due to the movement of

the highest intensity spot within the IA are included.
Certainty about the inference of Es from IAs is enhanced due to the facts that

l ) averaging was performed over several reference grids in order to find the initial
locations of the fluid particles which frees the HIS approach from its restriction to assign
coordinates of illuminated particles within the pixel pattern, and 2) the E detection
algorithm (HIS) examined arrays of 7 x 7 pixel to find highest intensity spots, thus,
including information of each pixel's environment yielding a higher accuracy than the

nominal one of a quarter of a pixel.

III- 3.1.3 Miscellaneous sources of error

In any data acquisition and reduction system unavoidable sources of error are

encountered of whose extent the investigator should be aware. Some of these sources of

error which enter V'Noise in equation I]I.2.b are addressed in this section.
When data are to be ensemble averaged, corresponding data blocks have to, in the EA

Procedure, exhibit identical engine speeds and have to be drawn from equal crank-angle

Positions of the piston. There are uncertainties encountered in this context for this study
due to instabilities in engine speed in the experimental phase and due to dimculties in
proper assignment of LEA images to crank-angles during the data reduction process
(mot e on these difliculties can be found in Appendix C). Also, problems arise when two
map 8 of instantaneous velocity vectors from difi‘erent engine cycles or difl‘erent sets of
data are to be compared. The results in the maps are ofl‘ marginally by engine speed and

Cllam-IICNangle.

 

61

Uncertainties about the absolute values of the velocity vectors arise due to reading
errors in the calibration process using a white square in the determination procedure of

appropriate scaling factors (pixel/mm; Appendix D). This error is unimportant in the
comparison of the results which stem fiom an application of one particular investigation

grid, e. g. laser grid for the study of the whole engine cylinder (fixed focus length of the

imager).

Further elements influencing the accuracy of LEA include the size and the shape of the
'Gaussian' bell used to detect Es in the HIS approach, depth of field, change in index of
refiaction of liquids due to change in ambient temperature and evaporation of constituents,

perturbation of the intake flow due to vibration of the experimental rig, etc.

The contribution of all these error sources can be considered negligible in the face of

those delineated in the preceding two sections, but are summarized here for the purpose of

completeness.

111-3 .2 Actual gualig of LEA - data

The HIS software itselfhas demonstrated to deliver suflicient accuracy, that is to say
the algorithm itself is exact, i.e. it finds the region with the highest occurring intensity
Within the assigned area (10 x 10 pixel). The decision and confirmation for HIS as being
the 11’ detection @proach of choice in this study, however, has to be made on grounds of
an analysis incorporating the issue of how well the detected highest intensity spot serves as
a representation for an intersection area. Or rather what is the degree of the variation of
detected Es for a particular IA in consideration of several reference grids. That leads
itulllecliately to the question: What is the degree of quality of the raw data -- LEA images
.. necessary so that the HIS approach yields results that are of acceptable accuracy? What

a
qua-11y is 'acceptable accuracy'?

62

For reference grids the intersection points deduced by HIS should be for one 1A
consistently at the same pair of coordinates, thus the detection of Es in images of
undistorted grids should essentially be fluctuation-free. That was not the case in this work

as could be seen in section 1113.12 (otherwise the discussion of issues describing the

accuracy of LEA would be a brief one).

With LEA, a quantification of the (in)accuracy encountered in the data reduction
process due to relocations of an intersection point within an intersection area can be
conducted based on the images that contain the undistorted grids. The fluctuation in the
detection of E locations for these grids should be zero so that any value difl‘erent from
zero encountered in the following analysis of reference grids can be considered inaccuracy
induced by the measurement technique. The quantification of the inaccuracy of reference
data - used for the generation of fields of velocity vectors in chapter V -- as it is reduced
by employing HIS, was based on the approach used to obtain fluctuation measures to
demonstrate the efi‘ect of the number of pixel in a LEA image on data reduction in
section 111.3. 1.2. The presented numbers in Figure III.3.2a and b are, thus, for each set of
engine parameters and direction (x and 2) an average of RMS fluctuations over all
fluctuations encountered at the individual intersection areas in the reference images. In
faCt, the evaluation of the RMS fluctuations was a by-product stemming from the
calculation that determined the mean coordinates for initial positions of fluid particles in a
LIPA grid conducted over 20 reference grids. Figure III.3.2a holds a comparison of

RMS fluctuation values for the uncertainty in connection with the relocation of an E
Within an IA for the laser grid with which the whole engine cylinder has been examined. In
Figul‘e III.3.2b the fluctuation numbers are exhibited for the laser grid with which the
intake jet has been examined (close-up grid).
Fl"om Figure III.3.2a which contains the fluctuations encountered in the reference grids

f0 . . . .
r the lnvestlgatlon of the entire engine cylinder one can see that the fluctuations in

63

connection with the x - direction are about four times higher compared to those in the

z - direction. (The sequence shown is for the three valve lift cases examined at 30 RPM
engine speed; Table 11.2.3.) This is an immediate reflection of the difference in length of
the major axes within intersection areas for the x - and z - direction even if the
encountered value of four does not match exactly the one of three found for the ratio of
the major axes of an IA earlier in this chapter. In the z - direction the inaccuracy is
marginally higher than the resolution with which HIS detects intersection points, namely a
quarter of a pixel (see section III. 1.4). In the x - direction the uncertainty is in the average
one pixel and a half, a considerable vagueness in the determination of a velocity vector.
The values describing the RMS fluctuation for the x - and z - direction are, to within the
accuracy obtainable in conjunction with a sample size of 20 reference grids, essentially
congruent when considering all nine sets of data stemming from the nine engine parameter
settings (Table 11.2.3).

The values for the RMS fluctuations associated with the close-up grid of laser lines in
the investigation of the intake jet in Figure III.3.2b are twice as high for the z - direction
and about 1.5 times as high for the x - direction when compared to the corresponding
values for the investigation grid applied to the entire engine cylinder. The difi'erence in the
Pr0portionality factors for the x - and z - direction is due to the fact that the major axes of
the IAs encountered in the laser grid that is employed for the investigation of the intake jet
do not coincide with the x - and z - direction of the reference frame of the engine cylinder
(see Figure 11.2.3b), thus decreasing the ratio of major axis in x - direction to the one in

z ‘ direction. The overall increase in uncertainty for the close-up laser grid is experienced

because an intersection area consists of about twice as many pixel in connection with this
do Se-~up grid when compared to the 'whole cylinder' investigation grid. There is a clear
correlation - almost linear - between the size of an intersection area and the degree of
a‘cclll‘acy, as should be anticipated. Uncertainties of around 2.6 pixel as stated in
Flg‘ll‘e III.3.2b for the x - direction is considerably high. A fluctuation of 2.6 pixel means

(,4

that the flow direction of fluid particles that convected a distance of less than 2.6 pixel

- which corresponds to 0.34 m/s for 20 RPM engine speed and 9 mm valve lift -- in the

digitized image cannot be resolved, i.e. 2.6 is the threshold for direction ambiguity. The

values for the three cases examined in connection with the close-up grid are for the x - and

z - direction congruent again to within the certainty achievable by a sample size of 20.
Overall, the accuracy in the z - direction, particularly for the 'whole shot' laser grid, is

satisfactory which is why velocity fluctuation data are being firrnished only for the

z - component.
The reader should bear in mind that the numbers for the fluctuation in the determination
of Es in reference grids in the previous paragraph are averages over the entire area of
investigation. If one breaks these average values up and discusses the distribution of
uncertainties in connection with individual intersection areas over the area covered by a
laser grid one notices quickly that the distribution of the inaccuracy is dependent on
several parameters. For instance, the uncertainty about the detection of the accurate E in
an IA increases with the penetration of the laser lines. For 20 RPM engine speed and
3 mm valve lift the distributions of the RMS fluctuation - representing the variation in
conjunction with the detection of intersection points in reference information - are for the
close-up laser grid exhibited in Figure [[133 for x - and z - direction. In the plots the laser
lines enter fi'om light. The gridlines in the x, z - plane of the 3D needle plot depict the

orientation for one set of parallel laser lines. The line-up of the needles in the plot under

450 denote the other set of laser lines. It can be seen that the uncertainty increases - in the

average - with penetration of the laser line, however, does not increase with penetration,

spealting in mathematical terms, in a monotone way. The ups and downs in the measure
1"
or the fluctuation for one laser line can be explained by the fact that not all laser lines

Within one set of parallel lines exhibit equal intensity, as outlined in section H.232 and

65

Appendix B. This generates IAs of difi‘erent intensity distribution along one laser line and

thus different values of fluctuation.
The characteristics of the fluctuation profiles for the laser lines with penetration are, as

one should expect, essentially the same for the x - and z - direction (Figure III.3.3a & b).
The average value of 'four' found previously describing by how much the uncertainty is
higher in conjunction with the x - direction compared to the one in the z - direction is
reflected by Figure [1133. When considering the individual uncertainties associated with
the individual intersection areas it turns out that there are proportionality factors between

the uncertainty in z and x in the range of about 2 to 5.

Some security check mechanisms have been implemented to avoid the most crude

mislocations of intersection points within intersection areas.
[During the E detection process a detected highest intensity spot that lies outside the

assigned area of 10 x 10 pixel is discarded. Recall that a convection of 10 pixel was

assumcd to be the limit within which the convection can be considered linear. This limit

Was set empirically by watching laser lines distort on a TV monitor.
ClAt the final computational level, a filtering procedure is applied in order to eliminate

Stray velocity vectors. If the velocity fluctuation of a local instantaneous velocity vector is
higher than 3 times the RMS fluctuation encountered at this location during the ensemble
a“'eraging process, this vector will be removed. The value of 3 was determined empirically.

\131 - 3,3 Estimation of errors in plots of velocity vectors

Following will be given a rough quantification of the order of errors encountered in the

cli‘Verse presentations of fields of velocity vectors within this work.

F or velocity plots containing instantaneous velocity vectors one has to add the
meettainty associated with the fluctuation of the detection of the intersection point due to

 

66

the movement of the highest intensity region within an intersection area - as introduced in
section 111.3 .2 - to that associated with image resolution. Two globally averaged - over
all examined intersection areas - numbers for the uncertainty in the detection of the E
due to relocation of an intersection point within an IA for the x - and z - direction are
determined from an analysis of reference grids for each of the two laser grid realizations
(‘whole cylinder' and close-up) independent of engine parameter setting. These numbers do
not, naturally, reflect the exact fluctuations in connection with each individual intersection
area. However, they can be considered good approximations for each IA For the 'whole
cylinder’ and the close-up investigation grids the measures for the uncertainty in the
detection of the E due to system noise are depicted in Table 1113.1. The summation of
the uncertainties for the two sources of fluctuation mentioned in the first sentence of this
paragraph is performed in this table. The numbers in row three are the actual measures for
the uncertainties of the velocity vectors in pixel which will be denoted along with a
1’ eference velocity vector in the maps depicting velocity vector distributions in chapter V.
From this row it can be inferred that the best accuracy attained in this application of LEA
has a dynamic range of about 17. Numbers in row four of Table III.3.1 show errors in
Percent when referring to the maximum convection allowed for fluid particles for each

direction of 10 pixel.

Table III. 3.1: Uncertainties in the maps of velocity vectors

 

 

 

 

 

RMS Uncertainty in pixel

Field of Uncertainty 'whole cylinder' grid close-up grid

\ due to x I z x z
‘ ‘ - instantaneous mov. ofE in IA 1.53 0.38 2.61 0.92
' e resolution 0.25 0.25 0.25 0.25
Ell-EB 1.78 0.63 2.86 1.17
error in percent 17.8 6.3 28.6 11.7
' ‘ ‘ elaserrfirle averaged.. 25% of row 3 0.45 0.16 0.72 0.29

- - - Velocity Vectors

 

 

67

In the EA domain both uncertainties, the one by limited image resolution and the one
due to movement of the highest intensity spot within the IA, are mainly filtered out during
the ensemble averaging procedure, however, certainly, not completely. As a first,
conservative approximation one can assign an upper limit for the error for ensemble
averaged velocity vectors - to be on the safe side - to about 25% of the original
uncertainty. This is a rather arbitrary, yet reasonable pick. The accuracy for an ensemble
averaged velocity vector results fi'om the values in the third row of Table 1113.1 and is
written down in the fifth row of the same table.

Although, e.g., vibration characteristics within the engine apparatus and thus relocation
issues of Es in IAs are difi‘erent - to say the least - when the piston of the engine is
moving as opposed to when the engine is quiescent, the values of the quantification of
RMS fluctuation encountered during E detection at individual IAs in reference grids as
delivered in section 1113.2 were used as a first approximation for V'Noise in an attempt to

Cleanse EA velocity fluctuations defined by equation III.2.b by subtracting the values
found in Figure [[133 (row 3). For a more accurate quantification of this uncertainty one

Will have to implement a model for relocation processes of an E within an intersection

area due to, among others, vibration in the apparatus on a length scale comparable to that
of 1111c major axes of an intersection area to find the exact correlation between relocation
mechanisms and the associated uncertainty about the intersection point, which then can
lead to an accurate removal of this contribution of system noise. In this study, however, a

first approximation must suffice.
The primary difi‘erence between the uncertainty in connection with image resolution and
the one with the motion of the highest intensity point within an IA is the fact that the
f()I‘IIIer is homogeneously distributed over the area of investigation (at intersections of
laser lines) and that the latter is dependent on the length of penetration of a laser line

to
8etimer with the light intensity exhibited by an individual laser line (keyword: size and

 

68

pixel intensity of intersection area). Whereas the influence of image resolution on the
uncertainty in velocity vectors increases the level of velocity fluctuation due to system
noise evenly by adding the same constant at each location within the area of investigation,
the influence of variation in the evaluation of E coordinates for one IA and several LEA
grids on the uncertainty will lead to a 'distortion' of velocity fluctuation information
throughout the field of investigation. In this work (section V.5.3) only the uncertainty in
conjunction with relocation processes of the E within an IA is removed fi'om EA velocity
fluctuation information in order to undo this 'distortion'. Since no precise measures for
particular uncertainties with respect to system noise are available for this investigation, no
efi‘ort has been undertaken to adjust for the exact level of EA velocity fluctuation

magnitudes.

Chapter IV
THE INTAKE FLOW FIELD IN THE AXISYMMETRIC
ENGINE EXAMINED BY FLOW VISUALIZATION

The fluid dynamical process which is examined in this study namely the intake stroke of
a four-stroke internal combustion engine is a reciprocating, i.e. periodic, hence unsteady
one. The intake stroke features very complex timing and geometry of fluid flow. It
therefore results in a fluid dynamically complicated flow field, which incorporates several
difi‘erent flow structures and instabilities of flow patterns which quantitatively -- and
sometimes to a certain extent even qualitatively - difi‘er from intake stroke realization to
intake stroke realization. The application of LEA results in velocity vector information in
two-dimensional maps of finite resolution. Thus, the turbulent motion on scales smaller
than the resolution of the mesh size of the laser grid (i.e. fine scale eddies) cannot be
detected in the flow fields of induction strokes when employing LEA However, the large
and intermediate scales generated (as we shall see) are resolVed directly from the
two-dimensional information. Therefore, instantaneous velocity vectors obtained do not
contain the small scale information, but as we hope to show using flow visualization, do
contain the essential information embodied in the important large scale structures that
develop.

The need to determine the dominant scales of the flow structures to gain an
understanding to the resolution required in the LEA grids was the primary motivation to
p erfOrm visualization. A secondary motivation was to use the visual data to help interpret
the Quantitative velocity and vorticity data that LEA would give. By considering velocity

maps as quantification of flow visualization, one is more likely to arrive at the correct

69

70

interpretation and understanding of the flow field. In general, it is always desirable to have

a visual image as analog information of a flow field.

Two visualization techniques have been used which are alike with respect to the
visualization concept they are based upon, and different with respect to the assignment of
the position of an observer who views the flow. The analogy is derived flom the fact that
both techniques are based on visualization by excitation of photoluminescent molecules
dissolved in the working fluid. The difl‘erence is mainly manifested by the distinction in the
frames of reference when describing the flow field: For one technique the approach is
Eulerian (reference flame rests relative to the flow), a fluorescent dye was used here; for
the other it was Lagrangian (reference flame moves with flow), a photochromic dye was
the marker for the fluid particles in this case. For the Eulerian approach -- where a
fluorescent dye is employed that emits radiation only during the time of excitation by an
energy source - the instantaneous situation of the flow field is recorded; with this
technique one fleezes the flow at a certain time by taking a snapshot of it. When using
photochromic dye -- which emits energy for a certain time afier the molecules have been
excited - as for the Lagrangian visualization approach it is possible to trace particular
fluid particles in the flow field as the flow field evolves (which is actually LEA's intrinsic
idea). The latter approach naturally yields a better understanding of the development and

restructuring of flow patterns inherent in the flow field.

The experiments involving the photochromic dye were performed at 1.5 RPM engine
Speed (25.5 RPM in air) and 3 mm VL. The experiments involving the fluorescent dye
were conducted at 20 RPM engine speed (340 RPM in air) and also 3 mm valve lift.

71

IV .1 Results flom Application of Photochromic Me

IV. 1.1 Experimental procedure

The chemical used was a photochromic dye, namely 1',3',3'-Trimethylspiro-8-nitro(2H-
l -benzopyran-2',2'-indoline) flom Kodak (reference-# 52590). This dye changes from
being transparent to emitting upon excitation, radiating visible, blue light for a lifetime of
several seconds. It requires an alcohol or an oil as its carrier for which several candidates
had been considered and tested. There were various demands towards the carrier of the
dye, one being that the index of reflection was to be close to that of the quartz engine.
Furthermore, the fluid had to be transparent in the UV range of the spectrum of electro-
magnetic waves, the dye had to dissolve thoroughly in it, and an applied excitation to the
flowing medium had to yield both high intensity radiation and high penetration. The
chemical complying with basically all cited requirements is Decahydronaphthalene
(C10H18)- The index matching agent was Hexane (C6H14) as had been outlined in
section 11.2.1.1. The concept of visualization in connection with the 'photochromic' dye
-- which is a chemical that consists of complex molecules - is to excite the dye molecules
dissolved in the working fluid with an energy input which is, for instance, implemented by
a laser. The molecules are twisted by the applied energy; in this state the chemical changes
color which can be made visible by exposing it to regular white light. As soon as an

excited molecule releases its energy by assuming its original shape it loses its ability to

'reflect' the blue in the incident white light.

F or the experimental series, excitation of molecules was mainly achieved at the throat
of the fluid reservoir (quartz chamber head, Figure 11.2.1a) by a focused laser beam
(erimer laser, 308 nm) which produced coloration in the dye by pulsing at a rate of
1 00 Hz. Since the valve is made of quartz the applied laser beam excites fluid particles in
two planes on either side of the valve which are 180° apart in circumferential direction.

 

72

With respect to the pictures in Figure IV. 1 . 1, the laser beam enters flom the right side and
is partially absorbed by the dye, generating a higher level of intensity on the respective side
of the cylinder. Prior to its application to the engine throat, part of the laser beam's energy
was split ofl‘, focused, and passed vertically through the piston head at the cylinder axis of
the engine. An incandescent light bulb provided the white light that visualizes the excited
molecules. Between light bulb and engine cylinder a rectangular piece of white glass was
installed in order to diffuse this white light more evenly over the test area. The
photochromic lifetime of the chemical is of the same order as the time necessary for one
realization of an induction stroke, so that all phosphorescence generated during the
induction stroke fades before the beginning of the next engine cycle, allowing data to be
taken in a reciprocating manner.

Results of the flow visualization experiments were recorded by the I-Ii-8 Canon L1A
camcorder. There was no controlled timing between the camera and the engine rig. The
camera took pictures of the flow field at its standard 30 Hz sampling rate. From the video,
the start and the end of the induction stroke can be determined by observing piston
motion, and with the knowledge of the data sample rate the engine speed can be evaluated
and the individual images can be assigned to crank-angles. Due to the low intensity level
of the reflected blue light flom the 'indirect' (direct laser light prepares the molecules for
excitation) white lighting, the exposure time had to be set at 33 .33 ms (i.e. shutter of

camera was open) to compensate for the low intensity level, thus valuable results could
only be obtained at very low engine speeds (1.5 RPM) - for low engine speeds the fluid
dynamics of the flow field are associated with large time scales. At higher engine speeds
these long exposure times yield firzzy images.
Although the engine speed of 1.5 RPM which corresponds to 25.5 RPM in air is too
low to be considered relevant to practical engine research, the results at this speed have,
nonetheless, shown some relevance. In particular, even at this low speed the overall flow

stl‘llcture, incorporating vortex rings and toroidal recirculation patterns, was found to be

73

the same as at higher engine speeds (see results flom employment of fluorescent dye at
20 RPM engine speed in section IV.2 and of LEA in chapter V). One major reason why
the results of the experiments with the photochromic dye at this low engine speed will be
discussed is that in the experiments employing the photochromic dye information of the
flow will be available throughout the entire induction stroke and within the whole engine
cylinder which is not the case in connection with the application of the fluorescent dye.
The photochromic experiments were conducted using a line of laser light above the valve
to mark fluid particles, whereas the fluorescent dye experiments were conducted using a
sheet of laser light (see section IV.2). This leads, in connection with the photochromic
experiments, on the one hand to illuminated flow within the entire engine cylinder, and on
the other hand to higher contrast between illuminated and non-illuminated phases of the
flow field.

IV. 1 .2 Flow visualization results

The experiments with the photochromic dye unveil flow structures of the flow field
during induction at 1.5 RPM engine speed that are similar to those discovered by other
investigators at a diverse range of engine speeds. All investigators found the global
structure of the flow field during intake to consist essentially of two toroidal recirculation
patterns. One recirculation pattern occupies most of the cylinder volume (main toroidal
recirculation pattern [TRP]), the other rotates in the corner between cylinder wall and
cylinder head (upper comer toroidal recirculation pattern). The latter is smaller and rotates
in the opposite direction when compared to the main pattern. The results of the flow
visualization experiments with the photochromic dye, an example of which is shown in the
pictures of Figure 1V. 1. 1, document the genesis of these features, as well as their role and

significance in the development of the flow field as a whole.

74

However, beyond the results of all investigations on flow visualization in axisymmetric
engines known to the author and mentioned in the introduction, this technique can obtain
a coherent picture of the flow field incorporating diverse flow structures on much finer
scales than has been previously available. For instance, the roll-up of the vortex rings of
the incoming jet as encountered in annular jet flows and described in section 1.1 of the
'Introduction' has, to the knowledge of the author, never been visualized before in engine
research and can be seen in Figure IV. 1. 1. The toroidal vortex rings start to shed flom the
axisymmetric intake jet at about 30° crank-angle (Figure IV.1.ld, Re=158).
Furthermore, Regan et.al. (1987) mention in passing that the angle under which the
intake jet enters the cylinder (induction angle) oscillates between 10° and 70°. In the
experiments using the photochromic dye a variation of the induction angle is clearly
perceivable: At first the angle inclines towards the cylinder axis due to the accelerating
piston (Figure IV. 1. la-e); the jet is then pushed back towards the cylinder head when the
main recirculation pattern visibly gains strength (a quantification of the fluctuation of the
induction angle will be given in section V. 1.2). Moreover, it can be observed, especially
for the 3 mm valve lift case shown in Figure IV. 1. 1, that what is referred to as toroidal
recirculation pattern in the upper corner between cylinder head and cylinder wall can only
be called a recirculation pattern after midstroke. Before midstroke the flow does not
recirculate in the upper comer. A confirmation for this circumstance is attained in
section V.5 when the intake jet is examined in more detail. Also, a distinguishable lump of
dye could be followed visually in one recorded engine cycle as it enters with the intake jet,
then rotates with the upper corner TRP, subsequently joins the main recirculation pattern
and finally ends up in a 'bufi‘erJ area (for more detail on the 'bufl‘er' area refer to
section V. 1. 1) above the piston head indicating that a fluid particle is not constrained to
one particular flow structure, and hence confined to the region of that structure's

evolution, in the flow field.

75

For a detailed description of the initiation of flow patterns and continuance of their
development for the entire engine stroke, the reader is referred to section V. 1.1 where for
20 RPM engine speed and 3 mm valve lift the flow structures encountered during intake
are presented in terms of maps of velocity vectors. The description includes the evolution

of the main recirculation pattern.

During the examination of the flow, an interesting vortex structure was observed: A
slim vertical vortex located at the cylinder axis and with its 'foot' or 'base‘ on top of the
piston head. Due to its visual appearance it was named a tornado vortex. The tornado can
be seen in Figure 1V. 1 . 1f through i. The vertical laser beam through the piston head was
implemented so as to visualize this tornado. According to observations, the rotating
tornado appears to be generated by a slight asymmetry of the annulus intake jet flow due
either to a non-perfect axisymmetry of the geometry of the model configuration (for
example, slight misalignment between the axes of valve and cylinder causes asymmetry),
or, possibly, an asymmetry in the growth rates of the instabilities forming at different
azimuthal positions. The tornado starts spinning as soon as the perimeter of the non-
axisymmetric main recirculation system reaches the cylinder axis at any circumferential
position 9; in the case of the pictures in Figure IV. 1.1 this occurs at a crank-angle of about
53° (IV. 1 . It). The spinning continues throughout the intake stroke indicating that the
vortex is stable for the entire intake stroke. To quantify the tornado, its number of
rotations was counted for difi‘erent engine speeds and valve lifis. Interestingly, it turns out
that the number of rotations stays constant for all cases at about 6 per induction stroke
realization; a definite rotational direction could, however, not be assigned for the tornado.
The constant number of spins performed by the tornado during induction means that the
vorticity of the tornado increases proportionally with increasing engine speed. The
evaluation of the number of rotations of the tornado, however, was a very crude

determination of spins by eye and the result has to be treated with caution. LEA studies

76

of the horizontal plane should depict and quantify the tornado. Eaton and Reynolds (1989)
observed a similar ('spiral') vortex in their flow visualization experiments in an

axisymmetric configuration at 1000 RPM in air between crank-angles 72° and 87°.

With these flow visualization results (Figure IV.1.1) it is possible to study
axisymmetrical aspects of the toroidal intake jet for the two-dimensional plane considered.
It can be asserted flom this study that at an engine speed of 1.5 RPM the two parts of the
intake jet which are 180° apart in the circumferential direction and visualized by
experiments, exhibit axisymmetry with respect to the overall behaviour of the incoming
annular jet (for example shedding of secondary vortices, induction angle, etc), however,
show considerable differences in size of flow structures and timing of fluid dynamical
events. The reasons for these asymmetries are mentioned above. They can never exist in
results stemming flom the application of numerical or analytic models to (potentially)
axisymmetrical flow fields.

In general, fine scale structure can be distinguished easily for the 1.5 RPM engine
speed, especially during the early intake phase and in the region of the intake jet,
throughout the whole induction stroke. A vortex ring created by the annulus jet flow on
the side of the jet facing the main recirculation pattern, for example, can be pursued by eye
flom its genesis to its collapse (this can be done because the photochromic dye has such a
long excitation time, and allows Lagrangian information to be obtained). Vortex rings roll
up at the interface of the intake jet with the surrounding fluid, grow, mix with the main
TRP, deform, and eventually are embedded and merged with the TRP where entrainment
is indicated by the fact that the dye depicting the vortex structure dissolves entirely. This
process essentially represents the large scale mixing that takes place during the intake
stroke.

The entrainment of an expanding vortex ring into the main recirculation pattern can be

seen in some particular cases as a source of an immediate supplement of angular

77

momentum to the main TRP - as opposed to the overall source emanating flom the
inbound turning of the inducted flow along the cylinder wall (see section V. 1.1). In order
to investigate how an immediate supply of angular momentum (that is stored in a vortex
ring) to the angular momentum of a TRP can be achieved, the video sequence visualizing
the intake jet for a valve lift of 6 m (not shown in this work) was thoroughly analyzed
and the following fluid dynamical mechanism could be observed: Qualitatively, the vortex
rings formed by the intake jet for 6 mm VL sustain their angular momentum - acquired as
the intake jet convects downstream - after the impact of the jet on the cylinder wall and
distribute their rotating motion in the TRP by expansion of their volumes in radial
direction.

Afier thorough analysis of the flow visualization results for all valve lifts examined it
tums out that the contribution to the angular momentum of the TRP stemming flom the
integration of the angular momentum of vortex rings seems -- at this low speed - to be
strongest for the 6 mm valve lift case. For 3 mm VL this effect is observable as well;
however, the majority of vortex rings get destroyed before they can contribute their
angular rotation to the main TRP.

At 9 mm valve lift the vortex rings are not fully developed before the intake jet turns at
the cylinder wall and thus, the observer does not see the efl‘ect of the coupling of the
vortex rings with the recirculation pattern as clearly. It can be observed in the flow
visualization pictures that the expansion of the vortex rings created in the case of a 9 mm
valve lift is suppressed by the rotating motion of the two recirculation patterns to either
side of the intake jet. For 9 mm VL the intake jet has its momentum most widely
distributed of the three cases examined due to the lowest centerline intake jet velocity at
the orifice, thus transferring the least amount of momentum onto its shedding vortex rings.

Therefore, the vortex rings associated with a 9 mm VL are more susceptible to effects of

the boundary conditions.

78

It would be very interesting to achieve the same visualization quality - that enables the
experimentalist to trace vortex rings - at higher engine speeds in order to document the
history of vortex rings and evaluate their behaviour as they deveIOp with the incoming
annular jet in regard to their contribution to the main recirculation pattern. A study at

higher RPM is presented in the next section.

IV.2 Results flom Application of Fluorescent we

Experiments with fluorescent dye were conducted at 20 RPM engine speed,
corresponding to 340 RPM in air, and a valve lifi of 3 mm. The dye used was 'fluorescein',
and the fluid was water mixed with KSCN, to reflactive index match water to the quartz
apparatus. Fluorescein dye emits, upon excitation by a source of energy, electro-magnetic
waves in the green range of the electro-magnetic spectrum of waves. The lifetime of green
emission - fluorescein being a fluorescent dye - is of the order of the time interval during
which the energy source is applied. The source of energy was for this series of flow
visualization experiments a sheet of laser light. The laser employed was the excimer laser
at 308 nm wavelength (UV, XeCl); recall that the time for one laser pulse - and thus
approximately the time of fluorescence - is 17 us. The sheet of laser light is spanned by
focusing the rectangular laser beam with a cylindrical lens to the site of investigation; it
intersects the axis of the engine cylinder. The sheet enters through the right wall (when
considering Figure IV.2.1), explaining areas of local overexposure of the film in the near-
wall region to the right of the cylinder axis. The fact that the fluorescent molecules emit
the electro-magnetic waves of the applied laser energy source results in a higher level of
illumination when compared to the case of the photochromic dye in section IV. 1. The
lifetime of the chemical (or the temporal window for the application of the laser sheet) is

here equal to the efi'ective exposure time of the imager system, independent of the time the

Shutter of the camera is in an open state. Flow visualization experiments could therefore

79

be conducted at higher speeds since short exposure times in connection with a high level
of photoluminescence guaranteed that fine flow structures could be resolved.

Flow visualization results were captured by a Photo-Sonics 4ML 35 mm movie camera
(series 2000). The rate at which flames were recorded was 60 Hz achieving a resolution of
2° crank-angle for 20 RPM engine speed. The camera triggered the laser externally in
order to ensure that the entire illumination radiated flom the scene was sensed by the film.
The medium used for storage of the visualization information was an ASA 1000 color
slide film (E6 process). The actual time the film was exposed to the scene of the evolving
flow field was 1.89 ms per flame.

Following are the revelations of flow structures as visualized by two difl‘erent
realizations of flow visualization experiments employing fluorescent dye, each of which

consisting of a different mode according to which flow marker were supplied to the site of

investigation.

IV .2.1 Flow visualization Mode I

Data acquisition was performed as follows: The engine cylinder was filled with a pure
index matched water/KSCN solution prior to the start of piston motion, and the reservoir
above the (closed) valve was filled with an index matched water/KSCN solution in which
fluorescein was dissolved thoroughly. At the beginning of piston motion marking the
initiation of the intake stroke the valve was opened and fluid flom the reservoir was able
to enter the cylinder. In chemical terms there is a two-phase flow where only one phase is
Optically sensitive to an energy input so that upon excitation the incoming jet flow can be
distinguished flom the non-sensitive fluid in the cylinder. Fluid dynamically, on the other
hand, the fluid consists of only one phase since the concentration of the dye is negligibly

low (10 drops flom a pipette into 800 ml aqueous KSCN), hence fluid dynamical mixing

80

of the two chemical phases is neither inhibited nor encouraged by concentration
differences.

The disadvantage with the visualization technique involving the fluorescent dye in an
application according to the concept described above is that after a certain crank-angle the
concentration of the dye in the cylinder becomes so high that the contrast between the
flesh incoming fluid and its surroundings, especially in the region where the toroidal
recirculation pattern rotates, becomes too low to resolve flow structures comfortably.

Since the approach with the fluorescent dye is Eulerian the images recorded can be
regarded individual 'snapshots' of the flow field, i.e. it is not possible to follow one
particular fluid particle as it convects in the flow. These snapshots are exhibited in
Figure IV.2.1. In Figure IV.2.1a a sheet of light can be discerned that spreads on top of
the piston head. Leakage of dye through the closed valve prior to piston motion settles on
top of the piston (quartz on quartz does not provide superior sealing). This fluorescent
chemical illuminates upon excitation by the laser and results in the sheet of light above the

piston head.

For early intake the images in Figure IV.2.1 confirm the revelations of flow field
structures attained flom the experiments with the photochromic dye discussed in
section IV. 1.2. The images exhibited by this higher engine speed (20 RPM) - featuring
the creation of coherent vortex rings by means of a roll-up of fluid at the boundary
between incoming jet and ambient fluid, the fluctuation of the angle under which the
incoming jet enters the engine cylinder, the impingement of the intake jet on the piston
head, the formation of the main toroidal recirculation pattern, and so forth -- support the
Statement made in section IV. 1.1 that the basic overall characteristics of the flow fields at
1 - 5 and 20 RPM are firndamentally identical. Difl‘erences can be found however, in the
time scales for the onset of the waviness of the incoming annular jet, and the time at which

Vortex rings begin to shed flom the edge of the jet.

81

The abovementioned sheet of light Spread on top of the piston head can naturally serve
as flow marker as well. One can observe this illuminated sheet as it gets compressed to a
lump of fluid located at the cylinder axis by the generation of the main toroidal
recirculation pattern which is supplied with momentum by the incoming jet. Interestingly,
through crank-angles 16° to 32° for which the sheet of light can be discerned as it gets
compressed (Figure IV .2. lb-e), there is at the cylinder axis no motion in negative 2 -
direction (towards the valve) relative to the piston head. The accelerating downward
motion of the piston seems to, for early intake, balance at every moment the upward
motion forced upon by the rotating recirculation pattern.

The vortex streets encountered in the experiments at 20 RPM engine speed have strong
resemblance to the vortex streets visualized by Dahm et.al. (1992; case Uo/Ui=4. 16 and
RF1133) and Rockwell and Niccolls (1972; for RFSZZO) if one filters out - for the
incoming jet as encountered with the engine examined in this study -- low fiequency
instabilities, i.e. waviness, of the jet about the jet axis which are asle due to

geometrical constraints, and thus pressure gradients imposed by the engine model.

