Thesis Final Julio 15 Emil

UNIVERSITAT POLITCNICA DE CATALUNYA
Departament de Fsica Aplicada
THE STRUCTURE OF TURBULENT JETS:

APPLICATION OF EXPERIMENTAL AND
ENVIRONMENTAL METHODS
EMIL SEKULA
Barcelona, July 2010
Director: Prof. Jos Manuel Redondo Apraiz

Codirector: Prof. Ana Mara Tarquis Alfonso
Dedicated to my girlfriend, Amaia and to my parents, Zbigniew and Lucyna.
Gratitude
To Professor Jos Manuel Redondo Apraiz for his constant help and ideas in the process of
redaction of this study. His impressive scientific experience in such difficult field as
turbulence had important influence on the presented work. His friendliness helps me during
these years not only in the research life.
To Professor Ana Mara Tarquis Alfonso being the coordinator of this thesis.
To Dr. Joan Grau for his help in the field of programming language and for offering
facilities of the Ima_Calc program.
To Dr. Alexei Platonov for the collection of Synthetic Aperture Radar (SAR) images and
his commentaries related to the satellite images analysis.
To Dr. Raffaele Marino and Dr. Luca Sorriso-Valvo for their cooperation.
To Prof. Allen Bateman for make accessible some useful data in present work and for his
cooperation.
To Departament de Fsica Aplicada de la Universidad Politcnica de Catalunya for facilities

founded during my academic period.
I would like to thank to all persons not mentioned personally here but having some
influence on this thesis.
Finally, but not meaning with less importance, I would like to thank to my girlfriend Amaia
for her invaluable support in both, research and personal life and to my parents, Zbigniew
and Lucyna for priceless help during all these years of my youth.
Thank you!
INDEX
GRATITUDE
CHAPTER 1. INTRODUCTION
1.1
State of the art
1.2
Thesis aims
1.3
Thesis structure
11
CHAPTER 2. BACKGROUND AND THEORETICAL CONSIDERATIONS
13
2.1
Turbulence and turbulent flows
13
2.2
Homogenous turbulence
15
2.3
Non-homogenous turbulence
23
2.4
Extended Self Similarity (ESS)
24
2.5
2D Turbulence
25
2.6
Turbulent jets
28
2.7
Buoyant plumes
32
2.8
Plane turbulent free jet
38
2.9
Circular turbulent jet
48
2.10
Plane turbulent wall jets
54
2.11
Other jet and plume configurations
62
2.12
The turbulent boundary layers
66
2.13
Geophysical turbulence
70
2.14
Mixing efficiency
71
CHAPTER 3. THEORY SAR AND FRACTALS
74
3.1
74
SAR (Synthetic Aperture Radar)
3.1.1
SAR applications
77
3.2
Fractal analysis
80
3.3
Self-affine fractals. Relationship between turbulence spectra and

fractal dimension
82
3.4
The spectral method
84
3.5
Fractal characteristics of intermittent turbulence
85
3.6
Box-Counting Method
87
3.7
Multifractal characterization
88
CHAPTER 4. METHODOLOGY AND EXPERIMENTAL SETUP
92
4.1
Introduction
92
4.2
Acoustic Doppler Velocimeter (ADV)
93
4.3
WinADV
98
4.4
Laser Induced Fluorescence (LIF) and Planar Laser Induced

Fluorescence (PLIF)
99
4.5
Particle Image Velocimetry (PIV)
102
4.6
DigiFlow
104
4.7
Particle Tracking Velocimetry
106
4.8
ImaCalc
106
4.9
Experimental setup
109
CHAPTER 5. EXPERIMENTAL RESULTS
117
5.1
Lower Reynolds number jet experiments
117
5.1.1
Velocity profiles
117
5.1.2
Jet velocities
118
5.1.3
Reynolds number
119
5.1.4
Standard deviation (r.m.s. turbulence)
121
5.1.5
Turbulence intensity
124
5.2
High Reynolds number wall jets experiments
126
5.2.1
Velocities
126
5.2.2
Turbulence and mean velocity parameters
127
5.2.3
Reynolds number
129
5.2.4
Standard deviation
129
5.2.5
Skewness
131
5.2.6
Kurtosis
132
5.2.7
Correlation
133
5.2.8
Covariance
134
5.2.9
135
5.2.10
Other results
136
5.3
High Reynolds number, two-phase Bubble jet
137
5.3.1
Velocities
137
5.3.2
Standard deviation
138
5.3.3
Skewness
139
5.3.4
Kurtosis
140
5.3.5
Correlation and Covariance
140
5.3.6
141
5.3.7
Other results
141
5.4
Data stability
142
5.5
Instrumental error
144
5.6
Comparison of the three cases of the jet configurations
144
CHAPTER 6. SPECTRAL MEASUREMENTS, STRUCTURE FUNCTION

TECHNIQUES AND SCALES ANALYSIS
147
6.1
Energy spectrum Fast Fourier Transform (FFT)
147
6.1.1
Normal jet, y = 4 cm case
149
6.1.2
149
6.1.3
150
6.1.4
Bubble jet, y = 4 cm case
150
6.1.5
151
6.1.6
151
6.1.7
Energy spectrum results discussion
152
6.2
Structure function analysis
159
6.2.1
160
6.2.2
160
6.2.3
161
6.2.4
161
6.2.5
162
6.2.6
161
6.2.7
163
6.3
Correlation and scales
168
CHAPTER 7. SAR AND EXPERIMENTAL IMAGES FRACTAL RESULTS
175
7.1
Introduction
175
7.2
River jets SAR images
175
7.3
Sea vortex analysis
179
7.4
Experimental jet results
182
7.4
Fractal dimension analysis conclusions
190
CHAPTER 8. CONCLUSIONS
192
8.1
197
Future works
REFERENCES
199
APPENDIX 1
207
APPENDIX 2
225
CHAPTER 1
INTRODUCTION
Scientific work on turbulence is difficult but fascinating, this field is still not totally
discovered and exist many unsolved and open problems. From the philosophic point of
view, a goal of every study should be based on the most possible conclusions drawn from
assumptions even they finally result to be erroneous. This falsity give us some idea and
serve as additional information for future works, it is like successive step to discover the
totality of existing problem. Before the detailed study is presented in this thesis we
familiarise with existing hypotheses and assume and test our own ones: then the next step is
to prove their correctness remembering that is not in the scientific spirit to think that our
final ideas are absolutely correctly and irrefutable. So to present this thesis we should ask
some basic questions: what is an object of our problem, why do we want to study it and
how we will conduct our work. In this chapter we will answer these questions and finally
specify the detailed objectives and general purposes of this study on the structure of
turbulent jets and plumes.
The basic and overwhelming, research problem of this study is based on the
understanding of real turbulence and its structure. Turbulence is a phenomenon that can be
found anywhere, in every field of life, for example, from the stirring of a coffee cup to the
wind in the atmosphere. Most practical flows occurring in nature and in engineering
applications involve non-homogeneous turbulent flows that are affected by boundaries or
body forces such as: currents below the surface of the oceans, the gulf stream as a turbulent
wall-jet kind of flow, the boundary layer in the earths atmosphere, jet streams in the upper
troposphere, strong cumulus clouds and deep turbulent convection in the ocean, boundary
layers growing on aircraft wings and propellers, combustion processes and mixing in
turbulent chambers, wakes of ships, cars, submarines and aircrafts. There are many more
existing examples. In our definition we can observe closely to one of the first feature of the
real turbulent flows, which has not been studied extensively: non-homogeneity. In physics,
homogeneity is the quality of having all properties independent of the position, i.e.
1
translational invariance. Obviously in non-homogeneity these properties depend on the

position.
The study of turbulence clearly is an interdisciplinary activity. It is important to know

that turbulence is not a feature of fluids but of fluid flows. If the Reynolds number is large
enough, the major characteristics of turbulent flows are not controlled by the molecular
properties of the fluid in which the turbulence occurred. Great progress has been made in
the last century (since Kolmogorovs work K41 and K62 theories (1941, 1962)) on the
structure and theory of homogeneous and isotropic turbulence, but non-homogeneous or
boundary affected flows still lack a comprehensive theory. Andrey Kolmogorov was a
Soviet mathematician who made major advances in different scientific fields (among them,
the probability theory, topology, intuitionistic logic, turbulence, classical mechanics and
computational complexity). Kolmogorov is widely considered one of the prominent
mathematicians of the 20th century. On the later part of scientific life he switched his
research interests to the area of turbulence, where his 1941 (K41) works had significant
influence on the field. In classical mechanics he is best known for the KAM theory
(KolmogorovArnoldMoser theorem).
Turbulent flow is the motion of a fluid having local velocities and pressures that
fluctuate randomly. In this movement subcurrents in the fluid display turbulence, moving in
irregular patterns, while the overall flow is in one direction. Turbulent flow is common in
nonviscous fluids moving at high velocities. Almost all flows, natural and man-made, are
turbulent. Occurrence of turbulent flows in many situations forces to next works on this
subject.
Tennekes and Lumley (1972) proposed a list of some basic characteristics of turbulent
flows:
Irregularity (or randomness): the turbulent flow is unpredictable.
Diffusivity: which cause rapid mixing and increased rates of momentum, heat and
mass transfer.
Mentioned before high Reynolds number
Three-dimensional vorticity fluctuations: turbulence is rotational and threedimensional.
Dissipation: turbulent flows are always dissipative.
Continuum: turbulence is a continuum phenomenon, governed by equations of fluid

mechanics.
Flows: turbulence is a feature of fluid flows and not of fluids as was commented
before.
Diffusivity is the diffusion coefficient. It is proportionality constant between the molar

flux due to molecular diffusion and the gradient in the concentration of the species (or the
driving force for diffusion). The higher the diffusivity (of one substance with respect to
another), the faster they diffuse into each other. This coefficient has the units of
(length/time).
Reynolds number [for Osborne Reynolds] is dimensionless quantity associated with the
smoothness of flow of a fluid. It is an important quantity used in aerodynamics and
hydraulics. At low velocities fluid flow is smooth, or laminar, and the fluid can be pictured
as a series of parallel layers, or lamina, moving at different velocities. The fluid friction
between these layers gives rise to viscosity. As the fluid flows more rapidly, it reaches a
velocity, known as the critical velocity, at which the motion changes from laminar to
turbulent with the formation of eddy currents and vortices that disturb the flow. The
Reynolds number for the flow of a fluid of density ; and dynamic viscosity through a
pipe of inside diameter D is given by
Re =
DU
where U is the flow velocity.
(1.1)
If the Reynolds number is not too large, the flow will be laminar. At higher Reynolds
number, the flow becomes chaotic in both space and time. The critical Reynolds number for
laminar flow in cylindrical pipes is about 1000.
It is not easy to solve turbulence problems because in fact it is mixed, random and not
regular process and it has not been possible to find a complete theory that describes the
phenomenon till now. There is not still a set of equations that could be used to efficiently
compute turbulent flows. The complexity of turbulence is also related to a large number of
scales of the flow playing an important role. The fundamental dynamical equations that
govern turbulent flow are the Navier-Stokes equations; their computational complexity
becomes hard task for large Reynolds numbers. It is a system of related non-linear partial
differential equations and must be supplemented by initial and boundary conditions and
always faced with the closure problem (a set on n-1 equations with at least n unknown
variables in it). Computational Fluid Dynamics tries to resolve this problem using oneequation models; the two equation models (for example, k model) and the second-order
closure models. There are some models existing for some specific flows because computers
have recently become more and more powerful but they are not universal.
In 1922 Richardson proposed fully developed turbulence as a hierarchy of eddies of
different size. He assumed a cascade process of eddies breaking down. At eddies of size L
energy is injected, then energy is transmitted to smaller and smaller eddies and finally it is
dissipated in small eddies of scale where a viscosity plays a dominant role. A central role
in this scheme plays the mean rate of energy transfer per unit mass. The Navier-Stokes
equation, which describes the evolution of the velocity field v of the fluid, is:
r
r r
v r r
1
+ v v = p + 2 v + f
t
(1.2)
where is the mass density, p is the pressure and the dynamic viscosity. The term
r r
v v is the nonlinear term and implies a breaking of bigger eddies into smaller eddies, the
r
term 2v is the viscous term and represents a dissipation of kinetic energy as internal
r
energy of fluid, while the term f represents the external forcing acting on the fluid.
Kolmogorov in 1941 studied (based on Richardsons cascade idea) fully developed

turbulence (turbulence which is free to develop without imposed constraints). Works of
Kolmogorov are mentioned in every standard textbook on turbulence or fluid mechanics.
Kolmogorov made assumption that turbulence should exhibit universal and isotropic
statistics for scales smaller than the integral scale L. Moreover, for scales larger than the
Kolmogorovs scale , the viscosity should play no dynamical role. There is a range of
length scales called the inertial range, in which the flow statistics are expected to be
universal, isotropic and independent of the viscosity. Since then extensive experimental and
numerical studies have attempted to describe the statistical behaviour of fully developed
turbulence. Landau was the first to point out that the Kolmogorovs theory (K41) could not
be true because he did not take into account intermittency. Landau stated that the energy
dissipation displays important fluctuations about its mean value. A consequence is that
Kolmogorovs theory must certainly be corrected in order to contemplate this intermittency
character. Taking note of Landaus suggestions, Kolmogorov and Obukhov introduced a
refined similarity hypothesis called log-normal model. In this version (K62) they assumed
that the energy dissipation is log-normally distributed.
After the refined similarity hypothesis, different types of intermittency models were
proposed to describe the turbulence cascade and particularly the behaviour of scaling
exponents (Frish 1995). The success of these models can be evaluated especially on the
basis of experiments and this is next reason why is so important to do experiments in this
field. However, there are no models that agree with all experiments, although each model
works quite well within a limited group of available data. For that reason it is not possible
generally recommend one model over the others. Some of the most popular models for
fully developed turbulence are:
She-Leveque model
p-model
-model or
random -model
Turbulent jet is a turbulent, coherent stream of material ejected from a nozzle into a
surrounding medium, or a nozzle designed to produce such a stream. Presented study refers
to a fluid jet particularly. The subject is the concern of engineers in calculating pipe flows,
jets or wakes, for examples.
The mechanics of the turbulent jets, although studied during the last decades, still is a
paradigm of flow behaviour, together with wakes and boundary layers and it is great
interest of researchers. In recent years, thanks to improved remote sensing we can observe
concentration on the environment, for example dilution and mixing of pollutants in water
bodies at many scales. For these reasons we need continue an investigation work in three
scientific fields: experimental, environmental and numerical to obtain a greater knowledge
of the existing research problems.
The turbulent wall jet configuration occurs often in many environmental and industrial
processes. The most popular applications come from the fields of aeronautics design, heating,
cooling, ventilation and environmental fluid dynamics.
There are many books precisely describing problems of turbulence and turbulent
phenomenon. We mention in Appendix 1 some more complete and used during the
evolution of this study.
1.1 State of the art

Apart from the well-known and basic information in the major books about turbulence, a
large amount of work and specific publications refer to this subject and in particular to the
structure of wall jets, both in laboratory conditions and in environmental relevant flows.
We will discuss here the wide range of specific publications used in this thesis, either from
experimental, technical aspects of measurement technique, in classical jet and plume theory
or in image processing or structure function and spectral techniques.
All the most important steps in turbulence of last century are summed up in the
publication of Lumley and Yaglom (2001). It is a brief, superficial survey of some very
personal points of view of the statistics and most important works of the last hundred years
in turbulence research, few clear conclusions are reported.
The most general treatment of multi-scale and fractal analysis of fluid turbulence is done
by, a well-known author in this area, the director of the Trieste Physics Institute:
Sreenivasan (1999). A few aspects of turbulence research in the last century are briefly
reviewed and a partial assessment is made of the present directions. There are two possible
scenarios. Our computing abilities may improve so much that any conceivable turbulent
problem can be computed away with adequate accuracy, so the problem disappears in the
face of this formidable weaponry. The other scenario which is common in physics is that a
particular special problem that is sufficiently realistic and close enough to turbulence will
be solved in detail and understood fully. There is well-developed body of knowledge in
turbulence that is generally self-consistent and useful for problem solving. However, there
are lingering uncertainties at almost all levels.
The phenomenology of small-scale turbulence is done by Sreenivasan and Antonia
(1997). Small-scale turbulence has been an area of especially active research in the recent
past and several useful research directions have been pursued. The authors selectively
review this work.
There are many possibilities of jets configurations from the basic ones to more
complicated. Here and in Appendix 1 we comment most relevant publications used in the
preparation of this thesis in the different research lines, as examples for wall jet discussion,
the paper by Craft and Launder (2001) explores, using different levels of turbulence
closure, the computed behaviour of the three-dimensional turbulent wall jet in order to
determine the cause of the remarkably high lateral rates of spread observed in experiments.
Their computations confirm that the strong lateral spreading arises from the creation of
streamwise vorticity, rather than from anisotropic diffusion. The driving vorticity source is
created by the anisotropy of the Reynolds stresses in the plane perpendicular to the jet axis
rather than to the bending of mean vortex lines. It may be observed too, since the threedimensional wall jet is an acutely sensitive flow for assessing turbulence models; it would
be desirable to establish definitive experimental or, possibly, les results for the fully
developed limit.
A more specific report on turbulence and mixing in geophysical flows was presented in
the book by Redondo and Linden (2000) Turbulent mixing in geophysical flows and in
Redondo and Linden (1998). In this report, the papers of workshop on Mixing in
Geophysical Flows in 1997 are summarized giving a state-of-the-art overview of present

research in geophysical turbulent mixing. The main topics discussed are stratified flows,
rotating stratified flows, gravity waves, instabilities and mixing, convection, experiments
and numerical simulations of geophysical flows and turbulent mixing. The mixing
processes are the key mechanisms for mass and momentum transfer in the oceans and
atmosphere and play a crucial role in determining the environmental conditions in which
mankind lives and operates. The approach almost universally adopted is to isolate one or
two particular processes and study these in detail using a combination of theoretical,
numerical, experimental and observational techniques.
More advanced data techniques and more sophisticated analysis leading to a better
understanding of the scale to scale transfer of energy may be obtained through the study of
the higher order structure functions and the intermittency. The problem of nonhomogeneous turbulence was investigated in this department (UPC), among others by
Mahjoub (2001). In this work a classification is proposed to determine the intermittency
and mixing ability. The variation of the structure functions and the scaling exponent in
decaying non-homogeneous turbulence produced by a grid and by a jet is measured with
sonic velocimetry, hot film and wire probes. In Mahjoub et al. (1998) the investigation
shows the advantage of using Extended Self Similarity (ESS) and also that in most cases
that 3 (absolute scaling exponent) is not one in complex non-equilibrium flows and
depends strongly on the separation distance, except near the source of turbulence at high
Re. The BDF model, after Babiano, Dubrulle and Frick (1995, 1997) was used to check
their experimental results and relate the values of the relative scaling exponents. The
different flows studied show different distributions of intermittency, which was defined in a
more general way. Mahjoub et al. (2001) presented in an environmental coastal flow a
statistical analysis of velocity structure functions for turbulent flows at 1m above the
bottom in a shallow (2m) bay in Denmark. High frequency (25 Hz) time series were
collected in different days. For the two times series studied, there was not clear inertial
range when the absolute scaling component had a scale-independent behaviour. The authors
also used the ESS concept to measure the relative scaling exponents. ADV measurements
of waves and turbulence captured the peaks in wind and current motions in the bay. The
difficulty in calculation of higher order structure functions in short time series lead us to
design specially quite long stable measurement periods of about 10 minutes (in some cases
up to 4 x 104 data points).
Intermittency was investigated by Anselmet et al (2001) in laboratory experiments with
turbulent flows, also Jou (1997) and Schertzer investigate multifractal cascade and
turbulence intermittency. The paper of Jou provides prelude to the consequences of
intermittency on the statistical properties of fully developed turbulence, mainly on scaling
laws for the different moments of velocity, energy distribution and diffusion behaviour. The
description of intermittency is carried out in the fractal model and in a more general
multifractal perspective. He emphasized that the energy transfer between eddies of different
scales strongly fluctuates from place to place, yielding intermittent bursts of turbulence.
The active regions occupied by the eddies do not fill the whole volume but only a subregion
of it, which in the simplest model may be characterized by a fractal dimension D 2.87.
The intermittency modifies the energy spectrum by attributing less energy to small scales in
comparison with that predicted by the standard Kolmogorov distribution; and it enhances
diffusion with respect to the usual Richardson law of non-intermittent turbulence. More
detailed analyses of intermittency must focus on the exponents of the scaling laws for the
different moments of the velocity. We strived to improve on the diagnostic methods in
order to study the structure of turbulence, in Appendix 1 a list of specific papers that show
the state of art of the present line of research is discussed, a recent update on environmental
turbulence may be found in the 31st issue of Nuovo Cimento C.
1.2 Thesis aims

Studies of the behaviour of relatively basic kind of jets (the free turbulent jet, wall jet,
buoyant jets and plumes) are an important basis for understanding more complex
configurations. Although the investigation of the above mentioned kinds of flow has been
done for many years aiming at the mean flow predictions, there are still some important
problems unsolved in the behaviour of the turbulent cascades and their structure because of
the past limits of measurements methods, Launder and Rodi (1983) or List (1982) .
We aim to understand the behaviour of turbulent jets in a deeper and more detailed way
incorporating the recent advances in non-homogeneous turbulence, structure function
analysis, multifractal techniques and extended self-similarity. Part of this thesis is based on
extending the measurements on turbulent structure on flows with similar wall-jet
configurations, but main aim is to get new results on the detail structure of the nonhomogeneous turbulent cascade processes and thus complement previous experiments,
which were mostly concern with mean structure and global fluxes.
One of the parts of this work includes multi-fractal analysis of SAR (Synthetic Aperture
Radar) and experimental jet (and plume) images. Scaling analysis allows us to investigate
the structure of ocean surface detected jets (SAR), to compare coastal and boundary effects
on the structure of river jets and to investigate the turbulent and fractal structure of non
homogeneous jets affected by different levels of turbulence upstream and downstream. An
innovative technique used to investigate the turbulent interactions within the inverse and
direct cascades near jets is to measure the spectral and fractal structure of the non
homogeneous jets and develop multi-fractal techniques which we believe will be useful for
environmental and industrial monitoring.
The present work is based on experimental and environmental jets configurations
showing similarity of the occurring phenomenon at different scales. Presented thesis has
been performed with experimental technique available at this time in the laboratory of Fluid
Dynamics of the UPC.
Experimental techniques develop very fast so we can use the new technology that will
increase our knowledge, even repeating some classical experiments under new light and
improved techniques. Moreover application of different experimental techniques affirm or
not a usefulness showing advantages and faults of each one and allows us to confirm the
previous results. Using on the same experiments more than one method of diagnostic
permits us to improve the understanding of the laboratory techniques; this is also usually an
important argument for research group. This thesis was performed in the laboratory
equipped, among others with Laser Induced Fluorescence (LIF), Particle Image
Velocimetry (PIV), Particle Tracking Velocimetry (PTV) and Acoustic Doppler
10
Velocimeter (ADV), these are very useful velocimetry methods measuring velocity of
fluids and used to solve fluid dynamics problems or to study fluid networks. Industrial and
process control applications or the creation of new kinds of fluid flow sensors are
advantages of these methods also.
It was also very useful for the development of this thesis the research period spent at the
Czech Technical University, being able to work at the Department of Technical
Mathematics.
Here we present the general and specific objectives of this thesis:
Understanding of the dynamics of non-homogenous turbulent motions (jets, plumes,

vortices)
Comparison of the experimental end environmental features of the jets (different

Reynolds numbers and scales)
Application and evaluation of the Acoustic Doppler Velocimeter in laboratory

experiments
Use of multifractal analysis of the experimental and environmental images of jets

and plumes
Statistical analysis of the experimental results in the turbulence phenomenon point

of view
Effect of the boundary layer in wall jet configuration
1.3 Thesis structure

The present thesis is divided into 8 chapters and 2 appendices. We start with a general
introduction to turbulence and describe the basic information about the thesis subject,
discussing the background of the thesis and mentioning previous works leading to
Appendix 1 on a description of the state of art and the most relevant bibliography.
Chapter 2 refers to all theory in order to familiarize a reader with the work subject. This
chapter is divided into different parts, explaining turbulence and turbulent flows theoretical
considerations, focusing special attention on non-homogenous turbulence, extended self
11
similarity, 2D turbulence, completing with information about various configurations of

turbulent jets and plumes.
Chapter 3 explains theory about Synthetic Aperture Radar (SAR), fractal analysis,
relationship between turbulence spectra and fractal dimension and description of the
different fractal methods.
Chapter 4 shows used experimental configurations with description of the applied
methods and techniques. Different laboratory configurations are explained.
In Chapter 5 experimental ADV results are presented for different configurations
including data stability and instrumental errors. Some conclusions are drawn.
Separate part (Chapter 6) of this thesis is related to spectral measurements, structure
function techniques and scale analysis results including final conclusions.
Chapter 7 contains SAR and experimental images fractal results for river jets, ocean
vortices and other sea surface structures.
In Chapter 8 we present overall discussion and conclusions and possible future works.
Additionally, the used bibliography is quoted in alphabetical order.
Two appendices are attached as supplementary information to give more detailed list of
existing works on the subject (Appendix 1) and utilized during the elaboration of this study
and description of BDF method (Appendix 2) to calculate intermittency.
12
CHAPTER 2
BACKGROUND AND THEORETICAL CONSIDERATIONS
2.1 Turbulence and turbulent flows

A general introduction on turbulence and turbulent flows was explained in Chapter 1.
Here we complete the specific and necessary theory about this thesis, more detailed theory
is available in any of the mentioned bibliography about turbulent jets and mixing.
As we mentioned before, it is very difficult to give a precise definition of turbulence.
There are many efforts to define it considering different aspects, as an example, proposed
Tennekes and Lumley (1972).
Some of the characteristics are: turbulent flows are always dissipative. Viscous shear
stresses perform work, which increases the internal energy of the fluid at the expense of
kinetic energy of the turbulence. Turbulence needs a continuous supply of energy to make
up for viscous loss. If energy is not supplied, turbulence decays rapidly. The major
difference between random waves and turbulence is that waves are non-dissipative (though
they are often dispersive), while turbulence is essentially dissipative.
Figure 2.1 shows some examples of turbulent flows occurring in real life.
13
Figure 2.1 Some examples of turbulent flows. Turbulence in the tip vortex from an airplane
wing (top-left) (source: Langley Research Center of the United States NASA). Turbulent
flow around an obstacle (top-right) (courtesy Wikipedia) and laminar and turbulent water
flow over the hull of a submarine (below) (source: Wikipedia).
Despite the importance and abundance of turbulent flows, the community of scientists
as Reynolds (1883) or Richardson (1922, 1929) has encountered many difficulties in
developing a satisfactory scientific general theory. One of the most important steps in our
understanding came in 1941 when Kolmogorov developed his theory (K41) (Kolmogorov
1941) about how the energy that is put into large turbulent motions cascades down to very
small scales where it is converted into heat by viscosity.
14
The cascade theory of energy proposed Kolmogorov in 1941. This energy cascade we
can observe in Figure 2.2 where, k is the wave number equal to 2/l, l is length scale, L is
the largest scale and U is velocity.
Figure 2.2 Energy spectra (a) and cascade of energy by Richardson and Kolmogorov (b).
The energy comes from big to small whirls and without any source of energy,
turbulence decay quickly.
Real turbulent flows such as geophysical flows are non-homogenous and non-isotropic.
In this work, we present the theories and assumptions related to both (with homogenous
turbulence) but more emphasis will lay on true-life cases.
2.2 Homogenous turbulence

Kolmogorov described his model for fully developed, locally homogenous and isotropic
turbulence using the concept of stationary and continuous energy cascade process. There
are many intermittency models prominent in such a simple turbulent regime based on
characterization of the random but homogenous nature of the energy dissipation field.
These models have limitations in real life, environmental and industrially relevant flows
where the turbulence is usually non-homogenous, non-isotropic and non-stationary.
15
In 1942 Kolmogorov introduced his theory (K41) for locally, homogenous, isotropic
and stationary turbulence using velocity structure functions. The velocity structure
functions of order p are defined in terms of the moments of velocity differences as:
r q
r r
q
S q (l ) = u x + l u ( x ) = (u l )
((
(2.1)
r
where <> stands for ensemble average and u is the velocity component parallel to l .
Kolmogorovs theory is based on the following similarity hypothesis:

r
For all distances l = l small compared with integral scale L, l <<L, the statistical
properties of the velocity differences ul for a distance l are uniquely determined by

the kinematic viscosity and the average rate of energy dissipation <> per unit
mass.
In the inertial range <<l<<L, when the distances l are large compared with the
scale , where the energy is dissipated, the kinematic viscosity should play no
dynamical role and the velocity differences ul are uniquely determined by the
quantity <>.
This model is based on the Richardson idea of energy cascade (Figure 2.3).
Figure 2.3 The energy cascade process following Richardsons idea.
16
In this model, the energy is transferred to small scales in steps. At eddies of size L
(scales are of the order of the flow width, contain most of energy and dominate the
transport of momentum, mass and heat) energy is injected, then energy is transmitted to
smaller and smaller eddies, until it is dissipated into heat at smallest eddies of size (small
scales responsible for most of the energy dissipation).
The essential hypothesis of this model is that for high Reynolds numbers Re and l
smaller than the integral scale L and larger than the Kolmogorov scale , the longitudinal
velocity structure functions satisfy the relation:
S q (l ) = C p ( l )
q/3
(2.2)
where Cp is universal constant.