IV.2.2 Flow visualization Mode II

For the pictures in Figure IV.2.2 the same experimental procedure as described in the
preceding section had been employed with the exception that the aqueous KSCN solution
containing the fluorescent dye was in the engine cylinder and the pure water/KSCN
solution was in the fluid reservoir. This method yields lower local concentration peaks of
the dye at the beginning of the intake stroke - thus, the intake jet is not overexposed by
the laser sheet as it was the case for 'Mode 1' after a few degrees of crank-angle - and,

therefore, results in higher contrast images for a greater range of crank-angles.
Consequently, the flow structures (e. g. the vortex street of the jet) can be visualized in

82

greater detail, and information of the flow field is available until about midstroke
(90° crank-angle).

In addition to the results obtained fi'om 'Mode 1', one can get, by employing 'Mode 11',
an improved qualitative perception of the order of axisymmetry exhibited by the flow field
of the induction stroke by comparing size and orientation of individual toroidal vortex
rings - that are created by instabilities experienced by the intake jet - in difierent planes
(see sequence of the pictures in Figure IV.2.2). As definition for a measure of the level of
axisymmetry of an annular intake jet one can thus use the degree of similarity between
'two' corresponding vortex rings separated circumferentially by 180° (to the left and to the
right of the valve in the examined plane). According to this measure, axisymmetry is
established quite well at some crank-angle positions of the piston as can be substantiated
from the images in Figure IV.2.2; namely at CAs 24°, 28°, and 32°.

When comparing Figure IV.2.2f and g at crank-angles 28° and 32° one perceives very
congruent flow patterns for the intake jet to the right side of the cylinder axis. The
difference in CA is 4° which corresponds to a time interval of 33.33 ms. A typical velocity
for a fluid particle moving along with the intake jet is 2 m/s for this particular setting of
engine parameters, as will be shown in chapter V. This means that for a time interval of

33.33 ms the jet convected about 70 m. If, in a mental experiment, the two images for
28° and 32° CA in Figure IV .2.2f and g are thought of being superimposed, the evaluated
distance of 70 mm matches roughly the distance between the center of the vortex ring at
28° CA that is just developing on the side of the jet that faces the corner between cylinder
head and wall, and the center of the second (geometrically with respect to the valve)
vortex structure at 32° CA facing the corner as well. The fact that the calculated distance
matches the one found in the images impr that the latter vortex structure (at 32° CA) is
the vortex structure that matured from the one that can be perceived as just being
developing at 28° CA Intermediate information available at 30° crank-angle (but not

83

included in Figure IV.2.2) strongly suggests the aboveoutlined assertion. Very similar
statements can be made for the development of the flow field between crank-angles 32°
and 36°. Note, however, that whereas there is a certain degree of congruity for the vortex
structures in the annuIUS intake jet at 28° and 32° CA to the left side of the cylinder axis
there is hardly any aflinity for those at 32° and 36° crank-angle. It can be concluded fi'om
the above discussion that it is possible to use the time interval of 33.33 ms as a typical
time scale for the development of vortex structures in the crank-angle range between 25°
and 40°.

As far as size, and in this context a typical length scale, of the vortex structures is
concerned one has to be carefirl in an attempt to just measure the extent of dye
representing vortex rings and to use this measure as characteristic length scale. As
mentioned in the beginning of section III.3 diffusion processes in conjunction with mass
- 5D ~ (Dt)1/2 - are of much lower order of magnitude than difl’usion processes in
conjunction with vorticity — 6,, ~ (vt)1/2. Thus, the area of fluid dynamical influence by a
vortex structure that is shed fi'om the intake jet extents firrther in C - direction than is
indicated by the dye (for a definition of C refer to Figure V2.4). Nevertheless, the size of
vortex ring structures have been evaluated for crank-angle 36° from the image in
Figure IV.2.2h since it can be shown that the spread of vorticity is proportional to the
spread of mass, namely 6v ~ (Sc)1’2 * on where Sc is the Schmidt number of the problem.
In Figure IV.2.2h the flow field at 36° CA (Re=2476) exhibits (on the 'lefi' side with
respect to the cylinder axis) a singular flow structure, incorporating two counter-rotating
vortex rings, that appears separated as a coherent flow pattern. From this pattern typical
dimensions of secondary vortices have been inferred. The diameter of the smaller vortex is
2.4 mm, the diameter of the larger vortex (this vortex was generated earlier in time than
the smaller vortex which then got knitted together into this particular, coherent flow

pattern) is 3.6 mm The distance between the axes of the two vortex rings is also 3.6 mm.

84

An interesting flow visualization technique had been unintentionally applied in addition
to the one based on fluorescence: Because the cylinder axis of the water analog model was
not exactly parallel to the direction of gravitation, an air bubble was formed at the cylinder
head when the cylinder was filled with the H20/1(SCN/fluorescein solution (this bubble
disappears of course during engine performance in LIPA experiments). The air bubble gets
entrained into the jet at some crank-angle during the intake stroke and disperses into
smaller bubbles. These small bubbles serve as flow marker in the flow field. Due to the
buoyancy of the air bubbles it can be expected that they are forced to the low pressure
center of the main toroidal recirculation pattern. Figure IV.2.2k verifies this prediction; it
shows the air bubbles accumulated where the center of the main TRP is located (for a

qualitative evaluation of the location of the center the reader be referred to section V. l . 1).

IV .3 Comparison of Flow Visualization Results at two different Enm' e Speeds

F low fields of the induction stroke of a water analog model were evaluated in the two
preceding sections fiom visual information at two different engine speeds, 1.5 RPM and
20 RPM. Although the overall fluid dynamical behaviour of the induction flow is alike for
the two engine speeds, there are significant difi‘erences in the presentations of the flow

fields during intake as displayed in Figures IV. 1 .1 and IV .2.2.

At 16° crank-angle, for instance, the annular intake flow at the higher engine speed
exhibits advanced waviness and vortex shedding, Figure IV.2.2c, whereas the intake flow
at the low speed can be seen to consist of a straight propagating fiee jet, Figure IV. 1.1a.
The incoming flow is much more unstable and complex during early intake at 20 RPM

engine speed. Single flow structures at the head of the jet separate which is not the case in
conjunction with the low engine speed where the head of the jet forms a whole complex

and coherent structure (like a classical 'starting jet').

85

For 20 RPM engine speed the angle under which the intake jet enters the cylinder
(induction angle) inclines closer towards the cylinder axis than the induction angle at
1.5 RPM. The reason for this is that the higher piston speed in the case of the 20 RPM
engine speed induces a higher pressure gradient between the piston head and the cylinder
head. At the 1.5 RPM engine speed the recirculation pattern that builds up between the
intake jet and the cylinder wall (Figure IV .1.1c and d) is much more developed, and
contains much more energy during early induction than the main recirculation pattern.
The more rapid rate of spread of flow structures in connection with the intake jet for
the 20 RPM case results in the flow field developing much earlier than for 1.5 RPM. For
instance, for 20 RPM engine speed the main toroidal recirculation pattern is taking shape
already at 32° CA (Figure IV.2.2g; see also Figure IV.2. 1e) and its extension almost
reaches the cylinder axis implying that the tornado will start spinning shortly after 32° CA.
At this crank-angle the main TRP for 1.5 RPM is not yet developed. Recall that for
1.5 RPM the generation of the tornado occurs at about 53° CA.
The size of the vortex ring structures enforced by the two engine speeds is essentially

equal as can be expected for identical nozzle width and constant difl'usion coefficient D.

Since an engine speed of 1.5 RPM is a very unrealistic case in the field of engine
research, the results of employing photochromic dye will not be discussed further.
However, it should be emphasized that overall fluid dynamical behaviour and
establishment of flow structures as encountered at this low speed was similarly observed in
the experiments where a fluorescent dye has been employed (20 RPM engine speed; as
discussed in section IV.2) and will be similarly encountered in the maps of velocity vectors

describing flow fields evaluated by LIPA (chapter V; 10, 20 and 30 RPM engine speed)
establishing the relevance and legitimacy of the discussion of fluid dynamical results at the

low engine speed.

86

On the basis of results on vortex ring dynamics described in section IV. 1.2 one could
speculate that the essential difierence between flow fields at small and large valve lifts as
well as at low and high engine speed is based on a mechanism set forth by the instability of
the annular jet flow, i.e. the roll-up and successive break-down of the vortex rings. Partial
prove for support of this hypothesis will be provided by quantitative fluid dynamical

parameters given in section V.2.3.

Chapter V
SCIENTIFIC FINDINGS IN RESEARCH OF ENGINES
OBTAINED FROM LIPA - DATA

The following sections contain results on fluid dynamics during induction of a water
analog model that were obtained fi'om the reduction of data acquired with LIPA

Firstly, the induction stroke as a whole will be discussed with the employment of the
'whole cylinder' investigation grid. This study comprises sections V.l through V.4. In
sections V.1 to V3, the development and characteristic features of the flow field in the
induction stroke of an axisymmetric water analog engine will be outlined based on the
projections of the three-dimensional field of velocity vectors onto the two-dimensional
imaging plane of the camera. Several fluid dynamical parameters will be defined in the text
which will serve to difi‘erentiate between flow fields induced by different settings of engine
parameters (engine speed and valve lift). A new hypothesis will be disclosed which will
help to examine the flow field of the induction stroke in a more focused way. In
section V.3 an attempt is undertaken to separate the influences of cycle-to-cycle variability
and of turbulence inherent in the flow field as sources for velocity fluctuations.

Section V.4 deals with vorticity distributions which are regarded a useful tool for the
evaluation of mixing performance.

Finally, in section V.5, by employing a 'denser spacing' version of a grid of laser lines to
the area between chamber and piston head when piston is at TDC, the intake jet will be
investigated more closely. Superposition of plots of velocity vectors onto corresponding
pictures provided by flow visualization confirm two things: there is strong congruity
between the fields of velocity vectors depicting structures of the intake flow with results
from flow visualization. And secondly, superposition of LIPA - data with flow

87

88

visualization results helps to understand the quantitative data: bridging the gap between
the very high resolution of dye flow visualization, which does not allow quantification of
the fields, and the lower spatial spacing of information provided by the employed laser

grid, which could lead to misinterpretation of the features of the flow field.

The main tool to display the flow field of the induction stroke in this chapter will be
two-dimensional maps of velocity vectors; velocity vectors being the most immediate
results from an application of LIPA Intrinsically, the plots denote distances and directions
that particular fluid particles convected during a certain time interval. To compare plots of
convection distances one has to make certain that the underlying time intervals are
identical. We will assume that the data imaging system acquired all of the images in the
field instantaneously, and thus we will speak of velocities instead of convection distances.

Vlfrthin a plot the velocity vectors are distributed according to the locations of
intersection points generated by the laser grid. In the figures displaying fields of velocity
vectors the setting of the plots comprises geometrical elements such as a valve, the
cylinder wall, and the piston head for orientation purposes. These plots will be referred to
as 'maps of velocity vectors'. Maps of velocity vectors will be utilized in sections 1, 3,
and 5 of this chapter. Each figure containing maps of velocity vectors shows a reference
vector denoting the magnitude of a vector of 1 m/s, and estimations for errors in each
direction (x and 2) according to Table 1113.1. The errors are depicted with respect to a
magnified velocity vector of 1 m/s.

A less common tool for data representation is a surface or contour distribution of
vorticity in a three-dimensional plot. These plots are used to portray mixing performance

during the induction process.

89

V.1 Description of Induction Stroke Flow Field in Terms of Maps of Ensemble
AveragpciVelocity Vectors

 

The flow field patterns during induction as represented by maps of velocity vectors at
selected crank-angles will be discussed in the following section for an engine speed of
20 RPM (340 in air) and a valve lift of 3 mm (highest Reynolds number at midstroke is
equal to 4213). This set of parameters was chosen so that the following results can be
compared directly to the flow fields revealed by the visualization technique discussed in
section IV.2. The maps of velocity vectors can be regarded a quantification of the
qualitative flow visualization results. The velocity vectors in the maps displayed in
Figure V. 1.1 are ensemble averaged (EA) over 42 engine cycles (EC); 1 will be referring
to these ensemble averaged results when I am speaking of the 'EA domain'. Recall from
section 11.2.2.4 that memory limitations of the SunWorkstation used for data reduction
was the reason for being restricted to 42 ECs. However, it turned out that a data base of
42 engine cycles seemed sufficient since considerable convergence of the fluctuation

intensity of the velocity components could be observed (section III.2.1.1).

V, 1.1 fimsis and fluid dmn'cal behavipur pf features of the flow field in water analog

model,

Following is a fairly comprehensive account of the flow characteristics of the induction
stroke as can be interpreted fi'om the velocity vector maps in Figure V. 1.1.

Prior to piston motion the fluid in the engine is, in the average, quiescent (very low
magnitude of velocity vectors, Figure V. 1.1a; in fact, fi'om the maps of velocity vectors of
individual cycles in section V.3 it will be apparent that there is initial motion in the flow
fields prior to piston motion). After the piston moved downwards a certain distance,

velocity vectors obtain magnitude and direction and depict motion of the intake jet and its

90

surrounding. The angle under which the intake jet enters the engine cylinder (induction
angle) is, in the beginning of the intake stroke, prescribed by the geometrical shape of the
valve. The intake jet impacts onto the piston head and separates into a flow towards (and
then up) the cylinder wall, and into a flow towards the cylinder axis (Figure V. 1. ld&e; for
more information refer also to Figure IV .2.2). The flow towards the axis will, in three
dimensions, create the main toroidal recirculation pattern (TRP) mentioned in chapter IV
which occupies most of the cylinder volume; the flow up the cylinder wall develops into a
toroidal recirculation pattern in the upper corner between cylinder wall and head. (The
two toroidal recirculation patterns present themselves as vortices in two dimensions, thus
encouraging the author to refer to the main TRP as the main vortex as opposed to upper
corner vortex, alias the upper corner TRP.) Since the intake jet - carrying linear
momentum - is forced to turn at the piston head and recirculate (main vortex), the linear
momentum of the intake jet becomes angular momentum with respect to the examined

x, z - plane of the engine cylinder.

One can infer fi'om the sequence of velocity maps in Figure V. 1.1 that as crank-angle
evolves and thus piston speed increases, the axis of the intake jet inclines towards the
cylinder axis until a crank-angle of about 44° is reached and the incoming jet enters the
engine almost parallel to the cylinder axis, Figure V. 1. 1f. (Note that the figure does not
clearly show this; one has to envision the intake jet by visually interpolating between
velocity vectors #7, 12, and 18, which presumably depict fluid dynamical behaviour of
parts of the intake jet; for assignment of velocity vectors refer to Figure II.2.3a.) The
reason for the inclination of the jet axis towards the cylinder axis is possibly an increasing
pressure gradient between piston and chamber head which is induced by the accelerating
piston (Note that throughout the whole stroke the flow of fluid is governed by the
pressure distribution within the engine.) As the momentum of the intake jet increases with
increasing piston speed, the supply of angular momentum to the main vortex becomes

enhanced. The main vortex grows in extent, degree of structure, and energetic content; at

91

72° CA (Figure V. l . 1 j) it assumes a roughly circular shape. As is denoted by the velocity
fields in Figure V. 1 . lg through k, the main vortex which becomes stronger by the
continual supplement of momentum starts to push the intake jet back up towards the
cylinder head until the induction angle assumes, fi'om 92° CA on, a constant value. The
convergence towards this final angle is faster for smaller valve lifts as will be illustrated in
section V. 1.2 'Profiles of induction angle'.

From crank-angle 72° (Figure V. 1.1j) on, a 'bufi‘er' zone is created in the region just
above the piston head. In this bufl‘er zone no particular flow structure dominates, nor is
there a consistent directional flow when taking 42 individual engine cycles into
consideration. Thus, the summation of velocity vectors over these 42 cycles yields in the
bufl‘er zone for the EA domain a low magnitude flow field with a random flow pattern.
Note, however, that discussed flow patterns are those perpendicular to the piston - there
could be significant swirl velocities. The investigation of the horizontal plane on top of the
piston head will yield conclusive data for this matter. The creation of the 'bufi‘er' zone on
top of the piston head could be caused by an establishment of a high pressure region above
the piston head where the fluid flow along the cylinder wall does not possess the necessary
momentum to overcome this pressure gradient as it approaches the piston. Also, curvature
effects of the toroidal recirculation pattern are important in the discussion of the creation
of the buffer zone. The same time scale encountered for the convergence of the induction
angle can be established for the time of initiation of the buffer zone: The smaller the valve
lift, the earlier the bufl'er zone can be observed in the velocity maps.

Based on the velocity maps in Figure V. 1.1 a qualitative observation can be stated
considering the center of rotation of the main vortex: This center moves with evolving
crank-angle down and towards the cylinder wall (in positive 2 - and x - direction) until
about midstroke (90° CA). This result gathered fiom the velocity maps is in accordance
with visual observations made during the LIPA experiments in individual realizations of

induction strokes.

92

After the incoming jet assumed a constant induction angle (about 100° CA,

Figure V. 1.1m) the flow field demonstrates a very stable, structured flow in the region of
examination (region spanned by the laser grid). One can almost speak of a 'fully developed'
flow field or of self-preservation of the flow field in the region examined. 'Fully developed'
paraphrases the fact that the orientation of a velocity vector at a particular location in the
EA flow field stays essentially constant throughout the remainder of the stroke, including
the velocity vectors around the axis of the intake jet. It seems that each velocity vector in
the investigated area can be mapped for different crank-angles according to a universal
formula. This holds even true for the case where the velocity fluctuations are highest -

30 RPM (510 in air) and 3 mm VL, see section V2.2. The center of rotation of the main
vortex stays at one location. It should be recalled at this point that ensemble averaged
results are the subject of analysis here; as will be demonstrated in section V.3 there is no
such thing as a stable flow field in individual engine cycles.

For this fully developed flow field the fresh incoming fluid through the valve flows
tangentially along the periphery of the main vortex and impinges on the cylinder wall. Only
a fiaction of the fluid flows up along the wall (negative 2), most of the fluid flows down
along the wall (positive 2). At a certain point (z - location) the flow along the wall in
positive 2 - direction turns inward towards the cylinder axis. This 'separation' point is
determined by the assumed - but meaningful - existence of an increase in pressure as the
flow approaches the piston head (adverse pressure gradient). This inward turning fluid
flow supplies the main vortex with momentum (this addition of momentum to the main
vortex can be called 'mediate' or 'indirect' supplement as opposed to an 'immediate'
supplement of momentum whose mechanism is described in sections IV. 1 .2 and V.2.3).
The boundary layer flow along the cylinder wall is confined to a small depth with a small
momentum thickness. As the floéw turns towards the cylinder axis the width of the flow

broadens and extends to even more volume as it enters into a region between the center of

93

rotation of the main vortex and the cylinder axis. (The pattern of the turning and
expanding boundary layer flow resembles that of a radial difi‘user.) Here, the flow is in
negative z - direction (towards the valve). Here, the flow decides whether -- depending on
valve lift and engine speed - to primarily mix the fuel/gas constituents of the flow in a fine
scale structured flow field and dissipate momentum, or to primarily conserve momentum,
stratify the flow, and redistribute the momentum within the main recirculation pattern. It is
this region - besides the region around the axis of the intake jet - in which the basic
difl'erences between the flow fields enforced by difl‘erent settings of engine parameters, i.e.
engine speed and valve lift, establish themselves through, e.g., characteristic length scales
in the flow field and degree of velocity fluctuation. This region which is a sub-area of the
main recirculation pattern will be subject to a more detailed investigation and is therefore
called 'area of scrutiny' (see Figure V. 1. lo for the location and description of this area).
Motivation of an investigation of the area of scrutiny is derived also from the fact that it is
desired to keep up air/fuel mixing in this region. Note that most of the mixing is
established by the intake jet. In the area of scrutiny it is wishful to obtain most

improvements.

By studying the intake stroke of an engine using information provided by maps of
velocity vectors it was possible to distinguish difi‘erent regions in the flow field
characterized by different flow patterns such as intake jet flow, toroidal recirculation
patterns, and boundary layer flow along cylinder wall. This dissection of the flow field will
be very helpful for its analysis. F urtherrnore, the importance of the 'area of scrutiny' was

pointed out. In section V.2 this region within the flow field will be examined more closely.

The flow field of the induction stroke of a water analog model as described above using
EA velocity vectors for 20 RPM engine speed and 3 mm valve lift is representative for all
nine flow fields as enforced by the nine difi'erent settings of the engine parameters

(Table 11.2.3): The flow features such as intake jet flow, toroidal recirculation patterns,

94

boundary layer flow along cylinder wall, and buffer zone above piston head were
encountered as flow structures in all nine flow fields. Difl‘erences in the nine engine cases
were manifested in the magnitudes of flow quantities, timing of initialization of flow
patterns and location of flow structures within the engine cylinder. For crank-angle 132°
the nine flow fields are exhibited for the EA domain in Figure V. 1.2. These nine flow fields

will be the basis for investigations made for the area of scrutiny in section V.2.

For crank-angles 16°, 20°, 32°, 36°, 44°, and 92° (Figure V.1.1b, c, d, e, f, and 1)
Figure IV.2.2c, d, g, h, i, and k exhibits the corresponding flow visualization results. One
can denote profound conformity between the qualitative and quantitative representations
of the flow field. The flow field in terms of velocity vectors quantifies the rate and
direction of convection of flow structures that are visualized in the images of
Figure IV .2.2. These results illustrate the usefulness and potential of a visualization-
quantification hybrid technique.

V. 1.2 Profiles of induction angle

A more detailed study of the angle under which the intake jet enters the engine cylinder,
induction angle or, has been conducted where the induction angle has been evaluated as a
firnction of crank-angle for each element in the matrix of experiments in Table 11.2.3. \Vrth
the help of the results fiom this study it will be possible to distinguish fluid dynamical
behaviour during induction as it is experienced in connection with the different settings of
the engine parameters. The basis for the evaluation of the induction angle were the maps
containing the ensemble averaged velocity vectors. Although the extraction of the angles
was performed by hand and although a considerable uncertainty exists about what velocity
vectors do actually represent the intake jet in the maps - thus, there is a substantial error
evoked by this data reduction process - it should be considered legitimate to infer some
trends fi'om the a—vs—CA plots. The results are exhibited in Figure V. 1.3. To emphasize

95

the rather qualitative nature of this set of data a spline fit procedure has been applied to

each curve in Figure V. 1.3 and the single data events have been omitted.

In general, the incoming jet enters at first under an angle prescribed by geometrical
design (45°; observable best for 10 RPM engine speed, Figure V. 1.3a). It then inclines
towards the cylinder axis, for some engine parameter settings up to 0°. Alter reaching a
local minimum - the definition of the induction angle or is given in Figure V. 1.3 -, the
induction angle increases towards a local maximum after which the angle decreases again
and assumes a considerably constant plateau value. For 20 RPM engine speed and 3 mm
valve lift flow visualization results have been presented in section IV.2.2. Ifone examines
the sequence of flow visualization pictures for this setting of engine parameters in
Figure IV .2.2 documenting the induction stroke from its start to midstroke one can
observe that the smallest, i.e. steepest induction angle (local minimum) is encountered at
about 44° CA (Figure IV.2.2i). This matches exactly the crank-angle found for the local
minimum of the induction angle in the maps of the ensemble averaged velocity vectors in
the case of 20 RPM engine speed and 3 mm VL (Figures V.1.1f and V.1.3b).

There are two obvious tendencies that can be deduced fi'om the plots in Figure V. 1.3.
Firstly, the smaller the valve lift, the earlier (crank-angle-wise) does the induction angle
reach the abovementioned plateau value. Secondly, the greater the valve lift, the lower is
the plateau value. This can be attributed to the fact that the actual mean angle of the
induction jet is slightly less as the valve lift increases because the axis of the intake jet is at
higher 2 - coordinates for higher valve lift. However, to a certain extent one can assert that
both tendencies mentioned in this paragraph also indicate that the main toroidal
recirculation pattern contains more momentum (as will be, in fact, verified in
section V.2.2) in the case of a smaller valve lift, thus being able for this configuration to
push the intake jet back up faster and higher.

96

The overall observations for the induction angle in the preceding two paragraphs are
independent of engine speed meaning that all the features exhibited by the induction angle
profile can be perceived at 10, 20, as well as 30 RPM engine speed; Figure V. 1.3a, b,
and c, respectively. Tentatively, a distinguishing factor for the three sets of induction angle
profiles corresponding to the three engine speeds can be established when a comparison is
performed in each set of the difl‘erence between the lowest and the highest plateau values
towards which the induction angle converges for each considered valve lift. This
difference between the lowest and the highest plateau value seems to decreases with
increasing engine speed. This can be explained by the fact that as the engine speed
increases, the plateau becomes less dependent on the valve lift. One has to be careful in the
interpretation of this data though, since the scatter in the data is too large to make definite

conclusions.

It should be noted that many of the results outlined in the previous two sections (V. 1.1
and V. 1.2) have been observed in the flow visualization experiments with the
photochromic dye at 1.5 RPM engine speed discussed in section IV. 1.2. This illustrates
that basic flow field components that are created at an engine speed of 1.5 RPM continue

to persist, with quantitative difl‘erences, of course, in the investigation of flow fields at

higher engine speeds.

97

V2 Discrimination of Flow Fields as enforced by different Parameter Settings of
the Bug e

 

In order to distinguish flow fields which are enforced by different settings of engine
speed and valve lift within the parameter space, three characteristic values were chosen to
provide measures to differentiate between the flow fields at 132° crank-angle. These three
characteristic parameters have been evaluated in the 'area of scrutiny' in order to establish
an understanding of the influence of the parameters engine spwd and valve lift. In the area
of scrutiny 22 velocity vectors were examined to evaluate the three parameters. The area

of scrutiny is circled in the velocity map of Figure V. 1. lo.

V2.1D 'tinf am r

First, the kinetic energy content in the area of scrutiny has been determined for each of
the nine flow fields. In the EA domain the magnitudes of the velocity vectors do not
properly reflect the content of kinetic energy contained in a specific region since positive
and negative contributions of velocity components tend to cancel each other out. Thus, an
average of the kinetic energy contained in the individual engine cycles in the area of
scrutiny has been calculated, with subsequent summation of these averages over the
number of engine cycles analyzed and division by the number of ECs. Kinetic energies are
scalars, the information on orientation has been stripped fiom the velocity vectors. The

kinetic energy content is evaluated as follows:

=#TCs =1[%‘=zl|ui2jl] (V.2.a)

where #ECs is the number of engine cycles over which the average is taken, N is the

number of velocity vectors in the area of scrutiny, namely 22, and u; is the local velocity.

98

From the equation it is obvious that a specific energy is being evaluated since KE is

normalized by mass.

Secondly, energy associated with velocity fluctuations of the z - component has been
evaluated for the area of scrutiny (the z - component has been chosen because of the
higher accuracy of velocity evaluation in this direction; see section 111.33), in a similar

manner as the kinetic energy, according to the following expression:

FE 1 “a l N( )2 V2b
= —— — u -— u . .

#ECs N; z, z, j ( >
where uz, ,3 EA is the local ensemble averaged velocity in z - direction. Equation V.2.b
defines a specific energy as well. Note that the fluctuation energy includes contributions
fi'om cycle-to-cycle variability as well as from velocity fluctuations due to turbulent

motion of the flow.

Thirdly, a measure for recirculativity has been formulated in terms of normalized
angular momentum. This measure indicates when applied to velocity vectors in the area of
scrutiny to what degree the fluid reenters the main vortex fi'om the area of scrutiny as well
as the order of regularity in this area (stratification). The argument goes that if the velocity
vectors in the flow field in the area of scrutiny are oriented such that they suggest that the
flow stays in an overall circular motion the fluid from the area of scrutiny reenters the
main recirculation pattern Angular momentum was used as parameter to describe
recirculativity because it is sensitive to whether or not the flow field is roughly circular
about a center. The angular momentum is normalized by local quantities of radius to
center of flow structure and velocity magnitude so that only the orientation of a velocity
vector with respect to the center of the recirculation pattern would enter the angular

momentum evaluation which is performed according to the following formulation:

99

1 ac: 1 N ri-It'iil-cosrr
z _ __ .2.
#ECsJ-=1[N§ r,-|rr,.| 1. (V c)

where r is the radius from the center of the recirculation pattern to the local velocity and
u is the angle that the local velocity vector forms with a line that is perpendicular to the

radius. Recirculativity equal to 'one' corresponds to a circle.

V.2.2 Information about difl‘erent flow fields in terms of defined parameters

A compilation of nine maps of velocity vectors - one for each engine parameter setting
- each at a crank-angle of 132° can be seen in Figure V. 1.2. The velocity vectors are
ensemble averaged. For 30 RPM engine speed the analysis was based on 63 engine cycles,
for 20 RPM engine speed 42 engine cycles were considered, and for 10 RPM 21 ECs as
mentioned in section 11.2.2.4. At a first approximation by eye one can readily detect
difl‘erences between the nine flow fields in the area of scrutiny in orientation and
magnitudes of the velocity vectors. In order to achieve objective qualitative measures that
unambiguously classify the flow field the above defined parameters (equations V.2.a
to V.2.c) were applied to velocity information from the nine sets of data in the area of

scrutiny.

The results of the evaluation of the parameters are plotted in Figures V.2.1 to V.2.3.
From Figure V.2.1a where the specific kinetic energy contained in the area of scrutiny for
132° CA is shown in a single logarithmic plot in dependence of the valve lift for three
difi‘erent engine speeds it is evident that the kinetic energy content in the area of scrutiny
decreases - non-linearly - with increasing valve lift for constant engine speed. It also
increases with increasing engine speed for constant valve lift as can be seen in

Figure V.2. lb where the same data as available for Figure V.2.1a are plotted as a function

100

of engine speed in an arithmetic coordinate system. These curves are as expected since one
anticipates that the momentum carried by the incoming annulus jet flow and transported to
the area of scrutiny increases with engine speed. This relation (engine speed ll :>
momentum ll) established fiom Figure V.2.1 receives support fiom an analytical calculation
by Miller and Comings (1957) as well as fi'om Tennekes and Lumley (1972) who reduced
the Reynolds equation in streamwise direction for two-dimensional jet flow -- two-
dimensional jet flow is a very reasonable approximation for the annulus jet flow
encountered in this research project. The formula the Reynolds equation was reduced to is,

for momentum flux J in the direction of the flow, in the notation of Miller and Comings,
J = prf (V.2.d)

where b is the opening of the nozzle, and U0 is the velocity at the centerline.

Figure V.2.1a contains for each of the nine data points an indication (bars) of how
much the specific kinetic energy fluctuates over the respective number of engine cycles
analyzed. The bars denote RMS fluctuation. The bars include errors due to mis-calculation
of velocity vectors from the area of scrutiny and can thus serve as a rough indicator for
system error. It can be noted that Figure V.2.1a shows that there is no 'overlap' between
the curves representing the kinetic energy for the three different engine speeds, i.e. the
Values for 9 mm VL for 30 and 20 RPM are higher than those for 3 mm VL and 20 and
1 0 RPM, respectively, even when considering the bars that denote error estimations. In

Figure V.2.1b the non-linear behaviour for the kinetic energy with respect to engine speed
becomes obvious. The increase of kinetic energy with engine speed is similar for the 6 and

9 mm valve lift cases. For 3 mm valve lift this increase is more emphasized.

The curves of the specific energy associated with velocity fluctuations in the z -
direction for the area of scrutiny at 132° CA are exhibited in Figure V.2.2, also in a single
logarithmic plot. The curves strongly resemble those of the kinetic energy discussed in the

101

previous paragraph and displayed in Figure V.2. la, however, on this (difi‘erent)
logarithmic axis. As a rule of thumb one can establish that the energy associated with
velocity fluctuations of the z - component is about one tenth of the kinetic energy. Keep in
mind that the velocity fluctuations discussed here include cyclic variations of the bulk flow
as well as velocity fluctuations due to turbulent flow within individual engine cycles. It can
be speculated at this point that after separation of the cycle-to-cycle velocity fluctuations
from those caused by turbulent motion of the flow, the trend suggested by Figure V.2.2
for velocity fluctuations in general, namely that velocity fluctuations increase with engine
speed as well as with decreasing valve lift, holds true for the velocity fluctuations due to
cycle-to-cycle variability. From Figure V.2.2 it can then be argued that the low velocity
fluctuations at high valve lifts are an indication that due to less cycle-to-cycle variability
the flow field in an engine is more predictable for large valve lifts. It can also be speculated
using the information provided by the curves in Figure V.2.2 that the turbulent energy is
higher in connection with the smaller valve lifts implying that enhanced mixing of fuel/air
phases can be achieved at smaller VLs. However, a confirmation of these speculations has
to be furnished on the basis of further analysis of the velocity fluctuations.
Figure V.2.2 also contains bars indicating the variation of the specific velocity

fluctuation energy with engine cycle. Vlfrth one exception these 'error' bars do not overlap
for different engine speeds at the three valve lifts indicating that the tendencies suggested

by the curves are considerably accurate.

In Figure V.2.3 the normalized angular momentum evaluated in the area of scrutiny as a
measure for the recirculativity is plotted versus valve lift for 132° crank-angle. Each point
sbv'mbolizes an average for the angular momentum over a crank-angle window of 10° and
an average over a certain number of engine cycles. For the evaluation of the values in
Figure V.2.3 a center of rotation had to be assigned. Interestingly, it turned out that

considering the accuracy achievable with the given spatial spacing of intersection areas in

102

the investigation grid the center of rotation moves only marginally when the setting of
engine parameters changes.

The curves in Figure V.2.3 illustrate for 10 and 30 RPM engine speed that the
recirculativity increases with increasing valve lift and decreasing engine speed. The slope
of increase with valve lift is steeper in conjunction with 30 RPM engine speed than with
10 RPM. For 20 RPM engine speed the available data points do not fall into a scheme
suggested by the results for the other two engine speeds. After reading section V.2.3 the
difl‘erence to the other two curves will become clear.

High recirculativity, if it is expressed in terms of normalized (by local quantities of
radius and velocity magnitude) angular momentum, means high regularity and degree of
order of the flow field (circular flow). Thus, it can be inferred from Figure V.2.3 that for
9 mm valve lift the flow field in an engine exhibits highest degree of regularity and the
fluid in the area of scrutiny is most likely to reenter the main recirculation pattern. At
9 mm VL rotation information is preserved. Note that a high degree of order expressed by
the flow field implies stratification of the flow field and thus inferior mixing performance.
On the other hand high degree of order (structure) enforces higher cycle-to-cycle
predictability which is why high recirculativity for large valve lifts will be a key to explain
the lower velocity fluctuations for 9 mm VL found in Figure V.2.2. For 3 mm valve lift
recirculativity suggests that the flow field in the area of scrutiny is least structured and the

fluid is least likely to reenter the TRP. With decreasing engine speed the likeliness of
recirculation increases.

These statements, arrived at by interpreting the plot denoting recirculativity for

different engine parameter settings (Figure V.2.3), can be confirmed by closer examination
Of the maps of EA velocity vectors in Figure V. 1.2.

A hypothesis will be presented in the following section which elucidates a relation
betheen the physics of the annulus intake jet flow and the trends of the curves of

 

103
recirculativity as well as the curves denoting energy associated with velocity fluctuations

- which are representative measures for a flow field in the area of scrutiny.

V.2,3 Discussion of the correlation between intake jet flow and the fluid dmamics of the
intake strpke with special reference to the area of scrutiny

 

In order to explain the low fluctuation energy and the high degree of recirculativity
associated with large valve lift and the high fluctuation energy and the low degree of
recirculativity associated with small valve lift the fluid dynamical process as a whole as

well as bilateral associations between flow structures have to be considered and evaluated.