There is an exact dynamical relation for the third order longitudinal velocity structure
function, which can be derived from the Navier-Stokes equations for homogenous and
isotropic turbulence. For incompressible turbulence, when u = 0 , then we have the
famous 4/5th law of Kolmogorov:
4
5
ul3 = l
(2.3)
The scaling relationship for structure functions in range between and L plays an
important role in experiments, for example in fixing the extent of the inertial range and in
estimating energy dissipation rate per unit mass <>, in turbulent flows:
isotropic
u
= 15
x
(2.4)
valid only for locally isotropic flows Hinze (1975).

It can be shown that the structure functions in homogenous, stationary and isotropic
turbulence which is in local equilibrium have a scaling behaviour:
17
S q (l ) l
(2.5)
where the scaling power q is usually called the scaling exponent of the structure function
of order q. For the Kolmogorov theory (K41):
q =
q
3
(2.6)
indicating that the scaling exponents of structure functions of order q are scale-independent
and universal quantities. It was then assumed that there is a dynamically determinate scale
which can be constructed just from the average rate of energy dissipation <> and the
kinematic viscosity as:

3
=

1/ 4
(2.7)
This length-scale which would represent the smallest eddy size not damped by
dissipation is called the Kolmogorov length-scale. Similarly we can rewrite Kolmogorovs
time and velocity scales defined as:
k = (
(2.8)
vk = (
(2.9)
1/ 2
1/ 4
Hence, the local Reynolds number with reference to the two scales and vk is equal to one:
vk
=1
(2.10)
Many experimental studies have been done to verify the relation for scaling exponent of
the structure function of order p predicted by Kolmogorov (1941), specially for the density
of turbulent energy per unit of mass at scale l for q = 2
18
E (l ) = C1 ' ( l )
2/3
(2.11)
and for its spectral equivalent
E (k ) = C2 '
2/3
k 5 / 3
(2.12)
Here, C1 and C2 in principle, are universal constants if the full Kolmogorov K41
hypothesis is met, l must be within the inertial subrange and k=2/l is the corresponding
wave number. Many works shows the approximately value of these constants, named also
after Kolmogorov.
This Kolmogorov relation basically predicts that (if Re is large enough) the energy
spectrum of fully developed homogenous turbulence is divided into three distinct wave
number regions (Figure 2.4):
1. The region of energy injection at largest scales

2. Inertial range where the energy is transmitted from large to small scales
3. Dissipative range, where the energy is dissipated by viscosity into heat from the
small scales, which are compared to the Kolmogorov length scale .
Scheme of energy spectrum as a function of the wave number scale is shown below
(Figure 2.4).
Figure 2.4 The shape of energy spectrum as a function of wave

number (modification of Redondo notes image).
19
Since the Kolmogorov theory K41 was published, extensive experimental and
numerical solutions attempted to confirm the validity of his -5/3 law, in the context of
three-dimensional homogenous turbulence. All these studies assumed that the power law 5/3 is observed in fully developed turbulence. The observed deviation in the k-5/3 scaling
law was invariably attributed to departures from the homogenous and stationary conditions
of the flow.
The experimental results show strong intermittent bursts instead of a steady behaviour
of the energy dissipation fluctuation, indicating that fluctuates strongly. These results
demonstrate this intermittency increases strikingly with increasing Reynolds number. This
phenomenon can also be shown in the probability distribution functions of the velocity
increments ul in the inertial range (Figure 2.6)
Figure 2.5 The probability distribution functions in the inertial range for different scales r
(PhD thesis of Herweijer 1995).
As it can be seen on this figure, the intermittency appears as a set of a few localized
spikes of very high activity leading to the non-Gaussian distribution of the largest but rarest
events.
Taking into account intermittency, Kolmogorov made previous suggestions more

quantitative. He introduced a refined similarity hypothesis relating the moments of the PDF
20
(Probability Distribution Functions) of the velocity increments ul to the moments of l.

Then, the corresponding prediction for the qth-order moment of the velocity increment ul
as function of the scale separation l is usually formulated as,
u lq ~ lq / 3 l q / 3 ~ l
(2.13)
Considering that there is a scale dependence of the dissipation as:
lq / 3 ~ l
q /3
(2.14)
now the scaling exponents are:
q = q / 3 + q / 3
(2.15)
where l is the locally defined energy dissipation per unit mass over a volume of size
r
r
l = l centred at space-position x , q / 3 is the scaling exponent of lq / 3 , <> refers to
r
averaging over all positions vectors x and q = 1,2 is the order of the statistical moment.
In a self-similar situation corresponding to a homogenous and isotropic turbulence

characterized by a scale uniform random dissipation field, 1 = 0 and the correction q / 3 for
q 3 in relation (2.15) is only induced by the intermittency correction. Then, this relation
guarantees also the basic results 3 = 1 for locally homogenous and isotropic turbulence.
Many experimental studies have been done to verify these relations predicted by
Kolmogorov. Here we present homogenous intermittency model called model
introduced by Frisch (1995). The idea behind this model comes from the Richardson
cascade, in which at each level of the cascade the energy of the large eddies L is uniformly
distributed over the other, eddies of size l as
ln = Ll n
21
(2.16)
where n = 0,1,2.. and 0 < l < 1.
We can define the energy per unit mass on scales ln as
l
En Pn vn2 = vn2 n
L
3 D
(2.17)
The parameter D is called a fractal dimension.

In this form Pn is the fraction of decreasing of the eddies, which has factor (0 < < 1),
defined as
l
Pn = n = n
L
(3 D )
(2.18)
where
3 D =
ln
ln l
(2.19)
vn is the typical velocity difference over a distance ln within an active eddy of size l and we
can define it as
1 1
3 3
n
l
vn l n
L
1
(3 D )
3
(2.20)
where n is the rate of energy transfer from n-eddies to (n+1)eddies in the inertial range
defined according to the Kolmogorov theory K41
n Pn
vn3

ln
Now, it is clear that the velocity field has the scaling exponent
22
(2.21)
h=
1 3 D
3
3
(2.22)
and the structure function of order p is written as
l
S q (ln ) = P v v n
L
q
n n
q
L
(2.23)
with
q =
q
q
+ (3 D )1
3
3
(2.24)
The energy spectrum is given as
E (k ) k
5 3 D
+
3
3
(2.25)
which is derived as a correction to the k-5/3 law of the Kolmogorov theory (K41).
More different intermittency models are explained in Mahjoub (2000).
2.3 Non-homogenous turbulence

In the previous sections we explained that in recent years many efforts have been made
to explain the intermittency phenomenon, particularly in homogenous flows. Some models
are mentioned in Chapter 1 of this thesis. In non-homogenous flows which are more
complex and have more practical interest, less attention has been given to the study of nonlocal dynamics which seems separated from intermittency and also seems to play an
important role in non-homogenous turbulence.
There are many models concerning the energy cascade scale to scale processes and the
dissipation rate, see Frisch (1995) for an account.
23
Kolmogorov and Obukhov in 1962 introduced the refined version of Kolmogorovs

similarity hypothesis taking into account intermittency. They assumed that for locally
homogenous and isotropic turbulence, the energy dissipation field strongly fluctuates in
both space and time. On the other hand, the average amplitude of the dissipation random
field scales quite uniformly in both space and time. In contrast, when the turbulence is nonhomogenous and non-isotropic, the dissipation random field is non-uniform in scale. This
means that the fluctuations and the amplitude of the variance of the energy transfer are
scale dependent quantities. In this case, the correction q/3 of relation (2.15) is associated
with both the intermittency phenomenon linked to the rarest events and the anomalous
dependence as a function of the length scale of transfer properties in the energy cascade
scales. Therefore, this relation is not strictly valid, because q are anomalous and scale
dependent.
2.4 Extended Self Similarity (ESS)

The Extended Self Similarity (ESS) is technique which is a key way to analyze
homogenous and non- homogenous flows. It is a property of velocity structure functions of
homogenous and non-homogenous turbulence. Instead of obtaining scaling exponents in
the usual way by plotting structure functions of the absolute velocity increments
ul
against l, we plot them against the third-order structure function of the absolute velocity
increment ul
and then we have:
ul
~ u l
3 q / 3
(2.26)
where q/3 is a relative scaling exponent and q is defined by
ul
24
(2.27)
where, q is now the absolute scaling exponent and may be different from q for odd values
of q because absolute values of velocity increments are used.
Some limitations of ESS were mentioned by different authors. They pointed out that the
ESS does not seem to work when the shear is strong, such as in the shear behind a cylinder
and in boundary layer turbulence. In contrast, some authors found that the ESS also works
well in these situations even. This suggests that the ESS may be also a specific and
convenient tool to analyze non-homogenous turbulence.
The use of
ul
instead of ul3 in ESS may be physically explained because it
refers to the scale by scale absolute balance of transferred energy at a given scale l and
includes both energy transfers from larger to smaller scales (normal cascade) and the
anomalous energy transfers from smaller to larger scales (inverse cascade). This fact
3
suggests that the ESS relation must be expressed in term of ul .
Some intermittency models for non-homogenous case are explained in Mahjoub (2000)
special attention focus on BDF model.
2.5 2D Turbulence
In the non-homogenous case, the energy spectrum can be steeper than k-5/3 and saturates
to k-3. This behaviour is illustrated in Figure 2.6, which shows clearly the transition from
homogenous and local dynamics k-5/3 to non-local and non-homogenous dynamics k-3. For
example, Babiano demonstrated in his numerical study for 2D non-homogenous turbulence,
that the spectral slope increases with degree of non-homogeneity and can reach up to k-3.
25
Figure 2.6 The evolution of energy spectrum E(k) from non-local to local dynamics
(Mahjoub 2000).
On the other hand, the scaling exponent of the second order velocity structure function
2 shows an important deviation from Kolmogorovs 2/3 prediction.

Kraichnan (1975) set forth, called the energy-enstrophy model, which was based on
spectral truncation of the underlying fluid dynamics equations, leading to the twodimensional turbulent energy-enstrophy cascades as an extension of Kolmogorovs K41
and K62 theories. He predicted an inverse cascade of energy (Figure 2.7) in twodimensional Navier Stokes fluid turbulence and proposed an inertial range with a k-5/3
power-law energy spectrum, just as in three-dimensional turbulence, but with a flux of
energy from small to large scales rather than the reverse. It is one of the most important
phenomena in fluid dynamics occurring, for example, in the atmosphere or ocean
In agreement with Kraichnan (1967) the energy and enstrophy conservation make that
energy actually flow to larger scales and this is a basic difference between 2D and 3D
turbulence (where energy flows toward small scales in a direct cascade). The enstrophy can
be interpreted as the quantity directly related to the kinetic energy in the flow model that
corresponds to dissipation effects in the fluid. It is defined as:
26
Z=
1 r
(2.28)
r
r
where = u is vorticity.
Figure 2.7 The inverse energy cascade characterized by k (5 3 ) in the power spectrum.
It is impossible to find in nature strictly 2D case. It exists only for numerical

simulations or for theoretical considerations.
Theoretically it is possible to resolve The Navier-Stokes equations from the 2D point of

view, where the flow velocity has only two components. The equation of the vorticity has
form,

2
+ u + v = 2 + 2
t
y
y
x
x
(2.29)
where, simplified z = .
The relation between mean enstrophy and energy spectrum is
Z (t ) = k 2 E (k , t )dk = (k , t )dk
where (k,t) is the enstrophy spectrum.
27
(2.30)
We consider a turbulent flow stationary, where the energy is injected by force f(k)
around ki, then the kinetic energy injection
f (k )dk
(2.31)
and the injection of enstrophy
= k 2 f (k )dk ki2
(2.32)
Kraichnan propose that if the Reynolds number is sufficiently high, there is a wave
number range ki < k << kd within the turbulence dynamic is a function of and k. Through
dimensional analysis we obtain the relation of the energy spectrum in this range as
E (k ) ~ 2 / 3k 3
(2.33)
The wave number kd correspond to the scale where the enstrophy is spread by the
molecular viscosity effect.
1/ 6

kd = 3

(2.34)
2.6 Turbulent jets

The present thesis focuses on turbulent jets and plumes and their turbulent structure and
mixing properties so it is indispensable to explain the classical basic theory about it. We
will concentrate first on turbulent jets in general and then explain different configurations,
such as the plane turbulent free jet, the circular turbulent jet and finally the turbulent wall
jets showing for most of the cases equations of motion, the integral momentum equation,
the integral energy equation and integral moment of momentum equation.
28
Turbulent jets are fluid flows produced by a pressure drop through an orifice. Their
mechanics, although studied for over fifty years, has recently received research attention that
has resulted in a much-improved understanding of the process by which they entrain
surrounding fluid.
Examples of turbulent jets are presented in Figure 2.8.
Figure 2.8 Different examples of turbulent jets. Vorticity magnitude at Reynolds number
3960 (source: Laboratory for Aero & Hydrodynamics Delft University of Technology)
(left). A simulation of a fully turbulent jet flow by David Glase (Purdue University) (right).
The greyscale contours represent mixture fraction. The lower half of the plot shows
velocity vectors, colored and scaled by their relative magnitude.
There is now overwhelming evidence that the initial growth of turbulence is a direct
consequence of large-scale motions generated at the jet boundaries. These large-scale motions
are primarily responsible for jet noise production and the initial entrainment of ambient fluid.
The basic sequence for axisymmetric jets seems to be as follows: in the immediate
neighbourhood of the orifice, the high-speed jet flow causes a laminar shear layer to be
produced. The shear layer is unstable and grows very rapidly; forming ring vortices that carry
turbulent jet fluid into the irrotational ambient fluid and irrotational ambient fluid into the jet.
It is clear that the production of vortices is the key element in initial jet dilution. Each vortex
wraps ambient fluid about itself, then, as the vortices pair, the fusion process mixes the
29
ambient and jet fluid. Circumferential instability, and possible interaction with helical modes,
leads to the apparent eventual self-destruction of the large-scale structures and this, in turn,
generates the subsequent small-scale turbulent mixing.
Different configurations of turbulent jets are shown below (apart from the further, more
detailed cases).
The radial turbulent jet is shown in Figure 2.9.
Figure 2.9 Definition sketch of radial turbulent jet (Rajaratnam 1976).
The compound jet is presented in Figure 2.10.
Figure 2.10 Definition sketch for compound jet (Rajaratnam 1976).

The confined jet is demonstrated in Figure 2.11.
30
Figure 2.11 Definition sketch for confined jet axisymmetric case (Rajaratnam 1976).
The jet in cross-flow is shown in Figure 2.12
Figure 2.12 Definition sketch of circular jet in cross-flow (Rajaratnam 1976).

The radial wall jet is presented in Figure 2.13.
Figure 2.13 Radial wall jet (Rajaratnam 1976).

The plane compound wall jet is demonstrated in Figure 2.14.
31
Figure 2.14 Definition sketch of a plane compound wall jet (Rajaratnam 1976).
The average equations of motion of several cases will be discussed in next sections.
2.7 Buoyant plumes

A plume (in hydrodynamics) is a column of one fluid moving through another.
Turbulent plumes are fluid motions whose primary source of kinetic energy and momentum
flux is body forces derived from density inhomogeneities. Several effects control the motion
of the fluid, including momentum, diffusion, and buoyancy (for density-driven flows).
When momentum effects are more important than density differences and buoyancy effects,
the plume is usually described as a jet. Usually, as a plume moves away from its source, it
widens because of entrainment of the surrounding fluid at its edges. Plume shapes can be
influenced by flow in the ambient fluid (for example, if local wind blowing in the same
direction as the plume results in a co-flowing jet). This usually causes a plume which has
initially been 'buoyancy-dominated' to become 'momentum-dominated'. This transition is
usually predicted by a dimensionless number called the Richardson number. The
Richardson number is the dimensionless number that expresses the ratio of potential to
kinetic energy
Ri =
gh
u2
(2.35)
where g is the acceleration due to gravity, h a representative vertical length scale and u a
representative speed.
A further phenomenon of importance is whether a plume has laminar flow or turbulent
flow. Usually there is a transition from laminar to turbulent as the plume moves away from
32
its source. This phenomenon can be clearly seen in the rising column of smoke from a
cigarette.
When high accuracy is required, computational fluid dynamics (CFD) can be employed
to simulate plumes, but the results can be sensitive to the turbulence model chosen.
Plumes have not been studied in the same detail as jets but nevertheless there have been
some recent gains in the understanding of their mechanics because they are of considerable
importance in the dispersion of air pollution. Some examples of different plumes are shown
on Figure 2.15
Figure 2.15 Different examples of plumes. Plume of the Space Shuttle Atlantis after launch
(left) (source: NASA APOD), industrial air pollution plumes (centre) (source: Wikipedia)
and large fire induced convection plume (right) (source: Wikipedia).
Basic to the understanding of all free turbulent flows is the process of entrainment
(Turner 1973) or mixing of outside fluid into the plume. It is observed that (like jets)
turbulent plumes have a sharp boundary separating nearly uniform turbulent buoyant fluid
from the surroundings. This boundary is indented by large eddies and the mixing process
takes place in two stages, the engulfing of external fluid by the large eddies, followed by
rapid smaller scale mixing across the central core. The vertical velocity and the turbulence
measured at a fixed point of the axis have an intermittent character. Though an
instantaneous profile across a plume is sharp-edged, the time averaged profile of velocity is
mother and can be well fitted by Gaussian curve. A detailed theory of the mechanism of
entrainment has been given by Townsend (1970).
33
As Batchelor (1954) pointed out, the increasing vertical flow in a plume also implies
that there is a mean inflow velocity across the boundary which varies as z-1/3. That is, the
linear spread of radius with height implies that the mean inflow velocity across the edge of
the plume is proportional to the local mean upward velocity.
1
r
w = F0 3 z 3 f1
b
2
3
g ' = F0 z
5
3
r
f2
b
b = z
(2.36)
(2.37)
(2.38)
where w is the vertical velocity,
F0 = 2 wg ' rdr
(2.39)
g ' = g ' / 0
(2.40)
g is the acceleration due to gravity, the difference of density, 0 constant
environment density, r is the radial distance from a vertical line above the source, z
height, f1 and f2 are functions explicated later, b is the radial length scale and is a
constant to be defined for particular profiles.
The actual form of functions f1 and f2 must be obtained using either more detailed
theories (with some questionable assumptions) or directly by experiment. The existing
solutions show that there is a flow from the environment into the turbulent plume, since the
mass flux is clearly increasing with height.
When the simplest entrainment assumption is made, so that the inflow velocity is taken
to be some fraction of the upward velocity, the equations of conservation of mass,
momentum and buoyancy can be reduced to the form
( )
d b2 w
= 2bw
dz
34
(2.41)
d (b 2 w 2 )
= b2 g '
dz
d b2 w g '
= b 2 w N 2 ( z )
dz
(2.42)
(2.43)
The velocity w and width b are defined by integrating the mass and momentum fluxes
across the plume:
w b = 2 w(r )rdr ,
w b = 2 w2 (r )rdr
2 2
(2.44)
N2 is the square of the local buoyancy frequency and equal to:

N 2 = ( g / 1 )(d 0 / dz )
(2.45)
1 is some standard density in the environment and 0 environment density.

When the environment is of uniform density, N = 0, so the third equation is a formal
statement of the fact that the buoyancy flux is constant. The multiplying constants such as
are now in terms of the entrainment constant :
6
b = z
5
(2.46)
1
5 9
3
w=
F z 3
6 10
(2.47)
5F 9
3
g'=
F z 3
6 10
(2.48)
Other, more general kinds of plumes can be treated by the same method.
The numerical value to be chosen for cannot be obtained theoretically and it must be
taken from laboratory experiments.
35
Lets calculate, as an example, two typical length scales for thermal. To gain insight in
the determining parameters, a dimensional analysis is executed, following the approach of
Fischer (1979).
First the volume flux Q, the momentum flux M and the buoyancy flux B are defined as:
Q=
D2w
(2.49)
M = Qw
(2.50)
B = Qg 0 '
(2.51)
in which w is the initial velocity and g0 the reduced gravity, defined as g 0 ' =
g.
Here, =
From these expressions, typical length scales can be derived:
LQ =
Q
M 1/ 2
LMB =
in which LQ
M 3/ 4
B1 / 2
(2.52)
(2.53)
represents the length over which the geometry of the injection nozzle
influences the propagation of the thermal and for lengths greater than LMB, the flow is
characterized by the buoyant forcing. For a thermal source buoyancy conservation
depends upon the temperature distribution in the environment and the temperature
dependence of the volumetric coefficient of thermal expansion. For a mass source, the
ambient fluid density must be constant. Thermals are discussed in detail by Turner (1973).
We will look now at the simple dimensional description of plumes using empirical
parameters to scale the volume (or mass) flux Q, the momentum flux M and the buoyancy
flux B. In a 3D flow the respective dimensions are:
[Q] = L3T-1, [M] = L4T-2 and [B] = L4T-3
36
but in a two-dimensional flow we have

[Q] = L2T-1, [M] = L3T-2 and [B] = L3T-3
We can scale the advance of a plume due to momentum only
z = Cm M 1 / 3t 2 / 3
(2.54)
z = Cb B1 / 3t
(2.55)
or due to buoyancy
It is easy to see that for distances greater than the jet length
Lm = Cl MB 2 / 3
(2.56)
the plume will behave as a pure buoyancy driven plume. In this example M is kept low. Cm,
Cb and Cl are constants.
Time-dependent plumes and jets with decreasing source strengths are investigated by
Scase et al. (2006) as an example and more examples are mentioned in Appendix 1.
As we told before, some part of this work is related to the multi-phase jet/plumes. As an
example, two-phase plume is shown on Figure 2.16
37
Figure 2.16 Denition sketch for a two-phase plume in a crossow. Socolofsky and Adams
(2002).
It illustrates two effects of separation for a bubble plume in a crossow. First, at some
height above the source, the crossow separates the entrained uid from the rising bubbles.
This occurs as the rise velocity of the entrained uid decreases with height allowing the
crossow to have an increasing effect.
Second, as observed, crossows transport bubbles having different slip velocities
(terminal rise velocities) differentially downstream; this is called fractionation.
Fractionation distributes the buoyancy over an increasing horizontal area with height. We
show here that the crossow separates the entrained uid from the bubbles at a discrete
height, hS, below which the bubbles and entrained uid behave like a mixed, coherent
plume and fractionation is negligible, and above which the separated uid may be treated as
a buoyant momentum jet.
2.8 Plane turbulent free jet
It is for example a jet of water coming from a plane nozzle of large length into a large
body of water or a jet of air into a large expanse of air. If we use suitable flow visualisation
techniques, we will find that the jet mixes violently with the surrounding fluid creating
turbulence and the jet itself grows thicker. Figure 2.17 shows a schematic representation of the
jet configuration discussed above, which is known as the plane turbulent free jet.
38
Figure 2.17 Definition sketch of plane turbulent free jets.
Experimental observations on the mean turbulent velocity field indicate that in the axial
direction of the jet, one could divide the jet flow into two distinct regions.
In the first region, close to the nozzle, known commonly as the flow development region,
as the turbulence penetrates inwards towards the axis or centreline of the jet, there is a wedgelike region of undiminished mean velocity, equal to U0. This wedge is known as the potential
core and is surrounded by a mixing layer on top and bottom. In the second region, known as
the fully developed flow region, the turbulence has penetrated to the axis and as a result, the
potential core has disappeared.
In the fully developed flow region, the transverse distribution of the mean velocity in the
x-direction, i.e. the variation of u with y at different sections, has the same geometrical shape.
At every section, u decreases continuously from a maximum value of um on the axis to a zero
value at some distance from the axis. Let us now plot u/um against y/b (Figure 2.18). The free
jets have top-hat velocity distributions in the potential core.
39
0.8
U/Um
0.6
0.4
0.2
0
-3
-2
-1
0
y/b
Figure 2.18 Velocity distribution for plane turbulent free jets.
Because a free jet entrains fluid from both sides, it spreads faster, and, therefore, it
centreline velocity decays faster than that for the wall jet in the flow development region near
the nozzle exit.
EQUATIONS OF MOTION
In this section we will show the equations of motion for the plane turbulent free jet. The
Reynolds equations in the Cartesian system are written as:
2 u 2 u 2 u u ' 2 u ' v' u ' w'

u
u
u
u
1 p
(2.57)
+u
+v
+w
=
+ v 2 + 2 + 2
+
+
t
x
y
z
x x
y
z
y
z x
2 v 2 v 2 v u ' v' v' 2 v' w'
v
v
v
v
1 p
(2.58)
+u +v +w =
+ v 2 + 2 + 2
+
+
t
x
y
z
y x
y
z
y
z x
and:
2 w 2 w 2 w u ' w' v' w' w' 2
w
w
w
w
1 p
+
+
+u
+v
+w
=
+ v 2 + 2 + 2
t
x
y
z
z x
y
z
y
z x
The continuity equation is written as:
40
(2.59)
u v w
+
+
=0
x y z
(2.60)
where the x-axis defines the axial direction of the jet, the y-axis is perpendicular to the x-axis
and is in the direction of the height of the nozzle and the z-axis is third axis of the co-ordinate
system; u, v and w and u, v and w are the turbulent mean and fluctuating velocities in the x,y- and z-coordinate directions, p is the mean pressure at any point, is the kinematic viscosity, is
the mass density of the fluid and t is the time variable.
Because the mean flow is two-dimensional, w = 0, / z of any mean quantity is zero;
u ' w' = 0 ; v' w' = 0 and since the mean flow is steady u / t = 0 and v / t = 0 . Further,
since the transverse extent of the flow is small, u is generally much larger than v in a large
portion of the jet and velocity and stress gradients in the y-direction are much larger than those
in the x-direction. With these considerations, the equations of motion could be shown to
reduce to the form:
u
u
2 u u ' v' u ' 2
1 p
+v
=
+v 2
x
y
x
y
x
y
0=
(2.61)
1 p v' 2
y y
(2.62)
u v
+
=0
x y
(2.63)
Integrating with respect to y from y to a point located outside the jet, differentiating and
substituting, we get:
u
u
1 dp
2 u u ' v' 2
+v
=
+v 2
u ' v' 2
x
y
dx
y
x
y
(2.64)
where p is pressure outside the jet.