As has been pointed out in section IV. 1.2 for low engine speed (flow visualization
results with photochronric dye) the supply of angular momentmn to the area where the
main toroidal recirculation pattern 'dwells' is imparted immediately by convection into this
volume of vortex rings that shed fiom the annulus intake jet on the side facing the main
TRP and that have the same sense of rotation as the TRP. The mechanism that
communicates the angular momentum of the vortex ring with that of the TRP can be
regarded as a kind of superposition of the toroidal axis of the vortex ring with that of the
main recirculation pattern (Figure V.2.4); it can be described as pairing of two vortical
Structures that have infinite extension in their axial direction. Presupposition for this

ingestion of one vortical structure into another to take place is that the coherent vortex
rings on the one hand contain a considerable amount of angular momentum, and on the
other hand are not yet broken up and dispersed into small scale, incoherent turbulent
motions, an occurrence that takes place after the incoming jet has convected a certain
diStance (see next section as well as the 'Introduction'). To determine the characteristics of
a toroidal vortex ring when the incoming annulus jet turns at the cylinder wall a study of
atmqus jet flow shall be undertaken.

104

V.2.3.1 Review of background of annulus fi'ee jet flow

Various investigations on two-dimensional, planar, or annulus fiee jet flows can be
found in the literature of which some have been briefly reviewed in the 'Introduction'.
Following is a quick summary on the physics of annular jet flow: An annular jet entering
ambient, quiescent fluid stays a stable jet with its core exhibiting potential flow for very
low Reynolds numbers (Re). With increasing Re, instabilities of the jet flow arise. The jet
first starts to propagate in a wave-like shape of long wave-length. VVrth further increase in
Re 'secondary' vortices start to roll up at the interface between the jet and its surrounding.
These vortex rings entrain fluid from the jet's surrounding, gain strength and extent as the

jet flow proceeds downstream and, eventually, break up. Over the distance between the
entering of the jet into the ambient fluid until the burst of the vortex rings the flow is
named 'transitional' (Yule 197 8; see Figure V.2.5). After their break-up the vortex rings
create small scale, incoherent eddies (small vortical structures) in a what is now called
'turbulent' flow. According to the investigations on fiee jets the length of the transitional
flow - the end of the transitional flow coincides essentially with the termination of the
potential core flow of the annular jet - varies with Reynolds number and width of the
nozzle at which the jet enters. The length of transitional flow has been evaluated for
annular jets to be approximately between 6 and 8 times the width of the nozzle (for
Reynolds numbers in the range of 1000 to 2000). This implies for the case of an induction
flow of an engine that the break-up of the coherent vortex rings which initiates turbulent
flow motion within the intake jet occurs further downstream for larger valve lift at
constant engine speed (the Reynolds number is approximately constant with constant
engine speed and varying VL).
Another important issue is the size of the vortex rings that are created by the intake jet.
Vortex ring structures initially scale with the width of the nozzle at which a free jet enters
aInbient fluid as can be confirmed by the flow visualization experiments from chapter IV.

105

In the context of vortex ring ingestion into the main TRP it seems intuitively appealing

that the larger vortex ring structures are more favorable to merge with the main toroidal
recirculation pattern by superposition of their axes - in the cases where the vortex rings
reach the cylinderwall -- since the length scale of their structures is closest to the length

scale associated with the main recirculation pattern.

V.2.3.2 Application of knowledge about annular fi'ee jet flow to fluid dynamical aspects
in engine

Let us consider the following case: Iffor a certain combination of engine speed
(Reynolds number) and valve lift the vortex rings associated with an incoming jet just do
not burst when the jet reaches the cylinder wall the contribution of angular momentum of
the vortex rings to the main toroidal recirculation pattern will be maximum. This is the

ideal case. All of the cases investigated in the preceding paragraphs deviated fiom this.

Interpretation of the curves in Figure V.2.3 based on fluid dynamical behaviour of the
intake jet flow will be attempted in this paragraph. Considering the curve for 30 RPM
engine speed. (Reynolds number at midstroke is equal to 6319) in Figure V.2.3 the degree
of regularity increases with increasing valve lift. This makes sense now after establishing
the physics of intake jet flow. The coherent vortex ring structures for the 3 mm valve lift
case get destroyed at the shortest E (for a definition of g refer to Figure V.2.4). The vortex
ring structures live longest for 9 mm VL, however, presumably break up as well at this
engine speed before reaching the cylinder wall since otherwise we would not expect the
curve to exhibit a generally linear characteristic. Ingestion of the vortex rings into the main
'I'RP does not take place for any of the examined valve lifts at 30 RPM. The increase in

regularity with VL at this engine speed is due to the fact that the vortex rings break up
later for larger VL, and that therefore the turbulent motion is less established when the

flow supplies the toroidal recirculation pattern. Furthermore, the vortex rings are larger in

106

extent for the 9 mm VL case and provide thus eddy structures on larger length scales in
the turbulent flow after the break-up of the vortex rings. Consequently, there is altogether
more large scale 'structure' in the main TRP and thus higher recirculativity for 9 mm valve
lift than for 6 and 3 mm. This is reflected by the curve in Figure V.2.3 for 30 RPM.

For 10 RPM engine speed the respective curve for recirculativity in Figure V.2.3 is
linear as well. Since the Reynolds number is three times smaller at this engine speed than
for 30 RPM, one can imagine that in this case vortex rings shedding from the annular
intake jet survive for this particular model engine at all three examined valve lifts until the
intake jet reaches the cylinder wall so that ingestion takes place for all vortex rings,
supplying the TRP with (immediate) angular momentum at all three VLs. This means that
the only difl‘erence for the three valve lift cases at 10 RPM engine speed is manifested
through the size of the vortex rings when the jet flow reaches the cylinder wall which
accounts for the increase in angular momentum with VL as found in Figure V.2.3, and
accounts also for the fact that the slope of increase of recirculativity with valve lift is less
steep for 10 RPM engine speed when compared to the slope for 30 RPM.

The case for 20 RPM engine speed represents an intermediate state between the two
cases analyzed in the two preceding paragraphs which is supposedly why the three data
points in Figure V.2.3 for 20 RPM engine speed do not follow the scheme proposed by
the curves derived fiom experiments conducted at 10 & 30 RPM engine speed. Data
available at this point are not suflicient to discuss the case of 20 RPM engine speed.

The above delivered hypothesis is a very simplistic one. One has to perform more

\Iisualization and LIPA experiments in order to verify the proposed mechanisms.

Altogether, it is possible to conclude from this section that high engine speed with small
valve lift is favorable for fine scale flow structures in the flow field of an engine, low
degree of stratification, and hence, best mixing performance during induction in the area of

Scrutiny. Large valve lift along with low engine speed establishes a structured flow field

107

which yields inferior mixing performance, however, on the other hand, also low cycle-to-
cycle variability. Qualitative verification of results about correlations between annular

intake jet and fluid dynamical behaviour in the area of scrutiny from flow visualization

experiments is discussed next.

V.2.3.3 Flow visualization results revisited

The ingestion of toroidal vortex rings, created by the incoming jet, into the main
recirculation pattern as hypothesized above was observed, as outlined in section IV. 1.2, in
the results of the flow visualization experiments with the photochromic dye. The
observations on the ingestion of the vortices as observed in the photochromic dye
experiments will be partially repeated here, now in the light of the findings about annulus
intake jet flow. The experiments discussed in section IV. 1.2 were performed at a low

engine speed of 1.5 RPM (25.5 RPM in air). Due to the low Reynolds number the length
of the transitional flow exceeds the distance between the orifice and the cylinder wall (I

(see Figure V.2.4 for the definition of (J), in particular for the 6 and 9 mm valve lift cases.
Thus the vortex rings associated with the intake jet can be traced in the video film fiom
their creation up to the cylinder wall, and their behaviour at §=£J where the free jet turns
into a constrained boundary layer flow can be studied. It was observed for the vortex rings
that the region that contains their angular momentum expands - entrainment of ambient
fluid - as the jet convects and that especially after the main flow of the jet turns to a flow
along the wall, the rotation induced by a vortex ring on the side of the jet that faces the
main recirculation pattern is merged into the rotation of the main recirculation system
Supplying it immediately with angular momentum. As mentioned in section IV. 1.2 the
entrainment of the angular momentum was most eficient for the vortex rings encountered
With a 6 mm valve lift. Note that the ingestion mechanism can only take place for a certain

Crank-angle range which is symmetric about midstroke since at the beginning and towards

108

the end of the intake stroke the Reynolds number is too small to destabilize the jet flow
and form toroidal vortex rings. It was learned in section IV. 1.2 that for 1.5 RPM engine
speed vortex shedding sets in at about 30° crank-angle (Figure IV. 1 . 1d). Coherent vortex

ring structures that can supply angular momentum to the main toroidal recirculation

pattern by ingestion can be observed from about 50° CA (Figure IV. 1 . It) on.

The generation, development, and break-up of vortex rings can be also observed in the
images which were generated employing fluorescent dye (Figure IV.2.2) for 20 RPM
engine speed and 3 mm valve lift (section IV .2.2). Since experiments were performed at a
higher Reynolds number than for 1.5 RPM the distance where the flow can be referred to
as transitional is shorter for constant valve lift, meaning that the break-up of the coherent
vortex rings formed by the intake jet into small scale turbulent motion occurs at shorter E,
when compared to the results obtained fi'om experiments at 1.5 RPM engine speed. Due
to the higher engine speed and small VL the length of transitional flow can be determined
from the images in Figure IV .2.2 to be less than 11/3. From these images it is observable
that the length of the transitional flow decreases as crank-angle evolves; the break-up of
the coherent vortex structures for crank-angles around nridstroke (Figure IV.2.2k) occurs
shortly after the intake jet enters the cylinder. This implies that during the engine stroke

the flow field experiences constantly changing initial conditions as far as the physics of the

intake jet flow are concerned.

In the section on comparison of the flow visualization results for the two engine speeds
( 1.5 RPM and 20 RPM) in section IV.3 it was established that waviness and vortex
Shedding occur much earlier for an engine speed of 20 RPM, and that the flow field
enforced by 20 RPM engine speed exhibits much faster spread of flow structures. This is
due to the higher maximum Reynolds number in connection with the higher engine speed
(maximum Re of 4213 for 20 RPM over 316 for 1.5 RPM) which lead to higher Re at any
given crank-angle during intake. The threshold Reynolds number to cause instabilities in

109

the annulus fiee jet is thus encountered at a lesser crank-angle. Also, the length of the
transitional flow decreases more rapidly with evolving crank-angle for an engine speed of
20 RPM than with one of 1.5 RPM. This implies that turbulent flow exhibited by the
intake jet is encountered earlier in the intake stroke for the higher speed. Early generation

of turbulent motion will spread out the flow structures faster.

Concluding fiom the above it seems fair to hypothesize that fluid dynamical behaviour
in the area of scrutiny can be predicted largely by consideration of the intake jet and
ultimately by the setting of engine parameters such as engine speed and valve lift. By
varying, for instance, (J (moving valve or changing valve/bore ratio) and valve lift at
constant engine speed (e.g. idling) the engine designer could manipulate the characteristics
of the intake jet at [I in order to accomplish a variety of objectives. One being the search
for a compromise at idling conditions between low cyclic variations in order to obtain
steadiness of engine performance at low engine speed, and high turbulent motion within an

engine cycle to achieve enhanced mixing; two ambitions that do not necessarily go hand in
hand.

The results fi'om these studies of an axisymmetrical assembly of a four-stroke IC engine
might be an explanation why off-center valves as well as an implementation of two valves
in commercial IC engines exhibit better engine performance. Both configurations should

Yield better mixing during intake through earlier break-up of coherent vortex structures.

110

V3 Data from Evaluations of Instantaneous Velocities

 

V.3.1 sof in tan us veloci vect rs

Considering velocity fields from maps containing instantaneous velocity vectors, it can
be asserted a priori - and will be shown later - that the main toroidal recirculation pattern
cannot be resolved distinctly at all crank-angles after midstroke. There is only a weak
indication of the main recirculation pattern in the-form the TRP was encountered in the
EA domain (for the TRP in BA see section V. 1.1). This result is comparable to visual
observations of the induction flow field during the actual LIPA experiments when the
human eye by integration with respect to time and space discems a dominating rotational
flow, i.e. the recirculation pattern, and a temporal analysis reveals that instantly there may
be no measurable large scale rotation. This should be considered a warning to be cautious
in an analysis where ensemble averaged data are interpreted. Furthermore, the degree of
mixing, and the turbulent structures, as well as their length scales, cannot be deduced fiom
EA data - unless the prerequisite of negligibly low cycle-to-cycle variability is given. This

situation served initially as motivation for the analysis of instantaneous velocity

information, where cycle-to-cycle variability would not result in loss of information.

It is an exhaustive effort to analyze the entire scope of instantaneous velocity data
available for the nine variations on the engine parameters (see Table 11.2.3 in chapter II).
Each of the nine sets of data comprises a number of engine cycles -- that contributed to
the ensemble averaging procedure in sections V.1 and V2 -- where each cycle covers a
crank-angle range of 180 degrees. To stay within the framework set forth for this work,

the author had to focus on topics in the discussion of instantaneous velocity information

that promised interesting results.

111

V3. 1.1 Results from individual engine cycles for 20 RPM and 3 mm valve lift

In Figure V.3.1 the entire induction stroke is exhibited for 20 RPM engine speed and
3 mm valve lift as the flow field evolves instantaneously (engine cycle #27). Features of
the flow field as recognized in section V. 1.1 in the maps of EA velocity vectors can be
identified in the maps that contain instantaneous velocity vectors as well - intake jet,
boundary layer flow along cylinder wall, recirculation pattern, variation of induction angle.
Also, the development of the flow field is comparable to that found in the EA domain; this
is emphasized by the fact that fiom about midstroke on the area of scrutiny is exclusively
governed by the flow that turns inbound away fi'om the cylinder wall. It is in the detail
where the maps of instantaneous data and the maps fiom the EA domain difi‘er. For
instance, in the EA domain the velocity vectors in the area of scrutiny display, after
Inidstroke, a clear and structured toroidal recirculation pattern as sole rotational motion;
the instantaneous velocity vectors depict for the same area and crank-angle range smaller
scale rotational flow where a TRP motion serves generally as a fluid dynamical embedding
(Figure V.3. lm-x). The degree of difference between results from instantaneous and EA
data is reflected in the value of velocity fluctuations as calculated by equation V.2.b in
section V.2.

With respect to congruity between the flow patterns encountered in velocity maps
stemming from instantaneous observation and ensemble averaging technique, it can be
said that - for this set of data - there is profound similarity concerning the overall flow
Structure. This justifies the application of ensemble averaging techniques in the analysis of

problems in fluid mechanics. In this context, the reader's attention should be focused on
the flow situation at crank-angle 48°, for instance. Comparing the content of
Figure V.3. lg with the content of Figure V. 1 . 1 g the overall correspondence between the
We flow fields becomes obvious; what constitutes the difl‘erences between the two flow

fields are the details of orientation and magnitude of the individual velocity vectors.

112

From Figure V3. 1, it can be found that for crank-angles after midstroke, when the
maps of ensemble averaged velocity vectors exhibit rather stable flow fields, this stability
does not exist in the case of the ‘real' instantaneous flow within an individual engine cycle.
In general, vortical structures with a length scale of the radius of the cylinder, i.e. the main
toroidal recirculation pattern, can be only weakly perceived after midstroke - to various
degrees - in the flow‘ fields of Figure V.3.1; at times a rotational structure of this length
scale can be barely observed (128° CA, Figure V.3. 1p). A high amount of irregularity is
encountered in conjunction with the vortical flow patterns, like cross-flows between the
layers of the main TRP. Only at certain crank-angles (e. g. 112°, 156°; Figure V.3.1n&u)
does the flow field comprise a distinct vortical structure with a length scale of the radius of
the engine cylinder, namely the projection of the main toroidal recirculation pattern onto
the examined x, z - plane. However, these are singular occurrences.

High variation with respect to the structure of the flow field within one engine cycle is
experienced as crank-angle evolves illustrating that the mechanisms of turbulence are
inherent in the flow fields.

Figure V.3.1 illustrates the actual complexity of the induction flow for an individual
engine cycle. Complexity of the flow field cannot be perceived in the maps containing EA
velocity vectors (section V. 1.1) since it is essentially filtered out during the ensemble

averaging procedure.

Following are some interesting issues in connection with instantaneous velocity
information discovered during the analysis of the velocity maps in Figure V.3.1
highlighting the notion of flow field complexity encountered with instantaneous velocity
data.

From Figure V.3. la (CA 4°) it can be noticed that there is fluid motion prior to piston
movement in an individual engine cycle that cannot be observed in BA velocity maps. The

'high' velocity magnitudes encountered at 4° crank-angle indicating preexisting motion

113

cannot stem fi'om piston motion which just started to accelerate. Magnitude and
orientation of this initial movement of the fluid prior to piston motion difi‘ers with engine
cycle. Hence, the initial conditions are different for every engine cycle which is one of the
main causes for cyclic variations of the fluid dynamic quantities encountered during
induction.

It can be observed that, at times, the flow field around the axis of the engine cylinder is
rather symmetrical in the examined plane where the fluid is essentially convecting parallel
to the z - axis in negative direction (CA 148°, Figure V.3. Is). On the other hand, during
intake it is also possible that this symmetry is broken up by fluid that convects across the
cylinder axis, as seen at crank-angles 128° and 140° (Figure V.3.1p&r).

At 132° CA (Figure V.3.1q), a very strong circular vortical motion can be observed
depicted by four velocity vectors that are marked in the corresponding map in
Figure V.3. 1. As will be seen in section V.4 this vortex accounts for a distinct peak in the
distribution of the corresponding vorticity vector (Figure V.4.5).

It is interesting that by the end of the intake stroke (CA 180°, Figure V.3. 1x), there are
local areas of considerably high kinetic energy. Also, no predominating flow structure can
be discerned. The end of the intake stroke marks the beginning of the compression stroke

and is thus important with respect to providing initial conditions for compression.

It can be concluded that the flow field within an individual engine cycle is actually
highly complex and evolves with significantly difl‘erent motions fi'om cycle to cycle. The
EA based analysis washes out this pseudo random detail. A message should be conveyed
to be cautious with results originating fi'om applications of ensemble averaged analyses as
well as from applications of its related computational technique CFD. The description of
the flow field in the EA domain as consisting of discrete toroidal recirculation patterns is a
model; for individual engine cycles distinct recirculation patterns vary a lot and rarely look

like the ensemble averaged pattern. The precise recirculation pattern exhibited in the EA

114

domain never actually exists in the individual realizations. The smaller the valve lift and the
higher the engine speed, i.e. the higher the cycle-to-cycle fluctuations (see Figure V.2.2 in
section V.2), the higher is the degree of variability from the EA.

V.3.1.2 Variation of flow field with engine cycle and change in engine parameters

In order to examine the difl‘erence in the instantaneous behaviour of flow fields for
three consecutive engine cycles at a particular crank—angle, maps of instantaneous velocity
vectors are shown in Figure V.3.2 for 20 RPM engine speed and 3 mm valve lift at
132° CA (ECs #27, 28, and 29).

The notion of cycle-to-cycle variability becomes clear when comparing the flow fields
for the three engine cycles in Figure V.3.2. For EC #27 a strong vortical motion
dominates the flow field. For engine cycles #28 and 29 no distinct rotational flow with a
length scale of the radius of the engine cylinder can be perceived. In fact, the flow fields in
connection with these latter two engine cycles exhibit considerable similarity to each other,
especially in the upper left corner of the area of scrutiny where in Figure V.3 .2b and c
velocity vectors denote a flow towards the center of the valve. This is the crucial region
where the flow has to decide whether to recirculate - as it is roughly the case for EC #27
at 132° CA - or to convect towards the valve - as it is predominantly the case for
EC #29 at 132° CA The velocity maps in the EA domain yield, naturally, a compromise in
this question of recirculation.

Furthermore, the flow in the lower right comer of the three fields in Figure V.3.2a - c
depicted by velocity vectors #73 and 74 (for location of velocity vectors #73 and 74
within grid of laser lines see Figure II.2.3a) exhibit for EC #27 a vertical flow towards the
piston head, for EC #28 a flow turned inbound towards the cylinder axis, and for EC #29
a flow towards the comer between piston head and cylinder wall. The variance in the flow

behaviour in this lower right corner of the engine cylinder exhibited in these three engine

115

cycles is the underlying factor that results in the existence of a bufl‘er zone above the

piston head - as it was described in section V. 1.1 - that is seen in the EA domain.

The degree of cycle-to-cycle variations as encountered in the previous paragraphs is
dependent on the settings of the engine parameters (see section V.2.2, Figure V.2.2). This
is demonstrated by an investigation of flow fields at one crank-angle for three consecutive
engine cycles fiom two additional, different data sets. Examined are the flow fields for
10 RPM engine speed and 9 mm valve lift, as well as for 30 RPM engine speed and
3 mm VL. Recall fi'om Figure V.2.2 (section V.2.2) that for these settings of the engine
parameters velocity fluctuations are the lowest and the highest, respectively. Note that the
sequence of Figures V.3.2 to V.3.4 shows velocity maps - or rather distances and
directions fluid particles traveled - after employment of time delays of 0.32 ms for
30 RPM (Figure V.3.3), 0.64 ms for 20 RPM (Figure V.3.2), and 0.96 ms for 10 RPM
engine speed (Figure V3 .4). The difi‘erence in underlying time delay implies that the
magnitude of velocity vectors in Figure V.3.3 is generically two times and three times
higher with respect to the magnitude of velocity vectors in Figures V.3.2 and V.3.4,
respectively.

It is apparent by visual examination of the lengths of the arrows in the velocity maps in
Figures V.3.2 and V.3.3 that these lengths are about as high for 30 RPM engine speed as
they are for 20 RPM indicating a factor-of-two difference in the velocity magnitude - due
to the difl‘erence in the underlying time delay. This means that while the engine speed
increases by a factor of 1.5 - where the valve lift stays the same - the velocity magnitudes
are doubled in the flow field.

It is obvious fiom the flow fields of the three consecutive engine cycles in connection
with 10 RPM engine speed and 9 mm valve lift (Figure V.3.4) that there are cyclic
variations. However, the three flow fields have very similar characteristics, such as the

tendency to recirculation. Cycle-to-cycle variability is rather small, leading to increased

116

predictability of cych motion in connection with low engine speed and large valve lift.
The flow fields in conjunction with 30 RPM engine speed and 3 mm VL, on the other
hand, exhibit at one crank-angle during induction profound cycle-to-cycle variability as
can be seen in Figure V.3.3a - c. There is hardly any similarity between the three flow
fields displayed in Figure V.3.3.

The essence from this section of an analysis of cycle-to-cycle variations during engine
performance for three difi'erent data sets is that a visual verification of the results
manifested by the curves in Figure V.2.2 (section V.2.2), which indicate increase of

velocity fluctuations with decreasing valve lift and increasing engine speed, is established.

V.3.2 Time history of a velocity vector -- an approach towards gcle resolved data
Mysis

 

This section will explore profiles of instantaneous velocity vectors - with respect to
crank-angle of the piston movement - at selected locations in the flow field. The
recording of time histories of velocity vectors at a specific point within the engine cylinder
on the basis of LIPA - data is legitimate because the marking of one particular fluid
'particle' in a flow field by the corresponding intersection in the grid of laser lines occurs
consistently at the same location within the engine cylinder for each laser pulse throughout
the intake stroke. (A more detailed analysis reveals that a slight fluctuation in the
coordinates of a particular origin location of a velocity vector is encountered; this is due to
the actual assignment of the vectors' origin by incorporating information about to where
the fluid particle convected after excitation, section 11.1.1.2; this 'mis'location is however
of much smaller order than the size of coherent structures studied in the flow field and,
thus, negligible.) In section V.3.2.1 several velocity profiles will be exhibited for the
induction stroke and fluid dynamical conduct of the flow field will be inferred fiom them.

117

So far, we have been dealing with velocity fluctuations in general without really
distinguishing the fluid dynamical nature of the fluctuations which is essentially twofold:
Fluctuation in orientation and magnitude of velocity vectors is due to (1) cycle-to-cycle
(CtC) variability and (2) to turbulent motion of the flow field within an individual
realization of an engine cycle. In the preceding section (V .3.1) we obtained a notion about
cycle-to-cycle variability by qualitatively comparing maps of velocity vectors for one
setting of the engine parameters at one crank-angle for consecutive engine cycles. From
the same maps one can also infer flow structures suggested by the orientation of velocity
vectors in a certain neighborhood and thus attaining a feeling for turbulent motions in the
flow field. However, it has not been possible in the study of engines thus far to achieve
qualitative measures. In this section an attempt is made to separate the two sources of
velocity fluctuations - cycle-to-cycle variability and turbulent motion in the flow field --
by means of a cycle resolved analysis based on velocity profiles at selected locations in the
flow field.

Cyclic variation during the performance of an engine is the result of different fluid
mechanical behaviour of the flow field (e.g. strength and location of vortex structures)
fiom engine cycle to engine cycle. This is caused by difl‘erent initial conditions for each
engine cycle prior to the movement of the piston, which are partly due to the dissimilarity
of initial motion of fluid particles in the flow field at the beginning of the intake stroke.
Turbulence inherent in the flow field is characterized by the strength and length scales of

vortex structures in the flow field.

In a cycle resolved analysis, with the goal to difl‘erentiate between the influences of
CtC-variations and turbulent motion as causes for velocity fluctuations, results have been
obtained in this work for 10 RPM engine speed and 3 mm valve lift. (A cycle resolved
analysis makes use of velocity profiles as will be outlined in section V.3.2.2.) An engine
speed of 10 RPM was chosen because at this engine speed data acquisition achieved the

118

highest temporal resolution within the volume of experiments performed, namely 2° crank-
angle. (Unfortunately, this still has to be considered a low resolution in the context of
cycle resolved analyses.) The cycle resolved analysis has been conducted in the vicinity of
the axis of the intake jet (location #07) and in the area of scrutiny (location #26); see
Figure II.2.3a for position of locations within engine cylinder. In the cycle resolved
analysis only the z - component of the two-dimensional velocity vector was subject to
examination because this component is, in this study, afiliated with the highest accuracy;

see section III.3.

V.3.2.1 Presentation of velocity profiles

Since in E4 an average is given of all the fluid dynamical action taking place within the
examined 21 individual engine cycles (for 10 RPM engine speed, ensemble averaging is
performed over 21 ECs; section 11.2.2.4) the task of velocity profiles from the EA domain
is to suggest a trend or tendency that cannot be apprehended in the examination of
individual engine cycles.

In the vicinity of the axis of the intake jet (location #07) the EA velocity profile for the
w-component of the 2D velocity vector exhibits an ascending curve towards a local
maximum that is at about 45° CA (Figure V.3.5). Then the velocity decreases almost
symmetrically to the value which is encountered at the beginning of the stroke although
the piston speed increases until crank-angle 90°. This property of the velocity profile at
location #07 for 10 RPM engine speed and 3 mm valve lift is due to the deviation of the
induction angle with crank-angle as outlined in section V. 1.2. From the respective curve in
Figure V. 1.3 one can see that the intake jet first inclines towards the cylinder axis (until
about 45° CA) and then backs up towards the cylinder head. Location #07 is fixed with
respect to the x, z - coordinate system. Hence, the axis of the inclining intake jet

approaches this location and for the furthest inclination, location #07 falls approximately -

119

for this setting of the engine parameters - onto the jet axis generating the observed high
magnitudes of velocity in the z - direction at location #07. When the incoming jet inclines
back towards the cylinder head, location #07 moves, with respect to the jet axis, towards
the outer region of the incoming jet (increasing 0 and thus into regions with lower
velocity magnitudes. After midstroke w increases slightly towards a second local maximum
at about 120° CA which is much smaller and barely recognizable. Then, w decreases
towards the end of the stroke.

For location #26 (Figure V.3.6) a sharply descending curve can be noted for the EA
velocity in the z - direction between crank-angles 40° (when this location is being passed
by the accelerating piston head) and about 68°. From then until the end of the stroke an
essentially ascending curve is encountered. The orientation of the bulk flow at
location #26 is such that it is directed towards the valve throughout the entire stroke
- thus in opposite direction to the piston movement. The velocity profile depicting the
magnitude of the z - component of the velocity vector reflects in BA for location #26 the
average strength of the main toroidal recirculation pattern. From Figure V3 .6 it can then
be maintained that the energy content of the main TRP is highest for 10 RPM engine
speed and 3 mm valve lift at a crank-angle of about 68°; the maximum occurs thus
considerably before nridstroke when the piston speed reaches maximum. This indicates
that in addition to the momentum flux of the incoming jet flow, geometrical constraints,
and thus boundary conditions, play an important role in the development and timing of the

main recirculation pattern.

Velocity profiles from instantaneous velocity data can, for some engine cycles,
demonstrate significant difl‘erences to the velocity profiles fi'om ensemble averaged data
discussed in the preceding paragraphs. The instantaneous velocity profile of the z -
component shows at location #07 (Figure V.3.5) for engine cycle #23, for instance, a very

distinct second maximum around 110° crank-angle. \Vrthin other engine cycles there is no

120

second maximum at all in the velocity profile at this location (see Figure V.3.? where
velocity profiles fi'om eight engine cycles are shown for location #07). The behaviour of
the velocity magnitude for crank-angles around the second maximum is summarized within
the profiles in EA and displayed by a sketchy hill. The first maximum, however, is
encountered in the profiles for all engine cycles investigated which is why this maximum
can thus be found distinctly in the velocity profile fi'om the EA domain. When considering
the velocity profiles fi'om EC #04 and 23 in Figure V.3.5 one can perceive profound
difi‘erences between the profiles, signifying cyclic variations of fluid dynamical behaviour
at location #07 .

For location #26 the two individual velocity histories in Figure V.3.6 demonstrate very
similar characteristics as the one stemming fiom the EA analysis. In fact, when
qualitatively studying all velocity profiles fiom individual engine cycles at this location less
fluctuation about the ensemble averaged curve when compared to the results at
location #07 could be observed. Nevertheless, there are suficient difl‘erences between the
individual velocity profiles which make it possible to group them according to curve
characteristics: for several engine cycles, the descending and ascending parts of the
velocity profiles possess about the same slopes (#10, 11, 17, 24, and 26). This idea of
grouping realizations of engine cycles according to similarities in the behaviour of fluid
dynamic quantities is the basis for conditional sampling, an approach that clusters results
and evaluates mean quantities fiom each cluster. Conditional sampling is another approach
used in engine research - besides cycle resolved analysis - towards discrimination of

velocity fluctuations.

V3 .2.2 Notes on cycle resolved analyses

The objective of numerous investigations in the field of reciprocating engines is the

separation of velocity fluctuations due to cycle-to-cycle variability from those due to

121

turbulent motion within an engine cycle based primarily on velocity information. There

have been difl‘erent methods to extract turbulence intensities fiom the velocity fluctuations.

One method which is used sometimes in firing engines is a conditionally-sampling
technique (Witze et.al. 1984, Foster and Witze 1988). In this case data are classified into
sets according to the value of a specified parameter, e. g. flame arrival time. Vlfrtze et.al.
(1984) employed a conditionally-sampling technique to avoid a smearing of the ensemble

averaged mean velocity profile.

The method that has received most attention in the engine research community is based
on the evaluation of a mean velocity profile within one engine cycle (cycle resolved
analysis) fi'om instantaneous velocity data. Since exclusively single probe measurement
techniques offered the velocity information in previous investigations of the fluid dynamics
in engines, data were - for one engine cycle -- available only at one location in the flow
field. This available velocity information is plotted as a firnction of crank-angle and takes a
form of a velocity profile or history. The difference between the instantaneous and a mean
velocity profile within one engine cycle signifies, in this approach, a measure for velocity
fluctuations due to turbulent motion of the flow field. The difl‘erence between the curve
from an ensemble averaging analysis and the mean velocity profile in one engine cycle is a
measure for cych variations. For the method of cycle resolved analysis, approaches differ
in how to determine the mean velocity profile.

In order to summarize work that has been done in an attempt to evaluate mean velocity
profiles within one engine cycle on the basis of instantaneous velocity profiles, the author
will lean on a review on cycle resolved analyses furnished by Catania and Mittica (1989)
who discussed the most important techniques - including one modeled by themselves - in
a paper and compared the performances of the difl‘erent techniques to each other. There
are, according to Catania and Mittica (1989), essentially five difl‘erent schemes. Of these,
Lancaster (1976), Rask (1981), Catania and Mittica (1985), and Daneshyar and Fuller

122

(1986) divided the velocity profile fiom instantaneous velocity data for the entire stroke
into crank-angle segments or windows and determined mean velocity profiles by time-
averaging (Lancaster 1976, Catania and Mittica 1985, and Daneshyr and Fuller 1986) and
cubic spline fitting (Rask 1981) procedures. The fifth method by Liou and Santavicca
(1985) constitutes the application of a Fourier transform to the complete instantaneous
velocity profile. In the fi'equency domain, the argument goes that low frequencies are
related to fluctuations in the bulk flow of an engine and thus, to cyclic variations; high
fiequencies are considered to be found in connection with velocity fluctuations due to
turbulent motion. A cut-off frequency, which is according to Lion and Santavicca (1985)
assumed to be related to a characteristic time of the engine combustion process, then
separates the influences on velocity fluctuations due to turbulence and CtC-variability.

More recently, Hilton et.al. (1991) suggested a routine tha -- according to them -- is
closer to the physics of the flow. The method they outline in their publication incorporates
discrimination on the basis of phase stability. Their 'unified method' allows the investigator
to adopt an entirely parametric representation of the various flow components in the
model where the parameters can be estimated by minimizing the difl‘erence between

estimated and theoretical autocorrelation functions in a least-square sense.

The author of this work decided to employ a cycle resolved analysis because of the
nature of the available velocity data. In order to establish the mean velocity profile within
one engine cycle, curve fitting by a polynomial of 5th order, Fourier transform procedures,
and data smoothing techniques have been experimented with and applied to the profiles of
velocity vectors acquired in this study. None of the above yielded satisfactory results.
Eventually, a simplified version of the approach suggested by Daneshyar and Fuller (1986)
emerged and was adopted for the separation of the sources of velocity fluctuation. An
adoption of one of the more sophisticated methods that were recapitulated above would
have been beyond the scope of this work.

123

V.3.2.3 Procedure and results of cycle resolved analysis

The intention of the application of a cycle resolved analysis in this study is primarily to
provide a blueprint of the concept and the demonstration of the possibility of cycle
resolved analysis in conjunction with LIPA Therefore, a simplistic approach was chosen
for this study with respect to the cycle resolved analysis in that the representations for
mean velocity profiles within one engine cycle consist of simple straight lines within
assigned crank-angle segments or windows. A straight line was chosen to curve fit the
instantaneous velocity profile also because a more sophisticated cycle resolved analysis
would be unwarranted in the light of the rather mediocre accuracy of the present data sets
(on accuracy of the data see section 111.3). The straight lines are evaluated through a linear
regression technique provided by a software package.

For the analysis at location #07 (area of intake jet) the set of 90 data points (resolution
of 2° CA) was divided into three subsets of contiguous crank-angle for which within a
subset the linear regression technique evaluates the curve fit constituting the representing
mean velocity profile. The assignment of crank-angle ranges for the three subsets was
done by hand through incorporation of information on flow behaviour (e. g. fluctuation of
induction angle) inferred fi'om the studied velocity profiles. The basic setup of profiles at
location #07 exhibiting essentially two ascending characteristics and one slightly
descending one as for example for EC #23 (Figure V.3.5), stimulated the investigator of
this study to dissect the curve into three sections. In Figure V3 .7 results from the curve
fitting process for the three segments in connection with eight velocity profiles fiom
location #07 can be seen. The fluctuation of the instantaneous velocity information about
the linear curve fits symbolizes velocity fluctuation due to turbulent motion in the flow
field.

For the analysis at location #26 (area of scrutiny) the established 71 data points could
be divided into essentially two subsets representing two phases of fluid dynamical conduct

124

(increasing and decreasing velocity magnitude, see section V.3.2.1). In Figure V.3.8 linear

curve fits for velocity profiles at location #26 are exhibited.

At location #07 the curve fitting straight lines were evaluated for each of the three
subsets of an engine cycle, successively for all 21 engine cycles (setting of engine
parameters: 10 RPM engine speed, 3 mm valve lift). Then, for each of the 63 subsets, the
fluctuation of the instantaneous velocity profile about the curve fit was calculated
employing an RMS approach as a measure for the turbulent motion encountered in this
crank-angle window. Note that the crank-angle windows are rather extensive [in the
average 30 degree] so that no momentary information on the turbulence inherent in the
flow field can be obtained; only a global value characterizing the overall turbulent motion
for crank-angle ranges of about 30 degree is being firmished for the flow field of an intake
jet. An arithmetic average for each subset (I through III) over the investigated 21 engine
cycles is then evaluated and included in Table V.3. 1. This number for the fluctuation due
to turbulence is evaluated by the following overall formula:

1 use: 1 N) 2 in
TUuuu =31}; Z(-N—Z(Wr “ch,,) ] (V-3-a)

k=l , i=1 k

where TU is the fluctuation due to turbulent motion in the examined flow field, #ECs is
the number of engine cycles examined (21), Nj is the number of data points (crank-angle
positions) in the respective subset I, II, or HI; w,- is the instantaneous velocity in an
individual EC, and we”, is the corresponding value of the linear curve fit.

In order to obtain a measure for the cycle-to-cycle variations (CV) an RMS value has
been determined for the difl‘erence between the curve stemming from an ensemble
averaging procedure over the instantaneous velocity profiles and the linear curve fit. The
evaluation scheme for CV is identical to that in equation V.3 .a for TU after replacing w,-

by w“ and reads

125

1 05C, 1 NJ( 2 “2
CVIJIJII zmé 717:2"? War—Worn.) k (V313)

Naturally, the lower the intensity of the cyclic variations over several realizations of
intake strokes (CV) the better is the representation of the mean velocity profile within one

engine cycle by the ensemble averaged velocity profile.

Fluctuation intensities were evaluated as well by normalizing TU and CV within each
subset I, II, and III. The normalization factors for the three sections are inferred from the
ensemble averaged velocity profile in Figure V.3.5 by taking an arithmetic average of the
EA values in the respective crank-angle sections 1, II, and III. For instance, for section I
the normalizing factor for the calculated velocity fluctuations is the arithmetic mean of the
first 30 data points of the EA velocity profile.

For location #26 the same formulas, equations V.3.a and V.3.b, and evaluation
mechanisms for the normalization factors are efl‘ective. The difi‘erence is that they are

applied to only two subsets, I and II.

In Table V.3.1 the results of the application of the two equations V.3.a and V.3.b to the
LIPA - data is exhibited for location #07; Table V.3.2 contains the results for location #26.

Table V. 3. 1: RMS fluctuation and intensifies in connection with turbulent motion and
cycle-to-cycle variability at a location in the vicinity of the axis of the intake jet.

 

 

 

 

I Location #07 I
fiection I Section II Section III
RMS Intensity RMS Intensity KM5 lnteneity
[in m/e‘l [in m/e] [in m/el
Turbulence 0.177 0.702 0.161 1.226 0.105 1.541

CtC - Variation 0.116 0.461 0.117 0.669 0.077 1.142

126

Table V. 3. 2: RMS fluctuation and intensities in connection with turbulent motion and
cycle-to-cycle variability at a location in the area of scrutiny.

 

Location #26

 

 

 

Section I Section II

RMS Inteneity RMS Inteneity

[in m/el l'in m/el
Turbulence 0.076 0.466 0.079 0.766
CtC - Variation 0.065 0667 0.052 0515

Table V.3.1 demonstrates that for 10 RPM engine speed and 3 mm valve lift the RMS
fluctuation as well as the fluctuation intensity of the z - component of the velocity vector
evoked by turbulent motion of the flow field is, at a location in the region governed by the
incoming jet (velocity vector #07), consistently -- throughout the induction stroke -
higher than the fluctuations and the intensities stemming from velocity fluctuations due to
cycle-to-cycle variability; a factor of about 1.5 can be deduced for each section. For both
sources of velocity fluctuation the ‘decrease' (there are only three representing values for a
crank-angle range of 180°, making it hard to speak of a real decrease) of absolute RMS
fluctuation with crank-angle shows the same characteristic: For both, turbulence and
cyclic variations, the RMS values for sections I and II are about the same; then, towards
the end of the stroke, RMS fluctuation decreases. For the fluctuation intensity --
normalized by ensemble averaged velocity information - a different picture is presented.
Turbulence intensity as well as fluctuation intensity based on CtC-variations increase
consistently with crank-angle.

The cycle resolved analysis reveals for the area of scrutiny (location #26, Table V.3.2)
that RMS fluctuations of w in connection with turbulence as well as turbulence intensities

are, for an axisymmetric engine configuration run at 10 RPM engine speed, 1.2 and 1.5

127

times, for sections I and II respectively, higher than RMS fluctuations and normalized
intensities caused by CtC-variability. Considering the averaged values in Table V3 .2,

there is no decrease of RMS fluctuation caused by turbulence as crank-angle evolves from
CA 40° to the end of the stroke (180° CA). Table V.3.2 witnesses, however, a decrease of
RMS fluctuation in conjunction with CtC-variability as crank-angle advances fi'om

section I to section II. Again, fluctuation intensities increase as crank-angle evolves for
both sources of velocity fluctuation which is, particularly for the turbulence intensities, a
very welcome fact since high turbulence intensities account for an enhanced mixing
performance in the flow field of an engine. Moreover, studies have shown that an
increased level of turbulence intensity improves combustion performance.

When comparing the results obtained in the vicinity of the axis of the intake jet
(location #07) with those obtained in the area of scrutiny (location #26), it can be inferred
that for the time scales considered (smallest time scale resolved is 33 .33 ms) velocity
fluctuations -- absolute as well as normalized - of the z - component of the velocity
vector due to turbulence as well as due to cyclic variations are higher in the region
governed by the intake jet than in the area of scrutiny by a factor between 1.5 and 2.

For a regular production engine run at 1600 RPM engine speed Catania and Mittica
(1989) performed cycle resolved analysis at a location 3.5 mm underneath the flat cylinder
head between intake valve I and II. The result of their analysis is that velocity fluctuations
due to turbulence are only a quarter of the fluctuations due to cycle-to-cycle variability.
This is in strong contrast to results found in this study and demonstrates the profound

influence of boundary conditions and engine speed on the fluid dynamics of the flow field.

It goes without saying that in order to understand fluid dynamical mechanisms that
influence the flow field during induction of an engine - particularly those of the turbulent
motion within the flow field - velocity fluctuations caused by turbulence have to be

evaluated for concise crank-angle windows whose time-equivalent intervals are, in an

128

optimal case, smaller than the smallest characteristic time scale encountered in the flow
field. This could not be accomplished in this study due to the quality (e. g. low temporal
resolution) of the LIPA - data at hand so that averages for the velocity fluctuations due to
turbulence had to be evaluated over ranges of between 30 and 100 degree (section II in
the analysis of location #26) crank-angle. However, LIPA has the potential to provide the
necessary velocity profiles for a thorough investigation of turbulent motion on any length
and time scale and at any location within an individual engine cycle.

At this point only some qualitative observations concerning the dynamics of turbulent
motion during intake can be deduced from the plots of velocity profiles in Figure V.3.7
and V.3.8. Firstly, within section I of location #07 (region of intake jet) there is - as can
be expected - an increase of turbulent motion fi'om crank-angle 0° on (increasing
fluctuation of instantaneous velocity profile about cycle resolved mean velocity). It is of
considerable importance to know the behaviour of velocity fluctuations in conjunction
with turbulence as intake approaches the end of the stroke (section III for location #07
and section II for #26), in particular for the area of scrutiny, since this knowledge can be
used as initial condition for the compression stroke. This leads to the second statement on
dynamics of turbulent motion during intake, according to which, inferred from examining
the plots in Figures V.3.7 and V.3.8, it can be stated that the velocity fluctuations caused
by turbulent motion of the flow field cease only gradually towards the end of the induction
stroke, and that there is considerable turbulent motion by the end of the stroke in the

region around the axis of the intake jet, as well as in the area of scrutiny.

It is clear that this technique (LIPA) has the potential to cleanly separate the turbulence
fi'om the cycle-to-cycle variability. It is the author's opinion to not go into firrther depth
with this investigation because it does not make sense to base a more thorough
investigation of velocity fluctuations incorporating cycle resolved analysis on the basis of

the present data base; especially at this low engine spwd which is of merely acadernical

129

interest. It is suggested that experiments using higher temporal resolution of the
acquisition scheme for LIPA - data and higher engine speeds, as well as longer ensemble
averages be made to yield significant information about the turbulence and cycle-to-cycle

variability in various engine designs.

V.4 Vorticity Contour Maps - Indicator for Mixing Performance

In the evaluation process of a vorticity vector - in conjunction with an application of
LIPA -- a boundary is placed around a certain set of velocity vectors and the behaviour of
these encircled velocity vectors with respect to magnitude and orientation is summarized
in one vorticity vector. This yields two - among others - characteristics of vorticity
distributions that have been taken advantage of in this work. (1) Distributions of vorticity
vectors can serve as indicator for the degree of structure encountered in a flow field
which, in turn, can be used to derive information about mixing performance. It is
intuitively appealing that the higher the degree of structure of the flow field the worse is
the resulting mixing performance due to the inhibition of molecular exchange between
layers of flow. (2) When using vorticity distributions, convection mechanisms of flow
structures can be observed qualitatively over ranges of crank-angle since flow structures
stand out distinctly in these distributions, e. g. a region of vortical flow is denoted by high
vorticity and a region of parallel flow by vorticity equal to zero. Also, axisymmetrical

aspects of the flow field can be studied easier in the vorticity domain.

Concerning data reduction of vorticity data, the same as what has been said about the
data available on instantaneous velocity vectors holds true for vorticity information: The
vast amount of vorticity data - characterizing the intake flow of an engine - made
accessible through LIPA cannot be analyzed within the flame of this work. However, a

initial investigation - which should serve as a starting point in the thorough study of the

130

vorticity information at hand at a later point - featuring vorticity distributions for the
whole engine cylinder at one crank-angle as they differ for the diverse settings of the
engine parameters, will be performed.

For this investigation each vorticity vector is calculated fiom three neighboring velocity
vectors. Thus, fiom the investigation grid in Figure II.2.3a 116 vorticity vectors can be
inferred as can be seen in Figure V.4.1. Since the two-dimensional x, z - plane of the
engine model at a particular location 9 has been investigated in this study only the
component of the vorticity vector perpendicular to the x, z - plane could be evaluated.
This component lies in the circumferential direction of the cylinder.

In Figures V.4.2 to V.4.10 the vorticity distributions are represented as contour
surfaces in two- and three-dimensional plots. In this investigation, there are nine sets of
data corresponding to the nine settings of the engine parameters (see Table 1123 in
chapter II) fiom which data will be presented. More specifically, data are available for the
three engine speeds examined (10, 20, and 30 RPM) along with the three valve lifts
investigated (3, 6, and 9 mm). Each Figure (V .4.2 through V.4. 10; all in all nine, one for
each data set) contains vorticity information at crank-angle 132° for one realization of an
individual engine cycle. The presentation of vorticity data will be delivered in two- and
three—dimensional contour plots where each plot confers a difl‘erent perspective of the
vorticity distribution to achieve a complimentary picture of the whole distribution. The
3D plots are exhibited for two difi‘erent views with respect to the azimuth.

The 2D plots in Figures V.4.2 to V.4.10 depict the geometrical setting in which the
circumferential component of the vorticity vector (are) has been evaluated. In the 3D plots
the magnitude of this component is denoted as a surface contour. For these plots, the
valve has to be pictured to be in the corner at small x - and small 2 - values. It has to be
noted that the perimeter locations of the array of locations where vorticity vectors are

being evaluated do not form a rectangle (see Figure V.4. 1). This induces dificulties for

131

the software used for the generation of these plots. The software can render only
rectangular grids for the determination of contour plots. This means that at areas for
which there is no vorticity data available the software extrapolates the existing data. These
extrapolations account for the steep vorticity distributions in the comers of the plots for

the two- as well as the three-dimensional vorticity contours.

Leafing vaguely through the contour plots in Figures V.4.2 - V.4.10 one perceives that
the vorticity distributions which are derived from instantaneous velocity information
exhibit surfaces in the two- as well as in the three-dimensional plots which can be
described as landscapes that include features like hills and valleys. Hills correspond to
areas of positive circumferential vorticity; positive vorticity encountered on the right hand
side of the cylinder axis is defined for the direction of the vorticity as possessing the same
orientation as the main toroidal recirculation pattern. Valleys, then, correspond to negative
regions of the circumferential vorticity.

From intuitive principles it makes sense to associate mixing performance -- of the fresh
entering fluid with the fluid already in the cylinder - during induction with the landscape
formed by the vorticity distribution. Vorticity distributions depicting high hill/valley
fluctuation indicating small scale structure in the flow field suggest enhanced mixing
mechanisms. Enhanced mixing mechanisms intermingle the new and old phases of fluid
present in the cylinder on the molecular level in order to provide a homogeneously mixed
fluid. These aspects of the vorticity distribution will be discussed in this section by using

the presented vorticity data.

There is one hill/valley formation that can be found for all data sets at the same region
within the examined x, z - plane. It is the hill/valley configuration characterizing the free
intake jet encountered during induction. The bill is located on the starboard side of the
intake jet, the valley on the port side. The existence of this configuration can be explained

when studying the Reynolds-averaged velocity profile of the streamwise velocity

132

component of the fiee intake jet. This profile is characterized by a Gaussian-like
distribution in the lateral direction where the maximum velocity of the profile coincides
with the axis of the jet. For an observer who is situated in the flow field on the starboard
side of the intake jet axis facing downstream, the high magnitude velocity vectors of the
jet center are to his/her left side. On his/her right side is an area of lower magnitude
velocity vectors. This setup yields vorticities of positive orientation on this side of the
intake jet and thus a hill in the contour plot. A similar argument explains the negative

orientation of the vorticity on the port side of the jet. This hill - valley combination can be

observed in the contour plots of the vorticity distributions in this section, especially in the
2D plots.

 

It is desirable to quantify the degree of hill/valley fluctuation along with the magnitude
of this fluctuation to achieve a measure with which it is possible to obtain insight into
mixing performance, and also to discriminate between landscapes of vorticity distribution,
particularly in the area of scrutiny, stemming fi'om the nine sets of data.

There are 27 locations in the area of scrutiny at which the circumferential component of
the vorticity vector has been evaluated in this study. (In Figure V.4.l these 27 locations
are connected by lines denoting the shortest distance between two locations of vorticity
evaluation; these lines form a hexagonal pattern.) This number of 27 does not sufiice to
perform accurate two-dimensional Fourier transform calculations which would yield the
desired quantification of the hill/valley fluctuation. For this work other solutions had to be
sought that are more fit for this particular problem. A solution was found in the form of
the first and second derivative of the vorticity distribution. An evaluation of first
derivatives (slopes in vorticity landscape) within a region of investigation together with

information about the variation - more precisely the sign changes - of the second

133

derivative (number of hills and valleys) within this same region provide information on the

magnitude and the frequency of the hill/valley fluctuations.

V.4. 1.1 Introduction of measures for the first and second derivative

Measures describing the first and the variation of the second derivative of the vorticity
distribution are evaluated for the area of scrutiny. Averaging procedures were employed
on the basis of these evaluations yielding single measures for one particular crank-angle
and one particular engine cycle and, subsequently, for one complete set of data
(corresponding to one setting of the engine parameters). The evaluation of the measure in
connection with the first derivative goes according to the scheme outlined in the
proceeding paragraph.

With the vorticity data in the area of scrutiny being available in discrete form at
27 locations - in networking these 27 locations where locations are only connected to
their nearest neighbors, a hexagonal pattern is generated (Figure V.4. l) - the first
derivative is evaluated by calculating the slope between neighboring vorticity evaluations
on the hexagonal scheme. To come up with a measure for the first derivative for one
particular crank-angle within one engine cycle the arithmetic mean of the slopes evaluated
for all vorticity pairs in the area of scrutiny is determined. This measure is then calculated
consecutively for all engine cycles within one set of data with subsequent application of an
averaging procedure over the results fi'om the individual engine cycles. This calculation
yields one point in the diagram in Figure V.4.ll representing the measure for the first
derivative (degree of slope) within the area of scrutiny for this set of data. In a

mathematical formulation the calculation reads for one particular CA:

 

05c: N _ .
sz=-—1— izmd “W (V.4.a)
#ECS i=1 N i=1 d j

134

where #ECs is the number of engine cycles over which averaging is performed, N is the
number of vorticity pairs in the area of scrutiny (namely 33; fiom hexagonal pattern), and
d is the spatial distance between the two locations of the two circumferential components

cool and (”c2 of the vorticity vector.

The evaluation of a measure for the variation - sign changes - of the second derivative
is with respect to the averaging process the same as the one for the first derivative. First,
the sign changes of the second derivative are determined in the area of scrutiny for
individual engine cycles at a particular crank-angle and these numbers are averaged over
all individual ECs for that data set. A sign change of the second derivative is determined
by comparing the signs of two neighboring evaluations of slope in the vorticity distribution
that is represented by the 27 vorticity evaluations. Ifthere is a difi‘erence in sign between
the two considered slopes either the top of a hill or the sole of a valley is detected and an
index in the algorithm is incremented. Comparison of two slopes for the detection of sign
changes includes three consecutive, 'chained' vorticities in the hexagonal pattern. Thus,

53 comparisons are performed in the area of scrutiny. The determination of SG being the
average of all sign changes occurring within all engine cycles for one data set is performed

for one crank-angle according to the following notation:

DECS

S0 = ICE—C; Z#sign_ changes, , where for one particular engine cycle
i=1
a) — (a a) — a)
#sigm_changes = #sign_changes + 1, if sip-(Li?) at sign[£i—53—)] (V .4.b)
I II

The results in conjunction with the evaluation of the above defined parameters (81 and
86) can be seen in Figures V.4.11 and V.4. 12.

135

V.4.1.2 Application of measures

Figures V.4.2 to V.4. 10 illustrate that for each engine speed examined (10, 20, and
30 RPM) the sequence for the three valve lifts (3, 6, and 9 mm) exhibits -- for 132° crank-
angle - a decrease of magnitude in the vorticity distributions with increasing VL. The
tendency suggested by this visual observation fi'om the 2D and 3D contour plots is
confirmed in Figure V.4.11 which contains at a crank-angle of 132° a measure for the
steepness (Sl, equation V.4.a) between hills and valleys in the landscape of the vorticity
distribution in the area of scrutiny. In this figure, where curves are plotted in a single
logarithmic coordinate system, the three curves for the three engine speeds demonstrate
clear trends with valve lift. It can be inferred that the larger the valve lift the less steep are
hills in the vorticity distribution. Also, 81 decreases for one valve lift with decreasing
engine speed. The curve for 30 RPM engine speed which suggests a linear behaviour in
the logarithmic plot must be considered the most reliable because averaging has been
performed over 64 engine cycles (20 RPM :> 42, 10 RPM :> 21).

Figure V.4.12 exhibits the measure for the frequency of the hill/valley fluctuation --
represented by SG -- in the landscape of the vorticity distribution in the area of scrutiny at
132° CA To within the margin of accuracy the three curves corresponding to the three
engine speeds not only align but are also constant with valve lift. This is an interesting fact
implying that the frequency of the hill/valley fluctuation is, in conjunction with the length
scale resolved by the applied LIPA laser grid, independent of engine speed and valve lifi.
To take it one step further, one might conclude on the basis of the results presented in
Figure V.4.12 that, on the length scale investigated, mixing performance is considerably
alike at all parameter settings of the engine. These results indicate how misleading an
analysis of mixing based on length scale alone can be. Clearly, the shape and magnitude of

136

the kinetic energy and vorticity energy spectrum must be taken into account, as well as the

range of the spectrum.

Although mixing is initiated on the scales the LIPA grids can resolve, ultimately mixing
is associated with small length scales which could not be resolved in this study. Further
investigation with denser spacing of laser lines in the investigation grid can yield insight on
what happens on those scales. As far as the large scales are concerned, the differences in
the nine investigated flow fields are within experimental error.

The constant values for SG along with increasing values of S] (for engine speed ll &
VL U) indicate that in the examined engine cylinder much stronger vortical motion exists
in connection with high engine speeds and small valve lifts on identical length scales.

V.4.2 Energy contained in vortical motions

A third quantity is established in this section, namely the energy contained in the
vorticity evaluated at the 27 locations in the area of scrutiny. This energy associated with
vortical motion - on the length scale examined - in the area of scrutiny was evaluated for
each of the nine settings of the engine parameters. The evaluation was performed at
132° CA in order to compare the results to the evaluations of the kinetic energy in
section V2. The nine values in Figure V.4. 13a and b depicting measures for the vortical
energy are averages of the energy associated with the circumferential component of the
vorticity vector over the area of scrutiny and the number of engine cycles examined for
each set of data. The calculation of the vortical energy (VB) is similar to that of the kinetic

energy in equation V.2.a and reads:

 

 

1 IEC: 1 N 2
VB = — a) . v.4.
#ECs Z[NZ " )1. ( c)

j=l i=1

 

137

where N is the number of vorticity evaluations in the area of scrutiny (27) and we is the

circumferential component of the vorticity vector.

Information on velocity magnitude and orientation fi’om several velocity vectors are
incorporated into the value of a local vorticity - and thus into the vortical energy defined
by equation V.4.c - as mentioned earlier. The kinetic energy, on the other hand, is a scalar
which denotes merely the order of the velocity magnitude. When comparing curves
containing energy information in connection with vorticity with curves containing kinetic
energy information in the same regions of interest, one can note - based on the statements
made in the preceding two sentences - the following: Ifthe structure (which is based on
orientation information from the velocity vectors) of the examined flow fields was to be
essentially the same for all nine flow fields resulting fi'om the nine settings of engine
parameters, orientation of velocity vectors becomes insignificant and it should be possible
to superpose curves for kinetic and vortical energy - as defined by equations V.2.a
and V.4.c - by appropriate scaling. In other words, if the orientation of velocity vectors
with respect to each other was, in the average, identical in the flow fields of the nine data
sets, the change of vorticity (energy) would be only dependent on the velocity magnitudes,
and, thus, the kinetic energy.

Figure V.4. 13 demonstrates that for 10 and 30 RPM engine speed the curves denoting
KB and VE could be overlapped by appropriate scaling, however, not for 20 RPM engine
speed. Employment of small data sample size and low resolution of the LIPA grid demand
caution, though, in the discussion of the vortical energy data, particularly for the higher
engine speeds and low valve lifis where small scale flow structure characterizes the flow
field which cannot be resolved appropriately (higher error in vorticity energy).

In Figure V.4.13a the actual energy magnitudes are plotted for the vortical and kinetic
energy in the area of scrutiny for the nine sets of data for 132° CA In Figure V.4. 13b
these curves have been normalized for each engine speed by the energy values for 9 mm

138

valve lift. The idea behind normalization was to try to find for one engine speed, the
difl‘erences between the two energy curves (KB and VE) with respect to valve lift. The
question after assimilation of the two energy curves at 9 mm VL is whether vorticity
energy increases proportionally with kinetic energy - if the two curves collapse it will be
seen as a verification of the fact that the flow structure is essentially the same for the three
valve lifts at one engine speed. Or whether there are differences encountered at the other
two VLs (3 and 6 mm) -- these difi‘erences must be due to differences in the structure of
the flow field.

The result is intermittent. While the two curves for kinetic and vortical energy in
Figure V.4. 13 for 10 and 30 RPM engine speed coincide to within the margin of error, the
two energy curves for 20 RPM engine speed exhibit conceivable difi‘erences. One notices
that for this speed the values for the vortical energy at 3 and 6 mm valve lifi decrease with
respect to the corresponding values for the kinetic energy, particularly for the normalized
curves in Figure V.4. 13b. One could speculate on slightly slippery grounds [since the
difl‘erences could be due to inaccuracy in the data reduction process; however, this
uncertainty will be put aside at this point for argument's sake] that the essence fiom the
preceding sentence implies that the structure of the flow field becomes difi‘erent with
decreasing valve lift on the length scale resolved. From the above observations, and taking
into account what has been said in section V.2.3 about the break-up of vortex structures
created by the intake jet, it is possible to conclude that for 20 RPM and 3 and 6 mm valve
lift the vortex rings of the intake jet break up before the jet reaches the cylinder wall - as
opposed to the 9 mm VL case where the vortex structures presumably survive until the jet
impinges on the wall - and thus creating the lower degree of structure responsible for the
non-unity ratio of kinetic and vortical energy for 3 and 6 mm VL.

Recall that the circumstance that the results for 20 RPM engine speed deviate from the
results for 10 and 30 RPM engine speed was also found in the discussion of the

139

recirculativity in section V.2.2. For an axisymmetric assembly with a bore/stroke ratio of
about '1' there must be some kind of transition in fluid dynamical behaviour at maximum

Reynolds numbers of around 4000.

V4.3 Miscellaneous topics involving vorticig distributions

V.4.3.1 Axisymmetry of the flow field

From the two-dimensional contour plots of the vorticity distributions in Figures V.4.2
to V.4.10 one can assert that the degree of axisymmetry about the cylinder axis exhibited
by the flow field is generally rather low for instantaneous vorticity data (recall that the
vorticity distributions in the upper and lower left corners are due to the extrapolation
procedure set forth by the used software). However, there are occasional times when
events exhibit some degree of axisymmetry, as for instance for 10 RPM engine speed and
3 mm valve lift (Figure V.4.2) where a hill to the left side of the cylinder axis has its
counterpart in a valley to the right of the axis. The marginal shift of the hill/valley
formation with respect to the cylinder axis might be due to slight asymmetries in the setup
of the engine model as has been experienced in the investigation for and outlined in
chapter IV.

For the highest engine speed (30 RPM), the author of this work perceived considerable
asymmetries of the flow field about the cylinder axis during the actual LIPA experiments
(taking advantage of the optical character of LIPA to provide a form of flow
visualization). Particularly for a valve lift of 6 m it could be observed that the setup of
the intake flow structure exhibits very high variations from cycle to cycle. For some cycles
the 'lefi' main recirculation pattern or ‘vortex' (in the observed plane) dominates, at others
the 'right' one. Then, during some cycles, considerable axisymmetry was encountered.

Other engine cycles witnessed convection of the left vortex to the right and again others

140

saw the right vortex convecting to the left, crossing the cylinder axis. Less frequently, a
stream of fluid crossing the cylinder axis diagonally towards the valve was noticed which
stems fi'om the flow at the outer perimeter of the main vortex (left or right). The fluid at
the perimeter of the vortex cannot follow the high curvature flow and breaks ofi‘.

In general, it was observed qualitatively during the LIPA experiments that axisymmetry
of the flow field essentially prevails until around midstroke and that the flow field then
becomes highly three-dimensional. The flow field during induction is naturally a function
of piston position. There is a steady build-up of the main toroidal recirculation pattern (as
has been shown in section V. 1.1). At midstroke the distance between piston and chamber
head is about equal to the radius of the main toroidal recirculation pattern. Thus, the
particular boundary conditions around 90° crank-angle are favorable for the main TRP
where the adverse pressure gradient the intake jet experiences above the piston head turns
the flow towards the cylinder axis and forces the fluid into a rotational flow (the cylinder
axis can in this context and in fluid dynamical terms be considered a plane wall);

Figure V.4. 14b. Also, at midstroke the incoming jet flow has the highest momentum flux,
resulting in the strongest rotational flow exhibited by the main TRP. After midstroke the
momentum flux of the intake jet decreases and the distance between piston and chamber
head increases with the consequence that the supply of momentum to the main TRP
becomes smaller, and that the adverse pressure gradient that serves as a mechanism to turn
the flow into the main TRP moves downwards and becomes less important;

Figure V.4. 14c. Consequently, after midstroke, the main toroidal recirculation pattern
cannot sustain its momentum against dissipation. Its two-dimensional appearance in the
laser grid plane diminishes, and out-of-plane motion can be perceived. This visual
observation has yet to be verified by LIPA - data.

141

V.4.3.2 Vorticity distribution in the EA domain

LIPA offers the opportunity to acquire planar data that is based on instantaneous
information of the flow field for both, maps of velocity vectors as well as for vorticity
distributions. This is one tremendous advantage that LIPA possesses over single-point
measurements performed by, e. g., Hot-wire and LDA techniques. Vorticity information in
the ensemble averaged domain, as offered by Hot-wire and LDA, cannot reflect, naturally,
the fluid dynamical mechanisms during individual engine cycles and therefore cannot
truthfully account for mixing performance. This shortcoming in conjunction with single-
probe measurement techniques is demonstrated in an impressive way by Figure V.4. 15
where a vorticity distribution evaluated fiom an ensemble averaged velocity field for
20 RPM engine speed and 3 mm valve lift at 132° crank-angle (Figure V.4. 15a and b) is
displayed side-by-side with a typical vorticity distribution from an individual engine cycle
with the same engine parameters (Figure V.4. 15c and d). [Note, that for the 2D plots in
this figure and for all 2D plots in the following figures, the 2D plots display vorticity
information for the same region as for the one in Figures V.4.2 through V.4.10.] The
3D plots show that the actual magnitude of the circumferential component of the vorticity
vector assumes much higher values within individual engine cycles than indicated by the
landscape of the vorticity distribution from the EA domain, where the vorticity distribution
presents itself as rather smooth. The hill/valley fluctuation appears to be higher in an
individual engine cycle as well.

In the beginning of section V.3.1 the author raised concerns about the validity of
conclusions based on results fi'om the EA domain. Here, these concerns experience
support. However, to be fair to results stemming fiom an analysis of EA data, information
from the EA domain has its legitimacy in connection with evaluations of, for instance, the

induction angle and the center of rotation of the main recirculation pattern.

142

V.4.3.3 Influences of crank-angle and engine cycle on vorticity distributions

Figure V.4.15 also contains a comparison of results at 132° CA (Figure V.4.15c and d)
and at 180° CA (end of stroke, Figure V.4.15e and f), both taken fi'om the same engine
cycle (#27) at 20 RPM engine speed and 3 mm VL. It is very useful to obtain information
about fluid dynamical quantities as they are encountered towards the end of the induction
stroke since this knowledge can be used as initial conditions for computational evaluations
of fluid dynamical quantities throughout the compression stroke.

For 180° CA one can see in the plots of Figure V.4.15e and f that the hill/valley
configuration characterizing the intake jet nearly disappears as can be expected since the
incoming flow through the valve ceases towards the end of the stroke. However, the
vorticity magnitudes in the distribution for the flow field at 180° CA seem to be not much
smaller when compared to those at 132° CA The hill/valley fluctuation even appears to be
about the same. In fact, calculations of 81, SG, and VE - which are defined above in
equations V.4.a to V.4.c and whose evaluations can be obtained from Table V.4.1 - for
180° crank-angle in the area of scrutiny indicate that the values of 81, SG, and VE for this
particular engine cycle are well in the range of the respective average-values encountered
at 132° crank-angle also provided in Table V.4.l (column 3). \Vrthin the engine cycle,

81 decreases moderately as the flow evolves fiom 132° CA towards the end of the stroke,
SG actually increases marginally, while the vortical energy decreases significantly. From
these evaluations of 81, SG, and VE obtained for an individual engine cycle one can
deduce that for 20 RPM engine speed and 3 mm VL the average vorticity magnitude in
the area of scrutiny decreases strongly towards the end of the intake stroke, and that the

spatial structure of the flow field remains appreciably constant.

143

Table V. 4.1: Comparison of SI, SG, and VE between two different crank-angle positions
for indirddual engine cycle as well as averaged over 42 ECS.

20 RPM, 3 mm VL, EC #27 Average at 132 deg
132 deg CA § 180 deg CA from 42 BCS
$1 (in l/smm) 39.69 32.22 31.79
SG (in %) 47.2 52.8 28.0
VE (in l/s"2) 32343 16097 18103

 

In Figure V.4.16 vorticity distributions for 20 RPM engine speed and 3 mm valve lift at
crank-angle 132° have been plotted for three consecutive engine cycles (#27, 28, and 29).
From Figure V.4. 17, where the value of 81 within individual engine cycles prior to the
application of the averaging procedure over the number of engine cycles for this set of
engine parameters (20 RPM engine speed and 3 mm VL) is shown as a function of engine
cycle, it can be seen that ECs #27, 28, and 29 exhibit considerably different levels of
slopes between neighboring locations of vorticity evaluations. This is, to some extent,
discernible qualitatively in the two- and three-dimensional plots in Figure V.4. 16a
through f. Slopes between hills and valleys in the distribution of the circumferential
component of the vorticity vector appear to be higher in the area of scrutiny in
Figure V.4. 16a and b than in Figure V.4. 16c and d, for example.

In an efi‘ort to point out differences and similarities between the vorticity distributions
encountered for the three consecutive engine cycles it can be noticed that one substantial
difference between the vorticity distributions for engine cycles #27 and 29 is established by
a long-stretched area of vorticity located below the hill/valley configuration that is created
by the intake jet. In engine cycle #27 this area contains negative vorticity, in EC #29
positive vorticity values are observed. In engine cycle #28 no such long-stretched region

can be observed at all. The differences in the landscapes of the vorticity distribution

144

between engine cycles for one set of data and one crank-angle illustrate and symbolize the

cycle-to-cycle variability that is so often the subject to research in the engine community.

It is, in the discussion of vorticity data, also interesting to study for one particular
engine parameter setting the evolution of the vorticity distribution with crank-angle within
one engine cycle. Using vorticity distributions, one can more directly answer questions
about development and dissipation of flow structures than it is fi’om a sequence of maps
containing instantaneous velocity vectors. As mentioned earlier, one vorticity vector
combines information on magnitude and orientation of several velocity vectors 'which
makes it easier to depict flow structures in distributions of vorticity. The study of the
evolution of flow structures with crank-angle on the basis of vorticity distributions has
been done for 10 RPM engine spwd and 9 mm valve lift. At 10 RPM engine speed, the
highest temporal resolution could be achieved for this work (2° crank-angle). A valve lift
of 9 mm was chosen because the structures in a flow field in connection with 10 RPM
engine speed and 9 mm VL can be expected to exhibit, in terms of fluid dynamics, the
most stable behaviour among the flow fields investigated due to the lowest momentum
flux of the incoming jet. Therefore, it is for this case the most likely that flow structures,
as denoted by vorticity distributions, can be identified in one plot and followed as crank-
angle evolves.