The last term in the above equation is smaller than the other terms and could be dropped.
Hence we obtain the reduced equations of motion as:
41
u
u
1 dp
2u u ' v '
+v
=
+v 2
x
y
dx
y
y
(2.65)
and:
u v
+
=0
x y
(2.66)
where p is simply written as p for convenience. We could rewrite the last two terms as:
1 u 1
1
+
u ' v' =
( + )
y y y
y 1 t
(2.67)
where 1 and t are, respectively, the laminar and turbulent shear stresses and is the
coefficient of dynamic viscosity. In free turbulent flows, due to the absence of solid
boundaries, t is much larger than 1 and hence it is reason able to neglect 1 and rewrite:
u
u
u
1 dp 1 t
+v
=
+
dx y
x
y
(2.68)
Further, because in a large number of practical problems the pressure gradient in the axial
direction is negligibly small and also to study the jet under relatively simpler conditions, let us
set dp/dx=0. Then:
u
u
u 1 t
+v
=
x
y y
u v
+
=0
x y
(2.69)
(2.70)
which are the well-known equations of motion for the plane turbulent free jet with a zero
pressure gradient in the axial direction.
The integral momentum equation may be deduced by multiplying now by and
integrating from y = 0 to y = :
u
0
u
u
dy + v dy = dy
x
y
y
0
0
Let us now consider the different terms of the above equation:
42
(2.71)
u
0
v
0
u
1
1 d
dy =
u 2 dy =
u 2 dy
x
2 0 x
2 dx 0
u
dy =
y
uv 0 u
v
dy =
y
u
dy
0 x
(2.72)
(2.73)
since for y = 0; u = um, v = 0 and for y ; u = 0, v = ve

where ve is a finite quantity known as the entrainment velocity.
Then:
d
u 2 dy = 0
dx 0
(2.74)
The equation tells us that the rate of change of the momentum flux in the x-direction is
zero; that is the moment flux in the x-direction is conserved (or preserved).
If the plane jet is issuing from an orifice of height 2b0 with a uniform velocity of U0, for
2
every unit length of the orifice, the momentum flux M 0 = 2 b0U 0 . If we imagine that this
momentum flux is emanating from a (fictitious) line source, located at the so-called virtual
origin from which x is measured, we have:
2 u 2 dy = M 0
(2.75)
The momentum flux M0 is an important physical quantity controlling the behaviour of the
plane jet. It effectively replaces individual values of b0 and U0.
Using the integral momentum equation, we will now develop a method of predicting the
variation of the velocity and length scales. For the plane turbulent jet, we have seen that the
velocity distribution in the fully developed region is similar. That is:
u / u m = f ( )
where = y / b
Let us assume simple forms for um and b as:
43
(2.76)
um x p
(2.77)
b xq
(2.78)
where p and q are the unknown exponents to be evaluated (do not confuse with pressure and
structure function order). Substituting last equations, we get:
d
u m 2 bf 2 d = 0
dx 0
(2.79)
where f2 stands for f 2 ( ) . Rewriting:
d
bu m 2 f 2 d = 0
dx
0
(2.80)
d
2
bu m = 0
dx
(2.81)
where
d is a constant, then:
We can say that bum2 is independent of x. That is:
bu m x 0
(2.82)
THE INTEGRAL ENERGY EQUATION

Let us multiply the first equation of motion by u and integrate it with respect to y from y
= 0 to y = . We get:
2
u
0
u
u
dy + uv dy = u dy
x
y
y
0
0
44
(2.83)
Let E = u 2 / 2 , the kinetic energy per unit volume.
2
u
0
uv
0
u
E
dy = u
dy
x
x
0
u
E
dy = v
dy
y
y
0
(2.84)
E
u
u
E
DE
0 u x dy + 0 uv y dy = 0 u x + v y dy = 0 Dt dy
where D/Dt stands for the particle derivative and DE/Dt is the total rate of change of the
kinetic energy.
0 u y dy = u
u
u

dy =
dy
y
y
0
0
(2.85)
u / y is the rate of production of turbulence, by the Reynolds shear stress working on the
mean velocity gradient. We have:
DE
u
0 Dt dy = 0 y dy
(2.86)
which says that the rate of decrease of the kinetic energy is equal to the rate at which
turbulence is produced.
For our present purposes, we will rewrite the above equation in a slightly different form.
u 2
E
u 2
u
dy
=
u
dy
=

0 x
0 x 2
x 2
E
u 2
u 2
v
dy
=
v
dy
=
v
0 y
0 y 2
2
Adding the above two expressions:
45
u 2 u
u
dy
2 x
u 2 u
2 x
dy
(2.87)
u
0
E
E
u 2
dx + v
dy =
x
y
x 2
0
0
d u 2
u dy =
udy
dx 0 2
(2.88)
We could now write:
d u 2
u
udy =
dy
dx 0 2
y
0
(2.89)
We see that the rate of decrease of the kinetic energy flux is equal to the rate at which
turbulence is produced. Using our earlier assumptions we could rewrite:
d
3
2
bu m f 3 d = u m gu m f ' d
dx 0 2
0
d
3 1
3
bu m f 3 d = u m gf ' d
dx
2
0
0
Letting:
1 3
0 2 f d = F1 and
(2.90)
(2.91)
gf ' d = F
Where F1 and F2 are constants, we could rewrite:
d
3
bu m
F
dx
= 2
3
F1
um
(2.92)
As before, the integral moment of momentum equation may be deduced as follows.

Let us multiply the equations of motion by y and integrate from y = 0 to y = . We obtain
then:
u
u
1
0 uy x dy + 0 vy y dy = 0 y y dy
46
(2.93)
uy
0
u
1 d
dy =
yu 2 dy
x
2 dx 0
(2.94)
u
1 d
2
0 vy y dy = 2 dx 0 yu dy 0 uvdy
Adding:
uy
0
u
u
d
dy + vy dy =
yu 2 dy uvdy
x
y
dx 0
0
0
y
0
(2.95)
1
dy = dy
0
y
(2.96)
The integral moment of momentum equation becomes:
d
u 2 ydy = uvdy dy
dx 0
0
0
(2.97)
v = u m b' f fd u m ' b fd
0
0
(2.98)
Substituting
where b=db/dx, f ( ) = u / u m , u m ' = du m / dx

Now:
uvdy = u f [u
m
where J 1 ( ) = f fd
b' J1 ( ) um ' bJ 2 ( )]bd
(2.99)
J 2 ( ) = fd
and
We may write:
d
u m 2 b 2 f 2 d = u m 2 bb' J 1 ( )d u m u m ' b 2 J 2 ( )d u m 2 b gd
dx
0
0
0
0
Letting:
47
(2.100)
2
f d = F3 , J 1 ( )d = F4 , J 2 ( )d = F5 , and gd = F6
where F3, F4, F5 and F6 are constants, we have:
F4
d
2
2
2
u m b 2 F5 u m bb'+ F6 u m u m ' b 2 + F7 u m b = 0
dx
(2.101)
2.9 Circular turbulent jet

Most jets are generated by flow outlet from a pipe as a point source, like the plume from a
volcano in geophysics, this configuration is described as a circular jet of diameter d emerging
from a nozzle with a uniform velocity of U0 into a large stagnant mass of the same fluid. If we
observe the jet, we would find that the size of the jet increases steadily as it travels away from
the nozzle as shown in Figure 2.19.
Figure 2.19 Definition sketch of circular turbulent jets.
If we use any time-mean velocity measuring device and measure the variation of the axial
velocity u with the radial distance r at different x-sections, we will obtain an interesting
description of the growth of the jet. We will find that up to section 1-1 (see Figure 2.19), there
is a core of flow with undiminished velocity equal to U0. At section 1-1, the turbulence
generated on the boundaries penetrates to the axis and the mean velocity on the axis begins to
decay with x. The core of fluid with the undiminished velocity is in the form of a cone and is
known as the potential cone or more familiarly as the potential core. This region from the
48
nozzle to the end of the potential core is known as the flow development region whereas the
region away from the end of the potential core is known as the zone of fully established flow.
In the region of fully developed flow we find that, at any section, u decreases continuously
from a maximum value of um on the axis to zero for large values of r. On figure 2.20, u/um is
plotted against an undimensionless distance r/b, where b is the value of r and U = Um/2 in the
case of the plane jet). This is done for Trupels data.
1
0.8
U/Um
0.6
0.4
0.2
0
0
0.4
0.8
1.2
1.6
=r/b
Figure 2.20 Velocity distribution in circular jets Trupels observations.
It is interesting to find that the velocity profiles are indeed similar.
In order to deduce the equations of motion, we start with the Reynolds equations in a
cylindrical system
(r , , z )
and apply the boundary-layer approximations since the jet
occupies only a small width in the transverse direction. The Reynolds equations in the
cylindrical system for steady axisymmetric flow may be written as:
2
2v
v r
v r v
1 p
1 v r vr 2 v r
+ vz
=
vr
+ v 2r +
+
r
z
r
r r
r r r 2 z 2
2
2
v ' 2 + v ' v ' + v r ' v '

r
z
r r
z
r
r
49
(2.102)
vr
2v
v
v
vv
1 v v 2 v
+ v z + r = v 2 +
2 +
r
r r
r
z
r
z 2
r
v 'v '
v r ' v ' + v ' v z ' + 2 r
r
r
z
2 v z 1 v z 2 v z
v z
v z
1 p
vr
+ vz
=
+ v
+
+ 2
r
z
z r 2 r r
z
and :
(2.103)

v 'v '
v r ' v z ' + v z ' 2 + r z (2.104)

z
r
r
rv r + rv z = 0
r
z
where v r , v , v z are the time mean velocities in the r, and z directions and v r ' , v ' and v z '
are the respective velocity fluctuations. For the circular jets without swirl, v = 0 and all the
terms containing v and its derivatives disappear from the above equations. Further, v z >>
v r ; gradients in the radial directions are much larger than those in the axial direction. Viscous
stresses could be assumed to be much smaller than the corresponding turbulent shear stresses
provided that the nozzle Reynolds number is greater than a few thousand. Further, turbulent
normal stresses are approximately equal in the radial and peripheral directions. With these
stipulations, the equations of motion become:
1 p
= vr '2
r
r
vr
(2.105)
v z
v
v 'v '
1 p
+ vz z =
v r ' v z ' + r z + v z ' 2
r
z
z r
z
r
rv r + rv z = 0
r
z
(2.106)
(2.107)
Integrating, substituting and simplifying in a manner similar to that of the plane jet, we have:
vr
v z
v
1 dp 1
+ vz z =
rv r ' v z '
r
z
dz r r
50
(2.108)
where p now is the pressure outside the jet. For convenience, let us now call the axial distance
x and let the velocity components in the axial and radial directions be u and v respectively and
let v r ' v z ' = . With these substitutions, the equations of motion become:
u
u
1 dp 1 r
+v
=
+
x
r
dx r r
(2.109)
ru + rv = 0
x
r
(2.110)
Since in most of the practical cases dp/dx is almost zero, let us consider zero-pressure
gradient flows. As a result gets further simplified to:
u
u 1 1 r
+v
=
x
r r r
(2.111)
These equations are the simplified equations of motion for the circular jet.
In the same way as for other configurations, the integral momentum equation is
described for a circular jet diffusing into a stagnant environment of the same fluid with
zero-pressure gradient; it is easy to see that the momentum flux of the jet in the axial
direction is preserved. We will now develop this criterion in an elegant manner.
Let us multiply last point equations by r and integrate with respect to r from r = 0 to
r = . We get:
ur
0
u
u
r
dr + vr dr =
dr
x
r
r
0
0
ur
0
vr
0
(2.112)
u
1 d
dr =
2rdru 2
x
4 dx 0
u
1 d
dr =
2rdru 2
r
4 dx 0
r
dr = 0
r
51
(2.113)
Hence:
d
2rdru 2 = 0
dx 0
The last equation states that the rate of change of the axial momentum flux in the axial
direction is zero or that the momentum flux in the axial direction is conserved.
Let us now develop equations for the velocity and length scales. Let:
u / u m = f (r / b ) = f ( )
(2.114)
um x p
(2.115)
and: b x q (do not confuse with pressure and structure function order).
With these substitutions, we have:
d
u m2 b 2 2f 2 d = 0
dx
0
(2.116)
u m2 b 2 x 0
(2.117)
and:
We need one more equation to evaluate these exponents. We will develop this second
equation firstly by considering the similarity of the equations of motion, secondly from the
integral energy equation and thirdly using the entrainment hypothesis.
For a circular turbulent jet, the integral energy equation is derived as follows.
We have:
u
u
r
0 ru x dr + 0 ruv r dr = 0 u r dr
2
52
(2.118)
2
ru
0
u 2
u
1
dr =
u
2
rdr
x
2 0 x
2
1
u
0 ruv r dr = 2
u 2 u
2 x
(2.119)
2rdr
Adding, we get:
2
ru
0
u 2
u
u
1 d
dr + ruv
dr =
2
r
dru
x
r
2 dx 0
2
0
(2.120)
The integrant in the above equation could be recognised as the kinetic-energy flux through
an element ring area. Let us now consider the remaining term:
u
0
r
1
u
dr =
2rdr
r
2
r
(2.121)
The integral represents the rate of production of turbulence so now we have the form:
u 2
d
u
2
rdru
= 2rdr
dx 0
2
r
0
(2.122)
This equation states that the rate at which the kinetic-energy flux decreases is equal to the
rate at which turbulence is produced. We have: u / u m = f ( ) and / u m2 = g ( )
Substituting these relations and simplifying:
d 3 2 1 3
b2
u m b f d = u m2 u m
gf ' d
dx
2
b 0
0
(2.123)
1
Letting: F3 = f 3 d and F4 = gf d becomes:
2
0
0
F
d 3 2
u m b / u m3 b = 4 = F5
dx
F3
F3 , F4 and F5 are constants.
53
(2.124)
Axisymmetric or axisymmetrical are adjectives which refer to an object having

cylindrical symmetry or axisymmetry. In this case we consider two dimensions (r and x).
For 2-D planar jets the coordinates x and y are considered.
2.10 Plane turbulent wall jets
The turbulent wall jet, even limiting attention to the topographically simple cases well
liked by academics, arguably provides more puzzles for those seeking an ordered set of
rules to describe turbulence than any other class of turbulent shear layer. Formally, we can
regard a wall jet as a boundary layer in which, by virtue of the initially supplied
momentum, the velocity over some region in the shear layer exceeds that in the free stream.
Wall jets are of great and diverse engineering importance and engineering applications
often feature 2D or 3D wall jets. The best known and most challenging applications lie in the
field of advanced airfoil design and in problems of heating, cooling or ventilating-areas.
There are major industrial applications such as the film-cooling of the lining walls of gasturbine combustion chambers and of the leading stages of the turbine itself. In both cases the
aim is to introduce a cool layer of fluid adjacent to a solid surface in order to provide
protection from a hot external stream. In practice, for reasons of constructional strength, it is
not possible to provide a continuous cooling slot extending over the full lateral extent of the
region to be cooled. A series of short slots or holes is thus employed. One wants, therefore, a
rapid lateral spreading of coolant to fill in the gaps but a low mixing rate in the direction
normal to the surface in order that the coolant is diluted only slowly. The most widespread
use of the wall jet for heat and (more importantly) mass transfer modification must be in the
automobile demister. On a larger scale, one may mention the deflectors used in conditioned
air-circulation systems in the home or factory. The design and positioning of such deflectors
becomes especially crucial in large-scale, one-of-a-kind applications such as a concert
auditorium. Clearly here, as in the other examples cited, a thorough knowledge of the fluid
mechanics of the wall jet-including the turbulence field it generates-can materially improve
the design of such devices. There is also the defrosting of vehicle windscreens in this area.
54
Various authors have discussed the remarkable spreading behaviour of the wall jet.
Early wall jet publications are found in Forthmann (1936), Sforza and Herbst (1970) and
Launder and Rodi (1983) presenting reviews of wall jet investigations up to 80s. Some of
the more recent studies, such as Zhou and Wygnanski (1993), Hsiao and Sheu (1994) and
Schneider and Goldstein (1994), have focused on the flow structures in wall jets. Recently,
Craft and Launder (2001) reported on a detailed computational investigation, which aimed
to identify the origin of the large lateral spreading mechanism.
The turbulent wall jet has been the subject of several experimental studies for both the
isothermal and the non-isothermal cases. The experimental approach is one way to extract
information to further understand the major effect of these different interactions. Wall laws
derived from experiments in the similarity region are thus used to complete the solution of
the problem, which implies that such a numerical resolution is possible only if experimental
results are available for the studied configuration.
Elevating an axisymmetric jet from a wall and adding a back plane is complex due to
the numerous interactions that determine the behaviour of the flow. This situation seems to
be most common in practical jet applications.
Because most practical applications involve the near field of wall jets, it is important to
study the distribution of velocities and turbulent stresses in these regions. For wall jets, the
viscous layer plays an important role in heat and momentum transfer.
Let us consider a plane jet of thickness b0 and uniform velocity U0 coming out of a
nozzle tangentially to a smooth flat plate which is submerged in a semi-finite expanse of
the same fluid as shown in Figure 2.21.
55
Figure 2.21 Definition sketch of plane turbulent wall jets.
The resulting flow can be viewed as a combination of an inner wall boundary layer,
where the velocity increases from zero at the wall to a local maximum, and an outer free jet
where the flow decreases from a local maximum to zero (or the free stream value in the
case of moving surroundings). The strong interaction between these two layers causes the
complexity of this type of flow. Evidences for this interaction is the low spreading rate (less
than for the corresponding free jet) and the influence on the flow near the wall is shown by
the displacement of the position of zero shear stress away from the point of maximum mean
velocity.
As the jet leaves the nozzle, due to the velocity discontinuity, a shear layer develops on
the fluid side. A boundary layer develops on the wall side. When the boundary layer meets
the penetrating shear layer, the potential core of the jet is consumed and beyond this
section, the flow is said to be fully developed.
If we make experimental observations of the distribution of the u-velocity in the ydirection at different x-stations (see Figure 2.22), that is, if we study the u(y) function for
various values of x, we find that the shape of the velocity profile is the same at all x-stations
in the fully developed flow region. We find that at any x-station, u increases from zero at
the wall to a maximum value of um at y = and then decreases to zero at some large value
of y. The region from the wall to the maximum velocity level is known as the boundary
layer and the region above this is generally known as the free mixing region. The velocity
scale is generally represented by um and the length scale b is taken as the value of y where
u=1/2um and u / y is negative. Let us find whether we could reduce these velocity profiles
56
at different x-stations to one profile by using our velocity and length scales. This is done in
Figure 2.22 using the experimental observations of Forthmann (1936).
On this figure =
y
.
b
1
U / Um
0,75
0,5
0,25
0
0
0,5
1,5
Figure 2.22 Similarity of velocity profiles in plane wall jets (Forthmann, 1934).
It is interesting to find that the velocity profiles are indeed similar. The phenomenon
described above is known as the plane turbulent wall jet.
Results for wall jet and free jet indicate that the jet spreading increases almost linearly
in agreement with the results of Tollmien (1926).
The equations of motion for a plane turbulent wall jet is described using a coordinate
system and notation similar to that of the plane free jet, we could show that the equations of
motion of the plane turbulent wall jet are:
u
u
1 dp
2 u 1 t
+v
=
+ 2 +
x
y
y
dx
y
and:
(2.125)
u v
+
=0
x y
Let us neglect the effect of any longitudinal-pressure gradient. Using the results of the
experimental observations, we have:
u / um = f ( y / b ) = f ( )
57
(2.126)
Further assume: t / um = g ( )
2
Using the continuity equation, we obtain an expression for v as:
v = um b ' f ' d bum

0
'
fd
0
where, as before, b=db/dx and um=dum/dx. Using these expressions we could reduce the
equation of motion to the form:
bu
g = m
um
'
'
f f ' fd b ' (ff ' f ' ) f ''
um b
0
(2.127)
where g ' = dg / d , f ' = df / d and f ' ' = d 2 f / d 2

In most of the wall-jet problems of practical interest, um b / is large and hence we
could neglect the last term in last equation and this approximation will bring in serious
error only in a very thin layer close to the smooth plate.
The integral momentum equation is derived multiplying the first equation of motion
by and integrating with respect to y from y = 0 to y , we have:
u
u
0 u x dy + 0 v y dy = 0 y dy
(2.128)
Following a procedure could be reduced to the form:
d
u 2 dy = 0
dx 0
(2.129)
where 0 is the wall shear stress.

As a first approximation, if 0 is neglected, we have:
d
u 2 dy = 0
dx 0
58
(2.130)
We could rewrite:
d
um 2 b f 2 d = 0
dx
0
which gives :
(2.131)
um b x 0
The presence of the shear-stress term is to reduce p (do not confuse with pressure) to
less than 0.5. If we now assume that:
0 xS
(2.132)
it is reduced to:
d
um 2 b f 2 d x S
dx
0
If we accept the first approximation 0 decreases inversely with x. Thus we have for the
plane turbulent wall jet:
um 1 / x ; b x ; 0 1 / x
(2.133)
The 2-D turbulent wall jet is a basic flow of fundamental interest for turbulence
researches because of its two-scale character. The interaction of large turbulence scales in
the outer layer with smaller scales in the inner layer creates a complicated flow field and
determines the development of the wall jet. In the ideal plane wall jet the flow is of infinite
extent in the transverse (z) direction, and unconstrained in either the streamwise (x) or
cross-stream (y) directions.
Three dimensional wall jet is shown on Figure 2.23
59
Figure 2.23 Schematic diagram of velocity structure of three-dimensional wall jet (modified
image of Craft and Launder (2001)).
Between the wall and the velocity maximum, U increases from zero to its maximum
value over a distance ym which is small compared with the ow width in the z-direction
(even if, for the moment, equal growth rates in the y- and z-directions are assumed).
Moreover, V may be expected to be significantly less than U since V / y is zero at the
wall, by continuity. Beyond the velocity maximum U / y changes sign, as does, in
consequence, the source of streamwise vorticity. The streamwise vorticity sources are thus
rather as shown in Figure 2.24 an arrangement which evidently induces a laterally outward
secondary ow.
Figure 2.24 Sense of streamwise vorticity source for laminar flow (Craft and Launder
(2001)).
60
Further details about the three-dimensional wall jet facility and the flow conditions can
be found in Abrahamsson et al. (1997)
A special case of wall jet is a wall jet flowing over a corrugated bed (such as the sketch
of Figure 2.25)
Figure 2.25 Definition sketch for wall jets on corrugated beds (Ead and Rajaratnam
(2004)).
Ead and Rajaratnam (2002) found that in plane turbulent wall jets in shallow tailwater
(1 ) 1 2
F0
= 2 2
(2.134)
where: = w / b0 ; tailwater depth ratio, equal to yt/b0; shear force coefficient equal
x = Le
to F/M0; F integrated bed shear stress defined as F =
dx = M
; M0 momentum
x =0
flux per unit width and Le length of the surface eddy.

A study of this equation show that for any given F0, decreases as increases and
shear stress on the bed also contributes to a decrease of . Further, increases as F0
increases. Using the experimental measurements of , , and F0 it was found that was in
the range of 0.120.15 (Ead and Rajaratnam 2002) for experiments conducted on wall jets
on a smooth boundary. For rough boundary, this coefficient is much larger (0,47-0,72).
Flown pattern and eddy length Le of this case of jet are shown on Figure 2.26.
61
Figure 2.26 Flown pattern for plane turbulent wall jets in shallow tailwater (Ead and
Rajaratnam (2002)).
2.11 Other jet and plume configurations
Here we present some other jet and plumes configurations describing shortly its
characteristics in order to remark the variety of possible configurations.
Two types of flow field induced by a vertical axisymemtric jet discharge in shallow
water are shown in Figure 2.27.
Figure 2.27 Different types of flow field induced by a vertical axisymmetric jet discharge in
shallow water: (a) strongly buoyant (stable); (b) weakly buoyant (unstable) (Kuang and Lee
(2006))
Many industrial and environmental effluent discharges into the water environment can
be categorized as buoyant jets in shallow water. Examples include thermal effluents from
once-through condenser cooling water systems of steam electric power stations, and sewage
discharges into riverine or brackish estuarine waters. In an early study on round buoyant
62
jets in conned depth, it has been shown that the ow pattern in the vicinity of such a
discharge can differ remarkably from that of a free buoyant jet ow; under high momentum
and shallow water conditions there is signicant interaction between the discharge and the
boundaries, leading to a breakdown of the jet structure and recirculating cells. Such a
situation is termed as an unstable near eld (Figure 2.27b); the bulk mixing or initial
dilution of the efuent discharge is reduced as a result of the recirculation and ow
instability. For a stable discharge, we can divided the ow into four regions of distinct
hydro-dynamic properties (Figure 2.27a): (1) Initial Buoyant Jet Region; (2) Surface
Impingement Region followed by a supercritical surface jet; (3) Radial Internal Hydraulic
Jump Region; and (4) Stratied Two-layer Counterow Region.
The negatively (the example above is positively case) buoyant jet with rigid horizontal
bottom boundary is presented below (Figure 2.28).
Figure 2.28 Sketch of vertical, turbulent, planar, negatively buoyant jet with rigid
horizontal bottom boundary (Cavalletti and Davies (2003))
The jet is generated at a two-dimensional slit source of width b, source velocity w0,
source density 0 into quiescent receiving uid of density a (<0) and effective depth H.
Subject to the assumption that the Reynolds ( R0 = W0b / ) and Grashof Gr0 = ( g ' ) 0 b 3 / 2 )
numbers of the source ow are both sufficiently high for viscous effects to be neglected, the
ow behaviour is conveniently described in terms of the geometrical ratio H/b and a source
Froude number Fd0 dened as
Fd 0 = W0 /[( g ')0 b]
1/ 2
63
(2.135)
where ( g ')0 = g ( )0 / a - modified gravitational acceleration at the source and ( )0 density difference between the source density 0 and the ambient receiving water density
a = ( 0 ( )0 ) .
More different cases of turbulent buoyant jets are presented in work of Jirka (2004),
between others, the three-dimensional buoyant jets or vertical buoyant jets.
The impinging jet is shown on Figure 2.29
Figure 2.29 Impinging jet flowfield and co-ordinate system (Knowles and Myszko (1998)).
The round normally impinging jet continues to raise considerable interest in the
research literature. Much of this is directed at surface heat transfer characteristics and stems
directly from industrial applications which use the very high levels of Nusselt number
which occur near the impingement stagnation point.
More complicated case is interaction between a plane wall jet and a parallel offset jet
shown in Figure 2.30.
64
Figure 2.30 Schematic diagram of the ow pattern (Wang and Tan (2006))
The origin of the coordinate system is set at the intersection of the jet exit plane and the
bottom wall, where x is along the wall. The two jets, with the same width w, are separated
in the vertical (y) direction by an offset plate with a distance of d (referred to as the offset
distance). Several shear layers exist in this ow configuration, including the outer layer of
the offset jet (denoted as Layer 1), the inner layer of the offset jet (Layer 2) and the inner
layer of the wall jet (Layer 3). In addition, unlike conventional unbounded parallel jets
where the outer layers are freely developing, the lowest layer in this ow conguration is a
wall boundary layer (where the no-slip condition applies). The existence of the low
pressure zone downstream of the offset plate causes the two jets to bend towards each other
(referred to as the converging region), and merge together at some downstream distance
(xmp), referred to as the merging point (mp) where the streamwise mean velocity (U) is zero.
Downstream of xmp it is the merging region which ends up at the combined point (cp),
where the deection point originally in the mid-part of the U- prole disappears. In the
combined region, the two jets can be considered to behave like a single wall jet. To
illustrate the jet development, the three jet half-widths i.e.,
1 1
y , y and
2 1 2 2
1
y respectively, for Layers 1, 2 and 3, together with the boundary layer thickness
2 3
1
along the bottom wall, are denoted. In the gure y is dened as the vertical distance
2
65
from the bottom wall to the point where U is half of its local maximum (Um), while is that
where U = 0.99Um.
Just mentioning, as an example of some numerical works, the mechanics of buoyant jet
flows issuing with a general three-dimensional geometry into an unbounded ambient
environment with uniform density or stable density stratification and under stagnant or
steady sheared current conditions are investigated by Jirka (2004). An integral model is
formulated for the conservation of mass, momentum, buoyancy and scalar quantities in the
turbulent jet flow. Integral models for the analysis and prediction of turbulent buoyant jet
effluxes into an ambient fluid environment are widely used in many fields of geophysical,
environmental and engineering applications. The model contains a number of important
features that in aggregate make it much more general, reliable and accurate. The model was
validated in several stages.
2.12 The turbulent boundary layers
There is a great interest in the boundary layers because as a source at shear turbulence is
non-homogenous. Still we have to find more precise methods to calculate properties of the
boundary layer. Amid many parameters such as the displacement thickness, the skin-friction
coefficient, it is helpful to uncover the prediction of mean velocity and shear-stress distribution
across the layer. There is also continuing research interest aimed toward finding an
understanding of turbulent structure due to shear flows of which the boundary layer is
probably the most interesting example.
How we can define the boundary layer? The boundary layer is a thin layer close to a solid
boundary within which vorticity varies rapidly as a result of the combined effects of viscous
diffusion and convection and outside which the vorticity is zero (or is non-zero and varies only
slowly). Other definition we can say is: boundary layer is a thin layer in which the effect of
viscosity is important however high the Reynolds number of the flow may be.
Across the boundary layer the flow velocity changes from the value zero at the boundary
to some finite value characteristic of an inviscid fluid. Also derivatives with respect to y
(perpendicular coordinate) of any flow quantity are in general much larger than those with
66
respect to x (longitudinal coordinate). Thus at points within the boundary layer we may use the
approximations:
u
u
<<
,
x
y
2u
2u
<<
x 2
y 2
(2.136)
where: u is longitudinal (streamwise) velocity and whence the equation of motion in the xdirection becomes:
u
u
u
1 p
2u
+u
+v
=
+v 2
t
x
y
x
y
(2.137)
where: t is time, v perpendicular (normal or spanwise) velocity, density and p pressure.