The results are presented in Figure V.4. 18 for a sequence of distributions of the
circumferential vorticity component right around mid-stroke (88°, 90°, and 92° CA).
Despite an overall similarity of the three plots it is not possible to distinguish
unambiguously areas of vorticity that stem fi'om an evolvement of one bill or valley as
crank-angle advances; no unequivocal statement can be rendered according to which a
particular hill (valley) in the vorticity distribution at one crank-angle can be identified in
the distribution at the proceeding crank-angle position.

145

Another sequence of vorticity distributions for three consecutive crank-angles is
exhibited in Figure V.4.19 at lower piston speed at crank-angles 150°, 152°, and 154°.
Here again, the vorticity distributions appear similar and it is again not possible to point

out specific vorticity structures and follow them through the sequence.

The time scale in connection with the rearrangement of the vorticity distribution
between examined crank-angles is obviously smaller than the data sample rate of 33 .33 ms
- even for this setting of low engine speed and large valve lift; thus, not allowing for an
unambiguous identification of a particular hill (valley) in vorticity distributions at two
consecutive crank-angles. There is possibly one exception, however. In the lower left
comer of the two-dimensional plot in Figure V.4. l9e a very distinct and prominent
hill/valley system can be discerned for 154° crank-angle. The early phase of development
of this system can be perceived vaguely in Figure V.4. 19c (152° CA). This system must
be, for several degrees of crank-angle of piston motion, so stable that it does not get
destroyed, or rotated so as not to be oriented normal to our data plane. In fact, it could be
observed at 156° CA (map of vorticity distribution for this CA is not shown in figure) as
well; between crank-angle 154° and 156° the system is however dissipating and thus

becoming very weak.

146
V5 Annulus Intake Jet

The LIPA grid of laser lines has been re-scaled to provide higher resolution data in a
region around the valve in order to investigate the incoming free annulus jet. Velocity
vectors could be determined at 15 locations fi'om this application of LIPA More detailed
information on the applied grid (Figure II.2.3b), its resolution, and the accuracy of the
measurements is available in chapters 11 and III. The intake jet is examined at three
difl‘erent settings of the engine parameters (engine speed and valve lift); two valve lifts
- 3 and 9 mm - are considered at an engine speed of 20 RPM, and the 3 mm valve lift
case is investigated at an engine speed of 30 RPM. Ensemble averaged results as well as

maps of instantaneous velocity vectors will be discussed in the following sections.

The most interesting fluid dynamical issue in the study of the intake jet 'under the
magnifying-glass’ are the interactions between the annular intake jet and the two toroidal
recirculation patterns bordering the jet, in particular that of the main TRP. The goal in this
section on the scrutiny of the intake jet is then to achieve a physical description of these
interactions and to measure the strength of the generated and manipulated flow structures
of the intake jet as they evolve in space and time. The interactions between the jet and the
periphery of the main TRP appear to be one of the major keys for the fluid dynamical

processes in the entire engine as has been suggested in section V2.

The maps of velocity vectors - or better, the maps containing distances that fluid
particles convected in the direction denoted by arrows - presented in the following
sections, depict the flow field as it is represented 0.17, 0.34, and 0.51 ms (for 30 RPM
engine speed and3 mmVL, for20RPMand3 mmVL, and for20RPMand9mmVL,
respectively) after the laser fired to build up the grid that illuminates the fluid particles.
Note that the time intervals are in integer multiples of 0.17 to simplify the act of

comparison of the plots for different engine speeds and valve lifts.

147

Except for a few issues that concern fluid dynamical descriptions of the global
behaviour of the intake jet, it is not feasible to compare the results obtained fi'om
experiments with 3 mm valve lift to those obtained with 9 mm valve lift, e. g. to relate
velocity vectors on a vector-to-vector basis. This is because the grid of laser lines marks
fluid particles at different locations with respect to the axis of the intake jet for the two
cases (the investigation grid was moved for the LIPA experiments examining the intake jet
at 9 mm VL in order to have intersections of laser lines at fluid dynamically interesting
spots in that flow field). Additional difliculties arise when trying to compare results
because of the fluctuation of the induction angle, which causes the locally fixed velocity
vectors to represent - as crank-angle evolves - difl‘erent areas within the flow field. This,
firrthermore, makes it diflicult to compare results for one engine parameter setting at
difi‘erent crank-angles.

V.5.1EAd main

The ensemble averaged results are obtained from 40 engine cycles for the 20 RPM
engine speed, and fi'om 60 engine cycles for 30 RPM. Again, reasonable convergence was
found. As in section V.1.1, the case of the 20 RPM engine speed with the 3 mm valve lift
will be examined in greater detail since flow visualization results are available for this set
of engine parameters, and the results are reasonably representative for those obtained with

difl‘erent engine speed and different valve lift.

V.5.1.l 20 RPM engine speed, 3 mm valve lift

In order to illustrate at what locations within the engine cylinder, especially with
respect to the fi'ee annulus jet, the grid of laser lines enables a quantitative analysis of the
flow field to be made, superpositions of maps of EA velocity vectors onto pictures

(instantaneous realizations) generated by flow visualization have been constructed (this

148

also reduces the chances of rnisinterpretations of flow structures). The superpositions have
the efi'ect of filling in the space between individual vectors with information. The
superpositions were done at crank-angles 32°, 52°, and 92°. The result is shown in

Figure IV.2.2g, j, and k.

In studying the congruity of velocity vectors with the flow structures, it is obvious that
the ensemble averaged velocity vectors do not and cannot describe in Figure IV.2.2g, j,
and k exactly the particular instantaneous flow situation denoted by the flow visualization
results, primarily within the region dominated by the intake jet [recall that EAs describe
preferred directions of the flow field]. Both the fact that the jet is turbulent and because of
cycle-to-cycle variability at the moment of data acquisition (whether via flow visualization

or LIPA), the imager faces different details of the flow structure.

As a first result from the sequence of the three superposed images in Figure IV.2.2g, j,
and k it is obvious that the length of the transitional flow decreases between 32° CA and
midstroke because of the increasing Reynolds number. This observation is congruent with

the results by Yule (1978).

Table V.5. 1: Assignment of individual velocity vectors to flow structures

 

 

 

I Veloci; Vector # Region of the flow field that is described by these velocity vectors]
2 and 4 Edge of intake jet close to inlet valve

 

1, 3, and 5 Corner between cylinder wall and cylinder head

 

7, 10, 13, and 15 Periphery of main recirculation pattern

 

 

 

6 and 8 Edge of intake jet that interferes with motion from upper corner flow
9 and 12 Edge of intake jet that interferes with periphery of main TRP
14 Around axis of intake jet

 

1 1 indecisive

 

 

 

 

149

Based on the superposed data in Figure IV.2.2g, j, and k one can assign the 15 velocity
vectors to certain flow regions and structures. (Figure II.2.3b contains the number
assignment for each of the 15 velocity vectors examined in this chapter. For
Figure IV.2.2g, j, and k the investigation grid was flipped.) Velocity vectors #7, 10, 13,
and 15 are located in the periphery of the main recirculation pattern. This flow structure
has a rotational bulk flow with turbulent motion associated with the smaller length scales
within the TRP. One expects these length scales to be smaller than the spacing density of
laser lines in the investigation grid indicating that these four velocity vectors must be
statistically independent fi'om each other. Velocity vectors #2 and 4 denote the upper edge
of the incoming jet close to the valve when the jet commences to exhibit instabilities, i.e.
waviness. Velocity vectors #1, 3, and 5 represent the fluid motion in the corner between
the cylinder head and wall, including that of the toroidal recirculation pattern in this area.
Velocity vector #11 cannot be assigned to any particular flow structure; to variable
degrees and intermittently, depending on the crank-angle, it describes the upper corner
TRP and the intake jet. The remaining five velocity vectors portray the intake jet after the
jet matured downstream. In Figure IV .2.2g (32° CA) velocity vectors #6, 12, and 14
depict coherent flow structures; in Figure IV.2.2j and k (52° CA and 92° CA) velocity
vectors #6, 8, l2, and 14 denote eddies in the turbulent jet after it experienced vortex ring
break-up. A summary of the role of each velocity vector in the description of the flow field
is given in Table V5. 1. From Figure IV .2.2g, j, and k one can conclude that the velocity
vectors appear to be very reasonable representations of the motion of the flow structure
that they quantify.

It is apparent that the coarse spacing of the grid of laser lines ofi‘ers very low spatial
resolution between velocity vectors. No fluid dynamic quantity defined by spatial
integration or difl‘erentiation, e.g. length scales fiom integration, can be determined for

particular flow structures with this setup since the procedure of integration or

150

difi‘erentiation over several velocity vectors in the present maps would include information
fiom other flow structures which leaves the calculation invalid. Also, since some of the
velocity vectors are statistically independent they cannot be the basis for the calculation
for one specific integration quantity. Furthermore, the temporal resolution (4° crank-angle
corresponding to 33 .33 ms) is too low to determine turbulence intensities or length and
time scales fi'om velocity vector profiles at one location by employing, for instance,
Taylor's hypothesis in an interpolation scheme. Hence, the way we will use the data will be
analogous to having 15 individual probes in the flow that are non-intrusive and which give

velocity information in any direction without bias.

There is a region in the flow field dominated by the incoming jet which can be called the
'interface zone'. In the interface zone the flow structures of the incoming free annulus jet
and the main toroidal recirculation pattern meet, interact, and change each other. More
specifically, in this area the azimuthal vortex ring structures created by the intake jet come
into contact with the turbulent periphery of the main TRP. The evolution of fluid dynamics
in this zone manifest the mechanism of direct transfer of momentum to the main TRP. In
the velocity maps for 20 RPM and 3 mm valve lift acquired for this work this zone
comprises basically, unfortunately, only two vectors, #9 and 12. From these, supported by
flow visualization results, it can be observed that the contact of the main TRP with the
intake jet is essentially tangential. Because of the low spatial resolution, however, it would
be rather risky to infer further manifestations fiom an observation of the two single
velocity vectors #9 and 12 for the purpose of a deduction of firndamental fluid dynamics
fi'om the interface zone. At this point it is worthwhile mentioning that the potential
ultimate goal as outlined in the beginning of section V.5 -- namely, a thorough
investigation of the interactions of the intake jet with the two TRPs - cannot be met in
this work due to the coarse spacing of lines in the investigation grid. One can see

however, that an analysis of the interface zone involving higher spatial resolution of the

151

LIPA grid, though, will yield remarkable insight into the fluid dynamics of the interactions
between the intake jet and the main recirculation pattern which in turn influences the main
TRP and thus the induction stroke as a whole. This, of course, presupposed that both a
high resolution LIPA grid and 'simultaneous' flow visualization results are obtained fiom
one and the same engine cycle. What the author can do in this study, nevertheless, based
on the data inferred from the applied coarse laser grid is deliver singular pieces of
information of a mosaic, where the mosaic is the comprehensive description and modeling
of the jet-TRP interactions, and possibly furnish speculations about fluid dynamics and
suggestions for the improvement of the measurement technique on grounds of these

pieces.

Hence, despite the coarse spacing, some interesting properties of the intake flow can be
unveiled by the ensemble averaged velocity maps in Figure V.5.1 where velocity
information is given for 20 RPM engine speed and 3 mm valve lift at selected crank-angle
positions.

In general, the maps exhibit considerably stable flow fields after midstroke, which is a
verification of results obtained in section V.1.1, although there are some velocity
fluctuations observable as crank-angle evolves in the periphery of the main TRP as well as
in the interface zone which can be expected due to highly turbulent motion and the limited
number of images over which the EA procedure was performed.

As can be seen in the pictures of Figure IV.2.2g, j, and k the core of the intake jet is
confined to such a small width compared to the spatial resolution of the grid of laser lines
that it is dificult to capture the intake jet with intersections of laser lines in order to infer
fluid dynamical quantities fiom the jet's core. The jet appears to be depicted, though, in
the velocity maps of Figure V.5. lg at 84° CA by velocity vectors #2, 4, 8, and 11,
however, it is not clear to what extent and where within the jet. Nevertheless, by

considering orientations of velocity vectors it is possible to 'construct' the intake jet by eye

152

and thus geometrically classify fluid dynamical conduct of the flow field in an engine. For
example, one can observe the inclination behaviour of the intake jet as it inclines first
towards the cylinder axis and is being pushed back up again (Figure V.5. la through g).

In studying the individual maps containing ensemble averaged velocity vectors in
Figure V.5.1 one notices that until shortly before midstroke (about 80° CA) velocity
vectors #7, 10, 13, and 15 which after midstroke depict the periphery of the main TRP,
document the inclination of the intake jet towards the cylinder head. Another interesting
observation is that fiom 80° CA on (after the intake jet has been pushed back up by the
main TRP) it seems to be possible to discern the flow of vortex ring structures in the
interface zone in the velocity maps (depicted mainly by velocity vector #12). The
presumably observable flow of vortex structures in the velocity maps is denoted by the
fact that the orientation of vector #12 prescribes a flow away from the axis of the intake
jet. However, information provided by this one velocity vector is too ambiguous to arrive
at qualitative conclusions about vortex structures generated by the intake jet.

It is an interesting occurrence that velocity vector #2 experiences a reorientation
towards the end of the induction stroke by about 180°. From Figure V.5.1 one can see
that velocity vector #2 assumes at crank-angles 172° and 180° a direction opposite from
that of what the intake jet prescribes throughout the stroke. This reorientation process is
encountered for an engine speed of 30 RPM (3 mm VL) as well; in this case it sets in
earlier.

The upper corner toroidal recirculation pattern also has an influence on the
development of the intake jet. A set of four velocity vectors (#1, 3, 5, and 11) essentially
represent the upper corner TRP as mentioned at the beginning of this section. They
evidence a very stable flow pattern in the upper comer after midstroke. The generation of
the TRP takes place in the upper comer shortly before midstroke as discussed in
section IV.1.2 and outlined in section V5. 1.3 in the discussion of the results for 20 RPM

153

and 9 mm valve lift. Unfortunately, the flow of the fluid transported by the upper corner
TRP cannot be unambiguously characterized when it is approaching the jet (velocity
vectors #6 and 8). This is because the actual intake jet cannot be discerned clearly from
the maps, and one therefore cannot estimate the space available to the TRP to turn into
the direction of the jet. It is evident again that the spatial spacing of IAs in the laser grid
for the examination of the intake jet is far fiom satisfying. One is compelled to
speculations about the fluid dynamics at the site where the upper comer recirculation
pattern and the intake jet join, in particular about the meaning of velocity vectors #6 and 8
in the velocity maps. Two interpolation scenarios are possible. Ifthe jet evolves close to
the two vectors (#6 and 8) the flow in the upper corner will impinge onto the incoming jet
and the contact of the upper corner flow with the intake jet will directly impart momentum
to the flow in the upper corner through shear forces. The upper corner flow does not
really recirculate (however, considering continuity aspects of the impinging flow, it seems
fair to say that it turns into the direction of the intake jet flow), and it would seem that the
flow in the upper corner forces, by impinging onto the incoming jet, a downward motion
onto the jet. If the intake jet is firrther away fi'om vectors #6 and 8 it would give the upper
corner flow more room to turn into the direction of the jet and recirculate. However, it
does not seem likely that there is enough room for the upper corner recirculation pattern
to reorient to such a degree that it would touch the intake jet tangentially as does the main
TRP on the other side of the jet. In this case, fluid carried by the upper corner TRP gets
entrained into the intake jet by the toroidal vortex rings shedding from the jet.

The effect of the upper corner toroidal recirculation pattern on the intake jet is of a
considerably different nature than the one of the main TRP. Although this is a qualitative
result, it is not the type that can be inferred from the flow visualization pictures in
chapter IV.

In the EA domain, when the field of velocity vectors indicate similar orientation in

connection with equally high velocity magnitudes, we can assume that there is a very

154

structured flow. The velocity vectors in the upper corner (#1, 3, and 5) exhibit just that,
suggesting that the recirculation pattern in the upper corner is much more structured than,
for instance, the periphery of the main TRP depicted by velocity vectors #7, 10, 13,

and 15. These indicate uncorrelated, low magnitude velocities. The turbulent flow field
constituting the periphery of the main recirculation pattern consists of flow structures that
convected a long way after their creation by the intake jet and, through that convection,
have undergone diverse interactions with the difi‘erent elements of the entire intake stroke,
such as the boundary layer flow along the cylinder wall, explaining the fine scale
structured flow field in the periphery of the main TRP. The eddy structures of the upper
comer recirculation pattern have a shorter history. Velocity vectors #1, 3, and 5 depicting
the upper corner TRP indicate that the TRP's bulk flow dominates over the motion of the
eddies.

The accounts of fluid dynamical behaviour outlined in the preceding paragraphs are
based on the results of 15 data points distributed throughout the region of investigation,
and rely heavily on the previous quantifications and flow visualizations. Even with 15 data
points in this region, it is very clear that the physics cannot be easily determined - this
speaks very loudly for the need to area information of the LIPA type. To verify the
suggestions made above a study is required where a grid of laser lines with much denser

spatial spacing is employed.

V.5.1.2 30 RPM engine speed, 3 mm valve lift

Velocity distributions look very much alike for 20 and 30 RPM engine speeds with
3 mm valve lift, which is why the 30 RPM case will not be discussed in depth; only a few
notes are worthwhile mentioning.

First, momentum of the intake jet increases with engine speed as has been outlined in
section V.2.2. The distances denoted in the maps in Figure V.5.2 for 30 RPM engine

155

speed are representative for the flow field after 0.17 ms, which means that the velocities
would be twice as high in a time frame that is associated with the results for 20 RPM
engine speed and 3 mm valve lift in Figure V.5.1 (0.34 ms). From Figure V.5.2 it is then
possible to infer that the velocity at the valve is indeed significantly higher for 30 RPM
(compare flow situations around midstroke, Figures V.5.1h and V.5.20).

Then, the increased inclination of velocity vector #4 (which depicts the upper edge of
the incoming jet for the 3 mm VL case) towards the cylinder axis throughout the intake
stroke in the velocity maps for 30 RPM engine speed as compared to the inclination of the
same vector for 20 RPM suggests that the induction angle for the intake jet associated
with 30 RPM engine speed is smaller, thus the intake jet steeper. This is a verification of
the corresponding curves in Figure V. 1.3 from section V. 1.2 (induction angle vs CA). This
occurrence makes sense from first principles according to which the higher engine speed
(30 RPM) imposes a higher pressure gradient which in turn leads to a steeper route of the

intake jet when it enters the cylinder.

v.5.1.3 20 RPM engine speed, 9 mm valve 1111

As mentioned in the beginning of this chapter, for 9 mm valve lift the investigation grid
had been adjusted to have intersection points, and thus velocity information, at potentially
interesting locations in the flow field. Unfortunately, no embedding of the velocity vectors
in a picture of flow visualization results can be achieved here due to the lack of flow
visualization results at this engine parameter setting. Hence, it is not possible to properly
assign the individual velocity vectors to flow structures and the interpretation of the flow
fields is, in conjunction with the low laser grid resolution, even more vague. However,
some general issues can be discussed on the basis of the results with this engine parameter

setting.

156

The flow patterns for the realization of 20 RPM engine speed and 9 mm valve lift as
displayed in Figure V.5.3 in terms of maps of velocity vectors is very stable after
midstroke until the very end of the stroke which is not, to this degree, the case for the
smaller valve lifts. This must be an indication for the more structured flow patterns in
connection with slower development of flow structures as typical for larger valve lifts
which has been discussed in section V.2.

For this setup of the laser grid, one laser line passes through the cylinder very close to
the evolving intake jet; see for example the flow situation at 96° crank-angle
(Figure V.5.3c) when the induction angle reaches a local maximum (see Table V.5.2).
Before CA 96° the strength of the main TRP increases with crank-angle and pushes the
intake jet upward. This had been observed in section V. 1 . 1. During this phase the flow
pattern in the upper comer does not exhibit strong momentum. From 92° CA on it could
be observed for 20 RPM engine speed and 9 mm VL that the upper corner flow of fluid
seems to gain strength. As a matter of fact, the motion in the upper corner can be
described a toroidal recirculation pattern only fi'om this moment on, as has been confirmed
by flow visualization results at low speeds (section IV. 1.2). The created upper corner TRP
forces the intake jet to move slightly back down towards the cylinder axis. From 108° CA
on the induction angle stabilizes and fluctuates marginally about the plateau value (see
Table V.5.2). The fluctuation is governed now by the interactions of the azimuthal vortex

structures of the intake jet with the recirculation patterns to either side of the jet.

The grid of laser lines did not reach far down enough in these experiments with the
9 mm valve lift in order to display what has been called the interface zone in the previous
sections. In fact, with the awareness of what the LIPA grid of laser lines did not
accomplish one can go back to the experimental phase and improve the grid. The

application of LIPA to a particular problem is a reiterative process.

157

V5 1.4 Comparison between investigated cases

In general, very good conformity can be established between the maps of velocity
vectors that present the intake jet, and the ones that present the overall fluid motion in the
engine cylinder which had been discussed in section V. 1 . 1. Furthermore, very good afi'mity
can be found when comparing the intake flow at 20 RPM engine speed and 3 mm valve lift
as represented by the velocity maps with the intake flow as represented in the flow
visualization pictures. The agreement of the two sets of LIPA results for the two different
views with each other and their agreement with the flow visualization results establishes an
additional sense of the accuracy of the LIPA experiments which leads to confidence in the
results.

A profile of induction angle versus crank-angle can be extracted qualitatively fiom the
plots of velocity vectors in Figure V.5.1 as the induction angle decreases towards a local
minimum, increases towards a local maximum, and then fluctuates about a fairly constant
plateau value (compare to results in section V. 1.2). A detailed quantitative measurement
of the induction angle is not undertaken because the velocity vectors are considerably
ambiguous about what the actual angle is under which the intake jet enters the engine
cylinder since one does not know where in the velocity maps the axis of the intake jet is
exactly. One is thus compelled to interpolate between velocity vectors which contain
velocity components fiom the vortex rings shed fi'om the jet in order to evaluate the
induction angle. A partial quantitative verification of the results obtained in section V. 1.2
is, however, accomplished if one determines the crank-angle at which the induction angle
assumes its highest value (highest drift back up towards the cylinder head) as well as the
crank-angle from which on the induction angle assumes its plateau value. The appropriate
numbers for the three data sets investigated in section V.5 can be found in Table V.5.2.
The accuracy of these numbers is believed to be higher than the ones in section V. 1 .2

158

because of the increased spatial spacing of laser lines in the LIPA grid with which the
intake jet is being examined. The correspondence to the results in conjunction with
induction angles as determined in section V. 1.2 is safely within the uncertainty associated
with the evaluation of the induction angle itself.

Table V.5. 2: Occurrence of specific events in profile of induction angle; the
numbers indicate the degree of crank-angle.

i=1? 1 ll 1 ' ‘i 1"} :Q’; ‘3' it") 1.?"

v .

angle at which
induction angle plateau
assumes local maximum is started

   

 

 

V.5.1.5 The intake flow in terms of non-linear chaotic system dynamics

Based on the results gathered in the preceding sections one can summarize and theorize
the following: The azimuthal vortex ring structures created by the the annulus jet come
into contact with the periphery of the main toroidal recirculation pattern in the interface
zone. The fluid dynamical interactions of the vortex structures with the recirculation
pattern determine the shape of the flow field downstream including that of the main TRP.
The induction stroke, thus, appears to consist, in terms of a model, of a reiterating control
system with one input - namely the fiee jet - where the main TRP challenges the intake
jet and the manipulated intake jet in turn influences the physics of the main TRP. The
system is governed by steadily changing boundary conditions, i.e. the vertically moving
piston head. In the EA domain this control system reaches equilibrium some time after

midstroke when the boundary condition becomes unimportant. At this time, according to

159

section V.5.1.3, the toroidal recirculation pattern in the upper comer is created. The upper
comer TRP also effects the intake jet which in turn influences the development of the
upper corner TRP. Hence, after midstroke, a second control system is set up. In this
context, a literature research on dynamics of non-linear chaotic systems might be helpfirl.
The consideration of the induction flow field in terms of models will be beneficial for
computerized analysis of the induction flow, in particular of the fluid dynamical
characteristics of the main toroidal recirculation pattern, e. g. in terms of mixing
performance. For this, of course, the interactions of vortex rings with the main TRP in the
interface zone have to be understood.

As a practical application of system philosophy according to which the intake jet flow
can be regarded as consisting of two reiterating systems (see preceding paragraph), a 'flip-
flopping' mechanism maintained by the two control systems is thinkable after midstroke
that explains the fluctuation of the induction angle. Flip-flopping could take place
according to the following scheme: As a result of the intake jet's inclination away fiom the
cylinder axis the upper corner TRP experiences an increased supplement of momentum
and, in return, pushes the intake jet back down towards the cylinder axis. Now, the main
TRP experiences an increase in momentum and, therefore, pushes the intake jet back up to
increase the supply of momentum to the upper corner TRP. The possibilities of fluid
dynamical interaction between the upper corner TRP and the intake jet comprise then both
options of flow field conduct delineated in section V.5.1.1, namely impinging of high
momentum flow onto the jet and recirculation of low momentum flow. This is merely a
hypothesis. But an intuitively appealing one. The time scale that is associated with the
increasing supplement of momentum to the upper corner TRP with subsequent inclination
of the induction angle towards the cylinder axis is of smaller order than the one associated

with the corresponding mechanism enforced by the main TRP.

160

V.5.2 Analysis of individual engr_r_1° e gcles

There are issues, such as momentum exchange of intake jet as well as magnitudes and
correlations of small scale motion, that cannot be discussed in the EA domain since the
procedure of averaging destroys instantaneous information about the flow field. Averaging
melts the diversity of information available fiom individual engine cycles about, for
instance, a flow structure at a particular location in the flow field - information such as
size and the difi‘erent states of development of that flow structure - into one average flow
structure - which never, of course, actually exists. This virtual flow structure summarizes
the behaviour of the instantaneous flow structures. Furthermore, the averaging process
levels out extreme values. The discussion of issues like the momentum exchange of the
intake jet requires cycle resolved maps of instantaneous velocity vectors that reveal
instantaneous in-cycle flow field information.

Unfortunately, as demonstrated in section V5. 1, the spatial spacing of the applied
investigation grid for the study of the intake jet is much too coarse in this work to achieve
a thorough picture of fluid dynamical mechanisms that involve a free annulus jet flow
interacting with toroidal recirculation patterns on either side of the jet. Hence, it is
impossible to infer a comprehensive, coherent fluid dynamics model of the flow within and
around the annulus intake jet. Furthermore, it turns out that intersection areas encountered
in conjunction with the LIPA grid from Figure II.2.3b are not small enough to resolve the
small scale motion inherent in the flow. But, they do resolve the large scale energy
containing motion. Thus, a few interesting features of the intake flow field, beyond those
found in the preceding section, can be deduced fi'om the analysis of individual engine

cycles where results stem fiom the application of the 'close-up' grid of laser lines.

In the following, some details from the maps of velocity vectors fi'om individual engine
cycles (EC) will be discussed. For the 20 RPM engine speed and 3 mm valve lift case,

161

engine cycle #49 is a good example of an individual realization of the intake stroke.
Results for this EC are displayed in Figure V.5.4. For 30 RPM and 3 mm VL, EC #24 is
analyzed (Figure V.5.5); for 20 RPM and 9 mm VL it is EC #14 (Figure V.5.6).

V.5.2.1 20 and 30 RPM engine speed, 3 mm valve lift

For 20 RPM engine speed and 3 mm VL it can be observed in Figure V.5.4 that the
intake jet is displayed by the velocity vectors in the maps at particular moments (crank-
angles) when the axis of the intake jet lies very close to one laser line that is part of the
laser grid (e.g. at 132° CA, Figure V.5.4q). At crank-angle 128° the axis of the incoming
jet coincides with the laser line as well, indicating that for two consecutive maps of
velocity information (crank-angle resolution for 20 RPM engine speed is 4°) the induction
angle possesses the same value. On the other hand, at 152° CA (not shown in figure) the
intake jet is depicted by velocity vectors #2, 4, 6, 8, and 11 also, however, four degrees
before as well as after 152° CA, the axis of the jet is away fi'om this laser line (for the
velocity field at 156° CA see Figure V.5.4r). The succession of these three maps serves as
an illustration for the fluctuation of the induction angle after midstroke. The sequence of
velocity maps in Figure V.5.4 shows that the intake jet is highest back up (local maximum
for induction angle) for 84° crank-angle (Figure V.5.4j). It also shows that at 132° CA the
induction angle has about the same value as for 84°. The results on the induction angle
inferred from EA velocity maps in section V. 1 .2 indicate that the induction angle does not
reach the value of the local maximum before the very end of the induction stroke. This is
another incidence where instantaneous data is not properly represented by results deduced
in the EA domain.

The flow visualization results in section IV.2.2 suggest a repetition time scale for the
toroidal vortex structures in the intake jet of about 33 ms (or 4° CA) between 25° and

40° crank-angle. If one assumes that there is a correlation between the repetition rate of

162

vortex rings and the fluctuation of the induction angle after midstroke, the time scale
associated with the fluctuations of the induction angle after midstroke should be of the
order of 33 ms as well. The range of 25° to 40° CA corresponds to 140° and 155° CA if
one assumes symmetry of fluid dynamical behaviour for the intake jet about midstroke due
to the symmetrical acceleration characteristic attained by the forcing movement of the
piston head. As observed in the preceding paragraph, at 128° and 132° CA the intake jet
assumes about the same induction angle indicating a time scale in association with the
fluctuation of the induction angle of 33 ms. This corresponds closely with what has been
found from the flow visualization results as a time scale for the development of azimuthal
vortex rings in the intake jet. However, the time scale associated with the fluctuation of
the induction angle has large error bounds since the temporal resolution of 4° crank-angle
for an engine speed of 20 RPM transforms into a sampling rate lower than the Nyquist
fi'equency for this problem. A closer comparison of time scales will stem from an
investigation employing a higher sampling rate.

The observed congruity of vortex structures at 28° and 32° CA in the flow visualization
results in section IV.2.2 cannot be observed unambiguously in the respective maps of
velocity vectors inferred fi'om LIPA - data provided by the close-up laser grid. Again
higher resolution quantitative data is needed. It would be interesting to see whether LIPA
results based on a field of velocity vectors derived fiom a LIPA grid with denser spatial
spacing of the laser lines would show the basic similarity seen in the flow visualization, but

still quantify the effects of the environment on the formation of these vortices.

Even with the higher resolution grid, Figure II.2.3b, it is not possible to study the
vortex structures that are generated by the intake jet. However, it was indicated in
section V.5.1.1 that there is evidence of some sort in the velocity maps for flow of vortex
structures in the vicinity of the core of the intake jet. More specifically, as could be

observed in section V.5.1.1, velocity vector #12 which is located in the interface zone was

163

found to exhibit at times, in EA, an orientation that cannot be expected to stem from a
flow that is parallel to the axis of the incoming jet. From the maps containing
instantaneous velocity vectors now (Figure V.5.4), it can be seen that the velocity vectors
representing the fluid dynamics in the interface zone - basically #9 and 12 - have
intermittently an orientation away from the jet axis; for vector #12 more often than for #9.
This fluctuation of orientation of these two vectors might denote the motion of a fluid
particle which is turned away fi'om the jet axis by either the motion of a toroidal vortex
ring, that of a fluid particle being entrained into the vortex ring, or that of a fluid particle
within an eddy in the turbulent flow phase of the intake jet - for the latter case velocity
vector #14 can be included in the study. This assumption that velocity vectors #9 and 12
depict the motion of a vortex structure obtains support fi'om the images generated by
superposition of a map of velocity vectors with the corresponding picture from the flow
visualization results where velocity vectors #9 and 12 are located at the edge of the intake
jet. An attempt to quantify the flow structures as represented by velocity vectors #9

and 12 is not being performed because of the uncertainty about the location of the vectors
within the flow structure at any particular moment due to the fluctuation of the induction
angle.

In this context, similar results can be obtained from a study of the velocity maps
stemming fi'om an engine parameter setting of 30 RPM engine speed and 3 mm valve lift
(Figure V.5 . 5).

There are two options to quantify flow structures using maps that contain velocity
vectors: Either the LIPA results obtained with the coarse investigation grid has to be
accompanied with a visual picture of the flow field that is recorded at the very same
instant as the distorted LIPA grid, or much closer spacing of the intersection areas in a
LIPA grid has to be accomplished yielding distances between knots in the laser grid far
smaller than the length scale of the flow structures that are being studied.

164

In an analysis of the velocity vectors in the periphery of the main TRP (#7, 10, 13,
and 15) in engine cycle #49 as exhibited in Figure V.5.4 it can be established from visual
analysis of the vector maps that after midstroke these velocity vectors seem to be
correlated at times (96° and 100° CA, Figure V.5 .4l&m) and uncorrelated at others
(120° CA, Figure V.5.4o). For 30 RPM and 3 mm VL (Figure V.5.5) correlation is
encountered at 108° and 132° CA (Figure V.5.5c&d); at 144° CA (not shown in
Figure V.5.5) the vectors appear to be uncorrelated.

As an example for an individual velocity vector that is located within the main
recirculation pattern, vector #13 describes the strong fluctuating motion characteristic of
the eddies in the main TRP, especially after midstroke when a consistent flow field is
established for the engine cylinder as a whole. This is the case for 20 as well as for
30 RPM engine speed and 3 mm VL.

V.5.2.2 20 RPM engine speed, 9 mm valve lift

For 20 RPM engine speed and 9 mm valve lift, engine cycle #14 is displayed in
Figure V.5.6. At midstroke (88° CA, Figure V.5.6b) the axis of the intake jet overlaps
with one laser line fiom the LIPA grid. The velocity vectors depicted in the map in
Figure V.5.6b to either side of the jet exhibit identical low magnitude characteristics, in
fact, they suggest the width of the intake jet for this setting of engine parameters; even the
orientation of these velocity vectors match roughly on a vector-to-vector basis when the
axis of the intake jet is taken as symmetry line. This suggests that the velocity vectors
generated by the laser line that overlaps with the intake jet depicts the jet at its centerline.
A very symmetric flow situation about the axis of the intake jet is encountered at this
particular moment. It is conceivable that the velocity vectors to either side of the intake jet
describe the interactions between the flow structures at the edge of the jet with the

recirculation patterns to either side. However, no unequivocal statement can be made

165

about the fluid dynamics that are described by the vectors to either side of the jet since
these velocity vectors cannot be assigned indisputably to regions in the flow field around
the intake jet due to the lack of simultaneous flow visualization results at this engine

parameter setting.

V.5.2.3 Momentum of the fi'ee annular intake jet

At around midstroke of induction, at least for one map of instantaneous velocity
vectors for each of the three data sets examined in this section, the axis of the intake jet
coincides more or less with a laser line. This enables a rough quantification of momentum
in the intake jet by using the magnitudes of the velocity vectors that depict the incoming
jet as velocities encountered at the jet axis Uc. The result is exhibited in Table V.5.3 where
the momentum flux J is calculated according to equation V.2.d in section V.2.2 and
normalized by density. The trend of the change of angular momentum with engine
parameter setting is as expected: The normalized momentum increases with decreasing
valve lifi and increasing engine speed.