The velocity component v normal to the boundary must also be small, and the massconservation equation:
u v
+
=0
x y
(2.138)
suggests that v and the boundary-layer thickness are of equal order of smallness.
1
x
y
tU
, y ' = Re 2 , t ' = 0 ,
L
L
L
If we introduce the next new variables:
1
u
v
p p0
2
, v' = Re
, p' =
,
u' =
U0
U0
U 02
x' =
(2.139)
where: L dissipation scale, Re Reynolds number, U0 mean velocity and p0 uniform

pressure, the complete equations of motion in the x and y-directions and the equation of mass
conservation become
2
2
1 v'
v'
v'
p '
1 v' 1 v'
+ u'
+ v' =
+ 2
+
,
Re t '
x'
y '
y ' Re x'2 Re y '2
u ' v'
+
=0
x' y '
u '
u '
u '
p ' 1 2u ' 2u '
+ u'
+ v'
=
+
+
,
t '
x'
y '
x' Re x'2 y '2
(2.140)
Aside from the usual consideration for technological applications, the turbulent boundary
layer is worth studying because it is especially rich in details and complications, although it is
67
a simple, well-defined flow. As the flow is bounded by one solid boundary and by one free
boundary, it exhibits the details one normally associates with confined flows, termed wall
turbulence. In these respects it is similar to channel and pipe flow. On the other hand, near to
the free boundary it incorporates also the features of free turbulence found in wakes and jets.
This combination brings us to the most fascinating question, namely exactly how the two
different types of flow match with each other.
Many of the ideas originated essentially with Prandtl and were based on the use of the
turbulent energy equation:
2
2
1 q
q
U p q 2
+V
= uv
U
v +
2 x
y
y y 2
(2.141)
where U and V are the streamwise and normal components of the mean velocity, q is the
absolute magnitude of the velocity fluctuation vector: q 2 = ui ui ( do not confuse with structure
function order) and is the viscous energy dissipation per unit volume.
The governing equation for the turbulent shear stress (Reynolds stress) uv can be
derived:
( )
( )
uv
uv
U p u v pu
V
= v2
+ +
+ uv 2 + v u 2u + v 2v (2.142)
x
y
y y x y

In order to obtain a tractable set of equations, different authors proceeded differently.

Bradshaw has assumed that:
(a) the shear stress is proportional to the turbulent kinetic energy:
= uv = a1 q 2
(2.143)
where: is shear stress and a1 guessed quantity (a1=0,15 was assumed as a good
approximation for wall turbulence),
68
(b) the viscous dissipation can be expressed in terms of the local shear stress as
1
=
L1
3/ 2
(2.144)
where L1 dissipation scale (was taken as a function of y / , the position in the boundary
layer):
L1 = f ( y / )
(2.145)
and boundary layer thickness,
(c) the troublesome pressure-velocity correlation and the velocity triple correlation can be
approximated as
pv +
1
r3
q 2 v = G1
2
(2.146)
The function G1 was assumed to be a function of y / and proportional to the skinfriction coefficient Cf:
G1 = C f g ( y / )
(2.147)
The two functions f and g were guessed on the basis of available experimental data.
This method has been quite successful in predicting boundary layers under a variety of
conditions. It has been less successful in attempting to predict free turbulence, probably
because then each class of flow would require at least one set of new universal functions f
and g.
We may also discuss the turbulent energy equation in the presence of buoyancy, then an
extra term due to gravity (or due to any other force, such as rotation or a magnetic field
appears). This is explained by Matulka (2009) or Tailleux (2009).
69
2.13 Geophysical turbulence
In a strictly two-dimensional ow with weak dissipation, energy input at a given scale

is transferred to larger scales, because these constraints stop vortex lines being stretched or
twisted. Physically this upscale energy transfer occurs by merging of vortices and leads to
the production of coherent structures in the ow that contain most of the energy. This
process generates the appearance of order from chaos.
This scenario is an appropriate model for geophysical ows which are known to contain
very energetic vortices mesoscale oceanic eddies and atmospheric highs and lows. This
upscale transfer of energy is inhibited at the Rossby deformation radius:
LR =
N
h
f
(2.148)
where h is the characteristic scale of the depth of the thermocline, N the Brunt Visl
frequency and f the Coriolis parameter.
The energy limitation is caused by baroclinic instability at larger scales, which accounts
for the dominant observed size of geophysical vortices detected in laboratory experiments
on annulus ow, where the ow is driven in a rotating annulus by dierential heating of the
lateral walls of the annulus, or by internal heating of the uid. A horizontal temperature
gradient is established which drives a zonal ow via the thermal wind balance. For
certain values of the parameters this ow is unstable to baroclinic modes that feed on the
energy in the temperature or density elds.
Many features have been identied with structures and phenomena observed in several
experiments, and understanding of atmospheric and ocean dynamics has been significantly
advanced. The experiments have provided new insights about the dynamics and have
revealed a wide range of nonlinear behaviours. Some performed experiments showed the
eect of mixing from the edge on a rotating stratified system. When the instability is
caused by dierential heating or by buoyancy there seems to be a range of very dierent
dynamic regimes. It was revealed the possible complex interactions between lateral (or
coastal) stirring and the rotating-stratied ow dynamics.
70
The investigation of such strongly non-homogeneous ow, which leads to intermittent

two-dimensional turbulence is believed to be very important if correct parameterizations of
pollutant dispersion (such as oil spills) in coastal areas are to be made. The availability of a
large-scale ow allows both to measure Eulerian velocities with precision, Lagrangian
ows using particle tracking as well as local measurements of diusivity by video
recording the dispersion of neutral tracers. Lagrangian buoys near an accident aid the
predictions of coastal currents (Redondo and Platonov (2001 and 2009).
One of the problems in order to identify oil spills is the possibility of confusion with
natural tensioactive spills, which may be due to plankton, algae or even wind pattern
reections on the ocean surface, one of the possibilities is to use the scaling properties of
the turbulence that advects and diuses the tracers (Seuront et al. (2001) and Bezzera et al.
(1997)).
2.14 Mixing efficiency
Turbulent mixing is a very important issue in the study of geophysical phenomena

because most fluxes arising in geophysics fluids are turbulent. The diffusion of physical
quantities in nature is due to turbulence.
Mixing across stratified fluids takes place often in geophysical situations. Important
mixing processes are associated with the formation of fronts, such as sea and mountain
breezes, tidal stirring, etc. In most cases, and due to the intermittency of energetic
processes, mixing is localized in time and space.
If we consider an experimentally study of mixing across density interfaces produced by

laterally heterogeneous turbulence where the turbulence is generated by a flow or air
bubbles rising through a density interface produced by brine and fresh water, the production
of turbulence is due to the velocity fluctuations induced by the bubbles (Redondo and
Cantalapiedra 1993).
The mixing efficiency is defined as the relation between the increase in potential energy
of the fluid PE and the kinetic energy that produces the mixing KE
71
PE
KE
(2.149)
We can define three scales for different types of instabilities at different ranges of the
Richardson number.
The Ozmidov scale
L0 = ( / N 3 )
1/ 2
(2.150)
where is the viscous dissipation, N is the number of holes producing bubbles.

The buoyancy scale
Lb = w' / N
(2.151)
where w` is vertical velocity fluctuation.

The third scale Lt represents a typical vertical distance travelled by fluid particles before
either returning to their equilibrium level or mixing,
Lt =
'
(2.152)
z
where is the density fluctuations, the reference density and z vertical direction.
For the stationary flow we can now write relation
w' u '2 / 2
1
' w'
= 2 Lb Lt
1
z
L0
' w'
(2.153)
where u is the r.m.s. turbulent velocity.
Gonzalez-Nieto et al. (2007) studied turbulent mixing due to convection using a

laboratory experimental model with two miscible fluids of different density with an initial
top-heavy density distribution. Mixing is produced by the evolution of a two-dimensional
array of forced turbulent plumes. They defined the non-dimensional mixing volume as
72

2h
*
VMIXING
= 1
3hL
(2.154)
where h is the depth of the lateral interaction between plumes start and hL is the height of
the light fluid layer. For more details see the paper.
73
CHAPTER 3
THEORY SAR AND FRACTALS
3.1 SAR (Synthetic Aperture Radar)
Part of this thesis is related to satellite (SAR Synthetic Aperture Radar) images
analysis. Environmental monitoring, earth-resource mapping, and military systems require
broad-area imaging at high-resolutions. Many times the imagery must be acquired in
inclement weather or during night as well as day. Synthetic Aperture Radar provides such a
capability. Synthetic aperture radar technology has provided terrain structural information
to geologists for mineral exploration, oil spill boundaries on water to environmentalists, sea
state and ice hazard maps to navigators, and reconnaissance and targeting information to
military operations. There are many other applications or potential applications. Some of
these, particularly civilian, have not yet been adequately explored because lower cost
electronics are just beginning to make SAR technology economical for smaller scale uses.
SAR is a form of radar in which multiple radar images are processed to yield higherresolution images than would be possible by conventional means. Either a single antenna
mounted on a moving platform (such as an airplane or spacecraft) is used to illuminate a
target scene or many low-directivity small stationary antennae are scattered over an area
near the target area. The many echo waveforms received at the different antenna positions
are post-processed to resolve the target. SAR can only be implemented by moving one or
more antennae over relatively immobile targets, by placing multiple stationary antennae
over a relatively large area, or combinations thereof. SAR has seen wide applications in
remote sensing and mapping.
SAR systems take advantage of the long-range propagation characteristics of radar
signals and the complex information processing capability of modern digital electronics to
provide high resolution imagery. Synthetic aperture radar complements photographic and
other optical imaging capabilities because of the minimum constraints on time-of-day and
atmospheric conditions and because of the unique responses of terrain and cultural targets
to radar frequencies.
Example of SAR image and SAR instrument are shown in Figure 3.1.
74
Figure 3.1 The surface of Venus, as imaged by the Magellan probe using SAR (left)
(source: NASA) and NASA's AirSAR instrument attached to the side of a DC-8 (right).
(Source: NASA).
A detailed description of the theory of operation of SAR is complex and beyond the
scope of this document. Instead, this chapter is intended to give the reader an intuitive feel
for how synthetic aperture radar works.
In a typical SAR application, a single radar antenna is attached to the side of an aircraft
(Figure 3.2)
Consider an airborne SAR imaging perpendicular to the aircraft velocity as shown in
the figure below (Figure 3.2). Typically, SARs produce a two-dimensional (2-D) image.
One dimension in the image is called range (or cross track) and is a measure of the "line-ofsight" distance from the radar to the target. Range measurement and resolution are achieved
in synthetic aperture radar in the same manner as most other radars: range is determined by
precisely measuring the time from transmission of a pulse to receiving the echo from a
target and, in the simplest SAR, range resolution is determined by the transmitted pulse
width, i.e. narrow pulses yield fine range resolution.
75
Figure 3.2 Synthetic Aperture Radar imaging concept (source: Sandia National Laboratories)
The other dimension is called azimuth (or along track) and is perpendicular to range. It
is the ability of SAR to produce relatively fine azimuth resolution that differentiates it from
other radars. To obtain fine azimuth resolution, a physically large antenna is needed to
focus the transmitted and received energy into a sharp beam. The sharpness of the beam
defines the azimuth resolution. Similarly, optical systems, such as telescopes, require large
apertures (mirrors or lenses which are analogous to the radar antenna) to obtain fine
imaging resolution. Since SARs are much lower in frequency than optical systems, even
moderate SAR resolutions require an antenna physically larger than can be practically
carried by an airborne platform: antenna lengths several hundred meters long are often
required. However, airborne radar could collect data while flying this distance and then
process the data as if it came from a physically long antenna. The distance the aircraft flies
in synthesizing the antenna is known as the synthetic aperture. A narrow synthetic beam
width results from the relatively long synthetic aperture, which yields finer resolution than
is possible from a smaller physical antenna.
Achieving fine azimuth resolution may also be described from a Doppler processing
viewpoint. A target's position along the flight path determines the Doppler frequency of its
echoes: targets ahead of the aircraft produce a positive Doppler offset; targets behind the
aircraft produce a negative offset. As the aircraft flies a distance (the synthetic aperture),
echoes are resolved into a number of Doppler frequencies. The target's Doppler frequency
determines its azimuth position.
Combining the series of observations requires significant computational resources. It is
often done at a ground station after the observation is complete, using Fourier transform
76
techniques. The high computing speed now available allows SAR processing to be done in
real-time onboard SAR aircraft. The result is a map of radar reflectivity (including both
amplitude and phase). The amplitude information contains information about ground cover,
in much the same way that a black-and-white picture does. Interpretation is not simple, but
a large body of experimental results has been accumulated by flying test flights over known
terrain. Early satellites provided a resolution in the tens of meters. More recent airborne
systems provide resolutions to about 10 cm, ultra-wideband systems (developed and
produced in the last decade) provide resolutions of a few millimetres, and experimental
terahertz SAR has provided sub-millimetre resolution in the laboratory. For purposes of this
work, a resolution of meters is more than enough considering, for example, sea vortices of
kilometres of length scales.
While this section attempts to provide an intuitive understanding, SARs are not as
simple as described above. Transmitting short pulses to provide range resolution is
generally not practical. Typically, longer pulses with wide-bandwidth modulation are
transmitted which complicates the range processing but decreases the peak power
requirements on the transmitter. For even moderate azimuth resolutions, a target's range to
each location on the synthetic aperture changes along the synthetic aperture. The energy
reflected from the target must be "mathematically focused" to compensate for the range
dependence across the aperture prior to image formation. Additionally, for fine-resolution
systems, the range and azimuth processing are coupled (dependent on each other) which
also greatly increases the computational processing.
3.1.1
SAR applications
This chapter discusses a few of the applications for synthetic aperture radar. These
applications increase almost daily as new technologies and innovative ideas are developed.
While SAR is often used because of its all-weather, day-or-night capability, it also finds
application because it renders a different view of a "target," with synthetic aperture radar
being at a much lower electromagnetic frequency than optical sensors.
Many applications for synthetic aperture radar are for reconnaissance, surveillance, and
targeting. These applications are driven by the military's need for all-weather, day-and-
77
night imaging sensors. SAR can provide sufficiently high-resolution to distinguish terrain
features and to recognize and identify selected man-made targets.
The ability to monitor other nations for treaty compliance and for the non-proliferation
of nuclear, chemical, and biological weapons is increasingly critical. Often, monitoring is
possible only at specific times, when over flights are allowed, or it is necessary to maintain
a monitoring capability in inclement weather or at night, to ensure an adversary is not using
these conditions to hide an activity. SAR provides the all-weather capability and
complements information available from other airborne sensors, such as optical or thermalinfrared sensors.
Interferometric synthetic aperture radar (IFSAR) data can be acquired using two
antennas on one aircraft or by flying two slightly offset passes of an aircraft with a single
antenna. Interferometric SAR can be used to generate very accurate surface profile maps of
the terrain. For example, Sandia has developed new mathematical techniques for relating
the radar reflection from the terrain surface to the time delay between radar signals received
at the two antenna locations. The techniques are directed at removing ambiguities in
estimates of surface heights and are referred to as 2-D least squares phase unwrapping.
Synthetic aperture radar provides the capability for all-weather, autonomous navigation
and guidance. By forming SAR reflectivity images of the terrain and then "correlating" the
SAR image with a stored reference (obtained from optical photography or a previous SAR
image), a navigation update can be obtained. Position accuracies of less than a SAR
resolution cell can be obtained. SAR may also be used to guidance applications by pointing
or "squinting" the antenna beam in the direction of motion of the airborne platform. In this
manner, the SAR may image a target and guide a munition with high precision.
Synthetic aperture radars offer the capability for penetrating materials which are
optically opaque, and thus not visible by optical or IR techniques. Low-frequency SARs
may be used under certain conditions to penetrate foliage and even soil. This provides the
capability for imaging targets normally hidden by trees, brush, and other ground cover. To
obtain adequate foliage and soil penetration, SARs must operate at relatively low
frequencies (10's of MHz to 1 GHz).
78
Recent studies have shown that SAR may provide a limited capability for imaging
selected underground targets, such as utility lines, arms caches, bunkers, mines, etc. Depth
of penetration varies with soil conditions (moisture content, conductivity, etc.) and target
size, but individual measurements have shown the capability for detecting 55-gallon drums
and power lines at depths of several meters. In dry sand, penetration depths of 10's of
meters are possible.
The motion of a ground-based moving target such as a car, truck, or military vehicle,
causes the radar signature of the moving target to shift outside of the normal ground return
of a radar image. Sandia has developed techniques to automatically detect ground-based
moving targets and to extract other target information such as location, speed, size, and
Radar Cross Section (RCS) from these target signatures.
A technique known as coherent change detection offers the capability for detecting
changes between imaging passes. To detect whether or not a change has occurred, two
images are taken of the same scene, but at different times. These images are then
geometrically registered so that the same target pixels in each image align. After the images
are registered, they are cross correlated pixel by pixel. Where a change has not occurred
between the imaging passes, the pixels remain correlated, whereas if a change has occurred,
the pixels are uncorrelated. Of course, targets that are not fixed or rigid, such as trees
blowing in the wind, will naturally decorrelate and show as having "changed." While this
technique is useful for detecting change, it does not measure direction or the magnitude of
change.
Finally, the most important application for purposes of this work is that synthetic
aperture radar is used for a wide variety of environmental applications, such as monitoring
crop characteristics, deforestation, ice flows, and oil spills. Oil spills can often be detected
in SAR imagery because the oil changes the backscatter characteristics of the ocean. Radar
backscatter from the ocean results primarily from capillary waves through what is known as
Bragg scattering (constructive interference from the capillary waves being close to the same
wavelength as the SAR). The presence of oil dampens the capillary waves, thereby
decreasing the radar backscatter. Thus, oil slicks appear dark in SAR images relative to oilfree areas. Related to this type of SAR application we will refer to work done in the
Departament de Fisica Aplicada of the UPC by Platonov (2002) and published in Redondo
79
and Platonov (2001), Platonov et al. (2008), Redondo et al. (2008), Redondo and Platonov
(2009), Redondo et al. (2009).
3.2 Fractal analysis

The fractals are geometric entities with some very meaningful properties, they present
self-similarity and they are the result of iterative processes such as turbulence. The selfsimilarity implies that if we accomplish observations from different scales the results are
similar. The other important aspect is that in most cases they are the result of iterative
processes with a very simple generator such as the Koch curve. The Koch curve (island) is
a mathematical curve and one of the earliest fractal curves described (see Figure 3.3)
Figure 3.3 The Koch island
When the fractal object under study is a real entity is very rare that it has a perfect selfsimilarity and in most realistic situations in environmental flows the observation and the
scaled observation are only statistically self-similar.
In order to maximize the information obtained from remote sensing (satellites, aircrafts,
balloons, buoys or other sensors) it is important to use geometry as an indicator of the
actual dynamics of the environmental flow which most of the times is dominated by
turbulent processes. A very useful tool to characterize these objects is the calculation of the
corresponding fractal dimension along the different coordinates. With the fractal analysis
80
we have a geometrical tool that is able to quantify certain aspects of the complexity of
features that are ubiquitous in environmental flows both in the atmosphere and in the ocean,
and even considering biological aspects.
The concept and the word of "fractal" are attributed to a 20th century mathematician,
Benoit Mandelbrot. His fractal theory revisited the concept of Kolmogorov capacity and
was used to try to quantify the immense complexity of nature in relatively simple
equations. His favourite example of such complexity was the craggy coast of Britain, which
was first studied by Richardson (1925), when seen from far above it looks somewhat
wrinkled and convoluted. Yet as an observer gets closer and closer to the shore, the
complexity of the coastline increases; smooth lines become jagged, more jagged and
complex until the observer is so close that he is observing the minute variation in the
positions of each individual grain of sand along the shore. Moreover, we can imagine this
observer measuring the length of the coastline with increasingly smaller rulers. As he takes
account of the added complexity as he measures with increasingly precise resolution, his
approximation to the length of the coast of Britain keeps increasing. In fact, he could very
well find that the length he is looking for diverges to infinity!
The fractal dimension generalizes the dimension concept of a geometric entity to the
fractional numbers and the fractal objects present these fractional dimensions.
For computing the fractal dimension of any strictly self-similar fractals we use formula:
D = log N / log S
(3.1)
where D is the fractal dimension, N the number of miniature pieces in the final figure to
cover it and S is the scaling factor.
The dimension is a measure of how completely these fractals embed themselves into
normal Euclidean space.
The calculation of fractal dimensions is one of the promising techniques for automatic
discrimination of the different seas surface structures.
81
3.3 Self-affine fractals. Relationship between turbulence spectra and fractal

dimension
A statistically self-similar fractal is by definition isotropic. In two dimensions defined
by x and y coordinates the result do not depend on the geometrical orientation of the x - and
y - axes. A formal definition of a self-affine fractal in a two-dimensional xy-space is that
f rx, r H y is statistically similar to f ( x, y ) where H is known as the Hausdorff measure

and r a scaling factor. Self-affine fractals are generally treated quantitatively using spectral
techniques. The dependence of x on t in the random time series is similar to the shape of the
Koch Island illustrated in Figure 3.3. There are variations in the time series at all scales.
The basic definition of the fractal dimension for a time series is:
H = 2D
(3.2)
It is very interesting to relate D to the frequency spectrum or to the spatial spectra

obtained from the Fourier transform of the time or spatial correlation functions, usual in
studies of turbulence. The reason is that from such frequency spectrum, the corresponding
fractal dimension may be derived, if the tracer scalar is passively advected by a turbulent
flow. Then the fractal dimension might be related to the energy of the turbulence with a
certain spatial or temporal dependence, and then the frequency spectrum exponent,
provided an inertial subrange exists, is a function of the box-counting fractal dimension as
demonstrated by Redondo (2008).
If we assume that the frequency spectrum has a well-defined shape over a significant
range of frequencies:
S ( f ) = f
(3.3)
we may obtain a relationship between the power and the fractal dimension D.
Using the variance of the signal (t) defined from:
V (T ) = < ( (t + T ) (t )) 2 >
82
(3.4)
where < > denotes the average over the entire period T.
The dependence for fractal time series V(T) T2H.
Here H is Hurst exponent. In fractal geometry, the generalized Hurst exponent is

referred to as the "index of dependence" and is the relative tendency of a time series to
either strongly regress to the mean or 'cluster' in a direction. The H has a relationship to the
fractal dimension D and to the Euclidean dimension E:
D=EH
(3.5)
The Hurst exponent H seems a complementary dimension to D, from this relation it is

usual to say that H is the fractal codimension.
Using T = 1/f and the description of the spectral density function S(f), we have
equivalently
S(T) T
(3.6)
and
T
S T 2 (t ) e ift dt T V
(3.7)
so we can relate
S ( f ) T V T 2 H +1 T 2 E +12 D
(3.8)
The relationship between the exponent of the spectral density function and the fractal
dimension is
= 2E + 1 2D
(3.9)
and inversely
D=E+
1
2
83
(3.10)
These geometrical-dynamical equivalences may be used to check different

methodologies of fractal dimension calculations.
Natural phenomena that yield continuous, power-law spectra do not have characteristic
frequencies and are scale invariant over the applicable range. Several examples in geology
and geophysics are shown by Turcotte (1992).
3.4 The spectral method
From the derivation discussed in last section, but applied to a spatial spectrum we may
also accomplish the calculation of the fractal dimension using directly the spectral analysis
on an image. With this methodology a unique value is obtained that characterizes the
overall fractal dimension of the system.
The steps are as follows: First we have to make an image segmentation to obtain the
interest region. Then we compute the FT (Fourier Transform) to obtain the frequency
spectrum representation (Iuv, u and v are the frequency discrete coordinates).
To compute the energy Suv, we use formula:
Suv = | Iuv |2
(3.11)
Then we have to obtain the radial distribution representation Sr of Suv. In this way, we
can find the from
Sr = r -
(3.12)
With a linear fit from log-log representation of Sr we may obtain the value using the
Euclidean dimension E = 3 and the fractal dimension relationship D = E +
D =3+
84
1
2
1
we have:
2
(3.13)
And thus we have a global, indirect measure of the fractal dimension from a radial
spectral energy. Note that only if the scalar used is velocity component, energy will have
the correct physical dimension, otherwise the energy spectrum will indicate the square of
the physical signal used.
3.5 Fractal characteristics of intermittent turbulence
Intermittent effects are one of the main topics of present research on turbulence. Here,
we introduce the so-called model, which takes into account the fractal characteristics of
intermittent turbulence.
We now that not all the regions of the fluid are active from the point of view of
turbulence. Such regions are so strongly convoluted that they are not completely spacefilling; a feature which is more and more accentuated as the scale size decreases.
As we told before, self-similar geometrical objects may be characterized by the socalled Hausdorff dimension or fractal dimension (Mandelbrot, 1983). The fractal dimension
D of an object embedded in a space of a given dimension is defined through the relation
N (l ) l
(3.14)
where N(l) is the number of hypercubes of edge l , for l0, necessary to cover the object.
To achieve a maximum of simplicity, we assume that each eddy breaks into eddies
which have half the previous scale length.
If we assume (Jou 1997) that the number of eddies at the n generation is given by
N (n) = N 0 2 Dn
(3.15)
Therefore, the volume filled by the eddies of the n generation is
V (n ) = N (n )L(n ) N 0 L0 2 (3D )n
3
where L0 is a scale of the biggest eddies.
85
(3.16)
It is seen that if D is different from 3, eddies do not fill the whole space but that smaller
and smaller eddies occupy smaller and smaller regions of the volume. D may be fractional
and here we call it the fractal dimension of intermittent turbulence.
We can write now the volume ratio of the eddies of n-generation to the initial volume as
(n) = V (n ) / V0 = 2 (3 D )n = [L(n ) / L0 ]3D
(3.17)
where L is eddy size.

The energy per unit mass in eddies of the n-th generation is given by E(n) (n)v(n)2.
The turnover time (n) corresponding to eddies of the n-th generation is (n) L(n)/v(n). It
then follows that the energy transfer rate is
(n )v 3 (n ) 1 (n )
(3.18)
We may express v(n) as
v(n ) L(n ) 1 (n )
1/ 3
(3.19)
and
E (n ) (n ) 2 / 3 [L(n ) / (n )]
2/3
(3.20)
It follows from here that

(3 D ) / 3
E (k )dk 2 / 3k 5 / 3 [kL0 ]
dk
(3.21)
This expression clearly shows the deviation with respect to the Kolmogorov -5/3 law
due to intermittent effects.
The consequences for higher moments of the velocity are more relevant. It may be
shown that the scaling laws for higher moments of the velocity are of the form
86
v(L )
q/3L q
(3.22)
where q is order of structure function and scaling exponent q given by
q =
(3 D )
q
+ (3 q )
3
3
(3.23)
3.6 Box-Counting Method

The similarity method and the geometric method of calculating fractal dimension
require you to measure the size. For many fractals its not hard, but for more complicated
shapes it is more difficult.
Generally we calculate the fractal dimension using either the Box Counting Method or
the Cluster Growing Method. First one is the most common methods for calculating the
fractal dimension of a self-similar fractal.
This method is based on the definition of the Hausdorff dimension where it is outlined
as the basic measure concept such as the univocal geometrical relationship between the
applications of coverage of simple (i.e. squares, circles, etc.) on a complex object in order
to measure its size.
To calculate the fractal dimension, this method proposes to make coverage of the object
and to characterize it with boxes of side E. For the plane these boxes will be square and for
an object in the space they will be cubes. The distribution of the boxes is accomplished in a
systematically manner, the intersection of these with the object carries the fact that we have
N boxes with a non void intersection. Already we have some values of N and E, but as they
are not exactly the result of the better coverage possible it may be applied the concept of
self-similarity and thus, the basic covering is accomplished repeating the process for many
different possible diminishing observation scales E (Figure 3.4).
87
Figure 3.4 The box-counting method. A sample of domain division in the different box
scale with a log-log representation and a linear fit that show the fractal dimension D
This leads us to have a functional characterization N(E). We are assembling, part by

part, coverage that it does not have a unique value or a well precise range of scales of the
maximum or minimum possible sizes. The iterations may be done for different ranges and
often the extreme values are corrupted due to bad resolution, therefore the optimum values
have to be calculated with an error criteria and the analysis of the statistics of the fits for
different scale ranges improves the process.
Finally, to extract the corresponding fractal dimension, the linear best fit calculation is
applied. The slope corresponds to the fractal dimension and the correlation indicates us the
kindness (lack of error) of the self-similarity obtained from different scales (usually fits
are required to have correlation coefficient better than 98%).
3.7 Multifractal characterization

Multifractal systems are common in nature, especially geophysics. They include fully
developed turbulence, stock market time series, real world scenes, the Suns magnetic field
time series, heartbeat dynamics, human gait and natural luminosity time series.
Embryogenesis is also multifractal system which represents a new type of physics, named
fractal mechanics.
88
A multifractal system is a generalization of a fractal system in which a single exponent

(the fractal dimension) is not enough to describe its dynamics; instead, a continuous
spectrum of exponents (the so-called singularity spectrum) is needed. Rather than being
point sets, multi-fractals are measures (distributions) exhibiting a spectrum of fractal
dimensions.
When we work with real images they do not have generally some perfectly defined
contours, but some wide ranges of scalar intensity values to process. If we group or filter
the available data and describe them by a single large set and calculate the fractal
dimension, then we lose the corresponding information due to the intensity variation.
To improve this situation, segmentation in many different intervals is needed, so each
contains a very well-defined intensity range. For each one of these ranges, the fractal
dimension calculation described above with the Box-Counting method is helpful and we
obtain the corresponding fractal dimension for each intensity level. The result of the
process will be a set of dimension values in a function of the chosen intensity, obtaining
sets of plots similar to that shown in Figure 3.5.
2
Fractal dimension D
1.6
1.2
0.8
0.4
0
0
100
200
300
Grey level
Figure 3.5 Fractal dimension as a function of grey level.
It is this different dimensional representation which gives us further information on the

studied process, and is able to describe the complexity observed for the different intensities.
89
This methodology permits to characterize by a set of dimensions, the fractals or

complex entities in which the fractal dimension calculation and the corresponding
multifractal characterization is different for the different intensities exhibited, so a
parameterization of the multiple, but different sets coexisting in the same area is possible.
Each of the intensity values may reflect different physical processes and lead to a different
value of its fractal dimension; this whole entity can be fractal or not fractal.
As an example, the study of wind speed time series data using two powerful techniques
is done by Tarquis et al. (2008). The methods provide a systematic means to identify and
more importantly quantify the multiple scaling exponents in the data. The main feature of
multifractals is that they are characterized by high variability on a wide range of temporal
or spatial scales, associated to intermittent uctuations and long-range power law
correlations. First, the structure function was applied obtaining (q) and from this the
generalized Hurst exponents H(q). Later, a detrended uctuation analysis was used to
estimate H(q) and then infer (q). Comparison of these two techniques suggests that both
methods allow a reliable multifractal characterization of multifractal not stationary time
series showing a general agreement in the results.
The Hurst exponents H(q) are defined as:
H (q ) =
(q )
(3.24)
where q is order of structure function and (q) is a monotonically non-decreasing function

of q if v(t) has absolute bounds.
In other work, Tarquis (Platonov et al. (2008)) uses the moment method to implement
multifractal analysis. The characterization of multifractal measures is the concept of
generalized dimensions Dq, which corresponds to the scaling exponents for the q-th
moment of the measure. They are dened as
n ( )
1
Dq = lim
0 1 q
log mi
90
i =1
log
(3.25)
where is the length size of the box and n() is the number of boxes in which mi > 0 (m is
mass of the measure).
Other example is investigation of turbulence intermittency, small-scale phytoplankton

patchiness and encounter rates in plankton done by Seuront et al. (2001). However,
previous estimates of contact rates are implicitly based on homogeneous distributions of
both turbulent kinetic energy dissipation rates and phytoplanktonic prey, while turbulent
processes and phytoplankton cell distributions have now been demonstrated to be highly
intermittent even on small scales. Turbulent kinetic energy dissipation rates and intermittent
(i.e. patchy) phytoplankton distributions can be wholly parameterized in the frame of
universal multifractals. A new field of marine research has recently been proposed, with the
introduction of multifractals, for the analysis and modelling of marine intermittency. The
analysis of high-resolution time series of temperature, salinity and in vivo fluorescence
recorded in the Eastern English Channel and the Southern Bight of the North Sea, showed
that these fields are not homogeneous, but rather present bursts of activity at all scales, also
called intermittent fluctuations, with skewed distributions far from Gaussian. These
fluctuations have been analyzed in the multifractal framework, with statistical tools that are
more general than standard methods (such as spectral analysis), which provide only limited
information on the statistics of the process. They generalize this approach to fully take into
account the intermittency of turbulent processes and phytoplankton distribution.
In this work we used the multi-fractal analysis of SAR and experimental jets (plumes)
images looking for relationship between these two kinds of jets with respect to multi-fractal
dimension analysis.
91
CHAPTER 4
METHODOLOGY AND EXPERIMENTAL SETUP
4.1 Introduction
The present work is based on experimental and environmental study of different types of
jet and plume configurations showing similarity of the occurring phenomenon at different
scales. The experimental part of this thesis has been performed with the experimental
techniques and the apparatus available in the laboratory of Fluid Dynamics of the Applied
Physics Department of the Universitat Politecnica de Catalunya.
Our aim was to extend the existing measurements on turbulent structure on flows with
self-similar wall-jet type configurations both with detailed 3D velocimetry and with flow
visualization and by extending the range of velocity and plume concentration
measurements, as well as their stability to allow better turbulence statistics, then be able to
get new results on the detailed structure of non-homogeneous turbulent cascade processes
and thus complement previous experiments, which were mostly concerned with mean
structure and global fluxes. The previous thesis of Mahjoub (2000) had investigated the
structure function analysis of grid wakes and specifically the wake of a 2D cylinder using
Extended Self Similarity in complex non/homogeneous flows. The application of structure
function and multifractal analysis allowed us to develop and test measurement techniques
that could be also used in the environment.
Experimental techniques develop very fast thanks to the increase in computer power and
the advances in sensor engineering; so we can take advantage of the new technology and
methodology that will increase our knowledge. We may also learn new things about
turbulence structure even repeating some classical experiments under the new light and
using improved techniques. Moreover application of different experimental techniques in a
synergic way is a check on their usefulness showing the advantages and faults of each of
the techniques, both in the laboratory and with field data and thus, allows us to confirm and
92
expand the previous results. Using on the same experiments more than one method of
diagnostic permits us to improve on the understanding of the different laboratory
techniques; this is also usually an important argument for a multidisciplinary research
group. Present work was performed in the Fluid Dynamic laboratory equipped, among
others, with Laser Induced Fluorescence (LIF), Particle Image Velocimetry (PIV), Particle
Tracking Velocimetry (PTV) and Acoustic Doppler Velocimeter (ADV). These are very
useful velocimetry methods measuring velocity field of industrial and environmental flows
and used to solve fluid dynamics problems or to study fluid networks. There are also
greater difficulties in measuring environmental turbulent flows than laboratory or industrial
or process control turbulent applications, and the possibility of exporting measurement
methods from the laboratory to the field and the creation of new kinds of fluid flow sensors
are also advantages of these lines of experimental fluid dynamics methodology.
In this chapter we will describe the main experimental techniques available (in the
laboratory of UPC Applied Physics department) and discuss the application and the
methodology of the experimental techniques that have been used during the elaboration of
this thesis. We present also some examples that have been useful in interpreting and
understanding the ADV velocity data presented in chapters 5 and 6.
4.2 Acoustic Doppler Velocimeter (ADV)

The Acoustic Doppler Velocimeter is measuring device from SonTek company able to
measure turbulence using ultrasounds. The ADV is a versatile, high-precision instrument
used to measure 2 and 3D water velocity components in a non intrusive way, as at the
actual measuring position there are no wakes or recirculation due to the sensor, there are
many possible configurations as seen in (Figure 4.1).
93
Figure 4.1 Acoustic Doppler Velocimeter and two different measuring probes (source:
SonTek).
The ADV is used to measure water velocity in a wide range of environments including
laboratories, rivers, estuaries and the ocean. It uses acoustic Doppler technology to measure
a flow in a small sampling volume of a few cubic millimetres located at fixed distance (5 or
10 cm) from the probe, there are emitting and receiving transducers and the 2 or 3
components of the velocity are obtained from the Doppler frequency shift due to the local
fluid motion, its small sampling volume and fast response frequency make the ADV an
excellent turbulence measuring instrument. The basis of the sound focusing is shown in
(Figure 4.2).
Figure 4.2 ADV probe configuration (source: SonTek).
The precise location of the ADV sampling volume is determined by the probe geometry.
In most flow regimes, a 5-cm distance to the sampling volume is sufficient to avoid flow
interference. In oscillatory flow such as breaking waves, a 10-cm distance may be
94
preferred. The drawbacks of using the 10-cm distance are a decrease in signal strength (6-8
dB) and an increase in the minimum water depth.
The ADV probe configuration is determined by a combination of four factors:
sampling volume location, coordinate resolution (3D or 2D), sensor mounting, and sensor
orientation. The probes can be constructed with almost any combination of these options.
The velocity range is programmable from 3 to 250 cm/s. The measure flow is
practically undisturbed by the presence of the probe. Data can be acquired at sampling rates
up to 25 Hz. With no zero offset, the ADV can measure flow velocities from less than 1
mm/s to over 2.5 m/s.
The ADV consists of three basic elements: the probe, the signal conditioning module,
and the processor. The probe is attached to the conditioning module, which contains lownoise receiver electronics enclosed in a submersible housing. The ADV conditioning
module and probe are connected to the processing module.
The ADV Field processor is a set of three printed circuit cards that operates from
external DC power and outputs data using serial communication or a set of analogue
voltages. The ADV Field can be operated from any PC-compatible computer or can be
integrated with a variety of data acquisition systems.
Typical ADV applications are: low flows, turbulence, small-scale structures, boundary
layer, acoustic altimetry, coastal oceanography, open-channel flow, flow-field mapping,
and sediment transport or water treatment.
ADV calibration factors are determined by the speed of sound and by the angles
between transmit and receive transducers. To ensure that the correct speed of sound is used,
the water temperature and salinity must be entered in the data acquisition software. The
calibration angles are measured at the factory and need only be changed when a new probe
is installed. Maintenance calibration is not required unless the probe is physically damaged.
The processing module performs the digital signal processing required to measure
Doppler shifts. This computationally intensive task is implemented on a PC-board. The
95
data are recorded to disk in highly compressed binary files, which can be easily converted
to ASCII format with the data conversion programs supplied with the system (Figure 4.3).
240
80
100
200
80
160
60
SNR [db]
Correlation [%]
Amplitude [cm/s]
60
x
y
z
x
y
z
40
40
120
20
20
x
y
z
80
0
10
Time [s]
0
0
10
Time [s]
10
Time [s]
80
x
y
Velocity [cm/s]
40
0
0
10
Time [s]
-40
-80
Figure 4.3 ADV software interface and different values recorded simultaneously (velocity
amplitude, correlation, SNR (signal-to-noise ratio) and velocity).
The amplitude of three velocity components, correlation, SNR and fluctuation

velocities are shown in this figure and they are explained more detailed in Chapter 5.
The water must contain some suspended particles. The relatively strong echo that this
particle generates per unit of concentration motivates the choice of size. The Doppler Effect
(or Doppler shift) is the change in frequency of a wave for an observer moving relative to
the source of the wave.
96
Although Acoustic Doppler Velocimetry has become a popular technique in laboratory

in field applications, several researchers pointed out accurately that the ADV signal outputs
include the combined effects of turbulent velocity fluctuations, Doppler noise, signal
aliasing, turbulent shear and other disturbances. In turbulent flows, the ADV velocity
outputs are a combination of Doppler noise, signal aliasing, velocity fluctuations,
installation vibrations and other disturbances. The signal may be further affected adversely
by velocity shear across the sampling volume and boundary proximity. Spikes may be
caused by aliasing of the Doppler signal. Wahl (2000) and other authors developed
techniques to eliminate aliasing errors called "spikes". Another technique for removing
measurement spikes in two phase flows when bubbles are entrained in the flow is described
by Rodriguez et al. (1999) when under wave induced turbulence the ADV was compared
with Electromagnetic velocimeters.
These methods were developed for steady flow situations and tested in man-made
channels. Using only "raw" ADV velocity data are not "true" turbulent velocities and they
should not be used without adequate post-processing. In this study we used WinADV
software in order to post-process our data described in next point.
Voulgaris and Trowbridge (1997) confirmed that mean flows measured with the ADV
agree within 1% with those measured by the LDV (Laser Doppler Velocimetry).
Turbulence intensity of the vertical component can be resolved accurately by the sensor,
while the intensity of the downstream components suffers from a high noise term that is an
inescapable feature of the geometry of the ADV. They accentuated, using a technique
tested for three typical environments, that the Doppler noise can change the true turbulence
characteristics significantly, even for high level turbulence flows.
More information about Acoustic Doppler Velocimeter you can find in Sontek
Webpage: www.sontek.com.
97
4.3 WinADV
WinADV (Figure 4.3) provides an integrated environment for viewing and processing
data collected using Acoustic Doppler Velocimeters (ADVs). WinADV was inspired by
the experiences of the Bureau of Reclamation's Water Resources Research Laboratory in
the use of ADV's for laboratory and field velocity measurements.
The main screen of WinADV is devoted to presenting ADV data in graphical form on
a time series velocity data graph and signal quality graph. Graphical controls allow the user
to select parameters for plotting, zoom in on portions of the data set, modify the graph
style, or select probes from files containing data for multiple ADV probes. Subsidiary
forms provide for review of ADV configuration data, adjusting filtering and sampling
options, previewing summary statistics, and processing files to produce data files and
summary statistics that can be imported to spreadsheets, graphics or other analysis
software.
WinADV provides the capability to quickly view time series graphs, histograms, or
FFT's of the various velocity data. Data may be filtered on-the-fly and sampling windows
may be established so that only relevant data are plotted. For data collected from moving,
or traversing probes, an equation describing the position of the probe as a function of time
may be entered so that ADV data may be directly plotted versus probe position. Velocity
data can be transformed from model to prototype scale using scaling factors. All of these
filtering, sampling, traversing, and scaling options may be saved and recalled for later use.
WinADV provides routines for processing entire ADV files or portions of ADV files
defined by sampling windows. Processing includes calculation of average velocities,
turbulence parameters, and aggregated measures of data quality. A single ADV file may be
viewed and processed, or collections of ADV files may be processed as a group, either
using the visual interface, or using command-line options.
Summary statistics for filtered and/or unfiltered data may be output in suitable formats
for import into Lotus 1-2-3, Microsoft Excel, or other graphics or analysis software.
Filtered and/or unfiltered time series data files combining velocity, signal amplitude,
correlation, signal-to-noise ratio, and analog/digital data may also be produced for use in
subsequent analyses external to WinADV.
98
WinADV was created for several purposes:
To permit easy review of ADV data files in the office following data collection
To simplify filtering of data and aid in the identification of anomalies caused by low
signal strength or overranging of the ADV probe
To simplify processing of data collected using flag markers, and facilitate various
schemes for the use of flag markers during data collection
To simplify the analysis of ADV data collected from moving probes
To simplify the analysis of ADV data collected under harsh field conditions where
marking of flag points and diligent attention to signal quality are more difficult
To permit batch processing of large numbers of ADV data files
To minimize the need for manipulating and analyzing large quantities of time series
data.
To simplify analysis of data from scale models and for situations in which the probe
orientation is opposite from the desired coordinate system.
4.4 Laser Induced Fluorescence (LIF) and Planar Laser Induced

Fluorescence (PLIF)
Laser-induced fluorescence (LIF) is a spectroscopic method used for studying structure
of molecules, detection of selective species and flow visualization and measurements (see
Figure 4.4). It is a full field and non-intrusive. The main advantage of fluorescence
detection compared to absorption measurements is the greater sensitivity achievable
because the fluorescence signal has a very low background. For molecules that can be
resonant excited, LIF provides selective excitation of the analyte to avoid interferences.
99
Figure 4.4 Laser Induced Fluorescence (photograph by Alistair Arnott)
The species to be examined are excited with a laser. The wavelength is often selected
to be the one at which the species has its largest cross-section. The excited species will after
some time, usually in the order of few nanoseconds to microseconds, de-excite and emit
light at a wavelength larger than the excitation wavelength. This light, fluorescence, is
measured. Fluorescent dyes (molecules) can absorb light at one frequency and subsequently
reemit (fluoresce) light at a different frequency.
Two different kinds of spectra exist, disperse spectra and excitation spectra. The
disperse spectra are performed with a fixed lasing wavelength, as above and the
fluorescence spectrum is analyzed. Excitation scans on the other hand collect fluorescent
light at a fixed emission wavelength or range of wavelengths. Instead the lasing wavelength
is changed.
An advantage of this method is that it is possible to get two and three-dimensional
images since fluorescence takes place in all directions (i.e. the fluorescence signal is
isotropic). The signal-to-noise ratio of the fluorescence signal is very high, providing a
good sensitivity to the process. It is also possible to distinguish between more species, since
the lasing wavelength can be tuned to a particular excitation of a given species which is not
shared by other species.
100
The LIF technique is used widely in research for a variety of applications. It has been
successfully applied for quantitative measurement of concentrations in fields like
combustion, plasma, spray and flow phenomena (such as Molecular tagging velocimetry).
Fluorescence imaging is a tool of increasing importance in aerodynamics, fluid flow
visualization, and non-destructive evaluation in a variety of industrial circumstances. It is a
means for producing two-dimensional images of real surfaces or fluid case-sectional areas
that correspond to data on properties such as temperature or pressure across the surfaces.
More advanced form of LIF is Planar Laser Induced Fluorescence (PLIF). It is a
nonintrusive optical diagnostic tool for making temporally and spatially resolved
measurements. For illumination, a laser beam is formed into a thin sheet and directed
through a test medium. The probed volume may contain a mixture of various gaseous
constituents and the laser may be tuned to excite fluorescence from a specific component.
Alternatively, the medium may be a homogenous fluid into which a fluorescing tracer has
been injected. An imaging system normal to the plane of the imaging sheet views the laserirradiated volume. Knowledge of the laser spectral characteristics, the spectroscopy of the
excited material, and other aspects of the fluorescence collection optics is required for
quantifying the parameter of interest.
A typical PLIF setup is shown schematically in Figure 4.5.
Figure 4.5 PLIF configuration (source: Dantec Dynamics)
Applications of PLIF can be found in process engineering (e.g. mixing in stirring

vessels, heating and cooling systems), biomedical engineering (e.g. transport of drugs in
101
biological flows such as in model veins) and fluid dynamics research (e.g. turbulent mixing
and heat transfer modelling, indoor climate etc.)
The basic equipment needed to carry out planar-LIF measurements is a laser source
with the appropriate optics to form a thin sheet of light, a fluorescent dye that marks the
fluid and that is traced during the measurements. This chemical compound absorbs the laser
light energy and re-emits light at a longer wavelength that can be detected by a
photodetector. Finally we need a CCD camera equipped with a sharp cut-off or narrowband filter, so that only the fluorescent light is recorded. This camera acts as an array of
light detectors (pixels).
Since its conception in the early 1980's, PLIF has grown to become a powerful and
widely used diagnostic technique.
General PLIF theory is described by Allison and Partridge, between others.
4.5 Particle Image Velocimetry (PIV)

Particle Image Velocimetry (PIV) is a whole-flow-field technique providing
instantaneous velocity vector measurements in a cross-section of a flow. Two velocity
components are measured, but use of a stereoscopic approach permits all three velocity
components to be recorded, resulting in instantaneous 3D velocity vectors for the whole
area. The use of modern CCD cameras and dedicated computing hardware, results in realtime velocity maps. The fluid is seeded with tracer particles which, for the purposes of PIV,
are generally assumed to faithfully follow the flow dynamics. It is the motion of these
seeding particles that is used to calculate velocity information of the flow being studied.
The main difference between PIV and other techniques is that PIV produces twodimensional vector fields, while the other techniques measure the velocity at a point.
During PIV, the particle concentration is such that it is possible to identify individual
particles in an image.
Typical PIV apparatus (see Figure 4.6) consists of a camera (normally a digital camera
with a CCD chip in modern systems), a high power laser, an optical arrangement to convert
the laser output light to a thin light sheet (normally using a cylindrical lens and a spherical
102
lens), a synchronizer to act as an external trigger for control of the camera and laser, the
seeding particles and the fluid under investigation.
Figure 4.6 Particle Image Velocimetry scheme (source Dantec Dynamics).
The features of this technique are that it is non-intrusive and measures the velocities of
micro particles following the flow with velocity range from zero to supersonic. We get
instantaneous velocity vector maps in a cross-section of the flow and all three components
may be obtained with the use of a stereoscopic arrangement and with sequences of velocity
vector maps, statistics, spatial correlations and other relevant data are available.
Results are similar to computational fluid dynamics, i.e. large eddy simulations, and
real-time velocity maps are an invaluable tool for fluid dynamics researchers.
In PIV, the velocity vectors are derived from sub-sections of the target area of the
particle-seeded flow by measuring the movement of particles between two light pulses:
U=
X
t
(4.1)
The flow is illuminated in the target area with a light sheet. The camera lens images the
target area onto the CCD array of a digital camera. The CCD is able to capture each light
103
pulse in separate image frames. Once a sequence of two light pulses is recorded, the images
are divided into small subsections called inter-rogation areas (IA). The interrogation areas
from each image frame are cross-correlated with each other, pixel by pixel. The correlation
produces a signal peak, identifying the common particle displacement, X. An accurate
measure of the displacement - and thus also the velocity - is achieved with sub-pixel
interpolation. A velocity vector map over the whole target area is obtained by repeating the
cross-correlation for each interrogation area over the two image frames captured by the
CCD camera.
Recording both light pulses in the same image frame to track the movements of the
particles gives a clear visual sense of the flow structure. Any particle that follows the flow
satisfactorily and scatters enough light to be captured by the CCD camera can be used. The
number of particles in the flow is of some importance in obtaining a good signal peak in the
cross-correlation. As a rule of thumb, 10 to 25 particle images should be seen in each
interrogation area.
When the size of the interrogation area, the magnification of the imaging and the lightsheet thickness are known, the measurement volume can be defined.
4.6 DigiFlow
DigiFlow has been designed from the outset to provide a powerful yet efficient
environment for acquiring and processing a broad range of experimental flows to obtain
both accurate quantitative and qualitative output, aimed specifically at fluid dynamics.
Conceptually it has its origins in an earlier system by the same author (DigImage,
originally released in 1992), but has been completely rethought and rewritten to provide a
far more flexible and easy to use package.
Key features in the package include:
Particle tracking velocimetry (Lagrangian)
Particle image velocimetry (Eulerian)
Synthetic schlieren (density/refractive index/surface deformation)
104
Dye attenuation (density/concentration/thickness)
Light Induced Fluorescence (LIF; correction for attenuation and divergence)
Powerful time series handling
General purpose image processing
Recipe cards
Advanced macro language
Code library
Optional direct control of digital video camera with real-time processing
It can be used for tasks as diverse as running numerical simulations of 2D and of

shallow water flows, and for providing a great timer for conference presentations.
Whereas most image processing systems are intended for analysing or processing
single images, DigiFlow is designed from the start for dealing with sequences or collections
of images in a straightforward manner. Efficiency is obtained through the use of advanced
algorithms (many of them unique to DigiFlow/DigImage) for built-in processing options.
DigiFlow builds on experience with DigImage from the user view-point to provide a
more powerful, more flexible, but simpler interface. A central feature of DigiFlow is a
powerful macro language (dfc) and interpreter. This provides users with an efficient and
flexible environment in which to automate and customise processing, as well as proving to
be a very useful general computational and plotting tool. Power and flexibility are obtained
through an advanced fully integrated macro interpreter (using DigiFlows dfc macro
language) providing a similar level of functionality to industry standard applications such
as MatLab. This interpreter is available to the user either to directly run macros, or as part
of the various DigiFlow tools to allow more flexible and creative use. Commercial versions
of DigiFlow include additional features such as partial compilation to further improve
performance.
Although not an essential component, DigiFlow retains the potential DigImage
released by the control of a frame grabber. Not only does this greatly simplify the process
of running experiments, acquiring images, processing them, extracting and plotting data,
105
but it also enables real-time processing of particle streaks and synthetic schlieren, for
example.
4.7 Particle Tracking Velocimetry (PTV)