In Table V. 5.4 two ratios are evaluated for the increase of momentum flux when engine
parameters are varied. The first ratio M denotes the increase of momentum when the valve
lift is lowered to 3 mm at constant engine speed; the second ratio III/II tells the increase
when engine speed is increased to 30 RPM at constant valve lift. The very same ratios were
calculated using the data available on the content of kinetic energy in the area of scrutiny
fi'om section V.2.2. It can be stated that for momentum flux as well as for kinetic energy in
the area of scrutiny the first ratio M is smaller than the second ratio III/II. These
tendencies assert that the increase of momentum with increasing engine speed from 20 to
30 RPM is stronger than the increase of momentum when the valve is being closed fi'om
9 mm to 3 mm. The fact that the momentum flux and the kinetic energy content in the area
of scrutiny both exhibit similar tendencies as speed is increased and valve lift decreased,

underline the notion of an existence of a proportional link between the level of momentum

166

carried by the intake jet and the level of kinetic energy in the area of scrutiny.

Table V.5. 3: Normalized momentum as a function of engine parameter setting.

 

Engine Parameter Setting

 

 

20RPM9mmVL 20RPM3mmVL 30RPM3mmVL

 

 

[in m"3/s"2]

 

 

 

I l 11 [[1
Maximum velocity
at center ofjet [in m/s] 1‘09 2-1 3.55
Momentum J / p
7 .56E-03 9.39E-03 2.67E-02

 

 

 

Table V.5. 4: Comparison of ratios as encountered for momentum and

 

 

 

for kinetic energy.
Ratio
II/I III/II
Momentum 1.24 2.85
Kinetic Energy 1.91 3.07

 

 

V.5.3 Velocity fluctuation within the intake jet

Miller and Comings (1 95 7) found a distribution of the streamwise velocity fluctuation
for a two-dimensional jet flow that can be descnbed as 'M'-shaped about the jet-axis with

round edges of the 'M‘. The local maxima of this distribution to either side of the jet can be

found, according to their study, at about n*=1 where the jet width b is at about n*=4

167

(11" is the normalized parameter of the distance Q perpendicular to the jet axis divided by
the jet width where the origin of C; is at the axis of the jet). Their velocity fluctuation
profile for the velocity component in C - direction exhibits an axisymmetric Gaussian-like
distribution about the jet axis.

To compare Miller and Comings' results to a velocity fluctuation distribution evaluated
fi'om LIPA - data the velocity fluctuation in z - direction is displayed in Figure V.5.7
where the singular evaluations of the fluctuations within the examined x, z - plane are
exhibited in a three-dimensional needle plot for the RMS value of the velocity
fluctuation wRMS as well as for its normalized counterpart wm/WEA. As mentioned
elsewhere the velocity fluctuation in z - direction has been chosen because of the higher
accuracy in connection with the data acquisition and reduction process for the z -
component of the velocity vector. Note that the velocity fluctuation in z - direction does
not represent the one in streamwise direction as was examined by Miller and Comings,
thus, the results of this study will not strictly match the curves by Miller and Comings.
However, in investigating the velocity fluctuation results in Figure v.5 .7, especially the
normalized fluctuation intensity distributions, one can discern a distinct valley (low values
of the fluctuation) with the sole of the valley being at the axis of the jet. This valley is most
distinct for the case of 30 RPM engine speed and 3 mm VL. Note that the error due to the
movement of the highest intensity spot within an intersection area has been taken out of
the data in Figure V.5.7 (see section 111.33).

It would be tempting to determine the edge of the annular jet from the fluctuation
intensity distributions taking into account Miller and Comings' scaling measure that the
maximum of the distribution has to be at 11"=1. But fi'om the results of the superposition
of a velocity map onto the corresponding flow visualization image for 20 RPM and 3 mm
valve lift (Figure IV.2.2g, j, and k) it is known that for this set of engine parameters
velocity vectors #9 and 12 represent the edge of the intake jet. This knowledge makes it
possible to locate the edge of the jet in Figure V.5.7B. From these diagrams it can be seen

168

that with a firrther increase in C beyond the edge of the jet the velocity fluctuation
increases. In this region of increased C, the locations of velocity vectors (#10, 13, and 15)
must be assigned to the periphery of the main toroidal recirculation pattern. It can be
speculated then that between velocity vectors #12 and 15, and between #9 and 13 there
ought to be a maximum in the fluctuations representing the interactions between the intake
jet and the periphery of the main TRP.

For 30 RPM engine speed, thus higher Reynolds number than for 20 RPM, and equal
valve lift the spread of the intake jet is slower meaning that the width of the jet is smaller
at any given 5,. This is reflected in the plots of Figure V.5.7y. Based on the argument for
20 RPM and 3 mm VL that the maximum velocity fluctuation ought to be found between
vectors #9 and 13 this very same maximum has to move to a smaller C due to the higher
Re and can thus be found now at a location close to velocity vector #9. This is exactly the

situation encountered in Figure V.5.7by.

When comparing the three plots in Figure V.5.7 containing the distributions of the
normalized velocity fluctuation intensity for the three engine parameter settings
(Figure V.5.7b) the expectation that the intensity increases with decreasing valve lift and
increasing engine speed is confirmed. The expected reciprocal behaviour for the width of
the intake jet - namely the decrease of the width with decreasing VL and increasing RPM

- is verified as well.

Contradictory to what had been conjectured in section V.5.1.1 on the basis of the
velocity maps in the EA domain for velocity fluctuations encountered in the two
recirculation patterns, the velocity fluctuations within the periphery of the main TRP and
within the upper corner TRP can be found to be, in the plots of Figure V.5.7, of the same
order of magnitude.

169

Due to the coarse spatial spacing of intersection areas and the mediocre accuracy
(section 111.33) in connection with the LIPA grid from Figure II.2.3b the main goal set
forth for section V.5, namely the description and quantification of flow structures in the
vicinity of the axis of the annular intake jet fell significantly short. Predominantly,
qualitative observations stemming from maps of velocity vectors were delineated where at
some points the study did not go beyond an achievement of fluid dynamical justifications
of specific velocity vectors. With a few exceptions no quantifications of fluid motion was
accomplished. However, this study is useful for a follow-up investigation where the

know-how acquired in this work can be used to design LIPA experiments.

Chapter VI
CONCLUSION AND OUTLOOK

The prime objective of this investigation was to apply the velocimetry technique LIPA
(Laser Induced Photochemical Anemometry) to study the induction stroke of a model
four-stroke engine in order to quantify the fluid dynamics of the encountered flow fields.
A secondary objective was to develop algorithms to enable accurate automation of the
data reduction technique for LIPA Also, as a consequence of the research, new flow
visualization has uncovered detail here-to-fore unexplored in very low RPM engines. The
research has paved the way to using whole field data by creating diagnostics that give
information about vortical structures -- their scales and intensities and number -- needed to

study engine modifications that lead to enhanced mixing in engines.

LIPA was successfully implemented for the examination of the induction stroke of the
water analog model. Data acquigition as well as reduction procedures were established
that assured the generation of LIPA - data (laser grids exhibiting distorted grid lines) and
ultimately velocity and vorticity information on induction. Similarities between flow
visualization images obtained for the induction stroke of the water analog model and fields
of velocity vectors corroborated the LIPA results. Flow visualization results also helped to
formulate new fluid dynamical conclusions about the flow field during induction

In order to determine the accuracy of the LIPA - data, several tests were performed to
achieve estimations of errors. The main source of error in the LIPA technique as employed
in this work occurs because of the fact that intersections formed by crossings of laser lines
of the LIPA grids possess considerable extent. \Vrthin these intersection areas, regions of
highest pixel intensity were able to move driven by, for instance, mechanical vrbration in

the apparatus. Thinness of laser lines and spatial resolution obtained by the lines of the
170

171

LIPA laser grid were found to be the essential issues in conjunction with the optical

implementation of the LIPA grid employed for flow field investigation.

Based on the planar information of two-dimensional velocity vectors at assigned and
consistent locations within a flow field, the fluid dynamics encountered during induction of
the water analog model were examined. Two engine parameters (engine speed and valve
lift) were varied to establish a dependency between the flow field and fluid dynamical
quantities encountered during the intake stroke of an engine, and boundary conditions.
Engine speeds of 10, 20, and 30 RPM were investigated as well as 3, 6, and 9 mm valve
lifts employing two LIPA grids with difl'erent spatial spacing of the laser lines: One grid
with which it was possible to obtain a field of velocity vectors whose region of study
covered most of a plane containing the axis of the engine cylinder; and another grid of
higher spatial resolution with which the incoming jet flow was analyzed.

In general, the information density in terms of spatial and temporal resolution in
connection with the LIPA images acquired was found to be adequate to examine the large
scale structures, but too low to examine the dissipative structure and the induction jet.
However, as mentioned elsewhere, it was beyond the scope of this study to achieve a
thorough, comprehensive understanding of the physics of the flow field. Shortcomings in

regard to resolution together with suggestions for their improvement are outlined.

On the basis of fields of velocity vectors (section V. l) and supported by flow
visualization results the flow field has been dissected into fluid dynamical elements; namely
a fi'ee annulus jet emanating fi'om the tip of the valve, two toroidal recirculation patterns -
one occupying most of the volume of the engine cylinder, and the other one in the corner
between cylinder wall and the intake jet -, a boundary layer flow along the cylinder wall,
and a bufi‘er zone above the piston head. Some bilateral interactions between these

elements as well as efi’ects of one element on another have been investigated.

172

Considering the intake jet first, the angle under which the intake jet enters the engine
cylinder (induction angle) as a function of crank-angle is dependent on the setting of the
engine parameters. For a constant engine speed it can be observed that the intake jet soon
after it is established, inclines fastest towards the cylinder axis, and then back up towards
the chamber head to assume a constant 'plateau' value for small valve lifts. The larger
valve lifts have the smaller plateau value. It can be observed, firrthermore, that the higher
the engine speed is the smaller is the difi‘erence between the highest and smallest plateau
value for a particular engine speed. It was argued in the text that the rates of change of jet
inclination are faster for small valve lifts because the built-up of the main recirculation
pattern is quicker in connection with smaller valve lifts, thus forcing the incoming annular
jet back up faster. A reason for the plateau value towards which the induction angle
converges being smaller for larger valve lifts is that the main recirculation pattern is less
strong in this case, and is not able to push the intake jet back up as far.

The 'strength' of the main recirculation pattern has been established through an
evaluation of kinetic energy in a particular region of the main TRP, the so-called 'area of
scrutiny' which is a sub-region of the main recirculation pattern. Clear linear characteristics
were obtained in a single-logarithmic plot indicating the dependency of the kinetic energy
with engine speed and valve lift. According to these curves kinetic energy decreases with
decreasing engine speed and increasing valve lift. This same characteristic is experienced
for the momentum flux contained in the intake jet. Quantitative evaluation of momentum
flux could be achieved for three settings of the engine parameters when inferring velocity
information fi'om maps of velocity vectors. The change of momentum flux - exhibited by
the jet - with setting of the engine parameters turned out to be very similar to the change
of kinetic energy when varying the same engine parameters. This was seen to be evidence
for a correlation between the momentum flux carried by the intake jet and the kinetic

energy content of the main recirculation pattern.

173

Evaluations of recirculativity (see sections V2] and V.2.2 for definition and results)
and vortical energy in the area of scrutiny for the nine data sets witness inconsistencies in
the results at an engine speed of 20 RPM (corresponding to 340 RPM in air) when results
achieved at 10 and 30 RPM engine speed are used as comparison. For 10 and 30 RPM
engine speed curves for recirculativity exhibit fairly linear characteristics - recirculativity
increases with valve lift. For 20 RPM engine speed the three data points do not fall on a
straight line. For 10 and 30 RPM engine speed it was more or less possible to overlap
curves depicting vortical energy with the corresponding ones depicting the kinetic energy
by scaling. This was not possible in connection with the curves for 20 RPM engine speed.
The character of the flow field encountered during the induction stroke of an
axisymmetrical engine appears to exhibit a transition when at 20 RPM engine speed the
valve lift is being decreased. It was argued in section V.4 that this transition might be
attributed to the fact that at 20 RPM engine speed toroidal vortex rings created by the
intake jet get destroyed for 3 mm valve lift, however survive until the jet reaches the
cylinder wall in the case for 9 mm valve lift, enabling for 9 mm VL ingestion mechanisms
between vortex rings and the main toroidal recirculation pattern to supply the main TRP
with angular momentum on convective time scales.

One major finding fi'om this study is that there have been strong indications that the
ratio of fill; describing the location where the intake jet turns turbulent has considerable
significance in connection with the fluid dynamics of the flow field in an engine. If fill] is
small the intake jet turns turbulent shortly after it enters the engine cylinder. In this case
the intake jet exhibits strong turbulent motion when it supplies momentum to the main
recirculation pattern. For large «fill; the turbulent motion of the jet is weaker resulting in

more structured motion of the main TRP.

A feeling for the complexity of the flow field has been conveyed in section V.3.1 by

means of fields of instantaneous velocity vectors. The manifold structure of the flow field

174

makes it very dificult to grasp the flow field computationally or even analytically. It was
concluded that each fluid dynamical element of the flow field as identified in section V. 1,
e. g. the intake jet, should be considered separately with initial and boundary conditions
supplied by the surrounding fluid dynamical elements. The property of complexity is
stripped from the flow field when ensemble averaging of velocity vectors is performed
over individual engine cycles. In the EA domain very stable and structured flow fields are
encountered in an axisymmetrical engine after midstroke. Hence, EA techniques have to
be applied with the awareness that after ensemble averaging fluid dynamical quantities,
these quantities do not reflect properly fluid dynamics of the flow field during individual
events, but that these quantities point out general tendencies inherent in the fluid flow
during induction. Ensemble averaging will be the technique needed to quantify difl‘erences
in flow structure that can be attributed to engineering design changes.

Velocity fluctuations were evaluated in this work in the area of scrutiny (see
Figure V.2.2). These fluctuations incorporate those due to cycle-to-cycle variability and
due to turbulent motion exhibited by the flow field. Evaluations of these velocity
fluctuations for nine variations of engine speed and valve lift showed strong velocity
fluctuations in connection with high engine speed and small valve lift and low fluctuation
for large VL and low engine speed. Since the velocity fluctuations contain those due to
CtC-variations and due to turbulence no conclusive statement can be drawn fiom these
correlations. Information on instantaneous velocity magnitudes at one location in the flow
field in form of velocity profiles was used for 10 RPM engine speed and 3 mm valve lift to
attempt the separation of velocity fluctuations due to cycle-to—cycle variability and
turbulent motion within the flow field in this study. Velocity fluctuation in connection with
turbulence was found to be higher than those in connection with cyclic variations by a
factor ofbetween 1.2 and 1.5, in the area governed by the intake jet as well as in the area
of scrutiny. In their cycle resolved study of engine data Catania and Mittica (1989)

evaluated the velocity fluctuations due to turbulence to be, in the vicinity of the jet axis,

175

one quarter of those due to cyclic variations. Their experiments were conducted in a
commercial engine at 1600 RPM. This demonstrates the profound influence of engine
speed and engine geometry on velocity fluctuations. In this study it was found furthermore
that turbulence intensities increase towards the end of the induction stroke in the region
close to the axis of the intake jet as well as in the area of scrutiny. This is a very welcome
fact since it has been shown by previous studies that an increased level of turbulence

intensity enhances combustion performance in an IC engine.

Parametric quantities were evaluated fi'om vorticity distributions in the area of scrutiny
for the nine sets of data representing magnitude and frequency of hill/valley fluctuations
encountered in the distributions. These quantities were evaluated to provide measures of
the number of rotating scales, their energy and size. - The calculation of these quantities
replaces an evaluation scheme where Fourier transforms vorticity information into the
frequency domain since the amount of vorticity data available for the area of scrutiny was
not suficient for the application of the Fourier transform. - The measure representing the
frequency of hill-valley variation in the vorticity distribution shows that neither engine
speed nor valve lift has an influence on the hill/valley fluctuation indicating that the overall
spatial structure of the flow field is similar - for the length scale investigated - for all nine
flow fields stemming from the nine settings of the engine parameters. The magnitude of
the hill/valley fluctuations, on the other hand, is highest for 30 RPM engine speed and
3 mm VL and decreases with decreasing engine speed and increasing valve lift. The
statement delivered in the two preceding sentences bear an indication that size-wise
comparable flow structures possess increasingly higher rotational flow as valve lift

decreases and engine speed increases.

The spatial spacing of laser lines in the investigation grid for the examination of the
intake jet was very coarse with respect to the vortices created by the intake jet, so that
only a few fiuitful results could be obtained fiom its application to the flow field in the

176

vicinity of the axis of the annular intake jet. One of the results of this quantification was,
that the toroidal recirculation pattern in the upper corner between the cylinder wall and the
intake jet, observed by several investigators, and detailed firrther in this investigation, does
not for an axisymmetrical engine come into significant strength until midstroke. This was
also supported by observations from flow visualization results performed in this study. It
was furthermore argued for after midstroke that the upper comer flow pattern recirculates
only intermittently.

Another discovery stemming fiom the closer examination of the intake jet is that for the
crank-angle range of about 130° to 155° typical time scales of about 33 ms exist for the
fluctuation of the induction angle for 20 RPM engine speed and 3 mm valve lift. Similar
time scales can be found for convection mechanisms of azimuthal vortex rings shed by the
intake jet in flow visualization results for a crank-angle range of about 25° to 40° for the
same setting of engine parameters. This suggests a correlation between the generation and
convection of toroidal vortex rings created by the intake jet and the fluctuation of the

induction angle.

Following is an itemized account of the most important fluid dynamical results

stemming from this investigation.

- The flow field encountered in an axisymmetrical engine configuration can be dissected
into an annulus free jet, a boundary layer flow along the engine cylinder wall, two
toroidal recirculation patterns, and a bufi‘er zone on top of the piston head.

- The characteristic of the induction angle comprises a decrease towards a minimum,
followed by an increase towards a local maximum, with a final approach towards a
plateau value; the characteristic is independent of the setting of engine parameters
between 10 and 30 RPM engine speed.

- Results from the EA domain and from analyses of instantaneous data have both

importance in scientific research of fluid flow. From EA data it is possible to determine,

177

e. g., induction angle, centers of rotational flow, structure of flow field; fiom
instantaneous data one evaluates in-cycle mixing, velocity and vorticity magnitudes, etc.

- The amount of kinetic, vortical, and fluctuation energy contained in the main TRP
increases with increasing engine speed and decreasing valve lift.

- A new parameter, 'recirculativity', was defined. It gives information on fluid flow,
about its structure and its tendency to recirculate. It was found that the recirculativity
basically increases with increasing engine speed and decreasing valve lift.

- The length of transitional flow of the intake jet, fill], is an important parameter in the

evaluation of the degree of structure of the main recirculation pattern.

Throughout the entire text improvements have been suggested for an application of
LIPA to a fluid dynamics problem in general, and to the induction stroke of an engine in
particular. Space for improvement is certainly encountered with the optical
implementation of the investigation grid. By setting up the optical system on an optical
bench and using micron-adjustable enclosures for optical elements one can achieve, for
instance, thinner laser grid lines, higher spatial resolution of the grid, and equality of line
thickness throughout the area of investigation

A suggestion for further research in the study of the intake stroke of the water analog
model employing LIPA includes the investigation of horizontal planes at difl‘erent
locations of z to quantify swirl and the tornado visualized in chapter IV, quantification of
flow motion on smaller length scales by using a LIPA grid of laser lines with higher spatial
resolution, extension of engine speed to more practical speeds, and attempts to accomplish
an ultimate hybrid technology where LIPA results are obtained from images that also
depict flow visualization information on the flow field to attain immediate quantification of
visualized flow structures. The basis for the latter project was laid in this investigation
where LIPA images were retrieved on film showing superimposed both a distorted laser
grid and visualized flow structures of the intake jet in one picture (Figure II.2.6b-e).

178

The author pointed out at various locations within this text that the length of the
transitional flow experienced for the annulus intake jet is an important parameter for the
development of the flow field during induction of an engine. Experiments including LIPA
and flow visualization incorporating higher temporal - as well as spatial - resolution
should be conducted on the intake stroke of a reciprocating quartz model. Then, it will be
possible to determine the extent to which the location of vortex ring break-up
demonstrates significance for the development of the intake flow field. It will also be
possible then to generate a model for the interactions between the annulus intake jet and
the two toroidal recirculation patterns.

It is important to understand to what extent there are interactions between the
azimuthal vortex structures of the intake jet and the main toroidal recirculation pattern in
the interface zone, and of what nature they are. We have seen that for 1.5 RPM engine
speed and 9 mm VL the vortex structures get suppressed, they cannot really develop; on
the other hand, we have seen for the same engine speed and 6 mm VL, that there seems to
be a mechanism that resembles vortex coalescing between vortex rings and the main TRP.
A laser grid with denser spatial spacing of intersection areas has to be spanned around the
valve in order to obtain fields with velocity vectors that would yield coherent information
about the flow structure in the interface zone in terms of magnitude and orientation of the
flow field at multiple locations. One can use the knowledge about these fluid dynamical
processes to generate new engineering modifications, as well as to describe a model of, as
was suggested, an iterating control system of bilateral influence between the intake jet and
the main TRP and try to, thus, ultimately predict the internal fluid dynamics of the main
recirculation pattern The apprehension about the fluid dynamics of the main TRP during
induction is the key to evaluate mixing performance and level of turbulence intensity
during the induction stroke of an engine for most of the volume of the engine cylinder

under investigation.

APPENDICES

APPENDIX A
Generation of Piston Motion

Since water is the basis for the working fluid, the engine setup has to account for the
consequences of using an incompressible medium. Therefore, the compression and
expansion strokes had to be avoided during engine performance. This was accomplished
by a specially designed cam-shaft. For a depiction of the cam's physical shape see
Figure Alb. The engine-speed vs. crank-angle characteristic of the cam-shaft can be seen
in Figure Ala.

Since one engine cycle comprises for the particular cam-shaft used for this research
project one revolution of the cam (3 60° CA incorporating two moving and two quiescent
strokes) and one engine cycle of a regular engine comprises two revolutions of its shaft
(four moving strokes, Figure Ala) the effective engine speed used for the presentation of
results in this study must be doubled when transforming fluid dynamical results gained
with the utilized engine model to real engine applications. For example, if the model
engine is run at an engine speed of 10 RPM, the equivalent, real engine speed as
encountered in four-stroke IC engines is 20 RPM. In a phase-space diagram (Figure Ala)
the difi‘erence between the engine-speed-vs-crank-angle diagrams as induced by the cam-
shaft and by the shaft of a regular engine is being pointed out. For one turn of the 'circle' in
connection with the cam-shaft used in this study (L 11, III, IV; at 11 and IV the piston is at
rest) one has to make two turns in conjunction with a regular engine (1, 2, 3, 4).

Because the optical encoder that is electronically timing the data acquisition process
(section [1.2.2.2 & Appendix C) puts out 360 pulses per cycle and thus achieving a crank-

angle resolution of 1° in the used data acquisition setup, one can concede on the grounds

179

180

of the above discussion on engine speed that the temporal resolution in this study cannot

be better than 2° CA when results are to be converted to the real engine domain.

During engine performance one encounters fluctuation of the engine speed of the order
of about :I:0.5 Hz for 10 RPM engine speed at which the crank-angle frequency is 60 Hz,
thus an uncertainty in engine speed of 0.8%. This instability can be ascribed to causes such
as uneven fiiction between the O-ring and the cylinder wall, to an unstable providing of

the driving torque, to fiiction in the mechanical elements transferring the torque, etc.

Sealing and Lubrication between Piston Head and Cylinder Wall

An attempt had to be made to achieve two goals concerning the O-ring that served as
interface between the piston head and the cylinder wall which are both made of quartz: On
the one hand the O-ring had to establish an impenetrable connection with the quartz of the
cylinder wall in order to inhibit the working fluid from leaking, on the other hand the
connection was not to be too tight in order to allow the piston head to move
harmoniously. The use of a plastic named 'Rulon' accomplished cardinal sealing of the
working fluid yet allowed the piston head to move smoothly within the cylinder. The
actual O-ring constellation consists of two Rulon rings of equal width. Tightness of the
O-ring was enhanced by placing a regular rubber band between the quartz of the piston
head and the Rulon rings. The Rulon rings were cut on the side facing the piston head in
such a way as to allow the rubber band to rest in a groove, a configuration which
according to the O-ring text-book allows for increased sealing. Lubrication between
O-ring and cylinder wall was assured by the fact that a thin layer of liquid builds up
between the two materials during engine performance.

One unpleasant side-efi‘ect of the Decahydronaphthalene used for the flow visualization
experiments in connection with the phosphorescent dye (section IV. 1) is its property to
expand regular rubber dramatically (which I experienced on a daily basis when DHN came

181

into contact with my rubber gloves). Hence, when the flow visualization experiments were
performed using Kodak's photochromic chemical where DHN was the working fluid the
regular rubber band in the 'Rulon O-ring' configuration had to be replaced by DHN
resistant Viton V14 O-rings. These O-rings needed machining to substitute the rubber

band properly.

182

APPENDIX B
Issues on remctive Index Matchig

Since the point in 2D space for a refiactive index matched compound is dependent on
temperature, the temperature in the measurement site has to be kept constant which was
assured roughly by a constant temperature of the ambient air in the laboratory.

Also, Hexane as index matching agent for Decahydronaphthalene exhibits a
considerable evaporation rate when the DHN/Hexane solution is exposed to air so that the
outer tank in which the DHN/Hexane solution constitutes the index matched fluid had to
be sealed from ambient air [still about 50 ml Hexane evaporated during a day of
experiments which, however, is just a fi'action of the 6000 ml DHN/Hexane solution in the
outer tank]. From empirical observation it can be maintained that no fast difl‘usion
(mixing) mechanism between Hexane and the DHN/Hexane solution is encountered when
Hexane is added to the top of the fluid reservoir. This means in inverting terms, that after
evaporation of Hexane from the free surface the DHN/Hexane solution stays index
matched at the site where the laser grid enters the outer tank.

As could be observed, a slight mismatch of the index of refraction is not that crucial in
the investigation performed for this work. F urtherrnore, the relevant time in conjunction
with data acquisition with respect to index matching issues is the time between taking the
reference grids and the recording of the actual data (distorted grids) during engine
performance. This time is generally less than 5 minutes since only the first engine cycles
recorded were used for data reduction. During this time neither a significant change in
temperature nor significant evaporation of the alcohol component in the LIPA - chemical
in the engine cylinder and of Hexane in the DHN/Hexane solution in the outer tank can be
experienced.

183

A considerable mismatch of the index of refiaction for the test facility as a whole was
noticed when a white paper square was recorded with the imager system for calibration
purposes (Appendix D). In order to record the square the imager was turned away from
the test facility with the focus length of the camera lens remaining unchanged. In air, the
square (and also a ruler) was then moved along the optical axis of the camera lens until the
object was in focus. As turned out after the object was in focus, the distance between the
camera lens and the object (8.125 in) was significantly smaller than the distance between
the camera lens and the cylinder axis (8.875 in) of the test facility (grid of laser lines)
during the LIPA experiments. It is assumed that this is essentially due to the fact that that
portion of the outer tank that is situated between engine and camera is made of Plexiglas
where the optical system of indices of reflection quartz - Plexiglas - air has the efi‘ect of a
concave lens. It is assumed furthermore that the effect of hypothetically index mismatched

fluids in the apparatus and the outer tank is negligible.

Optical Issues

Background noise of the grid of laser lines: The edges of the glass bodies on the basis
of the beam divider introduce difl‘raction lines between the generated parallel laser lines.
Not only are these difi‘raction lines responsible for enhanced background noise (decrease
in quality factor“, equation II.2.d), they also tie up energy in higher order lines, thus taking
energy away fi'om the lines of 'zero' order which are used to perform LIPA There exists a
certain distance between near and far field where difli’action is not experienced. It is
desired to move the grating that distance away fi'om the site of investigation. However, for
the experiments where the imposed grid of laser lines was to cover most of a two-
dimensional plane of the engine cylinder the distance between the beam divider and the
region of interest was determined by issues concerning (1) the achievement of highest

spread between the laser lines, (2) the implementation of the paths of the laser lines in

184

order to avoid geometrical occlusion, and (3) the imperative to intersect the two sets of

laser lines as close to 90° as possible.

Variation of intensity within grid of laser lines: The grid of laser lines with which most
of the cylinder volume was examined exhibits a constant increase - in z - direction of the
engine reference fiame - of energy intensity inherent in a laser line from line to line within
both sets of parallel laser lines where the upper lines close to the chamber head possess
lowest intensity. This variation in energy contained in the laser lines occurs since the
500 mm focusing lens is placed before the grating of the beam divider with respect to the
direction of the optical path of the laser beam. Because of this constellation the part of the
focused laser beam that gets split by the grating first is less focused and contains thus less
energy than the part that hits the grating later which accounts for the lines of higher
intensity in the lower right comer of the engine (see Figure 11.2.7).

Variation of the intensity distribution of the laser lines is also due partially to a non-

homogeneous distribution of energy within the rectangular primary laser beam.

Depth of field: The parallel laser lines are focused at the measurement site. Thus the
depth of field at the measurement site is of the order of micrometers. However, since the
area of interest comprises, for the examination of the whole engine cylinder, a large

region, laser lines can be slightly out of focus at the periphery of the area of interest.

Out-of-plane motion: LIPA is insensitive to out-of-plane motion assuming proper
camera alignment since it records the projection of the distorted laser grid onto the
examined plane. A key aspect in this context is the proper generation of planar sets of
parallel lines.

185

APPENDIX C

Synchronization Aspects, its Difficulties, and related Uncertainties

Because LIPA - data was to be taken at particular fixed crank-angles within engine
cycles - in order to perform ensemble averaging over a number of ECs at these selected
CAs - the facility providing the measure for data acquisition is the running engine
(motor). Fluctuations encountered in engine speed during engine performance
(Appendix A) made it impossible to use the 30 Hz timing scheme provided by a CCD
camera to time the data acquisition process; with this approach pictures cannot be taken at
consistent piston positions for the examined engine cycles. The width of a window around
a crank-angle position - at which fluid dynamics ought to be examined - into which
samples of LIPA information fall during data acquisition denotes the accuracy of the
application of an ensemble averaging technique.

The electronic timing procedure is initiated by an optical encoder that rotates in sync
with the cam-shaft (Figure 11.2.4) and which puts out an electronic signal with every
degree rotation of the cam-shaft (3 60 signals per cycle) as well as a single signal once
every cycle (index). In general, the former signal is transferred to the Excirner Laser and
forces it to pulse. The signal is also piped to a frequency generator. This device provides
TTL signals of selectable length which can be delayed relative to the triggering input
signal fi'om the optical encoder where length and delay of the TTL signal constitute
exposure and delay time, respectively, of the data acquisition procedure (section 112.3).
The conditioned signal from the frequency generator is applied to the external trigger
facility of the image intensifier lens (Figure 11.2.4; more on the imager system in

Appendix D).

186

In the experimental setup employed in this study, an 'electronic signal valve' is inserted
into the electronic circuit right behind the optical encoder (Figure 11.2.4). This valve
enables the investigator to select from the signals provided by the encoder those signals
which are necessary for electronic synchronization in the data acquisition process at a
particular engine speed. A technical drawing of the electronic realization of the electronic
device is shown in Figure C. l. The device was designed and implemented by John
Brandon at the National Cyclotron Laboratory at Michigan State University upon the
author's specifications. The two main features of this electronic valve are that (1) it allows
the experimentalist to transfer only every 2nd, 3rd, 4th, ..., or 10th crank-angle pulse
output by the optical encoder to the electronic network behind the ‘valve' and that
(2) there is an option to assign a window with a length to cover 100 degree crank-angle
during which only pulses are admitted through the electronic valve and provided at its
output. - Using the index pulse provided by the optical encoder this window can be set;
e. g. for the experiments conducted for this work the window was moved such that LIPA
images could be recorded during the intake stroke. - The remaining 260 pulses in one
cycle are blocked in order to save on the lifetime of the elements comprising the
experimental system, especially to save on laser energy. The option to choose every 2nd,
3rd, and so forth signal has been added to ensure timing between the gating of the image
intensifier lens (IIL) and the Vrdeo camera at particular engine speeds upon which will be
deliberated next.

Electronic signal conditioning, i.e. timing to acquire data with the imager, is
implemented as follows: The Video camera is running continuously to record what is
provided at the exiting plane of the image intensifier lens (refer to Figures 11.2.4 and C2
for graphical support for the following presentation of the timing procedure). The IIL is
gated and admits information at its exit to be recorded by the camera only during assigned
periods of time which are initiated by signals from the optical encoder mentioned above

187

(or ultimately by the investigator). The time slots during which the shutter of the IIL
opens and data is recorded by the camera commence after the chosen 'delay time' between
the firing of the laser grid and the intentional documentation of the distorted grids. Their
lengths, i.e. exposure time, are determined by the intensity of the laser light of the
investigation grid and the f-stop of the camera lens (for optimization of exposure time see
section 11.2.3.2). The conditioning of the gating signal for the IIL is, as outlined above,
performed by the fiequency generator. Upon reception of a signal from the 'electronic
valve' the fi'equency generator outputs a TTL signal after the assigned time delay with a
width of the TTL signal corresponding to the chosen exposure time which is transferred to

the IIL.

Since the Canon camcorder could not be triggered externally synchronization problems
must occur. With the camcorder running continuously and the IIL being gated, opening of
the shutter of the IIL occurs one time during the camera's 33.33 ms sampling intervals (for
appr. 0.1 ms in the average; exposure time) to reveal information from the investigated
area. Since during these 33 .33 ms sampling intervals two sets of lines of the camcorder‘s
CCD array have to be filled (R8170), the IIL however opening the gate only once for a
fi'action of the sampling rate, only one set of lines in an image obtains information from the
scene of interest. That means that spatial resolution of images is literally cut in half in
vertical direction. The loss of spatial resolution could, however, be compensated for
partially (section III. 1.4). Another problem evolves from this method of data acquisition in
that the setting of the engine speed and thus the implementation of the measure for data
acquisition cannot be achieved precise enough to match the robust 30 Hz sampling rate of
the camcorder. Also, as mentioned in Appendix A, the engine speed fluctuates. Therefore,
it can be stated that due to the uncertainty in engine speed it can happen that either two
LIPA images can be interlaced in one picture acquired by the camcorder or that a flame

can be empty. To avoid this uncertainty it was attempted to run the engine at an engine

188

speed that transforms into a crank-angle repetition rate which is slightly lower than the
30 Hz sampling rate of the camcorder. The consequence is that empty frames are
encountered. \Vrth the data acquisition system at hand only virtual synchronization is
achieved.

Improvement of the temporal resolution by a factor of 2 can be achieved by designing a
'crank—angle clock' that assigns the appropriate CA to an image (field) containing a
distorted laser grid. A set of horizontal lines carrying distortion information of one
investigation grid can then be interlaced in one LEA picture with information of the
distorted grid fiom the succeeding crank-angle position.

Another uncertainty enters through a bi—directional ambiguity (shifting) of the temporal
position of the start of piston motion in the electronic domain (index pulse fiom encoder)
in a sequence of LEA images for one engine cycle. This fluctuation accounts for a
vacillation of less than :I:1° CA Its effect is minimized by conditionally sampling those
engine cycles where the crank-angle position representing the beginning of piston motion
in the sequence of LEA images for one EC can be assigned to a marginally small crank-

angle window that defines the commencement of the induction stroke.

Due to the aboveoutlined problems (empty or double-filled frames, shifting of initial
crank-angle position in sequences of LEA pictures) the single LEA images had to be
assigned to crank-angle positions by hand.