Particle tracking velocimetry (PTV) is a velocimetry method, i.e. a technique to
measure velocity of particles. The name suggests that the particles are tracked, and not only
recorded as an image as it is suggested in another form, particle image velocimetry. The
three-dimensional Particle tracking velocimetry (3D-PTV) is a distinctive experimental
technique, based on multiple camera-system, three-dimensional volume illumination and
tracking of flow tracers (i.e. particles) in three-dimensional space by using
photogrammetric principles.
A typical installation of the 3D-Particle tracking velocimetry consists of three or four
digital cameras, installed in an angular configuration, synchronously recording the
diffracted or fluorescent light from the flow tracers, seeded in the flow. The flow is
illuminated by a collimated laser beam or by another source of light. There is no restriction
on the light to be coherent or monochromatic and only its illuminance has to be sufficient to
illuminate the observational volume. Particles or tracers could be fluorescent, diffractive,
tracked through as many as possible consecutive frames on as many cameras as possible. In
principle, two cameras in the stereoscopic configuration are sufficient in order to determine
the three coordinates of a particle in space, but in most practical situations, three or four
cameras are necessary.
4.8 ImaCalc
The theory of the fractal analysis was described in Chapter 3. It can be applied to
distinguish between the natural slicks and man-made features in SAR imagery, also in
particular for ship wakes. The fractal analysis of the structures formed by different tones of
grey may be determined in terms of a multi-fractal analysis. In this work we present the
application of the multi-fractal analysis in a user-friendly PC based program ImaCalc. The
differences in multi-fractality are detected using the multi-fractal box counting algorithm
106
on different sets of SAR images gives information on the age of the ship wake and jets
(plumes).
The theory and applications of fractal analysis is a rapidly evolving research field both
from a mathematical approach and in experimental and field applications. The basic theory
and method of box-counting used in ImaCalc tool Grau (2006) was followed and we used
self-similarity to identify different dynamic processes that might influence the radar backscattering from the sea surface. The image analysis algorithms are able to detect the selfsimilar characteristics for different SAR intensity levels i.
The fractal dimension D (i) is then a function of intensity and may be calculated using:
D (i ) =
log N (i )
log e
(4.2)
where N(i) is the number of boxes of size e needed to cover the SAR contour of intensity i.
The algorithm operates dividing the surface into smaller and smaller square boxes and
counting the number of them which have values close to the SAR radiation level under
study.
The work environment used for the multifractal characterization is an application
called ImaCalc created by Joan Grau. The program is guided to calculate the fractal
dimension of images using the Box-Counting algorithm. With this application we can
define in an image the region of interest, select an intensity range to analyze and execute
the multifractal characterization process in a simple iterative way.
Some of characteristics of the ImaCalc program are:
Work method
Layer definition
Properties of the work zone
Geometric calculations
Fractal analysis
107
- Box-Counting 2D
- Iterate Box-Counting 2D
In this software there is implemented a selection methodology that permits the use of
irregular polygons that delimit the region of interest, this tool works jointly with a zoom to
specify better the corresponding selection. On the other hand it is also possible to limit or
choose the intensity range (gray levels) to be used. In this way the analysis remains
delimited in a region so much in the physical space as in the intensity. Figure 4.7 shows an
example of a working region selected.
Figure 4.7 ImaCalc program interface. Different windows are shown presenting working
region selected, histogram of the image, zoom and fractal dimension result.
The used calculation algorithm is the Box-Counting method cited before. It is

permitted to use the different work selection methodologies and a wide choice of fit
constrains and intensity ranges. A single exclusive contour dimension calculation may be
performed, giving an only dimension value as a result, its application tends be centred
exclusively to plane curves characterization. Also it can be chosen to iterate the multifractal
analysis method with any intensity work range defined and the possibility of several
divisions in the calculation subintervals is also an option. As has been commented
108
previously, this methodology sections the entire interval in a set of subintervals with the
specified dimensions and on each one of them accomplishes the calculation of the
corresponding dimension.
The different related calculation options with the explicit methodology used in the
different subinterval may be obtained.
The application incorporates a results visualization window. Consist of a graph that
shows us the points obtained in the fractal dimension calculation using the Box-Counting
method jointly with the corresponding linear fit. There appears too other smaller graph that
shows us the observed dimension values as a grey-level function of the characterized
intervals.
In addition to the direct visualization of the data, the results also are kept in a file in
order to be analyzed independently with external tools. The formats of the images and other
import-export abilities of the program ImaCalc are very wide.
Example of the SAR image treatment using ImaCalc is presented in Figure 4.7
showing photograph of the analysis and corresponding histogram, regression, zoom and
fractal dimension windows.
4.9 Experimental setup
Experiments were carried out in the laboratory equipped with the tools and techniques
described in this chapter.
Experimental setup for different jets configurations is presented on Figure 4.8.
Workplace consists of Perspex tank of about 400 cm of longitude, 40 cm of height and 26
cm of width. We performed three groups of experiments with different outlet velocities
(outlet Reynolds number) changing the jets source (two kinds of pumps were used called
here small and big).
The flow parameters for experiments are:

a) Small pump: mass flow: Q = 64 cm3/s, outlet velocity: U = 32 cm/s, diameter of the
nozzle: D = 1,6 cm;
109
b) Big pump:
Single phase jet (called here shorter Normal jet) with mass flow: Q = 158,2 cm3/s, outlet
velocity: U = 166 cm/s, diameter of the nozzle: D = 1,1 cm
Two phase jet (called here Bubbly jet) with mass flow: Q = 204,2 cm3/s, outlet velocity:
U = 65 cm/s, diameter of the nozzle: D = 2 cm.
The Bubble jet case is obtained using special pump-nozzle connection guaranteeing
constant air water content.
Reynolds numbers may be calculated using these parameters and taking into account
the water viscosity with temperature between 15 and 20C.
The two pumps versions are shown on Figure 4.8.
Figure 4.8 Two cases of workplace (the small and big pump). Basic geometrical parameters
are shown.
There is water in the tank. The jets are produced by the pumps in a recirculation
system, hence constant parameters of the jet and ambient are guaranteed by presence of a
cooling system to keep relatively constant flow and ambient temperature. The level of the
water is 30 cm height. In the tank is submerged the pump, which produces the jet (in a
small pump case) and special submerged nozzle in the case of big pump (it is situated
110
outside of the tank). The pump (or nozzle) is situated 15 cm above the bottom of the tank (a
middle of the water level) and the ADV measuring probe is 5 cm higher because of the
ADV probe conditions (measuring point situation). Co-ordinate system is oriented as is
shown in Figure 4.8. Zero value of z coordinate corresponds with the axis of the jet (15 cm
above the bottom of the tank). We changed y distance from the wall in order to get different
jets configurations (from the free to the wall jet). We have to mention that geometrical
characteristics of the applied tank make our jets like channel free (or wall) jets, moreover
limited by the end wall situated for large values of the distance x and having clear
influence in the behaviour of them as will be shown further. Three distances of y were
taken into consideration: 4, 8 and 13 cm (the last one corresponds to the middle of the
tanks width) (Figure 4.9) to compare measurements. On this figure just examples of points
are shown. In the case of big pump, much more points were considered, to get more
detailed behaviour of the jet in different zones (close to the nozzle, zone of the boundary
layer appear and zone of the end walls recirculation effect for the farther downstream
distances x. The same y (4, 8 and 13 cm) values were considered both in small pump and
big pump cases.
Figure 4.9 Examples of points of measurements
Our results were performed for 110 nozzle diameters (in the case of higher Reynolds
number and single phase het) increasing the range comparing with other results (Law and
Herlina 2002).
The error associated with the local Reynolds number was calculated from the general
formula:
111
UD
Re = f
(4.3)
where :
U - streamwise velocity difference between previous and actual measurement for i-serie.
- kinematic viscosity difference for temperature change (2C for 15 minutes),
= 0,054 10 2 cm / s 2
Re =
Re
Re
U +
V xi
(4.4)
Re U
Re Re
=
+
Re
U
where Re - the mean Reynolds number.
The error of the Reynolds number is shown in Figure 4.10.

40
Re/ReAV [%]
30
20
10
0
0
10
20
30
40
x/D
Figure 4.10 Reynolds number error as an x/D distance for smaller Reynolds number
case.
Sekula and Redondo (2006) reported more detailed description of the lower Reynolds
number case experiment.
The roughness of the tank walls is ignored because of the slight value of the surface
roughness ( S = 0,001524 mm) taken for the fanning friction factor and correspond Reynolds
number.
Described before, the Acoustic Doppler Velocimeter was used to get velocity data and
other important magnitudes to calculate statistic and turbulent parameters. Measurements
112
were performed during 10 minutes for one point (10 minutes = 600 s * 25 Hz = 15000 data)
in order to get large data that allow us to calculate with more detailed analysis the turbulent
parameters and specially energy spectrums and structure functions. In total we have about
133 MB of the ADVdata results. Results are presented in next chapter.
There is not clear influence of circulation due to the finite tank width. Matulka (2009)
shows that relating kinetic energy and dissipation as (from the spectral relationship (K41)):
k C 3
(4.5)
and integrating
3
dk
= c( )k 2
dt
(4.6)
that depends strongly on initial conditions, in particular with k0.

In order to illustrate the used configuration some image capturing examples are shown
in Figure 4.11. The overall downstream jet range is shown to be about x/D = 110 and the
surface waves due to the end wall and recirculation effects are visible (top). At some
streamwise distance larger scales caused by the wall effect (finite width of the tank) are
noticeable apart from x/D 60. The complexity of this case is intensified by the presence of
the vortical structures (tornado structures) appearing randomly close to the surface
(bottom).
113
Figure 4.11 Images of experimental jet. The overall jet view (top) and zoomed close to
the nozzle region (bottom).
As an example of the ADV method used in our laboratory we mention here the work of
Ilse van de Voort. Graphical representation of the experimental setup is shown in Figure
4.12.
114
Figure 4.12 Graphical representation of the grid stirred turbulence setup (van de Voort).
Experiments were executed by making use of a rectangular tank near the bottom of
which a grid is stirred to produce zero mean flow turbulence. The ambient fluid in these
experiments was fresh water. From the top of the tank instantaneously a thermal with
negative buoyancy was released, with different densities. The thermal fluid was seeded
with particles. For a characterization of the flow generated by grid stirred turbulence, the
results from (Redondo and Metais 1995) have been used. The relations for the root-meansquare velocity u and the integral scale of turbulence l on the grid characteristics at a
distance z from the centre plane are as follows:
u ' = cM 1/ 2 s 3 / 2 z 1
l = z
(4.7)
(4.8)
where is the grid frequency, M its mesh size, s its stroke length and c and constants of
proportionality which vary with the particular grid used.
By using the ADV, all three velocity components were measured in one point with a
high sampling rate. This velocity information was used to obtain the velocity power
spectrum for different environmental turbulence levels. For the setup without
environmental turbulence, the slope of the velocity spectrum is more gradual than for the
115
setups with turbulence. For some latter parts of the energy spectrum, graphs display a
steeper slope and there appears to be a bending point in the characteristic, which is caused
by the presence of the environmental turbulence.
116
CHAPTER 5
EXPERIMENTAL RESULTS ADV MEASUREMENTS
5.1 Lower Reynolds number jet experiments

5.1.1
Velocity profiles
We present velocity profiles in two cross-sections (x = 4 (blue) and 11 cm (red

curve)). Decreasing of the mean velocity U is noticeable (Figure 5.1) with downstream
distance as proven in previous works. Notice that maximum value of U not correspond
with x-axis (y = 0cm). The effect of the nozzle modifies these profiles and in II crosssection the maximum velocity is moved farther from the jets axis. These effects are
detectable too if we compare our results with the classical Tolmien and Goertler
observations (Figure 5.1). We detect clearly the sharp decrease (for y/b between 0 and
1) in average streamwise velocity in the jet axis, near the nozzle, due to non
homogeneity and instability of the velocity fluctuations showing the influence on the
turbulence produced by the interaction of the nozzle and the axisymmetric jet. The
pump geometry could have other effect changing the velocity distribution. It is due also
to some vortex structures interacting with the basic jet flow. Some flow visualization
method may possibly confirm the existence of these phenomenons.
The Forthmann wall jet velocity profile is presented also in order to show difference
between the wall and free like jet cases showing dissimilarity for close to the jet axis.
1
10
I cross-section
II cross-section
0.8
y [cm]
U/UXMAX
Sekula - I cross
Sekula - II cross
Tollmien
Goertler
Forthmann - wall jet
0.6
0.4
0.2
0
0
10
20
30
y/b
U [cm/s]
Figure 5.1Velocity distribution in two sections (I. x = 4 cm (blue), II. x = 11 cm (red))

for the free jet case (y = 13 cm) and comparison of Sekula, Tollmien and Goertler
observations.
117
The measurement of velocity distributions in different cross sections for the free jet
let us to estimate the angle of the jet. Scheme of angle of the jet is shown in the sketch
of Figure 5.2.
Figure 5.2 Scheme of angle of the jet.
5.1.2
Jet velocities
In order to describe the effects of the wall we present three velocity components (U,
V and W) for three distances between the jet and nozzle centre and the wall (y = 4, 8 and
13 cm) (Figure 5.3). Obviously x-component shows much larger values that y and z and
in three cases the maximums correspond to the close to nozzle region. It is interesting
that for x-component we observe velocity peak at some distance from the outlet and this
peak appears closer to the nozzle as we are closer to the wall. It is related to the jet
geometry. We can observe also that this peak has less value as we are farther form the
wall, for y = 13 cm it is almost not detectable. For the very close to the nozzle
streamwise distances a not typical velocity distribution is exposed due to the potential
core and additional vortex structures occurring in this zone. Notice also, that in y = 4cm
case, streamwise velocity U decreases slower with x-distance than in other cases. Again,
the effect of the wall and boundary layer is observable. To see these effects we can use
log-log plots as shown in Figure 5.4. The notation used for the results is that total
velocity is a sum of the mean velocity U and the turbulent fluctuations (i or u ' (for x
component case)). Because a free jet entrains fluid from both sides, it spreads faster, and,
therefore, it centreline velocity decays faster than that for the wall jet in the flow
development region near the nozzle exit (see Figure 5.3).
118
Most of the existing works are based on the mean velocity results (Tachie et al.
2002 or Kwon and Seo 2005), here we focus our attention on the turbulent fluctuations
part.
40
40
4 cm
8 cm
13 cm
30
20
10
30
y = 4 cm
y = 8 cm
y = 13 cm
W [cm/s]
V [cm/s]
U [cm/s]
30
40
20
10
0
10
20
x/D
30
y = 4 cm
y = 8 cm
y = 13 cm
10
0
0
20
40
10
20
x/D
30
40
10
20
x/D
30
40
Figure 5.3 Jet velocity in x (left), y (middle) and z (right) direction for y = 4 cm, 8 cm
100
100
10
10
10
1
y = 4 cm
y = 8 cm
y = 13 cm
0.1
log W [cm/s]
100
log V [cm/s]
log U [cm/s]
and 13 cm.
y = 4 cm
y = 8 cm
y = 13 cm
0.1
0.1
10
100
y = 4 cm
y = 8 cm
y = 13 cm
0.1
0.01
0.01
0.01
0.1
10
100
0.1
log (x/D)
log (x/D)
10
100
log (x/D)
Figure 5.4 Logarithm U (left), V (middle) and W (right) for y = 4cm, 8cm and 13cm as
a logarithm (x/D) function.
5.1.3
Reynolds number
We can calculate the average Reynolds number describing the jet, in the usual way as:
Re =
DU
(5.1)
where: D is the diameter of the nozzle [m], U is the mean downstream velocity [m/s]
and - kinematic viscosity [m2/s]. Kinematic viscosities for suitable temperature are
shown in Table 5.1 and results of Reynolds number in Figure 5.5.
119
Temperature
Kinematic viscosity
15
m2
]
s
1,138 10 6
16
1,111 10 6
17
1,084 10 6
18
1,058 10 6
19
1,031 10 6
20
1,004 10 6
[C]
Table 5.1 Kinematic viscosity for water of different temperatures.
We can see that the Reynolds number is a function of velocity so the forms of
curves are similar to the velocity curves. For this reason we will limit here to present
only the y = 4 cm case. Usually only the VX component is used so the range of ReX for
these experiments would be 500 ReX 5000.
6000
4000
Re
x
y
z
2000
0
0
10
20
x/D
30
40
Figure 5.5 Reynolds number for the x, y, z component for y = 4 cm.
120
5.1.4
Standard deviation (r.m.s. turbulence)
Standard deviation (of data set or probability distribution), in probability theory and
statistics, is the square root of its variance. The variance of a random variable or
distribution is the expectation, or mean, of the deviation squared of that variable from
its expected value or mean. If a random variable X has the expected value (mean) =
E[X], then the variance of X is given by:
( X )2
(5.2)
Var ( X ) = E ( X )
(5.3)
The standard deviation is a widely used measure of the variability or dispersion,

being algebraically more tractable. It helps to detect tampering of data. A low standard
deviation indicates that the data points tend to be very close to the mean, whereas high
standard deviation indicates that the data are spread out over a large range of values.
In science, researchers commonly report the standard deviation of experimental
data, and only effects that fall far outside the range of standard deviation are considered
statistically significantnormal random error or variation in the measurements is in this
way distinguished from causal variation. We will observe here standard deviation peaks
showing spreading out of the velocity data (velocity fluctuations).
The form to calculate the standard deviation is:
2 = u ' 2 = u ' 2 B(u )du
(5.4)
B (u ) =
lim 1
T
T T
(5.5)
where: u or the Root-Mean-Square (RMS) or standard deviation, of the turbulent

velocity fluctuations (in this case for x-component), T is time and B(u) is the probability
density function of the turbulent fluctuations.
In Figure 5.6 we present an example of velocity data signal with the standard
deviation explanation.
121
Figure 5.6 Example of data signal with standard deviation explanation.
In Figure 5.7 we present the x, y and z component of the standard deviation or

turbulence r.m.s. values for our jet at different distance from the wall cases. We can
detect some interesting ranges of the jets. In all cases, close to the nozzle zone is having
large values. It is potential core zone where velocity fluctuations have important
significance. Then, at some downstream distance (about x/D = 10-20) there are standard
deviation peaks showing boundary layer effect on the jet developing. It is interesting to
see that in the close to wall jet case (y = 4cm) this effect is more visible and has more
influence on the jet length (for y = 4 cm is about 20 x/D, and for other cases about x/D =
9). The standard deviation for x and y direction are larger (these components are more
affected by the wall presence) than for z-direction. This is more visible if we compare
individual components separately (Figure 5.8) or (Figure 5.9).
40
40
sigma x
sigma y
sigma z
sigma x
sigma y
sigma z
20
20
10
20
x/D
30
40
20
10
10
10
sigma x
sigma y
sigma z
30
x, y , z - 13 cm [cm/s]
30
x, y , z - 8 cm [cm/s]
x, y, z - 4 cm [cm/s]
30
40
10
20
x/D
30
40
10
20
x/D
30
40
Figure 5.7 Standard deviation for x, y and z component for y = 4 cm (left), 8 cm

(middle) and 13 cm (right).
122
40
40
y = 4 cm
y = 8 cm
y = 13 cm
30
RMS [w'] [cm/s]
20
y = 4 cm
y = 8 cm
y = 13 cm
y = 4 cm
y = 8 cm
y = 13 cm
30
RMS [v'] [cm/s]
RMS [u'] [cm /s]
30
40
20
10
10
10
0
0
10
20
x/D
30
20
40
10
20
x/D
30
40
10
20
x/D
30
40
Figure 5.8 Standard deviation for x (left), y (middle) and z (right) component for y = 4, 8
and 13 cm.
The wall effect is more detectable if we plot together the mean velocity and r.m.s. of
velocity fluctuations for three distances y in separated mode (Figure 5.10). For the
streamwise component x, the rms has larger influence at some distance from the nozzle.
The velocity fluctuations have an important role for y component having larger values
that the mean value. The peaks between 10-20 jet diameters downstream are produced
by wall vortex wall as discussed in Benaissa et al.
100
100
100
log sigma x
log sigma y
log sigma z
10
log sigma x
log sigma y
log sigma z
log x, y , z - 13 cm [cm/s]
log x , y , z - 8 cm [cm/s]
log x, y, z - 4 cm [cm/s]
log sigma x
log sigma y
log sigma z
10
0.1
10
log (x / D)
100
10
0.1
10
log (x / D)
100
0.1
10
100
log (x / D)
Figure 5.9 Logarithm standard deviation for x, y and z component for y = 4 cm (left), 8
cm (middle) and 13 cm (right) as a logarithm (x/D) function.
123
40
40
40
y = 4 cm
U
x
30
30
y = 13 cm
y = 8 cm
U
x
20
U, x [cm/s]
x
U, x [cm/s]
U, x [cm/s]
30
20
10
20
10
10
10
20
30
40
x/D
10
30
y = 4 cm
V, y [cm/s]
y = 8 cm
20
10
20
30
40
40
V
y
20
10
0
0
30
y = 13 cm
V
y
10
40
30
30
10
30
40
V
y
20
20
x/D
V, y [cm/s]
30
10
40
40
40
V, y [cm/s]
20
0
0
10
20
x/D
x/D
30
40
10
20
x/D
Figure 5.10 The mean and r.m.s. of velocity fluctuations for the distances from the wall
and x (top) and y (below) component as an x/D function.
5.1.5
A uniform measurement scale of turbulence is needed to describe its level.

Turbulence intensity is that measurement scale.
Turbulence intensity is a scale characterizing turbulence expressed as a percentage
of the mean flow. An idealized flow with absolutely no fluctuations in velocity or
direction would have a turbulence intensity value of 0%. Such an idealized case never
occurs. However, due to how turbulence intensity is calculated, values greater than
100% are possible. This can happen, for example, when the average velocity is small
and there are large fluctuations present.
The turbulence intensity for i component is defined in the following equation:
Ii =
i (i )
U max
124
(5.6)
In some studies Umax is replaced by U but in this work we wanted to relate

turbulence intensity to the general flow system.
The level of turbulence (turbulence intensity) has implications in many different
fields. In the realm of aeronautics, drag on an airfoil (airplane wing) is related to
turbulence intensity affecting fuel efficiency and in extreme cases aircraft stability.
In the automobile industry, the measurement of turbulence intensity is used when
evaluating the aerodynamics of an auto body design. And it is not always low
turbulence intensity that is the goal: some race cars have flaps that deploy to increase
drag (higher turbulence intensity) when the car goes into a spin, in an effort to more
rapidly reduce the cars speed. When working with chemicals, sometimes high
turbulence intensity is a good thing (when intentional mixing is desirable) and
sometimes high turbulence intensity is not a good thing (at the face of a ventilation test
instruments where excessive turbulence intensity may induce unwanted spillage of
chemical fumes into a laboratory). In the realm of human comfort, heating or air
conditioning systems that are perceived to be too drafty are not comfortable. It has
been discovered that the perceived draftiness of air is a function of the turbulence
intensity of the moving air.
There are universal formulas for some flows to calculate turbulence intensity, for
example, for fully developed pipe flow at the core it can be estimated as:
I = 0.16 Re dh1 / 8
(5.7)
where Re dh is the Reynolds number based on the pipe hydraulic diameter dh.
Typically turbulence intensity is between 0.01 and 0.10 for many flows, but for
engines with often a very low average velocity, it could be higher and thus may not be
as relevant as plotting the turbulent rms u'. In environmental flows as a rule the
fluctuation are of the order of the mean so I 1.
It was confirmed in previous works that the flow structure of the sharp-edged orifice
plate jet in the near field is more three-dimensional than that of the smooth contraction
jet. The exit turbulence intensity of the sharp-edged orifice plate jet depends upon the
upstream conditions and these influences propagate downstream.
125
As we proven before, if the standard deviation is large, then the turbulent part of the
flow is large, i.e. the turbulence intensity is large. Turbulence intensity is related
directly with standard deviation so plotting curves will show similar results. We present
just them in order to show standard I values for our experiments (Figure 5.11). The high
values show that the jet at the nozzle was fully turbulent as also seen in Figure 5.1.
1
1
1
Ii = i(x) / Umax
Ii = i(x) / Umax
0.8
0.8
Ii = i(x) / Umax
y = 4 cm
0.6
x
y
z
0.4
y = 8 cm
x
y
z
0.6
Turbulent intensity
Turbulent intensity
Turbulent intensity I
0.8
0.4
0.2
0.2
0
10
20
x/D
30
40
y = 13 cm
x
y
z
0.4
0.2
0.6
10
20
x/D
30
40
10
20
x/D
30
40
Figure 5.11 Turbulence intensity I as an x/D function for y = 4 (left), 8 (middle) and 13
cm (right).
The conclusions are similar to these ones from the standard deviation analysis.
Effect of the wall and boundary layer is visible. It is interesting to observe that
turbulence intensities are very high, to not to say, extremely high comparing with
typical turbulent flows. Especially close to nozzle region is highly affected by the
turbulence level. Two reasons are possible, first one, the close to nozzle region is
sensitive for measurements quality (potential core) and second one, as was mentioned
before, values greater than 1 (or 100%) are possible and this can occur when the average
velocity is small and there are large fluctuations present. In our case the velocity is not
high but as mentioned above large turbulent fluctuations were measured by the ADV.
5.2 High Reynolds number wall jet experiments

5.2.1
Velocities
2
We present velocity results (U, V and VTOTAL ( VTOTAL = U + V )) of higher

Reynolds number and normal jet case (Figure 5.12). We can observe much larger
streamwise velocity values than for spanwise direction. The spanwise velocity has small
influence (comparing with U) in the total velocity VTOTAL. It is interesting to observe that
for very close to the nozzle jet region (x/D between 0 and 3) there is a very small value
126
of U. Then suddenly increase of this magnitude is noted but only for small range (x/D
3) and again much lesser (than for typical velocity decreasing profile) U is obtained.
Potential core in this range is observed making more sensitive conditions for
measurements. It was proven by different works that the ADV has some difficulties in
very turbulent flows and special post-processing data analysis is needed. Apart from
some downstream distance (x/D 22) the velocity decreasing profile is more stabilized
and in accordance with typical proven results (Webster et al. 2001). The boundary layer
and wall effects are not so perceptible like in small pump results for velocities results.
200
200
200
160
160
160
120
V [cm/s]
U [cm/s]
120
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
80
80
40
40
V TOTAL [cm/s]
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
40
80
x/D
120
120
80
40
0
0
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
0
0
40
80
120
x/D
40
80
120
X/D
Figure 5.12 Downstream velocity U (left), spanwise velocity V (middle) and total
velocity VTOTAL (right) for three distances y ( 4 cm black curve, 8 cm blue curve, 13
cm red curve) from the wall.
The WinADV program allows us to get, apart from the general mean and turbulent
velocities, more specialized turbulence statistics and other specific parameters; we will
define them here shortly presenting also the results for the different configurations used.
5.2.2
Turbulence and mean velocity parameters
Avg Vmag is the average of the individual velocity magnitude values for the whole
of the sampling time series (Figure 5.13). We can observe that velocity decay is more
stable than for each of the individual components and is more similar to the expected
proven results.
127
250
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
Avg Vmag [cm/s]
200
150
100
50
0
0
40
80
120
x/D
Figure 5.13 Avg Vmag - the average of the individual velocity magnitude values for the
time series for three distances y as a function of x/D.
Mag V- Avg shows the magnitude of the resultant of the two average velocity
components (Figure 5.14).
200
Mag V-Avg [cm/s]
160
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
120
80
40
0
0
40
80
120
x/D
Figure 5.14 Mag V- Avg - the magnitude of the resultant of the three average velocity
components for three distances y as an x/D function.
Span of 95 % Conf. Interval U, Span of 95 % Conf. Interval V are the width of

the confidence interval, centered on the average value of U or V (Figure 5.15).
128
Span of 95% Conf. Interval - V [cm/s]
Span of 95% Conf. Interval - U [cm/s]
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
40
80
120
40
x/D
80
120
x/D
Figure 5.15 Span of 95 % Conf. Interval U (left) and Span of 95 % Conf. Interval V
(right) - the width of the confidence interval, centered on the average value of U and V
for three distances y as a x/D function.
5.2.3
Reynolds number
The Reynolds number results are shown in Figure 5.16.

20000
16000
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
Re
12000
8000
4000
0
0
40
80
120
x/D
Figure 5.16 Reynolds number for three jet wall distances y as an x/D function.
5.2.4
Standard deviation
In the first part of this chapter we have explained definition of the standard
deviation. WinADV program has possibility to plot different magnitudes related to this
term. There are:
RMS [u], RMS [v] are the root-mean-square of the turbulent velocity fluctuations
(the square root of the mean of the deviations from the mean velocity) (Figure 5.17).
129
The RMS turbulence is equal to the standard deviation of the samples. For example, the
RMS turbulence for the x velocity component is defined as:
RMS [u '] =
(u ')
U (U ) / n
2
(5.8)
n 1
|RMS[V']| is the magnitude of the resultant formed from the individual RMS values
for each component (Figure 5.17).
RMS[Vmag'] is the root-mean-square value of the time series of individual velocity
200
160
160
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
120
80
300
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
200
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
120
RMS [V] [cm/s]
200
RMS [V Y] [cm/s]
RMS [VX] [cm/s]
magnitude values (Figure 5.17)
80
100
40
40
0
0
40
80
120
0
0
40
x/D
80
120
x/D
40
80
120
x/D
120
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
RMS [Vmag`] [cm/s]
100
80
60
40
20
0
0
40
80
120
x/D
Figure 5.17 RMS [u] (top left) , RMS [v] (top middle), |RMS[V']| (top right) and
RMS[Vmag'] (bottom) for three distances from the wall y as an x/D function.
Analysing the set of standard deviation plots we can observe various clear
characteristics of the jet turbulent velocity. For x component the strongest fluctuations
are observed for range x/D = 18 28. We should remember that the range before of
these values is questionable due to the potential core of the jet and sensitiveness of the
ADV measurements. It is interesting to observe that for y-component velocity
fluctuations reveal strongest changes; the effect of the wall is more perceptible for this
130
component. If we look more in detail on the previous plots (velocities and turbulence
parameters) we can perceive peaks of the boundary layer appearance. The effect is less
significant with larger values of y. We have:
For
y = 4 cm, x/D = 9
y = 8 cm, x/D = 23
y = 13 cm, x/D = 57
Having more measurements for different distances y we could plot the boundary
layer appearance point x/D as a y/D function.
5.2.5
Skewness
Skewness is the asymmetry distribution ratio around an average which contains

some information on eventual differences between positive and negative deviations
from the mean value. It is the third statistical moment. A skewness of zero indicates a
distribution that is symmetric about the mean. With WinADV software is possible to
calculate skewness value for each direction component. Its definition is:
n
3
2
3
Skewness X =
U 3 U U 2 + 2 (U )
3
n
n
(n 1)(n 2 )s
(5.9)
where s is the standard deviation (definite before as RMS or )
A critical aspect of ADV use is the choice of an appropriate velocity range for the
instrument. If the velocity range is set too low, aliasing of the velocity data may occur
when velocities exceed the maximum range, causing occasional velocity "spikes" in the
data. These spikes are caused when the phase difference being measured by the
instrument exceeds 180 degrees. At this point the instrument cannot differentiate
between a velocity that has exceeded the maximum positive range, and a velocity that
has nearly reached the maximum negative range. Thus, a spike from positive to negative
or vice-versa is produced. This biases the average and instantaneous velocity data.
Unfortunately, aliasing may occur without being detected by the user if close attention
is not paid to the ADV screen during data collection. Also, if data are being collected at
low sampling rates, occasional aliasing may be difficult to detect, since it gets averaged
in with otherwise good velocity readings.
131
In theory, a dataset containing significant aliasing should have a skewed

distribution, with a few samples located far away from the mean of the distribution. To
assist in identifying overranging and aliasing WinADV computes the skewness of the
velocity distribution for each probe beam and reports the maximum skewness (Figure
5.18). A rough guideline from limited experience suggests that data with skewness
values greater than about 1.5 should be more closely investigated. For x - component
case 97 % of results fulfils this condition, for y component 100 % is approved. Our
results seem to be not affected by aliasing.
Skewness is related to the third order structure function, which is related to the scale
to scale energy transfer.
3
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
2
Skewness y
Skewness x
0
0
40
80
120
x/D
40
80
120
x/D
Figure 5.18 Skewness of x (left) and y (right) component for three distances y.
5.2.6
Kurtosis
Kurtosis is the fourth statistical moment and characterizes the relative peakedness or
flatness of a distribution compared to the normal distribution (Figure 5.19). Positive
kurtosis indicates a relatively peaked distribution. Negative kurtosis indicates a
relatively flat distribution. The Kurtosis is defined as:
2
4
4
6
3
Kurtosis X = K1 U 4 U U 3 + 2 (U ) U 2 3 (U ) K 2
n
n
n
(5.10)
where:
K1 =
n(n + 1)
(n 1)(n 2)(n 3)s 4
3(n 1)
K2 =
(n 2)(n 3)
(5.11)
132
(5.12)
The kurtosis results confirm relatively peaked distribution of our data. Specially the
parts between x/D = 2 and x/D = 30 for x component and x/D = 5 and x/D = 50 for y
component are affected by larger values.
16
16
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
12
Kurtosis y
Kurtosis x
12
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
0
0
40
80
120
40
80
120
x/D
x/D
Figure 5.19 Kurtosis for x (left) and y (right) component for three distances y.
5.2.7
Correlation
The correlation parameter provides a measure of the quality of each sample (0

indicates the lowest quality to 100 the highest quality). It can be used to filter out bad
samples of the data file. A typical setting for this filter is to remove samples with a
correlation less than 70. WinADV permits the user to specify this criterion, a different
filter criterion or combination of criteria. Furthermore, WinADV permits filtering to be
done using the average correlation for all three channels or the minimum correlation
from amongst the three channels. Filter criteria are expressed in terms of the samples to
be retained. Low correlation values may indicate problems related to turbulence, signal
strength, scatterer density, excessive air bubbles, or problems with the probe itself. We
should mention here that the recommendation of Avg COR > 70 is difficult to interpret.
Some works prove definitely situations in which COR values much less than 70 were
the best could get, and yet the velocity data seemed to be accurate, judging from their
correlation with other independent measurements and their consistency with ADV data
from adjacent areas of the flow in which we could get COR>70. It can be difficult to
obtain COR>70 in highly turbulent flows that are near the upper limit of the velocity
range setting for the ADV. In some situations in which increasing the velocity range
133
setting to the next higher range produced a dramatic increase in COR values, but the
velocity data stayed about the same. This indicates that good velocity measurements can
be obtained even when the correlation is low. This we will see for the case between 5
and 40 x/D where the turbulent intensity is largest.
Looking for our correlation results we can observe an initial part of jet (x/D = 1 to
x/D = 56) with lower than recommended values (Figure 5.20). Since, other qualitative
parameters confirm correctness of them; we suppose the problem of high turbulence
level affecting this region. There is some solution. The velocity-spike filter attempts to
identify individual velocity readings in the time series that exhibit consecutive large and
opposite changes in velocity direction and magnitude relative to adjacent velocity data.
This filter may be well suited to data sets that have good signal-to-noise ratios, but
suffer from low correlation scores due to turbulence. The filter is implemented in
WinADV
and
is
described
in
the
program
manual
by
Wahl
(http://www.usbr.gov/pmts/hydraulics_lab/twahl/winadv/).
100
Avg COR [%]
80
60
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
40
20
0
0
40
80
120
x/D
Figure 5.20 Average correlation ratios for three distances y.
5.2.8
Covariance
Cov-XY is the sample covariance for x and y velocity components computed by

WinADV for use in analysis of Reynolds stresses (Figure 5.21). It is a measure of the
correlation between two variables. The covariance of the x and y velocity components
may be expressed as:
134
Cov XY = u ' v' =
UV U V
n 1
(5.13)
n(n 1)
WinADV computes the sample covariance, hence the n-1 terms in the denominator.
Covariance XY [cm2/s 2]
1200
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
800
400
0
0
40
80
120
x/D
Figure 5.21 Covariance XY for three distances y.
5.2.9
The turbulence intensity definition was described in the first part of this chapter. We
present turbulence intensity results for x and y components (Figure 5.22). For the
streamwise x component we see the concentration of highest values in first range x/D =
4 - 29. We can observe more important influence of velocity fluctuations for y
component in range between x/D = 2 and 32. Effect of the wall existence is more
perceptible in a case of this component.
135
1.6
1.6
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
1.2
Turbulence intensity Iy
Turbulence intensity Ix
1.2
0.8
0.4
0.8
0.4
0
0
40
80
120
x/D
40
80
120
x/D
Figure 5.22 Turbulence intensity for x (right) and y (left) component for three distances
y.
5.2.10 Other results

In order to calculate the percentage of samples that are retained after a user-selected
filtering operation was performed, % Good magnitude is implemented with WinADV
program. We present results in Figure 5.23. Most of the results are considered as good
having values above 80 % level. Some ranges with lower quality data are perceived,
where the higher turbulence intensity affects the jet or effect of the end wall
recirculation zone is present.
Avg AMP is signal amplitude (Figure 5.23).
Avg SNR - Signal-To-Noise-Ratio is the ratio of signal strength to the background
acoustic noise level inherent in the ADV instrument (Figure 5.23). The values are given
in dB relative to the noise level. For collection of instantaneous velocity data a signalto-noise ratio of 15 dB or higher should be maintained. For measuring mean velocities,
the signal-to-noise ratio should be 5 dB or higher. Low signal-to-noise ratios usually are
caused by a low concentration of scatterers in the sample volume. SNR is calculated by
subtracting the noise level from the AMP, and then reporting the result in dB. Thus, the
name "signal-to-noise ratio" is not really correct, since it is not a true ratio, but a
difference. It indicates the amount of signal strength in excess of the noise level. In our
case instantaneous and mean velocity data are object of interest. Most of the data meet
the conditions of good signal strength. Some exceptions are noticeable for different
regions for the three cases of a distance y.
136
100
200
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
160
% Good
60
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
40
40
Avg SNR [dB]
Avg AMP [counts]
80
60
120
80
20
20
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
40
0
0
40
80
x/D
120
40
80
120
x/D
40
80
120
x/D
Figure 5.23 Percent of good results - % Good (left), Average Amplitude - Avg AMP
(middle) and Average Signal-To-Noise-Ratio - Avg SNR (right) for the three distances
y as an x/D function.
5.3 High Reynolds number, two-phase Bubble jet

The same results as for the single phase (Normal) jet case are presented in this
section in order to have the point of reference in comparison of classical type of jet with
a two-phase jets. The descriptions of magnitudes are explained in previous sections.
5.3.1
Velocities
The x,y-component and VTOTAL velocity are shown in Figure 5.24. The much higher
downstream velocity is logic comparing with V. Range of the lower velocities is
observable at distances between x/D = 1 and x/D = 15 but within a peak of x/D = 6
(between 6 and 12) is visible. The effect of fractionation (Figure 2.16) is detectable by
ADV probe. Other, related to velocity magnitudes, results are presented on the same
figure as, Avg Vmag, Mag V-Avg, Reynolds number Re and Span of 95 % Conf.
Interval for x and y component. In all cases sharp peaks are noticeable related to the
separation of jet phases, here the effect of buoyancy is clearly noticed between 10 and
20 x/D.
137
80
80
y = 4 cm
y = 8 cm
y = 13 cm
40
20
40
20
0
20
40
60
20
40
x/D
20
40
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
16000
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
40
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
12000
Re
60
60
60
x/D
80
Mag V-Avg [cm/s]
Avg Vmag [cm/s]
60
x/D
100
80
40
20
0
0
y = 4 cm
y = 8 cm
y = 13 cm
60
VTOTAL [cm/s]
60
V [cm/s]
U [cm/s]
60
80
y = 4 cm
y = 8 cm
y = 13 cm
40
20
8000
4000
20
0
0
20
40
60
0
0
20
40
x/D
20
40
60
x/D
Span of 95% Conf. Interval-V [cm/s]
Span of 95% Conf. Interval- U [cm/s]
60
x/D
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
2
0
0
20
40
60
x/D
20
40
60
x/D
Figure 5.24 Velocity component x (top row left), y velocity component (top row
middle), VTOTAL component (top row right), Avg Vmag (middle row left), Mag VAvg (middle row middle), Reynolds number Re (middle row right), Span of 95 %
Conf. Interval for x (bottom row left) and y component (bottom row right) for three
distances of y as an x/D function.
5.3.2
Standard deviation
The standard deviation plots are presented in Figure 5.25. We can observe the
region of higher influence of velocity fluctuations in both, x and y component but is
more accentuated in first case. Very narrow and strong higher standard deviation zone
for the x component and less strong and much wider for y component indicating a firmer
impact of bubbles than the wall, affecting more rapidly the jet structure, x/D = 8 - 14.
The wall occurrence affects the jet structure on a larger downstream distance, x/D = 3 20. Composed effect is shown on the |RMS [V']| plot.
138
80
80
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
40
20
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
80
(RMS[V']) [cm/s]
60
RMS[v'] [cm/s]
RMS[u'] [cm/s]
60
100
40
60
40
20
20
0
0
20
40
60
20
x/D
40
60
x/D
20
40
60
x/D
100
RMS [Vmag'] [cm/s]
80
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
60
40
20
0
0
20
40
60
x/D
Figure 5.25 The standard deviation for x (top left), y (top middle), total standard
deviation (top right) and the root-mean-square value of the time series of individual
velocity magnitude values - RMS [Vmag'] (bottom) for three y distances as a function
of x/D.
5.3.3
Skewness
The skewness results for x and y component are shown in Figure 5.26. For the x
component 97 % of data are not affected by aliasing and for y component all the data
warrant to be lesser than 1.5 conditioned by WinADV manual. For the larger distances
x/D (40) for downstream component, higher skewness values indicates not
asymmetric distribution.
139
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
Skewness y
Skewness x
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
0
0
20
40
60
20
x/D
40
60
x/D
Figure 5.26 Skewness for x (left) and y (right) component for three distances y as an x/D
function.
5.3.4
Kurtosis
The kurtosis results for x and y component are shown in Figure 5.27. More irregular
distribution is noted for x component showing for some points strongest peakedness
than for y component where distribution is more regular.
16
16
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
12
Kurtosis y
Kurtosis x
12
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
8
0
0
20
40
60
x/D
20
40
60
x/D
Figure 5.27 Kurtosis for x (left) and y (right) component for the three distances y as an
x/D function.
5.3.5
Correlation and Covariance
Correlation and covariance results are presented in Figure 5.28. The limitation of
70% Avg COR shows at initial distance (up to x/D = 23) problems with turbulence and
excessive air bubbles effects as some of existing troubles of ADV measurements.
140
200
80
160
Covariance XY [cm2/s2]
Avg COR [%]
100
60
40
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
20
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
120
80
40
0
0
20
40
60
20
x/D
40
60
x/D
Figure 5.28. Correlation (left) and covariance (right) for the three distances y as an x/D
function.
5.3.6
The turbulence intensity result are shown in Figure 5.29.

1.6
1.6
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
1.2
Turbulence intensity IY
Turbulence intensity Ix
1.2
0.8
0.4
0.8
0.4
0
0
20
40
60
x/D
20
40
60
x/D
Figure 5.29 Turbulence intensity for x (left) and y (right) component for the three
distances y as an x/D function.
5.3.7
Other results
The percent of good results % Good, average amplitude Avg AMP and the ratio of
signal strength to the background acoustic noise level inherent in the ADV instrument
Avg SNR are shown in Figure 5.30. Most of the data are with high % Good ratio, only
some exceptions are perceptible for x/D = 12 18. Avg AMP and Avg SNR have
almost linear distribution as an x/D function. The signal-to-noise ratio meets >15 %
condition for all jets ranges.
141
100
200
80
160
80
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
% Good
60
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
40
20
Avg SNR [dB]
Avg AMP [counts]
60
120
80
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
40
20
20
40
60
x/D
40
20
40
60
20
x/D
40
60
x/D
Figure 5.30 % Good (left), signal amplitude Avg AMP (middle) and the ratio of signal
strength to the background acoustic noise level inherent in the ADV instrument Avg
SNR (right) for three distances y as an x/D function.
5.4 Data stability

In order to verify data stability we done several tests dividing all sampling data for a
one point by 5 ranges and calculating then the mean. The results for the mean
streamwise velocity U and RMS for x component for both, the single and two-phases jet
are shown in Figure 5.31. The firmness of the data is proven confirming no existence of
time dependence phenomenon during the data recording. The errors results for the same
cases are shown in Figure 5.32.
142
40
30
30
U [cm/s]
U [cm/s]
Normal Jet, y = 4 cm
x = 0 cm
Y = -0.00944 * X + ...
x = 62 cm
Y = -0.624 * X + ...
x = 123 cm
Y = -0.13889 * X + ...
20
20
Bubbly jet, y = 4 cm
x = 0 cm
Y = -0.307 * X + ...
x = 52 cm
Y = -0.069 * X + ...
x = 109 cm
Y = 0.02184 * X + ...
10
10
0
1
20
20
16
16
Normal jet, y = 4 cm
x = 0 cm
Y = 0.0686 * X + ...
x = 62 cm
Y = -0.016 * X + ...
x = 123 cm
Y = 0.0662 * X + ...
12
Samples serie
RMS [u'] [cm/s]
RMS [u'] [cm/s]
Samples serie
12
x = 0 cm
Y = -0.013 * X + ...
x = 52 cm
Y = -0.349 * X + ...
x = 109 cm
Y = -0.0119 * X + ...
0
1
Samples serie
Samples serie
Figure 5.31 Data stability for the mean streamwise velocity x (top) and root mean
square for x component (below) for different x distances from the nozzle as a samples
series (or time) function for single (both left) and two-phase (both right) jets.
80
80
Normal Jet, y = 4 cm
x/D = 0
x/D = 56
x/D = 112
60
x/D = 0
x/D = 26
x/D = 55
60
40
U [cm/s]
U [cm/s]
40
20
20
-20
-20
200
400
600
Time [s]
200
400
Time [s]
Figure 5.32 Errors for jet cases under consideration.
143
600
5.5 Instrumental error

For the ADV the instrumental error is 1% of the measured velocity down to a
minimum of 0.25 cm/s. We present this error results for single and two-phase jets in
1.6
1.6
Instrumental error [cm/s]
Instrumental error [cm/s]
Figure 5.33.
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
1.2
0.8
0.4
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
1.2
0.8
0.4
Minimum instrumental error
Minimum instrumental error
0
0
40
80
120
x/D
20
40
60
x/D
Figure 5.33 The instrumental error of ADV for single (left) and two-phase (right) jet
case.
5.6 Comparison of the three cases of the jet configurations

It is interesting to compare results of the three cases in order to conclude effects of
different configurations (the free and wall jet), Reynolds numbers (different outlet
velocities), phenomenon (water jet, multiphase jet).
Here we present the most important findings of done analysis of the standard
velocity ADV data. It is important to compare non-dimensional values (for example,
x/D, U/UMAX, etc.) in order to evaluate correctly the different jets configurations for the
same magnitudes.
Potential core of the smaller Reynolds number jet case is not affecting ADV
measurements in this region; they fulfil typical downstream velocity decaying
distribution described in previous works. For higher Reynolds number cases (Normal
and Bubble jet) the effect of the close to the nozzle zone is deforming velocity profiles
that are much slower that for theoretical and experimental considerations. For the
Bubble jet case separation effect is notable. The downstream velocity is decaying
slower for normal jet than for bubble jet case because buoyancy is more important in
144
second case and it is causing slowdown of the mean velocity. The outlet Reynolds
number has influence on a jet length. The length of the normal jet is higher (x/D = 112)
comparing with the bubble jet (x/D = 58).
The velocity component V has less influence on the jets behaviour however it is
helpful component to detect wall and bubbles effects, for higher Reynolds number cases
there are peaks exhibiting these phenomenons. A log-log plot is a useful instrument to
see more in detail, existing curves behaviours.
The standard deviation magnitude is interesting statistical tool to detect conducting
of velocity fluctuations. For the smaller Reynolds number case, the peaks range of
higher values is larger than for higher Reynolds number cases. Especially for Bubble
jet case, the sharp and narrow range of higher root mean squares is noticeable showing
rapidity influence of the separation zone for the two-phase flows. For the smaller
Reynolds number case the wall effect is more perceptible because of the lower mean
velocities.
Comparing average velocity magnitudes Avg Vmag for the normal and bubble jet
cases we can observe that in first case the peaks of the wall effect are discernible, the
fractionation phenomenon for the bubble jet vanishes this effect. Smaller values of this
quantity are for the bubble jet case.
If we compare the magnitude of the resultant of the two average velocity
components Mag V-Avg we observe potential core zone affecting data measurements
till some distance from the nozzle, for normal jet case it is about x/D = 23 showing
difference depending of y distance and for bubble jet it is about x/D = 16 independent
on y value. Apart from these distances a velocity decaying profile preserve typical
observations. More regular distribution is noticeable for the two-phase jet and it is
detectable on more plots, for example Span of 95% Conf. Interval-U and Span of 95%
Conf. Interval V point more importance of the wall effect for normal jet case than for
bubble jet.
The skewness results for x component show similar asymmetry distribution ratio
around an average for the normal and bubble jet affecting with non asymmetry initial
and final regions of the jets. For y component effect of the wall is more detectable by
ADV device analysing skewness magnitude.
Kurtosis x indicate similar distribution for both cases, normal and bubble jet and
more peakedness is observable for some distance from the nozzle, about x/D = 22 and
x/D = 13 respectively. Component y is similar for both cases having congruent values.
145
The average correlation points recommendable values from x/D = 50 for the normal jet
and x/D = 24 for the bubble jet and the ADV measurements problems in close to the
nozzle range due to high turbulence levels and excessive air bubbles for the bubble jet
case. For the normal jet case there is a difference of correlation distribution as an x/D
function between three distances y, for the bubble jet it is not detectable and data are
more regular.
The covariance between x and y component has much larger values for the normal
jet than for the bubble jet case.
The turbulence intensity presents stronger turbulence effects for the bubble jet case
but on shorter x/D range than for normal jet case where turbulence is not so strong but
appears on larger range of x/D. The turbulence intensity results indicate strong
turbulence scales.
The percentage of considered as good results is high for both cases (normal and
bubble jet) and only for some ranges data are affected with slower % Good ratio values.
The average amplitude (Avg AMP) is more regular for the bubble jet where almost
linear function as an x/D would be applied, for the normal jet case data exhibit more
irregular curves. The same situation happens with average signal-to-noise ratio where
some data for the normal jet case are affected by lower than conditioned values, for the
bubble jet case all the data fulfil recommendations.
It is clear that the production of vortices is the key element in initial jet dilution and the
geometry of the pump (the jet source) is affecting the close to the nozzle region.
146
CHAPTER 6
SPECTRAL MEASUREMENTS, STRUCTURE FUNCTION
TECHNIQUES AND SCALES ANALYSIS
6.1 Energy spectrum Fast Fourier Transform (FFT)

WinADV software has option to plot FFT graph. The FFT graph shows the
amplitude of the Fast Fourier Transform (FFT) of all three velocity components. The
maximum frequency displayed on the graph can be adjusted, but cannot be higher than
the Nyquist frequency (half the sampling rate). The amplitudes are plotted on a log-log
scale. The FFT is computed so that a pure sine wave with peak time-domain amplitude
of 1.0 will have FFT amplitude of 0.5.
It computes Fast Fourier Transform (FFT) using the algorithm described in Press,
Flannery, Teukolsky, and Vetterling (1989, Numerical Recipes: FORTRAN Version).
For a time series vector V containing n = 2 m samples, the FFT vector F contains j =
1 + 2m-1 elements satisfying the equation:
1 n1
F j = Vk e 2 ( j / n )ki
n k =0
(6.1)
The amplitude of F is independent of the sampling period and rate. A pure sine
t
wave with amplitude 1.0, such as f (t ) = sin 2 , will have resulting FFT amplitude
T
of 0.5 at frequency 1/T.
The large data sample for a one measurements point was performed to be able to
obtain detailed energy spectrums and structure functions.
We present results of energy spectrum for different x/D (only some, the most
significant), to show energy distribution for different scales and y distances in Figures
6.2 6.7 as a frequency function. Examples with more noisy energy spectrums are
showed to expose different jet zones. As a curiosity, we realized polynomial fits of fifth
degree for every spectra in order to observe energy peaks for different scales (red
curves) using equation:
147
Y = a 0 + a1 X 1 + a 2 X 2 + ... + a 5 X 5
(6.2)
This method was devised to display in tentative way some relation between energy
curves behaviour for different jet cases and distances y. Even this approach may leave
some doubts, there were found some logic connections amidst these parameters. The
fifth order was found to be the most adequate to observe the peaks because higher
orders tends to straighten the curves. We are awake that this method involves some
errors due to the used fit. More detailed and correct method would be using the running
average fit.
Running average fits are generated by taking the average of the data within a
specified range on either side of a given point. The window width controls the range of
data used in the calculation. Using this method we may analyze different energy
spectrum peaks, in Figure 6.1 we present example with different window width (a 5, b
17, c 31). For larger values of the window width, the different energy spectrum
curve peaks are perceived. Further, more detailed work on this point would reveal more
exhaustive results on the scale to scale energy transfer.
100
100
10
100
10
10
x/D = 65
x/D = 65
x/D = 65
0.1
Energy [cm 2 /s 2 ]
Energy [cm 2 /s 2 ]
Energy [cm 2 /s 2 ]
0.1
0.1
0.01
0.01
0.01
0.001
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
Frequency [Hz]
10
100
0.0001
0.001
0.01
0.1
Frequency [Hz]
10
100
0.001
0.01
0.1
10
100
Frequency [Hz]
Figure 6.1 Energy spectrums with running average fit implemented for Normal jet
case, y = 4 cm and x/D = 65. Three cases of window width are presented: a 5, b 17
and c 31.
148
6.1.1
100
100
100
10
10
10
x/D = 65
x/D = 51
x/D = 82
1
E [cm2/s2]
E [cm 2/s2]
E [cm2/s2]
0.1
0.1
0.1
0.01
0.01
0.001
0.001
0.01
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
0.001
0.01
Frequency [Hz]
0.1
10
0.001
100
0.01
100
0.1
10
100
Frequency [Hz]
Frequency [Hz]
100
10
10
x/D = 98
x/D = 112
1
E [cm2/s2]
E [cm 2/s 2]
0.1
0.1
0.01
0.01
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
0.001
0.01
0.1
Frequency [Hz]
10
100
Frequency [Hz]
Figure 6.2 Energy spectrum for normal jet case and y = 4 cm for different distances x/D.
6.1.2
100
100
10
100
10
10
x/D = 59
x/D = 69
x/D = 80
0.1
E [cm2/s2]
E [cm2/s2]
E [cm2/s2]
0.1
0.1
0.01
0.01
0.01
0.001
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
0.0001
0.001
0.01
Frequency [Hz]
0.1
10
100
0.001
100
0.1
10
100
Frequency [Hz]
100
10
10
x/D = 101
x/D = 91
1
E [cm2/s2]
E [cm2/s2]
0.01
Frequency [Hz]
0.1
0.1
0.01
0.01
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
Frequency [Hz]
0.001
0.01
0.1
10
100
Frequency [Hz]
Figure 6.3 Energy spectrum for normal jet case and y = 8 cm for different distances x/D.
149
6.1.3
100
100
10
100
10
x/D = 63
0.1
E [cm2/s2]
E [cm2/s2]
E [cm2/s2]
0.1
0.1
0.01
0.01
0.01
0.001
0.001
0.001
0.0001
0.0001
0.001
x/D = 76
10
x/D = 50
0.01
0.1
10
100
0.0001
0.001
0.01
Frequency [Hz]
0.1
10
100
0.001
100
0.1
10
100
Frequency [Hz]
100
10
10
x/D = 103
x/D = 91
1
E [cm2/s2]
E [cm2/s2]
0.01
Frequency [Hz]
0.1
0.1
0.01
0.01
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
0.001
0.01
0.1
Frequency [Hz]
10
100
Frequency [Hz]
Figure 6.4 Energy spectrum for normal jet case and y = 13 cm for different distances
x/D.
6.1.4
100
100
10
100
x/D = 36
10
10
x/D = 30
x/D = 42
0.1
E [cm2/s2]
E [cm2/s2]
E [cm2/s2]
0.1
0.1
0.01
0.01
0.01
0.001
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
0.0001
0.001
0.01
Frequency [Hz]
0.1
10
100
0.01
100
10
100
Frequency [Hz]
100
10
10
x/D = 55
x/D = 48
1
E [cm2/s2]
E [cm2/s2]
0.1
Frequency [Hz]
0.1
0.1
0.01
0.01
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
Frequency [Hz]
0.001
0.01
0.1
10
100
Frequency [Hz]
Figure 6.5 Energy spectrum for bubble jet case and y = 4 cm for different distances x/D.
150
6.1.5
100
100
10
100
10
10
0.1
E [cm2/s2]
E [cm2/s2]
E [cm2/s2]
x/D = 42
x/D = 33
x/D = 24
0.1
0.1
0.01
0.01
0.01
0.001
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
0.0001
0.001
0.01
0.1
Frequency [Hz]
10
100
0.001
0.01
0.1
Frequency [Hz]
100
10
100
Frequency [Hz]
100
x/D = 55,5
10
10
E [cm2/s2]
E [cm2/s2]
x/D = 58
0.1
0.1
0.01
0.01
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
0.001
0.01
0.1
Frequency [Hz]
10
100
Frequency [Hz]
Figure 6.6 Energy spectrum for bubble jet case and y = 8 cm for different distances x/D.
6.1.6
100
100
10
100
10
10
x/D = 32
x/D = 23
x/D = 41
0.1
E [cm2/s2]
E [cm2/s2]
E [cm2/s2]
0.1
0.1
0.01
0.01
0.01
0.001
0.001
0.001
0.0001
0.0001
0.001
0.01
0.1
10
100
0.0001
0.001
0.01
0.1
Frequency [Hz]
10
100
0.01
0.1
10
100
Frequency [Hz]
100
100
10
10
x/D = 50
x/D = 59
1
E [cm2/s2]
E [cm 2/s 2]
0.001
Frequency [Hz]
0.1
0.1
0.01
0.01
0.001
0.001
0.0001
0.0001
0.001
0.001
0.01
0.1
10
0.01
0.1
10
100
Frequency [Hz]
100
Figure 6.7 Energy spectrum for bubble jet case and y = 13 cm for different distances
x/D.
151
6.1.7
Energy spectrum results discussion
Analysing energy spectrum of different jet cases and distances y we can observe
energy distribution for different scales (here frequency). It was found that the spectrum
of the velocity field is completely flat when the probe is very close to the nozzle and it
evolves toward a power law as the distance from the nozzle increase. This is very
interesting because shows that when the probe is far enough from the nozzle the
turbulent structures can interact allowing for energy cascade. For farther x/D distances
the inertial range of Kolmogorovs -5/3 law is more detectable and for very close to the
nozzle cases almost horizontal distribution is due to the signal noise and not fully
developed turbulence, these last results have less importance for our analysis.
Intermediate slopes are affected by not reached inertial range turbulence and occurring
processes. This non-universality in the spectral slopes seems to be due to the coherent
structures and their contribution to the transfer dynamics. It is interesting to observe that
for some regions energy spectrum is steeper than k-5/3, which clearly indicates that the
dynamics in these regions tends to be nonlocal. Two different phenomena contribute to
the deviation from Kolmogorovs self-similar prediction: the nonlocal dynamics and the
intermittency. Finally, different energy amplitudes are perceptible depending on the
points position.
For larger scales energy peak is caused by appearance of additional vorticity scale
(related to the wall effect) and two different slope ranges are perceptible as portrayed in
Figure 6.8, modifying the turbulence cascade. The experiments show how external
velocity fluctuations change significantly the energy of fluctuations in the mixing outer
layer, causing a rapid growth of the wall jet. The energy cascade and dissipation process
is a common characteristic of all of the turbulent flows (Jaw and Chen 1999).
We can observe that the occurring processes modify the energy spectrum by
attributing less energy to small-scales in comparison with that predicted by the standard
Kolmogorov distribution. The energy curve slope has value -1,68 instead of the -1,67
for the x/D = 60 case, as an example. Even for the ranges where the energy spectrum
meets universal -5/3 Kolmogorovs law the homogeneity and non-intermittency is not
warranted. The deviation from the Kolmogorov theory K41 for the low order moments
is small. The corrections to the spectrum due to intermittency are very small and thus
require largely converging statistics to be evaluated. The effects of the experimental
setup (recirculation zone, the wall effect) are larger than the small intermittent
152
correction. This confirms that the use of the velocity structure function (done further in
this chapter) is a sensitive tool to understand the complex dynamic of the turbulent jets
and that higher order moment information is much more useful than the spectra.
100
10
-0,63
1
-5/3
E [cm2/s2]
0.1
0.01
-1,68
0.001
x/D = 60
0.0001
1E-005
1E-006
0.001
0.01
0.1
10
100
Frequency [Hz]
Figure 6.8 Energy spectrum for normal jet case, y = 4 cm and x/D = 60 showing two
different slope ranges.
Here we present only plots for selected points but examining energy spectrum, point
by point; there are observable two dominant scales existence and the energy curve slope
affected by the occurring processes.
The energy of small scales is affected by the sampling rate (the Nyquist frequency).
The Nyquist frequency is half of the sampling frequency of a discrete signal processing
system. It is sometimes known as the folding frequency of a sampling system. The
sampling theorem shows that aliasing can be avoided if the Nyquist frequency is greater
than the bandwidth, or maximum component frequency, of the signal being sampled.
Last works on turbulence are more concentred on small scale turbulence about
Kolmogorov scales, here is why we focus our attention on large-scales where the most
of energy is present; we call it here macro turbulence (the scales are between large scale
L and Taylor microscale ). Energy spectrum is presented as a frequency function but it
is easy to convert it to time or length scales:
f [Hz ] =
1
T - Time scale
T
153
T V = L - Length scale
where V is the mean velocity at a point.
Polynomial fits of fifth degree were applied in order to observe peaks for different
scales revealing local energy increasing and ranges of the signal noise. We can plot
energy peaks frequency as an x/D for two jet cases; results are shown in Figure 6.9.
The running average fit was used to observe maximum energy peak frequency and
corresponding distance from the nozzle. For the normal jet case, the energy peak
frequency appears at distance closer to the nozzle if y is smaller, for bubble jet case this
is not perceptible. The wall jet effect is clearer visible using the energy spectrum
analysis for the different streamwise distances.
'Bubble jet'
y = 4 cm
y = 8 cm
y = 13 cm
Energy peak frequency [Hz]
Energy peak frequency [Hz]
2.5
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
1.5
0.5
0
0
40
80
120
20
40
60
x/D
x/D
Figure 6.9 Energy peak frequency as an x/D function for normal (left) and bubble right
jet cases.
It is interesting to analyse the peak energy value as a streamwise distance, it is