An engine speed of 10 RPM for the investigated water analog model (which transforms
into 20 RPM if one considers the cam-shaft factor of 2; Appendix A) corresponds to a
time interval between two degrees of crank-angle of 16.67 ms. Having a prescribed data
acquisition rate of 33 .33 ms by the Video camcorder, images are thus acquired at every
other degree CA Since 10 RPM for this water analog model relates to 20 RPM for a 'real'

engine (in this application a whole cycle consists of all four strokes whereas in real engine

189

applications one cycle consists of two strokes) the crank-angle resolution is cut in half
again Hence, for 10 RPM (20 RPM) there is a crank-angle resolution of 4°. In this text

statements of engine speeds always refer to the engine speed in conjunction with a 'real'

four-stroke engine.

The imager system is turned by 90° for the experiments in the investigation of the
whole cylinder - so that the quartz valve is located in the LEA images on the right side
(Figure 11.2.7), oriented vertically - in order to make most use of the 500 x 820 lines
ofl‘ered by the CCD array of the camcorder. Thus, it is in the x - direction of the engine
reference fiame that LEA information suffers fi'om the dividing in half of spatial
resolution due to the acquisition of only one set of horizontal lines by the Video

camcorder as was outlined earlier in this section. As it turns out this loss of resolution is

nainor in the aspect of accuracy.

190

APPENDIXQ

Specifications of Hardware

MGER SYSTEM

For the recording of LEA images an assembly consisting of two electro-optical
components was used; Figure II.2.4b and c. The first component is a commercial color
Video Camera & Recorder from Canon Inc., L1A Hi8 (high resolution, 8 mm). This
camcorder works on the basis of the interlacing (TV) standard RSl70, thus, sampling
'complete' images at a rate of 30 Hz. The R8170 standard defines that for a complete
image two sets (fields) of horizontal lines are filled with information from the scene of
interest at a frequency of 60 Hz (see Figure C.2), one after the other, and that these two
sets of lines are then interlaced, yielding a complete image (fi’ame) at a fiequency of
30 Hz. The camcorder's array of photodiodes consists of 500 horizontal and 820 vertical
lines. Since the facility that digitizes LEA images (MegaVision) can only digitize frames
of up to 512 by 512, data reduction could not take fully advantage of the potential of the
horizontal resolution. The Canon camcorder LIA offers many features such as variation of
exposure time and control of aperture which can be inferred fiom the Service Manual for
the LlA However, it is unfortunately not possible to externally trigger the camera, thus,
no locking of the camera into the electronic measure provided by the running engine, i.e.
the optical encoder, could be established.

Mounted in fiont of the camcorder is the second electro-optical element, a lens/image
intensifier component that was custom-made for the 'Turbulence Structure Laboratory' by
Electro-Optical Services, Inc. (EOSI). This image intensifier lens (IIL) - basically two
tandem image intensifiers which are equipped with P-46 high speed phosphors - yields a
photon magnification of 50,000 while ofi‘ering a resolution of 30 line pairs per millimeter.

191

The [IL can be gated where electronic shutter input ranges from 100 ns exposure to
continuous operation. Its receptive S-20 photocathodes exhibit a peak of 'sensitivity' for
electro-magnetic waves at a wavelength of 480 nm. The image intensifier system requires

a 12 Volt DC power supply.

MOE DIGIHZING FACILITY

The facility to digitize the data that is stored on Video 1118 tape is the '1024xm'
designed by MegaVrsion. The MegaVision (as the digitizer is referred to from now on)
ofi‘ers a resolution of 1024 x 1024, however, only 525- or 87 5-line images can be read in.
This means that when acquiring 500 x 820 format (x, y) images as is the capacity of the
employed data recording facility (LIA camcorder by Canon), MegaVision truncates as
well as compresses information in the horizontal plane. MegaVision discriminates
information into 256 grey scale levels. There are 12 flames of memory in which each four
512 x 512 images can be stored. It has been apperceived that images can move -
essentially in y - direction - within a quadrant of a flame during the digitization
procedure. This movement might be due to synchronization problems of the used VCR.
Using a calibration procedure for LEA images where a fixed spot of high pixel intensity
(reference point) is added into the images it is possible to adjust for this movement of the
image within the memory fi'ame. This fixed spot assures that the coordinates of the origin
of the engine assembly (Figure II.2.3a, x, z - reference frame depicted at the valve) are at
(0,0) in all LEA images which contain the undistorted and the distorted grids. The
reference point is implemented by a continuously emitting LED whose light is projected
onto the photodiode of the imager by a mirror. The combination of LED and mirror can
be seen in the picture of the setup of the data acquisition system contained in

Figure II.2.4b between the imager and the outer tank of the engine apparatus. The

192

reference point can be seen as a circle in Figure II.2.3a and as a white point in the upper

left corner in the pictures of Figure 11.2.7.

A software package provided by MegaVrsion comprises many image processing
algorithms. However, only data management routines were employed within the data
reduction process for this work since it was more convenient to incorporate image
processing software in the programs which were discussed in chapter [11. The MegaVision
system works within the flame of this investigation in conjunction with a Sun 3/260

Workstation.

In order to scale the distances - that fluid particles convect - as they are found in the
digitized images to the distances convected in real physical space one has to find out how
MegaVision maps the 500 x 820 images provided by the Vrdeo camera into its ovim
memory. It was found, as mentioned above, that during the digitization procedure
MegaVision truncates part of the image in the horizontal plane - the left side - and then
compresses the rest to 512 pixel. A test object - a ruler was recorded with the Vrdeo
camera and its mapping into memory by MegaVrsion was then assessed -- showed that
25 x 25 mm is represented by 400 x 438 pixel after digitization yielding a compression
factor of 1.095 in the horizontal direction.

To determine in LEA images the correct number of pixel that represents a distance in
the physical space, to thus find conversion factors for the two different investigation grids
applied in chapter V (grids for the investigation of the engine cylinder as a whole and for
the more scrutinizing examination of the intake jet), illuminated square sheets of paper
were recorded by the Canon camcorder onto tape and the images holding these squares
digitized. Now, the number of pixel for each side of the square was evaluated in the

digitized domain to determine the necessary conversion factors.

193

W

FORTRAN Pro am

Following is the listing of the FORTRAN program that implements the HIS approach

as delineated in section 111.12.
program WindowIADet

Description:

The algorithm seeks Intersection Areas using
a Window based approach A window is moved over the
input image and the pixel values lying inside this
window are added yielding an increasing number
when the window approaches an intersection area

Author: Bernd Stier, Jan-19-93 {(b&s)}
Modified versions: Aug-15-93 {(b&s)}
Sep-lS-93 {(b&s)}

Input : -parList.datI
-assign.datl
-imgXXXpic
-imgXXX.REF

Output : -imgXXX.LOC (into Nision/LEA/Engine/RPMmm/lmtion)
4&8!

Comment: This program is called by "Frame” and
shouldberunonlybythatshell.

 

OOOOOOOOOOOOOOOOOOOOOOOO

c Section 0 - Define fields, variables, and parameters
c
parameter (ima=1024,kma=1024)
common / array / LGma,kma),LA(jma,kma)
common / IntAreaAver I AintA(100,100)
dimension jE(100),kE(100)
character firame*70,fname2*70,fname3 *70,fname4*8,number*6
character DN(512,512)
logical exist
character header(512)
character Runs(300)*6
character answer‘3
integer flagl

194

write (‘,'(I//A/A/A)')
& ' DETECTION OF INTERSECTION AREAS USING A WINDOW‘,
& ' INTEGRATION APPROACH’,

 

& I I

write (*,'(/A/AIA)')

& 'Algorithm moves a 10x10 pixel window over',
& 'the intensity array of an image, determines‘,
& “the weight of the window at coordinate'

write (*,'(AlA/A)')

& 'points by integration and assigns local',

& ‘maximum weights to Intersection Areas.',

& 'Evaluate parameters for intensity slopesl 1'
write 021000)

c
c
c

cSectionI—Reeallthe coordinates ofthe undistorted

c lasergridandscrutinizeareaaroundthem
c

 

open (45,file=‘test')
open (46,file='assigndat1')
open (47 ,file=‘parListdatI')

c --Thissectionassignsthedatadirectory—-
read (463') mAnz
do 20 inz=l,mAnz
read (463) Win)
c write (‘3') Runs(inz)
20 continue
close (46)
if (mAanQJ) goto 27
open (46,file='assign.datl')
write (463) mAnz-l
do 25 inz=2,mAnz
write (46,*) Runs(inz)
25 continue

 

c -- TAUFE (assigns filenames)
27 flagl=0
read (47,'(110lA)') kAnLnumber
read (47,'(AIA)') fname4,answer

c BIG LOOP 30
do 30 krunch=l,kAnz
firame='/usr3/stierl'/IRuns(1)//'l'/Inumber/I‘.pic‘

= Msion/IEAIEnginef/Ifnamflll'l‘l/

 

 

195

&Rnns(l)//'/Location/‘//number//'.LOC'
c —- Find appropriate reference grid for 5 RPM --
if (fname4(l:2).EQ.'05') then
jindex=krunchl2+1
if (jindex.LT.10)
& fname2='/usr3/stier/img00'//CHARGindex+48)//'.REF
if (jindex.GE. 10 .AND. jindexLT.20) then
jindex=—jindex-10
fname2='/usr3/stierlimg0l'l/CHAROindexM8)/I'.REF'
endif
if (jindex.GE.20 .AND. jindex.LT.28) then
jindex=jindex-20
firame2='/usr3/stier/im302'l/CHAROindex+48)//‘.REF'
endif
if (jindexGE.28) fname2='/usr3lstier/img2845.REF'
goto 35
endif
c -- Find appropriate reference mid for 15 RPM --
if (fname4(l:2).EQ.'15') then
if (fname4(8:8).EQ.'I') goto 33
if (ICHAR(number(6:6)).GT.55 .AND. flag1.EQ.0 .AND.
& ICHAR(number(5:5)).GT.48) flagl=1
if (flagl.EQ. 1) then
firame2='/usr3/stier/img1830.REF'
else
fname2='/usr3/stier/'//numbcr//'.REF'
endif
goto 35
endif
c -- Find appropriate reference grid for 10 RPM -—
33 if (fname4(8:8).EQ.'I') then
if (ICHAR(number(6:6)).GT.50 .AND. flag1.EQ.0 .AND.
& 1CHAR(number(5:5)).GT.48) flagl=l
if (flagl.EQ. 1) then
fname2= lusr3/stier/img 1345.REF‘
else
firame2='/usr3/stierl‘/Inumber/I‘.REF
endif
goto 35
endif
if (ICHAR(number(6:6)).GT.55 .AND. flag1.EQ.0 .AND.
& ICHAR(number(5:5)).GT.49) flag1=1
if (flagl.EQ. 1) then
fname2='lusr3/stierlimg2845.REF
else
firame2='/usr3/stier/‘llnumbcr//'.REF
endif

 

c
35 write (453*) firame,fname2,fname3,ICHAR(number(5:5))
grits (‘.'(/A/A,A,A,A.A/A)')
& '-—— WORKING ON ',Runs(1),'l',mimber,'.pic
& I O
i=ICHAR(number(6:6))

 

 

 

196

if (i.EQ.57) then
number(6:6)='0'
number(5:5)=CHAR(ICHAR(number(5:5))+l)
else
number(6:6)=CHAR(ICHAR(numbcr(6:6))+l)
endif
INQUIRE (file=fname,exist=exist)
if (exist) goto 40
write (‘3‘) ' - FILE DOES NOT EXIST ll --’
goto 30

 

40 continue
open (2,file=fname2)
open (3,ACCESS='DIRECI",F1LE=fnarne,FORM='UNFORMA'I'I'ED’,
& RECL=512)
open (4,file='higSpotdatO')
open (8,file=fname3)

c ~Determinewhichlineismissing—
Lsum1=0
Lsum2=0
mode=1
read (3,rec=l) (headeraok),lok=l,512)
read (3,rec=250+l) (DN(i,250),i=1,512)
read (3,rec=251+l) (DN(i,251),i=l,512)
do 333 i=300,500
L(i,250)=1C1-1AR(DN(i,250))
if (L(i,250).LT.0) L(i,250)=L(i,250)+256
L(i,251)=ICHAR(DN(i,251))
if (L(i,251).LT.0) L(i,251)=L(i,251)+256
Lsuml=Lsum1+L(i,250)
Lsum2=Lsum2+L(i,251)
333 continue
if (Lsuml.GT.Lsum2) mode=2

write (‘3') ' -- READ IN PIXEL INTENSITIES --'
do 80 k=1,256
if (mode.EQ. 1) then
read (3,rec=2‘k-l+l) (DN(j,2*k-l),j=1,512)
elseif (mode.EQ.2) then
read (3,rec=2"'k+1) (DN(i,2*k),j=l,512)
endif
80 continue

197

 

c
0 Preparing Image for Analysis
e

 

write (*,"') ' — FILLING LINES -—-'
do 90 k=l,256
do 90 j=1,512
if (mode.EQ. 1) then
L(j,2*k-l)=ICHAR(DN(i,2*k-l))
if (L(j,2*k-l).LT.0) L(i,2*k-l)=L(j,2*k-1)+256
L(i,2‘k)=L(j,2*k-l)
elseif (mode.EQ.2) then
L(i,2*k)=ICHAR(DN(j,2*k))
if (l.(i,2*k).LT.0) L0,2*k)=L(j,2*k)+256
L0.2‘k-l)=L(i.2‘k)
endif
90 continue

write (*,"') ' -- PERFORMING AVERAGING --'
do 350 j=2,511
do 350 i=2,511

LA(i.i)=(L(i-lJ-IHMiJ-IHLGHJ-IH

& Mi-IJ)+L(iJ)+L(i+l.i)+

& L(i-l,j+1)-t-L(i,j+1)-1~L(i+l,j+1))/9
DN(iJ)=CHAR(LA(iJ))

350 continue

write (*,*) ' —- WRITING -—'

OPEN (unit=l 1, file='caXXXmapic', form='UNFORMAT'TED',

+ access='DIRECT‘, recl=512)

write (1 1,rec=1) (headerflok),lok=l,5 12)

do 400 j=l,512

write (1 l,rec=j+1) (DN(i,j),i=l,512)

400 continue
c
c Preparation finished I
c

 

 

read (2,*) [Am

write (‘,’(A,12,A)')

& ' -- FINDING HIGHEST INTENSITY SPOTS (**',IAanz,'") --'
if (answer.EQ.‘yes’ .OR. answer.EQ.‘y')

& write (*,"') ' -- by DOUBLING RESOLUTION —'

do 500 i=l,IAanz
c —anndkEaretheooordinatesforEsinreferencegrids-
read (2."‘) J'IP(i).klP(i)
if (jE(i).EQ.0) then
PIntj=0
PIntk=0
jIntP=0

198

kIntP=0
goto 580
endif
DOUBLE RESOLUTION (DYNAMIC RANGE) OF IA ----
if (answer.NE.‘yes' .AND. answer.NE.'y') goto 530
write (’.'(A$)') 'X'
krux=0
do 520 kIntP=kE(i)-15,kE(i)+15
jn1x=0
krux=krux+1
do 520 jIntP=jE(i)-15,jE(i)+15
jrux=jrux+l
AintA(2*jrux-l,2*krux-l)=(4*LA(jIntP,klntP)-I-2*LA(ilntP-l,kIntP)
& +2*LA(jIntP,kIntP-l)+LA(jIntP-l,kIntP-l))/9.0
AintA(2'jrux,2"krux-l)=(4 ‘LAOIntP,kIntP)+2*LA(iIntP+l,kIntP)+
& 2*LA(iIntP,kIntP-1)+LA(jIntP+1,kIntP-1))/9.0
AintA(2"'jrux-l,2*kruxH4‘LA(iIntP,klntP)+2*LA(iIntP-l,kIntP)+
& 2*LA(jIntP,kIntP+l)+LA(jIntP-l,kIntP+l))/9.0
AintA(2*jnrx,2*krux)=(4"LA(iIntP,klntP)+2*LA(iIntP+1,klntP)+
& 2’LA(jIntP,kIntP+l)+LA(jIntP+l,kIntP+l))/9.0

C

 

520 continue
c
530 Wmax=0
if (answer.NE.‘yes' .AND. answer.NE.‘y') then
jmi=jE(i)-13
jmx=jE(i)+7
kmi=kE(i)-13
kmx=kE(i)+7
else
jmi=8
jmx=48
krni==8
krnx=48
endif
do 550 k=krni,kmx
do 550 j=jmi,jmx
CALL Intemate(j,k,W,answer)
if (W. GT.Wmax) then
Wmax=W
jhelp=j
khelp=k
endif
550 continue
if (jhelp.EQ.jmi .OR jhelp.EQ.jmx .OR
khelp.EQ.kmi .OR khelp.EQ.kmx) then
write (45,*)
& i,’ Out of range! Intensity: ',LA(jhelp+3,khelp+3)
PIntj=0
PIntk=0
jIntP=0
kIntP=0
goto 580
endif
W=Wmax

199

c -- TRANSFORM BACK TO ORIGINAL COORDINATE SYSTEM --
if (answer.EQ.'yes' .OR. answer.EQ.‘y') then
PIntj=jE(i)-l5+II~1T(((jhelp+3)-l)/2)
if (MODtjhelp+3,2).GT.0) PIntj=PIntj-0.25
if (MODGhelp+3,2).EQ.0) PIntj=PIntj+0.25
PIntk=kE(i)-l5+INT(((khelp+3)-l)/2)
if (MOD(khelp+3,2).GT.0) P1ntk=PIntk-0.25
if (MOD(khelp+3,2).EQ.0) PIntk=PIntk+0.25
goto 580
endif
c write (*,'(A,12,A,I6,110/8X,A,15,I10)') 'IA# ',i,': initial',
0 & jE(i),kE(i),'accurate',jhelp+3,khelp+3
jIntP=jhelp+3
kIntP=khelp+3
write (4,'(/A)') 'Area of highest intensity'
do 560 k=khelp,khelp+6
wn'te (4.'(1214)') (L0.k)J=J'h¢1PJhelp+6)
560 continue
580 continue
if (answer.EQ.'yes' .OR answer.EQ.'y') then
write (8,'(2(F10.3))') PIntj,PIntk
else
write (8,'(2(15))') jIntP,kIntP
endif
500 continue

do 700 klax=480,500
if (LA(450,klax).LE.60) then
write (8,'(A,110,112)') ‘ 450',klax,LA(450,klax)
write (45,'(A,110,112)') ' 450',klax,LA(450,klax)
goto 710
endif
700 continue
710 continue
close (2)
close (8)

30 continue
c BIG LOOP 30

 

 

1000 format (ll,
& ' ')

 

end

200

 

00000

Appendix AI: Integration

subroutine Intematefimi,kmi,W,answer)

Description:
Subroutine to sum intensities within window

Author: Bernd Stier, Jan-l8-93 {(b&s)}

 

0000000

parameter (jma=1024,kma=1024)

common / array / L(jma,krna),LA(jma,kma)
common / IntAreaAver / AintA(100,100)
character answer‘3

W=0
n=-4
m=-4
do 540 k=kmi,kmi+6
n=n+1
m=-4
do 540 j=jmi,jmi+6
m=m+1
if (n.EQ.-3 .OR n.EQ.3 .OR m.EQ.-3 .OR m.EQ.3) then
alpha=1
elseif (nEQ.-2 .OR nEQ.2 .OR m.EQ.-2 .OR mEQ.2) then
alpha=4
elseif (n.EQ-l .OR nEQJ .OR. m.EQ.-1 .OR mEQJ) then
alpha=l4
else
alpha=l7
endif
if (answer.NE.'yes’ .AND. answer.NE.'y') then
=W+alpha‘LA(i,k)
else
W=W+alpha'%intA(j,k)
endif

540 continue

C

return
and

BIBLIOGRAPHY

BIBLIOGRAPHY

A‘r‘ta, S., Tabbal, A, Munck, G., Fujiwara, K., Hongoh, H., Tamura, E., and Obana, S.
1990 Numerical simulation of port-valve-cylinder flow in reciprocating engines,"
SAE Paper No. 900820.

Am, 8., Tabbal, A, Munck, G., Montrnayeur, N., Takenaka, Y., Aoyagi, Y., and
Obana, S. 1991 Numerical simulation of swirling port-valve-cylinder flow in Diesel
engines, SAE Paper No. 910263.

Arcoumanis, C., Bicen, AF., Vlachos, NS, and Whitelaw, 1H 1982a Efl‘ects of flow
and geometry boundary conditions on fluid motion in a motored IC model engine,
Proc. Instn Mech. Engrs, 196, pp 1-10.

Arcoumanis, C., Bicen, AF ., and Whitelaw, J.H. 1982b Measurements in a motored four-
stroke reciprocating model engine, .1. Fluids Engng, 104, pp 235-241.

Arcoumanis, C., Green, H.G., and Wong, KY. 1987 Flow structure visualisation in
reciprocating engines, Flow Vrsualization IV Proceedings of the Fourth
(International) Symposium on Flow Vrsualization, Paris, France, August 26-29,
1986, 705-711, Hemisphere Publ. Co., Washington, DC.

 

Arcoumanis, C. and Whitelaw, J.H. 1987 Fluid mechanics of internal combustion engines
- a review, Proc. Instn Mech Engrs, 201, pp 57-74.

Ballard, DH. and Brown, CM. 1982 Computer Visiop, Prentice - Hall, Englewood
Clifi‘s, NJ.

Bicen, AF., Vafidis, C., and eritelaw, J.H. 1985 Steady and unsteady airflow through
the intake valve of a reciprocating engine, J. Fluids Engng, 107, pp 413-420.

Buchner, S. 1992 Flow field and mixture separation measurements in a dense solid-liquid
two-phase vertical pipe flow, Diplomarbeit, Dept. Mech. Engrg, Michigan State
University.

Catania, AE. and Mittica, A 1985 Cycle-by-cycle, correlation, and spectral analysis of

LC. engine turbulence, Flpws in 1mm Cpmbusp’on Engrp' es - m, FED-Vol. 28,
T. Uzkan, W.G. Tiederman, and J.M. Novak, eds., ASME, New York.

201

202

Catania, AE. and Mittica, A 1989 Extraction techniques and analysis of turbulence
quantities fiom in-cylinder velocity data, Basic Processes in Internal Combustion
Engines, ASME, Proceedings of the Twelfth Annual Energy-Sources Technology
Conference, Houston, Texas, ICE-Vol. 6.

Champamre, RH. and Wymransld, 1.]. 1971 An experimental investigation of coaxial turbulent
jets, Int. J. Heat Mass Transfer, 14, pp 1445-1464.

Chan, WT and Ko N.W.M. 1978 Coherent structures in the outer mixing region of annular
jets, J. FIuidMech, 89, pp 515-533.

Chu, CC. 1987 A study of turbulence production and modification in boundary layers using a
new photochromic visualization technique, PhD Thesis, Dept. Mech. Engrg, Michigan
State University.

Chu, C.C., Hsieh, C.S., Shi, ST, and Wang, CT. 1991 A quantitative study of the flow near
surfaces induced by vortex motion, AIAA 91-1816.

Dahm, W.J.A, Frieler, CE, and Tryggvason, G. 1992 Vortex structure and dynamics in the
near field of a coaxial jet, J. Fluid Mech., 241, pp 371-402.

Daneshyar, H. and Fuller, DE. 1986 Definition and measurement of turbulence parameters in
reciprocating I.C. engines, SAE Paper No. 861529.

Eaton, AR and Reynolds, WC. 1989 Flow structure and mixing in a motored
axisymmetric engine, SAE Paper No. 890321.

Ekchian, A and Hoult, DP. 1979 Flow visualization study of the intake process of an
internal combustion engine, SAE Paper No. 790095.

Falco, RE and Chu, CC. 1987 Measurements of two-dimensional fluid dynamic
quantities using a photochromic grid tracing technique, Proceedings of SPIE, 814,
pp 706-710, San Diego.

Falco, RE. and Chu, CC. 1988 Experimental determination of skin fiiction modifications
due to elastic compliant surfaces using quantitative visual techniques, Report
ISL-8&3, Dept. Mech. Engrg, Michigan State University.

Falco, RE., Chu, C.C., Hetherington, MH., and Gendrich, GP. 1988 The circulation of
an airfoil starting vortex obtained fi'om instantaneous vorticity measurements over an
area, AIAA 88-3620.

Falco, RE, Gendrich, CR, and Chu, CC. 1989 Vorticity field measurements using Laser
Induced Photochemical Anemometry (LEA), Seventh Symposium on Turbulent
Shear Flows, Stanford University, August 21-23.

203

Falco, RE. and Nocera, D.G. 1991 A study of flow properties of wet solids using Laser
Induced Photochemical Anemometry (LEA), Liguid-Solid-Flows-I991, M.C. Roco
and T. Masuyama, ASME Fed-Vol. 118, pp 143-145.

F alco, RE. and Nocera, D.G. 1993 Quantitative multipoint measurements and
visualization of dense solid-liquid flows using Laser Induced Photochemical
Anemometry (LEA), P1articulpte Two-Phase Flow Chapter 3, MC. Roco,
Butterworth-Heinemann.

 

Foster, DE. and \Vrtze, PO. 1988 Two-component laser velocimetry measurements in a
spark ignition engine, Combust. Sci. and Tech, 59, pp 85-105.

Fox, RW. and McDonald, AT. 1973 Introduction tp Fluid Mechanics, John Wiley &
Sons, New York, Santa Barbara, Chichester, Brisbane, Toronto.

Gieselrnann, M. 1991 Implementation of and measurement with the LEA - technique in a
subsonic jet, Diplomarbeit, Dept. Mech. Engrg, Michigan State University.

Gosman, AD., Tsui, Y.Y., and Watkins, AP. 1984 Calculation of three dimensional air
motion in model engines, SAE Paper No. 840229.

Hadded, O. and Denbratt, I. 1991 Turbulence characteristics of tumbling air motion in
four-valve SI engines and their correlation with combustion parameters, SAE Paper
No. 910478.

Henriot, 8., Le Coz IF, and Pinchon, P. 1989 Three dimensional modelling of flow and
turbulence in a four-valve spark inmrition engine - comparison with LDV
measurements, SAE Paper No. 890843.

Hilbert, HS. 1991 Measurements of flows during scavenging in a two-stroke engine,
MS. Thesis Dept. Mech. Engrg, Michigan State University.

 

Hilbert, HS. and Falco, RE. 1991 Measurements of flows during scavenging in a two-
stroke engine, SAE Paper No. 910671.

Hilton, AD.M., Roberts, J.B., and Hadded, O. 1991 Autocorrelation based analysis of
ensemble averaged LDA engine data for bias-fiee turbulence estimates: a unified
approach, SAE Paper No. 910479.

Jain, AK. and Dubes, RC. 1988 Algorithms for Clustering Data, Prentice - Hall,
Englewood Cliffs, NJ.

Kent, J.C., Mikulec, A, Rimai, L., Adamczylg AA, Mueller, S.R, Stein, RA, and
Warren, CC. 1989 Observations on the efl‘ects of intake-generated swirl and tumble
on combustion duration, SAE Paper No. 892096.

204

Ko, N.W.M. and Chan, W.T. 1978 Similarity in the initial region of annular jets: three
configurations, J. Fluid Mech, 84, pp 641-656.

Lancaster, DR 1976 Efl‘ects of engine variables on turbulence in a spark-ignition engine,
SAE Paper No. 760159.

Lee, K., Yoo, S.C., Stuecken, T., McCarrick, D., Schock, H., Hamady, F.,
LaPointe, L.A, Keller, P., and Hartman, P. 1993 An experimental study of
in-cylinder air flow in a 3.5L four-valve SI engine by high speed flow visualization
and two-component LDV measurement, SAE Paper No. 930478.

Lempert, WR, Magee, K., Ronney, P., Nosenchuck, D.M., Miles, RB., Gee, K., and
Haugland, RP. 1993 Flow tagging in water by Photo-Activated Nonintrusive
Tracking Of Molecular Motion (PHANTOMM), AIAA 93-0517.

Liou, T.-M. and Santavicca, DA 1985 Cylce resolved LDV measurements in a motored
IC engine, J. Fluids Engng, 107, pp 232-240.

Miller, DR and Comings, E.W. 1957 Static pressure distribution in the fi'ee turbulent jet,
.1. FluidsMech, 3, pp 1-16.

Mohr, B. and Ertel, H. 1986 Brechungsindexanpassung fiir Geschwindigkeitsmessungen
mittels Laser-Doppler-Anemometrie, Internal Stuay by the 'Lehrstuhl fi'rr
Stromungsmechanik', Universitat Erlangen-Nr‘irnberg, Erlangen.

Moriyoshi, Y., Karnimoto, T., Ohtani, H., and Yagita, M. 1991 Stratification of swirl
intensity in the axial direction for control of turbulence generation during the
compression stroke, SAE Paper No. 910261.

Narnazian, M., Hansen, 8., Lyford-Pike, E., Sanchez-Barsse, J., Heywood, J., and Rife, J.
1980 Schlieren visualization of the flow and density fields in the cylinder of a spark-
imrition engine, SAE Paper No. 800044.

Omori, S., Iwachido, K., Motomochi, M., and Hirako, O. 1991 Efi‘ect of intake port flow
pattern on the in-cylinder tumbling air flow in multi-valve SI engines, SAE Paper
No. 910477.

Ponce, A, Wong, P.A, Way, J .J ., and Nocera, D.G. 1993 Intense phosphorescence
triggered by alcohols upon formation of a cyclodextrin ternary complex,
J. of Physical Chemistry, 97, pp 11137-11142.

Popovich, AT. and Hummel, RL. 1967 Light-induced disturbances in photochromic flow
visualization, Chem. Engrg Sci. 29, pp 308-312.

Potter, MC. and Foss, JP. 1982 Fluid Mechanics Great Lakes Press, Inc., Okemos.

 

205

Rask, RB. 1979 Laser Dappler anemometer measurements in an internal combustion
engine, SAE Paper No. 790094.

Rask, RB. 1981 Comparison of window, smoothed-ensemble, and cycle-by-cycle data
reduction techniques for laser Doppler anemometer measurements of in-cylinder
velocity, Fluid Mechanics of Combustion Systems, T. Morel, RP. Lohmann, and
J .M. Rackley, eds., ASME, New York.

Regan, C.A, Chun, KS, and Schock, H.J. 1987 New developments and applications in
gas lasers, Proceedings of SPIE, 737, pp 17-27, Los Angeles.

Rockwell, DO. and Niccolls W.O. 1972 Natural breakdown of planar jets, J. of Basic
Engng, 72-FE-5, pp 720-730.

Semenov, ES. 1963 Studies of turbulent gas flow in piston engines, Tech. Transl. F-97
NASA, pp 122-147.

Shack, DH. and Reynolds, WC 1991 Application of particle tracking velocimetry to the
cyclic variability of the pre-combustion flow field in a motored axisymmetric engine,
SAE Paper No. 910475.

Stier, B. and Falco, RE. 1993 New developments in LEA (Laser Induced Photochemical
Anemometry), Proceedings of SPIE, San Diego, July.

Stier, B. and Falco, RE. 1994 Application of LEA (Laser Induced Photochemical
Anemometry) to a water analog model of a four-stroke IC engine, SAE Paper
No. 940282.

Tennekes, H. and Lumley IL. 1972 A Firg Course in Tu_r’pulen<_>e, The MIT Press,
Cambridge, Massachusetts, and London, England.

Trigui, N., Afi‘es, H., and Kent, J.C. 1994a Use of experimental data as initial conditions
for in-cylinder CFD simulations, SAE Paper No. 940280.

Trigui, N., Kent, J.C., Guezennec, Y., and Choi, W.C. 1994b Characterization of intake
generated fluid flow fields in IC engines using 3-D Particle Tracking Velocimetry,
SAE Paper No. 940279.

Vafidis, C. and Whitelaw, J.H. 1986 Intake valve and in-cylinder flow development in a
reciprocating model engine, Proc. Instn Mech Engrs, 200, pp 143-152.

White, F .M. 1991 Viscous Fluid Flow McGraw-Hill, New York.

 

Winsor, RE. and Patterson, DJ. 1973 Mixture turbulence -- a key to cyclic combustion
variation, SAE Paper No. 730086.

206

Witze, PO. 1977 Measurements of the spatial distribution and engine speed dependence
of turbulent air motion in an IC engine, SAE Paper No. 770220.

Witze, P.O., Martin, J.K., and Borgnakke, C. 1984 Conditionally-sarnpled velocity and
turbulence measurements in a spark imrition engine, Combustion Science and
Technology, 36, pp 301-317.

Yoo, SC. 1994 An in-cylinder air flow study in a four-valve SI engine by three-
dirnensional LDV measurements and high speed flow visualization, PhD Thesis,
Dept. Mech. Engrg, Michigan State University.

Yule, AJ. 1978 Large-scale structure in the mixing layer of a round jet, J. Fluid Mech., 89,
pp 413-432.

Yurechko, V.N. and Ryazantsev, Y.S. 1991 Fluid motion investigation by the photochromic
flow visualization technique, Expt'l. Thermal and Fluid Sci. , pp 273-288.

FIGURES

207

'Y

 

 

 

 

 

 

 

 

 

 

(C)
Vbrticity
Evaluation

 

 

 

 

 

Figure II. 1.1: Evaluation of velocity and vorticity vectors from convection of
investigation mid with the fluid flow.

Mean Pixel Intensity

Mean Pixel Intensity

208

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

120 a\\
100- N
80*-
Backgroundlntensity
T"rl""l""l"TT—I""l
0 10 15 20 25
DelayTime[inms]

Figure 11.1.2: Degradation of intensity of radiation emitted from

investigation mid with time; influence of bleaching
and shelf-life.

 

110

 
   
 

+ CLVI
“‘9— CL VII
-9- CL VIII

  
   

 

  

—B—CL|X
+CLX

  
 

 

 

 

 

80

 

Background Intensity

 

 

 

r
i

Time Delay [in ms]

Figure II. 1.3: Results of attempt for refinement of used LEA -

chemical by adding supplements of the alcohol
component t-BuOH

209

 

Table 11.2.1: Engine geometry

 

 

 

 

lne Dimension
Parameter

(3) Bare 82 mm
Stroke 82.4mm
Clearance 18 mm
Valve-B 52 mm

Valve-Elbow 0.59

Bore/5mm 0.995

 

 

 

 

 

Figure 11.2.1: (a) Schematic of the water analog model showing the essential quartz
pieces [included in (a) are the measures of the engine geometry].
(b) The quartz cylinder, head, piston, and valve along with diverse other
components of the engine configuration. Also shown are the four metal
'riders' for the four sides of the outer tank - used for the calibration of
axisymmetry -, that hold the fluid reservoir at the right distances from
the outer tank after the collar established axisymmetry.

210

    

Figure 11.2.2: Apparatus used to create the mid of laser lines in the quartz engine.
(a) Schematic showing the paths of light and grid location.
(b) Photograph of the setup, with the image acquisition system in the
backm'ound.

211

Investigation of Fluid Dynamics
(a) in whole Engine Cylinder

 

(b) ofIntake Jet

 

 

b
>
>
to
s‘
,5

 

f

-705;I'xdinint¢'so¢im-u
due-atom [Grizzlies-dim

 

Location #8
(«any firm In Mung

    

330pixalin
Murals

Location #38
Figure 11.2.3: Grids of laser lines for investigation of (a) most of the cylinder volume

and (b) intake jet. For both mids a typical intersection area is shown
magnified and scrutinized.

 

 

 

 

 

 

 

 

 

 

Figure 1124: (a) Schematic of the timing system; (b) Photo of the image intensified CCD
camera in data taking position; (c) Schematic of the overall data acquisition
system.

213

 

Figure H.2.5: Black and white images holding LEA grids generated by the optical system.
(a) The reference mid (light enters the cylinder wall on the right, and can be
seen coming out of the other side of the quartz valve -- near the center of the
picture).

(b - 6) Four independent photos showing distortions in the laser mid larger
than useful for data acquisition (long delay time). Engine speed is 20 RPM.

214

 

(0)

Figure 1126: Color images of LEA grids showing a sequence of four data photos and the
calibration grid used with them from preliminary tests. (a) Calibration grid.
(b - 6) Four successive data frames; these frames also reveal the flow pattern
of intake vortices because the LEA - chemical irradiated in the previous laser
pulses is still glowing after it has been warped around the vortices (20 RPM).

(6)

215

(a)

(b)

(c)

 

Figure H.2.7: Examples of the undistorted (a) and distorted (b, c) LEA grids acquired
with the IIL/camcorder combination. To optimize the CCD format, the

camcorder was turned 90° so that the valve is on the right side of the
image, and the cylinder wall on the bottom.

216

 

 

 

 

 

 

Figure I111: Comparison between results of inference of coordinates of illuminated
'particles' fi'om LEA images as deduced by hand (cross, square,
and diamond) and by machine (dotted lines).

          
    

,:-:- 9,53%?

 

dcpcuns"Bdses

 

Figure 111.1.1: Nomenclature for
intersection area

 

217

36.6% Breton.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

in an IA according to which 'HIS' chooses the highest intensity spot.

Figure 1111.2: Weight distribution used for the evaluation of intensity distribution

 

 

- _ q q d

. . u o

o e o o

n u ms ”

u u .I “
IsOs'IIIOOIOssssssss” ssssssssssssssss .0 ssssssssss “003.0! sssssssssssss ”0‘

. . o .

u u I ”M. a

n n V. n n

u n .D "d u

. o. u e .

. o. d o t o

u ..u e "a u

o I. .. n o u n

o .I. o "1 u
I .................... A .......... “4.1.3. ........... m .3..“ ............. «.1

n u e . e .

. I . t u u

u -3. «a. ”u u

u .- r...” P . u

. .o: .... I .
‘lltallaIlls.slslslasahIDaa.-IIslstlIsOI" IIIIIIIIIIIIIIIIIIIIIIIIIIIII ”ll

. H m ...m. r m
I...‘....I..s..'.~..l...h...‘.'l.'...'..l-‘ ............. J. ................. ...l
i .................... .H ............... a ............... .H ................ mi

N . u H
Issac-IllslssalsssssssmlIIsau!.ssusssvs‘aslsssssslsasso‘“ IIIIIIIIIIIIIIII ”CIA

p b b - _

  

 

 

applications of 'IAR' and 'HIS'

Figure 111.1.3: Comparison between the results of
, to four intersection areas.

218

 

 

, . I
. MM 'M ii“ ‘MIVMMMMI I,"
M ‘1 MM H ’MIIIM
LII,y MM‘ /...

   

 

(a)

 

 

 

Figure 111. 1.:4 Depiction of an intersection area in
terms of pixel intensity; (a) shows the pixel
intensity distribution for an IA as it is delivered
at the output of the CCD camera, (b) exhibits the
same distribution after the image was conditioned.

 

 

219

 

mama?“ 05885 «e $280532 HQE Semi

 

“FWWVMN—‘

 

"3.295

ofiwcm eoaawumoiz
=a Lo>o ofifiwififi
can 5:33 28 an
owab>< oSEomem

RMS Fluctuation [in pixel]

220

RMS Fluctuation [in pixel]

 

1024x1024 5121512 256x512 256x512
pixel image

Figure 1113.]: Effect of change in number of pixel per image and extent of
object in image in x - and z - drrection.

 

 

 

 

 

 

/

 

 

 

 

 

 

 

 

 

 

1.5.
,/

0.5/ .—

o . .
30RPM,3mmVL 30RPM,6rrmVL 30RPM,9n‘mVL

(a)

 

 

 

 

 

 

 

 

 

(b)

 

O
20RPM,3mnVL 20RPM,9rm1VL 30RPM,3rmIVL

Figure III.3.2: Velocity fluctuations exhibited by reference laser 'ds employed
1n investigation of (a) whole enmne cylinder and ) the intake
jet as close up.

 

221

 

 

 

 

no -
U7

 

Uncertainty in

 

 

 

 

Uncertainty in P/xe/

 

 

 

 

Figure III.3.3: Uncertainty about determination of initial positions in reference mids
for 20 RPM engine speed and 3 mm valve lift as encountered for the
close-up investigation mid;
(a) in z - direction, (b) in x - direction.

222

 

 

 

 

(h) 124.10

 

(1) 155.2
Figure IV.1.1: Flow visualization of the intake stroke using photochromic dye (color
change upon irradiation). The piston head is visible near the bottom of
each photo, and the cylinder head and intake valve are visible near the
top. It is interesting to note a spiral up-flow of fluid that has reached
the piston head exists throughout the sequence.

. ...ne e ..

‘Van. , , “M.,...

      

(a)
12'

(b)
16'

(C)
20"

((0
28°

(e)
32°

 

223

”""“‘“~!

Figure IV.2.1: Flow visualization of the intake stroke using a fluorescent dye in the

reservoir above the valve (MODE I). A laser sheet of 308 nm light is slicing
the cylinder both above and below the valve (the impurities in the quartz of
the prston head make it clearly visible). Results are at 20 RPM and 3 mm
valve lift. The numbers are in demees crank-angle. By 40° CA the cylinder
is filled with dye and the experiment is stopped. Some air bubbles have
formed from air in the water and partly obscure the view.

(a),
8

(C)
16'

(d)
20'

(e)
24'

 

224

 

‘rrw—n-

225

 

Figure IV.2.2: Flow visualization of the induction stroke, where the inducted fluid rs dye-
free and the fluid 1n the cylinder has been dyed (MODE II). Engine speed
is 20 RPM and valve lift 1s 3 mm. The LEA results for CA3 32, 52°, and
92° (see section V.5. l) are superimposed on the visualizations.

226

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

—o

Ira/s
Velocity Vector

(WMas‘uu-abmmudwl

227

Depiction of error.
area of ..
ecrutlny ‘ x

 

Reference

 

z Velocity Vector

Figure V. l . 1: Maps of ensemble averaged velocity vectors (over 42 ECS)
at selected crank-angle positions for 20 RPM engine speed
and 3 mm valve lift illustrating the development of intake flow.

228

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I I f r v v V r—— I V ‘
1° RP":’"9""‘ 'M' 3 ‘ “'2 ' 1° "Hicnqme speed. 5 U VL; 10 ”meagre. speed. 9 an V3.
‘1?“ "’ “"‘Ed‘l‘y’ E ‘ ”-17 as name-10v) ; 11:17 as tramdolsyl
- J I ‘ : . 1 .
r s ' ' I k e I t
.................. >., \11 "
. ‘I , V ’ :8 # .............................. ?.y...........u .......
' «s 5' r: (E "' : - w
r J V . .y. §
‘ - V . - . v-
9 . . '1 . ) .
r 9 F . . s so . ,
P 3 ' w ' J 9 ’ .
............ A ~e..............,e.?.--~o¢---ea.a- on. r l t ) - '0 h ‘ ’ ’-
A . ......................... , ............. ,. ................... ‘ ............... ' ______________
D A J M K .‘ A H 2 y r ... N a‘ V
a A s k : ‘I : w p- A '3
A 9 - R J A ‘9 I
r I A n- : 3 fi , 1 T
‘ , fi~ J " Q : o. ’ 3 .1 5' T a. . .4 7"
............. x...--...o....r..-‘..‘..--.e...-sloucu-uuo o_..,,.,,_...~,.:_.,.._..,,......x.‘uug....--.o0.oo-.oocc _...‘-.,_._.r¢.............-ur-o‘.....-.....s....-----
_ ,, . .
‘ I J V ‘ 3 a v * ‘ A ‘2 +
‘1 A t '7 - " “ " " v
1 4 P ~ F- P k:
«I J . r . e ‘ P ' ¥
" V 4 r 1 (- y (-
.......... r.Ar>1. r-A\‘.‘ r,r-a,
r e . v :r «a ‘ v ' c‘ ° ‘
‘ ( v: ) ; k f' 1 .4
t I. 5(- k ‘ I»
. - 1s
.............................. r - e
1 1 1 1 1 1 1 1 1 '1
r V T I r r v ' v v v
20 Rpmenqine speed. 3 m VL: '1 20 IPnzenqrm speed. 6 .- VL: 20 RPHjenqtne speed. 9 in VL;
(03.62 ms t1me:de1ayl : (0.72 as ttmmelsy) (0.82 ms trajdelsyr
1 ' q : ' : : (- ‘ g s.
w .
r ....u ............ a ........................................ s ............................. ‘ """"""" f """""" ' """" “
........ ‘ ,,...,.,..._,,'._. w :’ : r x ‘ z} a r
‘ I. 1 \ s A~ A y ).
r 1 3 ; a ‘5 y . 8,
. V . ’ n- _ ) . a A 4' '8 I P
n- “ ' K : 1 : a -
r 3 . a as. r \1:
R 'l _____ R : r- : , ...... "‘,‘1 ........... ;..
fi ............. +....+..., ......... L)...-s--! . * .......... ~ ......... *cooo' ......... ’. I...:.-8..... , fl 1 i : s
R a a 5 k 1 a- : a '
R ) : ‘r ,. : - R ,, a
T Q _ ' 1\_’ Q ; ’ R 1‘
“ , ¢ : v 1‘ _ E If . 7 E.» . . 1t : , :3
........... N.r...----~-v------"0*-"'"""'f""""' s...........~..I.........--.....bo+o.....-....ansnuo-o-o _""""""+<""""""é"?'4""'""",."""".
‘1‘ 4‘ : 2 ' ’8 . ‘1‘ - - ‘1‘ : ‘ : :
: q . b . Jr - : R . 4 : )1 g - 1‘ 3 g ; N
‘1‘ 1 : r- 3 . R: R1 K : e 1 a' it .
‘ ‘5 3 . ' : : : . ': ;
3 7‘ : u .' 't . 1* : ’t : s. 2‘ . 4‘ : q ' ‘
4 ' r- : 4- - tr - 9s - F -
k - K 1 e . R f a
......... ikpir k‘tvb .. A,
P .L : r K; ' q ; F. . x
v. F : I. : “ : 1::
L ; (- ; ; :
V 2" : 7* :
w . u . . .’
......................................... s.............I v v
1 l L 1 1 4 1 1 1 1 1
s s 1 fl 7 v v .T *1 fl 1
30 ”Hzongtne speed, 3 m VL: . 30 RPngangina speed. 6 In W 2 30 ”1450119111. speed. 9 ran VL;
10:32 as unsidslsy) ‘ : (0,42 as trauma-lay) ; g (0?" as ungaisy)
; ‘ (- 1 - F E i' ' ‘ 5 °
'- I . . . ' b
u. . - l‘ I ............................ it; ......................
........ inn-0:.--a----)-------~Y--nun-“5.”--'-' '----"-"-"-':--" ’ \ S , ) y y L
a . j 1 . .
x "I: a \I - j: ‘3 x3 A y‘ ’:
' fl , 5‘ : V : 11 3 9 1 3 3 a
k “ 1 5 J T s : . a: 1
k .4 6 A : y ’ I p ‘“ I ? ‘ii
.. "...”K. ________________ “.1 ___________ ' ..................... , ............... ', .............. ; ....... ""1 .......... a ......... ‘n... ............. ;.., .....
" 7- " 9 : ” * s .. ° 5 e 5 '9 r 9 5
¢ 1 : k . . v 1 0 k ‘ “A .
tr ( T .1 ‘P 9 t k T a > :
1 94 1 . ‘ : .
r- - R : it .y I? . 1‘ 3y
‘1‘ ‘ J 2“ .: ‘5 ......... . ...... "....L ......... ' 4‘ 9
...-u ............. a ......... ~ """ ‘ . ' ---------- *4. --------------- s.’ ........... s .........
............ ,Onno-hoolcl- , . ¢ . a : $ . $ . I
K a : 4' . a d, ; '5 . : It : : Av
K ‘ ‘ 3 “Z 1 "' ' ’ 11 I R J :
“ ‘ . ° ‘ . 3 s r s i *2 at
Q T g " : rt ‘ t’ q. : ... '- :
K , . .. : 39 ............... F. ............ fi- ................. ‘.....1 ...... _ ......... 3”“ _____________ I: ____________ 3.9! ......
--------------- TOO-Ionor-UI-IC'CIIDIs-u .‘ up. ' s f * . e .
'R ‘ : r d» R p; r :
t r T *' + I t '1'
‘ Iv. \- . K.
6 ° 1. ‘ g, . :r 5
........................... x..?...... .. v- 1"
1 1 1 I 1 1 1 i r 1 1 1

 

Figure V. 1.2: Ensemble averaged velocity vectors at 132° crank-angle
for all mne settings of the engine parameters.

.1: Lu

- ”...,-..” T_1r

229

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

é

(a) 1% 45.0
10 RPM '3
engine g
speed “a

0.0

2

i

3.: 45.0
(b) g
20 RPM «...
O

°-° ( — 3 m Valve 11a
......... 6 m
“~— 9m

3

3 45.9
(c) g
30 RPM 5
3

 

 

 

 

 

 

 

0 45 90 135 180 225

Crank-A0310 [in 4081150]

Figure V.1.3: Profiles of induction angle in dependence of engine speed
and valve lift

 

230

 

 

 

 

 

 

 

 

 

 

g 10 011 £\
a \r———i
E
.i
'3
IE. 10 "°
tn
1 1 l (a)
3 6 9
Valve Lift [in mm]
a
a
‘a
a o ,
B .o
i
g 0.4
3 :
0.0 I l I (b)
10 20 30
Engine Speed [in RPM]

Figure V.2.1: Specific kinetic energy inherent in area of scrutiny at
-angle as a function of valve lift and engine speed;
(la)2 KE versus valve lift, (b) KE versus engine speed.

 

...-mm

231

 

4.0

 

40

 

 

 

Fluctuation Energy in z-Direction [in m"2/s"2]

 

 

I l
6

Valve Lift [in mm]

0.1

Figure V.2.2: Energy content in velocity fluctuations for z - direction
inherent in area of scrutiny at 132° crank-angle in
dependence of valve lift and engine speed.

 

 

 

0.7 - / m

 

Normalized Angular Momentum (negative)

 

 

0.5 l l l
6
Valve Lift [in mm]

Figure V.2.3: Comparison of angular momentum in area of scrutiny for different
settings of valve lift and engine speed at 132° crank-angle.

 

232

 

..............................
.........

1‘ Figure V2. 4: Mechanism of vortex ring
ingestion

 

 

        

Potential tore

 

 

_.’ Smooth interface
I
I Q ’I \\ convoluted eddies.”
£0) \‘
Nut/11‘ J\\\\\
/4 Rotational flow
I nrrammcnt «\
m wedges
1 , Turbulent flow
lmnsrtmnul flow "WI-1W1! “my

nuts I’ “1le

Figure V.2.5: Schematic decomposition of an axisymmetric jet flow field
into two characteristic regions of flow. [Sourcez Yule 1978]

233

  

17.7.
(wwwmmtwwfif Velocity Vector

234

Depiction of error.

—.

  

X

Reference

x
Frame and...

Imls
Velocity Vector

(Wiser-MW)

235

Reference

1:
Frame and...

 

 

 

 

 

 

  
  

 

180' CA

 

Imls
Velocity Vector

Figure V.3.]: Maps of instantaneous velocity vectors for one engine cycle
at selected crank-angle positions for 20 RPM engine speed
and 3 mm valve lift illustrating the development of intake flow.

 

236

$8 36 “>22. 25C .3 EB . .
. 8 am Re «293 2:95 . o -
A> E m 98 SEM on 8 835 cameo o>w=oomeoo 085 com 298.» 362...; escomwwrmflflowmwm .m.m.> 059m

can... we agar—on

.3»qu 498?;
fi—

 

 

ougkum

 

237

as «3 $28 25.5 .9. e5 .3 can sec 8&8 ..NS 035-0380
..H> SE n can 2: cm 3 ”no—26 ofimeo 39.838 85 com £083 360—? msooeflcfimfi me an: “00> oSmE

e E

III

mono no cognac

83$ ~983—
{E—

      

EV§ NELK
N U

 

 

350.0%

  

238

as 86 are 25.5 .: Ba .2 an. see came ..3 0353.55
..5 88 m new 2.2 2 3 86.3 2.95 033828 8:: com 3863 romeo—o.» 9523.863 .«o 3.32 ”v.92 23E

x e

.11

name we :otofioo

an»... e83.
“E

 

      

ES quak
a

 

uo§b\u-

 

239

 

 

 

 

 

 

 

1200 : . _*__| Ecm I;
400 5 555‘ ' ‘5'"

 

 

 

u. 55 555’ 5555!. .: 5m

J; L ::+
.. .... 3
123-5
35; 3
“ u
_ ‘b— ::

*4“! ..—
W

 

 

Velocity Magnitude, w - Component [in mm/s]

 

 

 

 

 

n “Mu
v * l
- VI I
. E i
I I
I E E
-400 5 i i i n
o 45 90 135 180 225
Crank-Angle [in degree]

Fig. V.3.5: Velocity profile for 10 RPM engine speed and 3 mm valve lift;
location 18 in the vicinity of the axis of the intake jet.

 

 

 

 

 

 

 

 

 

 

555 555 555155, 5'. \

 

-200

 

I351 I555

0 45 90 135 180 225
Crank-Angle [in degree]

Fig. V.3.6: Velocity profile for 10 RPM engine speed and 3 mm valve lifi;
location is in the area of scrutiny.

.... 40° T I I
m ‘\ 5
E + Ecmo 5
fl. —- EA
200-. ............
—9— 50m '
5 ‘ I ‘
. ,L .I‘ 1117“ 5 IIII'III5 5
3 v 3'.) ugly I
g F“ 555555, 5IIIlhI II .
’3‘
§
:>

 

 

 

 

 

db--

 

 

240

Vdnu'y [a rank]
Velody [h min]

   

Oink-Ald- [n «ho-l amt-Md- [In den-l

Vdod‘y ['n will]
Vdoony' [h WI]

   

oak-And. rn deg-1 Gut-Aw- r- than)

Vdodtyfhnnh]
Vdody [hum]

   

ant-Anu- [n duo-o] Out-And- [iI «have!

VIM [in m]
Vuha'y [h m)

   

WM I'll m] Gilt-Md! ['II 4"]

Figure V.3.7: Profiles of the z - component of the instantaneous velocity vector located
in the region governed by the intake jet for eight difi'erent engine cycles;
the plots also contain linear curve fits for the profiles.

241

Vdodly [h min]
Velocity [in ma]

 

Cut-Awe [m age.) Oink-And: [n ml

Vdotily [in WI]
Velodty [‘11 mil]

 

45
(hnk-Aldilh m1 oak-wmml

way [in my.)
Vdodty [in my.)

   

’0
out-Anus rm agree] Grunt-Angle [In deuce]

deeity fl: nun/a]
VMy [‘n m]

 

Gilt-M I'll 4'9“] cub»;- ['II 49"]

Figure V.3.8: Profiles of the z - component of the instantaneous velocity vector located
in the area of scrutiny for eight difl‘erent engine cycles; the plots also
contain linear curve fits for the profiles.

 

 

 

‘3 unis aways—1W

 

242

\ cyumler wall

/\ Y

\7\/\/\

 

\/

/ \f \/ \/

//3< A/\/\/\/\ /X

f

 

 

/

 

 

Figure V.4.]: Investigation scheme for vorticity distributions.
Scheme includes hexagonal pattern formed by vorticity vectors
from area of scrutiny. The 116 vorticity vectors are derived
from the 74 intersections in the grid of laser lines exhibited in
Figure II.2.3a

 

 

243

2D contour plot

 
  
 
 
 

 

/

View for
azimuth = 0'

Figure V.4. 2: Difl‘erent perspectives showing
the distribution of the
circumferential component
of the vorticity vector.

The area covered is about 50x50 mm.

3D contour plots

Vbrticity [in Us]

244

2D contour plot

3D contour plots

azimuth -20°

  
    

  
   

Mk5;

   

   

\\ \\. _
\ " 7
\i‘lun "INF, 1'", '4'. °° :
‘lllllh fl 1
“‘ll‘lil
t 'i/ 'E
!!t 1%
\lvlll’i"; 9
/vl‘:vfor 200
azimuth = 0'
Figure V.4.3: Different perspectives showing \
the distribution of the x\‘ ‘/2
circumferential component

of the vorticity vector. azimuth 55'
The area covered is about 50x50 mm.

245

2D contour plot

3D contour plots

azimuth -20°

4" $

"ql'l‘l\1III: \ ,\\\\ $:\‘\‘

»\\\\\\\\
\\\\\\\

l M.

I; :lfgz‘. ..II’ \\\ “I \ \\\\\\\\
t ‘ w ~
\ v"“‘:“~

 

elevation 30

     

 
       
   

  

   

Ml

  
 
  
 

    

     

    
   
   

i. °° E
\\\,llllll‘ir W": .:n 5
' \ 3.1!. 'Q\ —‘
“I I " 111.51%": 'I’.‘t'l.«\'. .‘f:../" I?
Wig. fig. "'4" .
‘ 'Iztliiijix'ti °° 9
View for I,” 2
azimuth = 0' 0°
Figure V.4.4: Difi‘erent perspectives showing \ \
the distribution of the x
circumferential component
of the vorticity vector. azimuth 55°

The area covered is about 50x50 mm.

246

\
~ I"

ml

 

 
    

    

my 'ii.
it" 4' MI' “I I

r li‘N ‘,‘3‘

, IIIIII ”hi \i“ ,

i ////i
«t \\

 
  

  

    
  

azimuth = 0'

Figure V.4.5: Difi‘erent perspectives showing
the distribution of the
circumferential component
of the vorticity vector. azimuth 55'
The area covered is about 50x50 mm.

Vbrticity [in Us]

247

W

W

 
   
   

    

ii “iii. iiii

Hi
i {Slit
\\\\\ \\ ..."? 'N\\\\\\

        
 

   

\\\\\\\

./

   

I.

 
  

 

elevation 30

     
  

 
           
  
  
   
  

,ll‘i .
”Mu”: "1W"

 

\brticity [in Us]

  
 

     
  

 
 

“I
iii ’1th 'I'I'it’”' ii,IIIrI
, I f M
/v|’ f ’q‘“ “VI” Efflil’ll’l"'l\\\\“\\\ \\~ 200
m = \

Figure V.4.6: Difierent perspectives showing
the distribution of the
circumferential component
of the vorticity vector. azimuth 55°
The area covered is about 50x50 mm.

248

2D contour plot

3D contour plots

azimuth -20°

i 11150.2,“
3 \\\\\<VII'I/’(.~ ‘.
i‘ \\\

 

elevation 30

   

 

 
   
   
      
 
 

     
  

   

    
 
 
 

   

    
 
 

\\\<‘\ E
N“ 11"”1’03
\ "Vii ”’h'h‘.‘ ‘ 5
‘TI!’ ii"."ii\£"”'ii\ :3;
@211. = 0' \\\\\\'v \ r 3°°
l

Figure V.4.7: Difl‘erent perspectives showing
the distribution of the
circumferential component
of the vorticity vector. azimuth 55°
The area covered is about 50x50 mm.

249

  
 
  
  
     

3D contour plots

(til 4
if ,
WW \.\W

W?

«1 “\W

\“0\\

 

till» iii
‘diW VI
\\ v

’1"

',II': «it
t. 1""

  

Vorticity [ in Us]

FigureV.4.:8 Diefire entpers pective sh .
the distribution fothe
(fix-cumferentialc component
olfthev mcivectoty azimuth55°

The eareacove veisred ab outSOxSOmm.

250

2D contour plot

3D contour plots

azimuth -20°

 
 
   
    

    
 
 

 
  

‘34
'

I

f”

, it“\\\ "5‘ «W
in W
1W

1““ it

MM
Iii

  
  

A

\

——

 

7';

/
/r4

    
 

:;(
/

:’
///—

ity,
iii

A;

   
    

   

    
  

    
   

    
    

       
  

   
 

 

 

 

\t‘rtg (”I "I, Lift ‘

\f '1‘“ .2: 1"" («1’ iii LE:

15% = a ““y W“ Wm“ :2
\v \

   

Figure V.4.9: Difi‘erent perspectives showing

 

the distribution of the /z
circumferential component
of the vorticity vector. azimuth 55°

The area covered is about 50x50 mm.

Vorticity [in Us]

251

2D contour plot

3D contour plots

    
  
   

     
  
  

v\ “m

1'" W ‘\\\‘i

”W

\

it

 
  

  
 
 

4%

     
 
   

 

 
   

"unit! 3;

”Nil:

[’1
a?
hi "'i “W

“W.“ mu I3“
imiul/

Figure V.4. 10: Different perspectives shwing

 

§§§§"8858
Wrticity [in Us]

   
         

the distribution of the
circumferential component
of the vorticity vector. azimuth 55‘

The area covered is about 50x50 mm.

 

252

Slope of Vorticity [in l/smm]

   

6
Valve Lifi [in mm]

Figure V.4.11: Evaluation of slope in circumferential vorticity distribution
in area of scrutiny as a ftmction of valve lift and engine speed
at 132° crank-angle. Error bars are evaluated 1n the same manner
as for Figures V.2.1a and V.2.2.

Sign Switch [in %]

 

3 6 9
Valve Lifi [in mm]

Figure V.4.12: Sign switches of slope 1n vorticity distribution 1n area of
scrutiny as a function of valve lifi and engine speed
at 132° crank-angle. Error bars are evaluated In the same manner
as for Figures V.2.1a and V2. 2. Error bars for 20 RPM are
representative for other two speeds.

253

 

 

 

 

 

 

 

 

 

 

        

'— an,
N —I IV
a 2' . ‘ (a)
g 10“°q E ‘ $\.\
g E :_‘ _‘ \ —A
a ‘ 9 ‘ \V
j a
Lg m - \
8 g m4 \
3 3 av j
0 .

9 10.1.o_ "g- r
E- m “ b 4\*

7 l l

3 6 9

vortical energy

 

.......... ..'.'».‘-.
............ .....‘.

 

 

2.0

 

1.5

 

1.0

 

 

2.0

 

 

 

 

 

Normalized Energy

 

 

 

 

 

 

2.0
1.5
1.0 \
l I l
6 9
Valve Lifi [in mm]

Figure V.4. 13: Comparison of kinetic and vortical energy stored in area of scrutiny as
a function of valve lift and engine speed at 132° CA; (a) denoting
magnitudes, (b) afier normalization by values for 9 mm VL.

 

 

254

 

.Eofiaa noun—8:02 R283 :38 05

mo mom—5:3. 23¢ 05 no Soho Egon mo scumbag: ”3+5 oaammm

 

oxomeE bad 13% :3»: $39;

 

86528 2:53

 

 

 

 

 

 

\\\

528m 5ng
.3 35528
a 28955

 

 
 

 

 

 

 

 

 

 

omens—56 .3 252.»

w\\\\

   

 

 

 

 

 

255

.88 095m Sosa mm 33 wouawumgfi Anna. DE 296 oammco 25 55m? 395. 8935 .«o 33 com“ 98 .mfl 3.0

88m mcouznfimmc 33:9, 5232. 83388 8..» can ...—on .302? 3083535 88m 23 3955 03885
88m £32 90?? 5253 ‘5 88 m 23 3on ofiwco 25— on com 8335,“? bBEo> .«o acumen—Eco ”2 .v.> 059m

         

      

        

  
    

,

   

  
   

          
   

 

   
 

          

  
 

       

 

  
     

   

  
   
 
  

     
   

    

    

  

 

  

av
§J ’1 “ — £1 sat
..., , .....ué.....,..a........,.. ,1 ..., a, ,, .... .3? . .
. a._..m..§....e=.w.«..‘... . a .... «...... :1 (gs.
s ......s = ., g . . ‘\ ... ... . ,.
. ,. .. ‘ 3:: \ _ ... a. a, . .... {amuse ... .
. . a Y., S < . 4 .2 w. n i=3, .. . E . E:
... 42.5 .- 4.2,..31S .. ...... ,, ... ... s.............,.,.,«,“....mu..
L _. (......‘s? ... é... ... ..
8. .

 

 

 

 

 

  

 

 

 

<0 .3—

I? (Q s.

1‘ p"
§
Ot‘

a

A

a \ till!
\ 2w) Andy $§ft§l

 

256

.88 095m Bone mm 8.8 Baawumgfi .oEfiéEEu .9: “a a: o>_d> SE m
23 30% 335 Sam on com 86.6 898 328850 025 8m 5% 36?? mo cocoacom ”3 .3» oBmE

      

             

 

 
      
   

 
   

 

 

 

\ ,. 2...... ... E.
.... ...: 5
P7”. =me V
/_ 5% “:9
a .\

 

 

 

,

....sz .w..._\\‘..,.......,

, A .. ..
a.

 

    

3V

 

 

3% 0m mg 0m 3% 0m

3 ,2 WW 0 (l

.
'

0.2.

a, 8 ”J -
o9 ‘ a.”

  

 

Slope of Vorticity [in l/smm]

257

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

7 Ti H M 1mg .3}
:21 1 .5 r 11‘ l
.1) .0 20 . H50 .,

Engine Cycle

Figure V.4.17: Slope determined from the distribution of vorticity in the
area of scrutiny at crank-angle 132° for individual engine
cycles for 20 RPM engine speed & 3 mm VL.

258

as open 38“ a 8: cesarean ..8 23 ..8 ..ww safi-§s a exchange
2503 15 88 a 28 2mm 2 8m 95. um: 29$ oemmuo 28 c8 Sewn—£36 36E?» mo 383.com ”E .v.> 952m

         
 

            

    

   

       
      

     
  

        

        

    

 

           

 

  

 

 

 

 

 

 

 

 

.... ewwa...............n .3., ., iza h...........fl......._._\._=;’=_ s
a =2 a , ~ . \z: 1. 3’
. as: .1... a”... . .._._.;.,r...a...,.._._._..a.~._.‘....wE. a:
L x 8-
<U .Na <0 .8 <0 .ww

A3 3

0.06r

xx

 

259

 

 

.88 open 80% 2 83 33333 #2 Ba N: “on $35355
«a 15 SE a 28 2mm 2 com 9;. DB 298 ofiwao 28 com magnesium. £099, mo coconuom ”3 .v.> oBmE

       

                        

   
     

     

                 
             

      
       

  

 

 

 

 

           

 

8? . S. 8? v .m:=‘wt '«g_—==— .
.... E sea.  : ...... .. .
8. .5: dh‘ufi 8. k: 1” ”EN “é. Moflflofih—wmfi‘.
. .V
3 5 ......is:

<0 .Nfl <0 .3—

 

 

260

1111/:
Velocity
Vector

Depiction
of error:

     
   

172' CA
G)

180' CA
(1)

Figure V.5.1: EA velocity vectors depicting the
intake jet as close-up for 20 RPM
and 3 mm VL; time delay: 0.34 ms.

261

1min
Velocity
Vector

Depiction
of error.

 

Figure V.5.2: Fields of ensemble averaged velocity vectors depicting the intake jet
as close-up for 30 RPM engine speed and 3 mm valve lift at selected
crank-angle positions; time delay: 0.17 ms.

—>

In]:

Velocity
Vector

 

Depiction

of error.

 
    

132' CA
(d)

 

Figure V.5.3: Fields of ensemble averaged velocity vectors depicting the intake jet
as close-up for 20 RPM engine speed and 9 mm valve lifi at selected
crank-angle positions; time delay: 0.51 ms.

262

 

lm/s
Velocity
Vector

Depiction

of error.

 

263

128° CA

 

 

 
     
      

  

132' CA
. \ (q)

 

 

156' CA
\ "‘ (r)
\ 1 \
_. " I
lmls
Pekxfioz
Vector

 
     

  

108° CA
\ / (n)

 

 

 

Figure V.5.4: Maps of instantaneous velocity vectors from one individual engine cycle
illustrating the dynamics of the intake jet at selected CAs for 20 RPM
engine speed and 3 mm valve lift; time delay: 0.34 ms.

264

 

108' CA
(G)
lmls
Velocity
Vector

 

  
 
      
    

  

102' CA
G»)

 

 

 

Figure V.5.5: Maps of instantaneous velocity vectors from one individual engine cycle
illustrating the dynamics of the intake jet at selected CAs for 30 RPM
engine speed and 3 mm valve lifi; time delay: 0.17 ms.

128' CA
\ (C)

In]:

Velocity
Vector

 

132' CA

 

........

Figure V.5.6: Maps of instantaneous velocity vectors from one individual engine cycle
illustrating the dynamics of the intake jet at selected CAs for 20 RPM
engine speed and 9 mm valve lift; time delay: 0.51 ms.

265

(a) RMS Fluctuation Wms (b) Fluctuation Intensity wm/WEA

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   
 
 

 

 

 

[ in mm/s]

 

 

 

 

 

 

 

 

 

 

 

 

(1)
30 RPM, 3 mm VL

Figure V.5.7: Velocity fluctuations for (a) RMS fluctuations and (b) fluctuation intensities
at 132° CA for 20 and 30 RPM engine speed and 3 and 9 mm valve lifi.

 

266

 

 

 

 

 

 

Compression Exhaust
Intake. 13 _
1:2‘3g4iregular
. ' ’ Engine
Piston)
Speed
dszt 1 Phase-
1,3 Space
N m
11 z
2.4 11]
Intake - Exhaust - gfim'f’tfit ti Difference of
= 1 rr 111 . W ”5:52;: m: G the two characteristics
' ' i
Study 1n the phase (1 agram
rio' 270' = , (a)
0 V
360 Crank-
Angle
(b)

Figure A. l: (a) The difl'erences between the
piston motion of a normal four-
stroke engine and the water
analo model used in this work.
The fferences, required because
of incompressibility of water, are
also shown in phase space.

(b) A picture of the cam-shaft
used to create the piston motion.

 

 

I)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

g E
l
l 2 3
gear ‘
z a 3 5 3
'1 5 -
1 .0; I, .5. 3
1‘1] r. i
EH}: i: g
5 .
. . I a i E
i +
88885815565! 86888835883 85386838853 56833865883
£5 . . £5 £5 1: £5 5
#l

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure C.l: Drawing of circuit of 'electronic signal valve'

268

 

 

 

 

 

 

 

 

 

 

 

 

 

horizontal Lines (500) in CCD Array
of Hi8 Canon Camcorder
1 2
aetl g 4 set 11
1 6
9 8
n 10
13 12
3 3
‘9’ 23
$3303} Acquisition
scanner; °f
shift of 1.0.“! one set (250) . other set (250) ;
L of hori- mm! lines
Camcorder H q H 1 II 1
T i 2 . >
M 3 5 Q mne
’ N 2
30 Hz f .
" one image 5
optical \ I
Encoder '- : \ - *
Timing k time
’ e
> 33.33 ms momfififfi
between two images

Figure C. 2: Virtual synchronization between timing on engine side (optical encoder) and
continuously running camcorder on camera side.