presented in Figure 6.10. We see that profiles seem to be similar to the downstream
velocity profiles taking into account modifications of closeness of the nozzle region
comparing with the classical distribution. For the bubble jet case the peaks energy
distribution is more regular. The energy value is higher for the two-phase jet case in the
separation region.
154
0.8
0.8
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
0.6
Peak's energy [cm2/s2]
Peak's energy [cm2/s2]
0.6
0.4
0.2
0.4
0.2
0
0
40
80
120
20
40
x/D
60
x/D
Figure 6.10 Peaks energy value as an x/D distance for normal (left) and bubble (right)
jet case.
Let us introduce adimensional parameter
f y
where f is the energy peak frequency
E
and E the value of peaks energy and plot it as an x/D function as shown in Figure 6.11.
An increase of the fit line slope is observed with augmentation of the y distance, the
slopes values are presented in Table 6.1. Much higher values of the lines slopes are
observed for the bubble jet case where buoyancy produced by bubbles is affecting the
jet structure.
200
200
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
160
160
120
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
f*y/
f*y/
120
80
80
40
40
40
80
120
x/D
Figure 6.11 Non dimensional magnitude
20
40
60
x/D
f y
for normal (left) and bubble (right) jet
E
case for three distances y as an x/D function.
155
Jet case
Normal
Bubble
y = 4 cm
0,2264
0,4023
y = 8 cm
0,5683
1,083
y = 13 cm
1,025
2,168
Table 6.1 Line fit slopes for non dimensional
f y
magnitude for the two cases of jets
E
and different y distances.
We can plot different energy spectrum as a function of different scales in order to

look for different aspects of the energy and the scales distributions; we present as
example, some possible options for the same point (x/D = 65, normal jet case, y = 4 cm)
in Figure 6.12. On this figure k is the wave number defined as k =
2
. Again we can
L
detect two energy curve ranges with different slopes or affected by different effects. The
polynomial fit was applied for these cases in order to exhibit different ranges.
156
x/D = 65
0.1
E*k [cm/s2]
E*k(5/3)
0.1
0.01
0.01
0.001
0.001
0.0001
x/D = 65
0.0001
0.01
0.1
10
0.01
0.1
-1
10
10
k [cm-1]
k [cm ]
10
10
1
1
x/D = 65
0.1
0.1
E*k3
E [cm 2/s2]
0.01
0.01
0.001
0.0001
0.001
x/D = 65
1E-005
0.0001
1E-006
1E-007
1E-005
0.01
0.1
0.01
10
0.1
-1
k [cm-1]
k [cm ]
10
0.1
0.01
0.001
x/D = 65
0.0001
1E-005
0.1
10
T [s]
Figure 6.12 Different energy spectrum plots for the same point of measurements (x/D =
65, normal jet case, y = 4 cm)
It is interesting to see the energy spectrum behaviour using average method. For
example, if we have 15000 data, first, lets take data from 1 to 14995, then from 2 to
14996, then from 3 to 14997 and so on respectively, taking always the same number of
157
data number. Afterwards we can average all of our data series (for each one, we can plot
individual plot) in order to see behaviour of the energy transfer distribution, we present
example in Figure 6.13 comparing original plot with 10 and 25 plots averaging.
10
0.1
0.01
Normal jet, y = 4 cm, x/D = 65

Original
10 plots average
25 plots average
0.001
0.0001
0.001
0.01
0.1
10
100
Frequency [Hz]
Figure 6.13 Original and averaged energy spectrum plots for the normal jet case, y = 4
cm and x/D = 65.
We can observe that energy fluctuations amplitude is decreasing with increasing

number of plots to average. Repeating this iteration we would acquire one energy
spectrum curve but with the same slope as original, comparison of the polynomial fits
for three cases are shown in Figure 6.14. Very small differences are noticeable between
these curves confirming the usefulness of the original energy spectrum.
100
Normal jet, y = 4 cm, x/D = 65

Polynomial fit - original
Polynomial fit - 10 averages
Polynomial fit - 25 averages
10
E [cm 2/s2]
0.1
0.01
0.001
0.0001
0.001
0.01
0.1
10
100
Frequency [Hz]
Figure 6.14 Comparison of the original and averaged polynomial fits plots for the
normal jet case, y = 4 cm and x/D = 65.
158
6.2 Structure function analysis

In chapter 2 we described definition of the structure function. Here we present
results for the two jet cases and different x/D and y distances in time. From 2nd to 6th
order structure functions where calculated, where:
2nd order green curve
3rd order yellow curve
4th order red curve
5th order purple curve and
6th order black curve on the plots
The results are presented in Figures 6.15 6.20, for different jets configurations and
distances y. As was mentioned before, the third order structure function of the velocity
difference deserves special attention, because of its proportionality to the separation
distance l in the inertial range and it is a standard procedure in the analysis to define an
inertial range where the third-order scaling exponent should have value 1 as predicted
by Kolmogorov (1941). It was found later in experimental works on turbulent jets that
the slope of the log-log plot of the third order structure function as its function is
different from unity. Moreover, in non-homogenous and non-isotropic this
proportionality is not warranted.
We calculate absolute value of the structure functions and regular monotonous
increasing behaviour is observed for all fully developed turbulence ranges cases. The
absolute value of the third order structure function is indication of the balance of the
transferred energy at scale l to either larger or to smaller scales. The normal value of
the third order structure function indicates the transfer dynamics in energy spectrum and
possible inverse energy cascade, more probable in non-homogenous flows. These two
quantities (absolute and not) differ increasingly in non-homogenous turbulence, as
shown in Mahjoub (2000) and each characteristic refers to distinct statistical
information.
159
6.2.1
100000000000
10000000000
1000000000
1000000000
100000000
100000000
10000000
10000000
1000000
1000000
100000
100000
1000000000
100000000
Sp
Sp
1000000
100000
Sp
10000000
10000
10000
10000
1000
1000
1000
100
100
10
2
3
4
5
6
10
1
100
x/D = 65
x/D = 51
0.1
0.1
10
100
1000
2
3
4
5
6
0.1
0.01
x/D = 82
10
2
3
4
5
6
0.1
0.01
0.1
10
100
1000
0.01
0.1
10
100
1000
1000000000
100000
100000000
10000
10000000
1000
1000000
100
100000
Sp
Sp
10
10000
1
1000
0.1
100
x/D = 112
2
3
4
5
6
0.1
0.01
0.1
2
3
4
5
6
0.01
x/D = 98
10
0.001
0.0001
10
100
1000
0.01
0.1
10
100
1000
Figure 6.15 Structure functions for different x/D distances for normal jet case and y = 4
cm.
6.2.2

1000000000
1000000000
100000000
100000000
100000000
10000000
10000000
10000000
1000000
1000000
1000000
100000
100000
100000
10000
Sp
Sp
10000000000
1000000000
Sp
10000000000
10000
10000
1000
1000
1000
100
100
100
x/D = 59
10
2
3
4
5
6
10
1
0.1
1
0.1
10
100
1000
0.01
0.01
0.1
10
100000000
100000000
10000000
10000000
1000000
1000000
100000
100000
10000
10000
Sp
Sp
0.1
2
3
4
5
6
0.1
0.01
x/D = 80
x/D = 69
2
3
4
5
6
10
1000
100
100
1000
0.01
0.1
10
100
1000
1000
100
10
10
x/D = 91
x/D = 101
2
3
4
5
6
0.1
2
3
4
5
6
0.1
0.01
0.01
0.01
0.1
10
100
1000
0.01
0.1
10
100
1000
cm.
160
6.2.3
100000000
10000000000
1000000000
10000000
1000000000
100000000
100000000
10000000
10000000
1000000
1000000
100000
100000
10000
1000000
100000
1000
Sp
Sp
Sp
10000
10000
1000
1000
100
100
10
100
0.1
0.01
0.1
2
3
4
5
6
10
2
3
4
5
6
0.01
x/D = 76
10
2
3
4
5
6
x/D = 63
x/D = 50
1
0.1
0.1
10
100
1000
0.01
0.01
0.1
10
100
1000
0.01
0.1
10000000
1000000
1000000
100000
100000
10000
10000
Sp
100000000
10000000
Sp
100000000
1000
10
100
1000
1000
100
100
10
10
x/D = 91
2
3
4
5
6
0.1
x/D = 103
1
2
3
4
5
6
0.1
0.01
0.01
0.01
0.1
10
100
1000
0.01
0.1
10
100
1000
cm.
6.2.4
10000000000
1000000000
1000000000
1000000000
100000000
100000000
100000000
10000000
10000000
1000000
1000000
100000
100000
10000
10000000
1000000
Sp
Sp
Sp
100000
10000
1000
1000
100
10000
1000
100
2
3
4
5
6
2
3
4
5
6
1
0.1
0.01
0.1
100
x/D = 36
10
x/D = 30
10
0.1
10
100
1000
2
3
4
5
6
0.01
x/D = 42
10
0.1
0.01
0.1
10
100
1000
0.01
0.1
10
100
1000
1000000
1000000000
100000
100000000
10000
10000000
1000
1000000
100
100000
10
Sp
Sp
10000000000
10000
1000
0.1
x/D = 55
2
3
4
5
6
0.01
100
x/D = 48
10
0.001
2
3
4
5
6
1
0.1
0.01
0.1
0.0001
1E-005
10
100
0.01
1000
0.1
10
100
1000
Figure 6.18 Structure functions for different x/D distances for bubble jet case and y = 4
cm.
161
6.2.5
10000000000
1000000000
1000000000
1000000000
100000000
100000000
10000000
10000000
100000000
1000000
1000000
10000000
100000
1000000
100000
10000
10000
100000
Sp
Sp
Sp
1000
1000
100
10000
100
10
1000
x/D = 33
1
100
x/D = 24
0.1
2
3
4
5
6
10
1
0.01
0.1
0.01
0.1
10
100
10
2
3
4
5
6
0.001
0.01
0.0001
0.001
1000
0.01
0.1
x/D = 42
2
3
4
5
6
0.1
10
100
1000
0.01
0.1
10
100
1000
100000
1000000000
100000000
10000
10000000
1000
1000000
100
100000
10
Sp
Sp
10000
1000
100
0.1
10
x/D = 58
x/D = 56
0.1
2
3
4
5
6
0.01
2
3
4
5
6
0.001
0.0001
0.01
0.01
0.1
10
100
0.01
1000
0.1
10
100
1000
cm.
6.2.6
10000000000
1000000000
1000000000
1000000000
100000000
100000000
100000000
10000000
10000000
10000000
1000000
1000000
1000000
100000
100000
100000
Sp
Sp
Sp
10000
10000
10000
1000
1000
1000
100
100
100
10
10
x/D = 23
0.1
2
3
4
5
6
0.01
0.1
10
100
1000
2
3
4
5
6
1
0.1
0.1
0.01
x/D = 41
x/D = 32
10
2
3
4
5
6
0.01
0.01
0.1
10
100
1000
0.01
0.1
10
100
1000
10000000000
10000000000
1000000000
1000000000
100000000
100000000
10000000
10000000
1000000
1000000
100000
100000
Sp
Sp
10000
10000
1000
1000
100
100
10
10
x/D = 50
1
0.1
x/D = 59
2
3
4
5
6
1
2
3
4
5
6
0.1
0.01
0.01
0.001
0.01
0.1
10
100
1000
0.01
0.1
10
100
1000
cm.
162
The plots with almost plane curves for different structure functions orders
correspond to the ranges where turbulence is still not fully developed or the turbulence
intensity is decaying (farther downstream distances x/D). We show them in order to
remark this phenomenon. We use these results to calculate scaling exponents to observe
discrepancy with the Kolmogorov theories.
6.2.7
In homogenous turbulence, a standard and straightforward procedure in the analysis

of the structure functions is to select in a log-log plot of structure functions the interval
bounds where the slope for the third order structure function versus scale is unity. This
implies self-similar scaling of the third-order structure and by extension self-similar
scaling of all structure functions. The bounds of this range are taken to be the bounds of
the inertial range. The inertial range is physically defined as a range of scales where
both the forcing and the dissipation process are irrelevant because they are locally in
equilibrium.
In non-homogenous flows, the scaling exponent may also depend on the separation
distance l in a complex fashion. Therefore, the fundamental result that third-order
structure function is equal to 1 of the Kolmogorov theory that allows us to define an
inertial range is not observed. It was proven experimentally that the scaling exponents
depend on l, even for third order and the scaling laws of the velocity structure functions
cannot be clearly detected following the Kolmogorov theory. Even the Reynolds
number is large enough; there are these deviations from the Kolmogorovs theory (K41)
in non-homogenous flows.
In order to calculate scaling exponents q and compare it with Kolmogorov theory
we will use Extended Self Similarity technique described in Chapter 2. Here we will
explain shortly the steps to obtain the scaling exponent for a one example; the same
steps are applicable for all points. We present structure function plot for the normal jet
case, y = 4 cm and x/D = 60 showed in Figure 6.21.
163
x/D = 60
1
2
3
4
5
6
1000000000
100000000
10000000
1000000
Sq
100000
10000
1000
100
10
1
0.01
0.1
10
100
1000
Figure 6.21 Structure function for x/D = 60, normal jet case and y = 4 cm.
If we now plot the structure functions Sq as a function of third order structure

function S3 we have results shown in Figure 6.22.
1000000000
100000000
10000000
x/D = 60
1000000
1
2
3
4
5
6
Sq
100000
10000
1000
100
10
1
100
1000
10000
100000
S3
Figure 6.22 Structure function for x/D for normal jet case, y = 4 cm as a third order
structure function S3 function.
The scaling exponent is the slopes value of every curve of q order, for our example
these values are shown in Table 6.2.
164
0.38
0.72
1.23
1.41
1.55
Table 6.2 Scaling exponent values of different q orders.
Plotting, we have (Figure 6.23) and the difference between standard K41 theory and
true life results is observable.
2
1.6
x/D = 60
K41
1.2
0.8
0.4
0
1
Figure 6.23 Scaling exponent q as an order q function for x/D = 60 (black curve) and
K41 (blue curve) results.
Results of the scaling exponent as a q function for six configurations (two jet cases
and three different distances y) are presented in Figure 6.24.
165
x/D = 65
x/D = 82
x/D = 98
K41
1.6
1.6
Scaling exponent p
Scaling exponent p
2.4
x/D = 59
x/D = 69
x/D = 80
x/D = 91
x/D = 101
K41
1.2
0.8
1.2
x/D = 63
x/D = 78
x/D = 91
x/D = 101
K41
Scaling exponent p
2.4
0.8
1.6
1.2
0.8
0.4
0.4
0.4
0
1
0
1
Bubble jet, y = 4 cm
x/D = 30
x/D = 36
x/D = 42
K41
x/D = 33
K41
0.8
0.4
1.2
0.8
0.4
0
3
1.2
0.8
0.4
0
2
x/D = 32
x/D = 41
K41
1.6
Scaling exponent p
1.2
1.6
Scaling exponent p
1.6
Scaling exponent p
0
1
Figure 6.24 Scaling exponent q for different jet cases (normal jet top, bubble jet
bottom) for three distances y and different x/D as a p order.
The scaling exponent plots show that it was very difficult to obtain a statistical
estimation of them even a quite high Reynolds numbers was intended, because it was
difficult to detect inertial range. There is not or it is very small inertial range where the
slope of the third-order velocity structure function is equal to 1 even using ESS tool.
Very thoughtful and scrupulous analysis is needed to obtain more valuable information
on the jet characteristics. The high level of no homogeneity and no isotropy hamper in
data analysis. These results suggest that the inertial range and the scaling laws of the
velocity structure functions in such kind of flows cannot be detected simply according
to the phenomenological theories suggested by Kolmogorov (K41 and K62) even for
high Reynolds numbers. The scaling exponents vary more in the region of maximum
interaction between the jet and wall.
Now we can use different methods to measure the intermittency in the turbulent
flow. One of them, the most direct, is taking into account the sixth-order scaling
exponent of the velocity structure function and defined as
166
= 2 6
(6.3)
As an example, other one is using the intermittency factor (Mahjoub 2000)

indicating for the = 1 practically homogenous and non-intermittent type of the flow
and it was demonstrated its dependence on the structure function order q. Comparing
different forms of the intermittency estimating measures: (K62), Frisch (1995) and
Mahjoub et al. (1998) with fractal topology of the flow we may distinguish regions of
the higher mixing efficiency taking into account four different basic conditions: high or
small intermittency and high or small fractal dimension.
Looking for the scaling exponents figures it appears that the intermittency parameter
depends on the location in the flow but its decreasing with downstream distances it is
not so clear visible like in previous works due to the farther streamwise distances
recirculation zone and effect of the different processes that do not occur in a typical free
jet behaviour. They depend also on the flow configuration. It was found before, that
close to the jet nozzle the flow is more intermittent due to the distribution of the
coherent structures near the jet nozzle. For this reason, in this work, we focus special
attention on the intermediate ranges (x/D between 59 and 101 for the single phase jet
case and x/D between 30 and 42 for the two phase flow). Mahjoub proved that for the
wall jet case the boundary layer affects the large-scale structures and the breaking of
these large-scales structures provides a mechanism for a decrease of the intermittency.
But this is valid only for the large scales and it is not sufficient to take into account only
one intermittency factor. It may be different for different scales and be not linear. Our
results indicate that the intermittency is higher if we are closer to the wall, for the single
phase jet case. For the two phase jet case the wall effect is not so noticeable due to the
strongest effect of the bubbles buoyancy. There is distinct intermittency in the close to
the nozzle range comparing with farther distances where the boundary layer effect
dominates.
It was found that relative scaling exponents q/3 do not have constant values and
depend on the downstream distance for non-homogenous flows and they deviate from
the Kolmogorov law K41.
The deviation of curves from the standard Kolmogorov K41 theory for all cases is
observable showing less scaling exponent values for larger q orders. The latter are
167
characterized by very irregular scaling exponent, contrasting the theoretical results for
homogeneous, isotropic and fully developed turbulence. They suffer from the
experimental conditions (relatively small size of the tank), being affected by some
recirculation large scale structures present in the flow for farther downstream distances
especially where the end wall perpendicular to the jet axis is modifying its geometry.
They are similar to the (K41 and K62) and present theories estimation for fully
developed region of the jets (Figure 6.23).
The visualization methods would reveal these features for different jet ranges, for
example, we know that the geometry of the nozzle acts physically on the initial zone of
the jet structure.
The description of how to use structure functions to measure intermittency using the
beta-model and the role of locality in higher-order exponents is described by Mahjoub
(2000). In general it has to be remembered that the third-order structure function scaling
exponent does not have to be one, in non-homogeneous or even in non-local flows. The
obtained direct structure functions (without using ESS) results confirm this argument. A
compensation mechanism may act between the most energetic, but rare events or eddies
and the common but weaker (and smaller) eddies.
6.3 Correlation and scales

In statistics, correlation is any of a broad class of statistical relationships between
two or more random variables or observed data values. It is one of the most common
and most useful statistics.
The correlation is defined as:
R=
N xy ( x )( y )
[N x ( x) ][N y
2
( y )
(6.4)
where
N number of pairs of scores,
of x scores,
xy - sum of the products of paired scores, x - sum
y - sum of y scores, x
- sum of squared x scores and y .

2
In Figure 6.25 the correlation R for different jet cases and different x/D is shown.
168
1.2
x/D = 0
x/D = 22
x/D = 44
x/D = 65
x/D = 89
x/D = 111
x/D = 0
x/D = 23
x/D = 43
x/D = 65
x/D = 90
x/D = 101
0.8
Correlations R
0.4
0.4
-0.4
2
10
-0.4
0
Time [s]
10
x/D = 0
x/D = 13
x/D = 24
x/D = 36
x/D = 50
x/D = 55
x/D = 0
x/D = 13
x/D = 24
x/D = 36
x/D = 50
x/D = 55
-0.4
0.4
Time [s]
10
10
10
0.4
-0.4
4
x/D = 0
x/D = 13
x/D = 24
x/D = 36
x/D = 50
x/D = 55
0.8
1.2
0.8
0.4
Time [s]
1.2
Correlations R
Correlations R
Time [s]
1.2
0.8
0.4
-0.4
0
x/D = 0
x/D = 23
x/D = 44
x/D = 66
x/D = 91
x/D = 102
0.8
Correlations R
Correlations R
0.8
1.2
Correlations R
1.2
-0.4
0
10
Time [s]
Time [s]
Figure 6.25 Correlations function for normal (top) and bubble (below) jet cases for
different y and x/D distances.
We can observe that the correlations functions are decreasing more rapidly as a
function of time if we are closer to the jet nozzle. At some distance x/D we have
maximum and again it decreases faster after some farther distance. It is interesting to
see more rapid reduction of the correlation distribution for the bubble jet case. For
initial, close to the nozzle region, there is no correlation because there is no sufficient
accuracy to 25 Hz, as we go farther, the correlation begins to appear (in the scales of
recirculation).
Averaging the correlation functions we get similar to the original distribution, three
different examples are presented in Figure 6.26 proving not significant influence of
averaging number.
169
x/D = 42
1 average
2 averages
3 averages
4 averages
5 averages
'Bubble' jet, y = 13 cm
x/D = 36
1 average
2 averages
3 averages
4 averages
5 averages
0.8
Correlation R
0.8
Correlation R
1.2
1.2
0.4
0.4
-0.4
2
10
0.4
-0.4
0
x/D = 50
1 average
2 averages
3 averages
4 averages
5 averages
0.8
Correlation R
1.2
-0.4
0
Time [s]
12
16
20
Time [s]
Time [s]
Figure 6.26 Averaged correlation functions for three different cases.
More detailed analysis of the correlation functions demonstrates the two different
scales influence in some cases; the difference between one and two scales appearance
cases on the correlation distribution is shown in Figure 6.27. The additional scale effect
is visible for x/D = 76 case for about T = 2,7 seconds (a curve peak).
1.2
1.2
0.8
0.8
x/D = 76
Correlation R
Correlation R
x/D = 55
0.4
0.4
-0.4
-0.4
0
Time [s]
Time [s]
Figure 6.27 The correlation functions for two different x/D values for single phase jet
case and y = 4cm.
We can relate the correlation function with integral scale as shown in Figure 6.28.
Note that in this case the integral scale and Taylor microscale refer to the time.
170
Figure 6.28 Correlation as a time function.
The integral time scale is defined as:
Lt = R(t )dt
(6.5)
Analysing the correlations results (Figure 6.25) we can distinguish different time
scales for the various x/D distances. Close to the nozzle, the small (and more rapid)
scales are evident and they are increase as a downstream distance increases, and then for
farther values (x/D = 101-111 for single phase jet and x/D = 50 for two-phase jet)
decreasing of the scales.
The aim of this work is to investigate the experimental jets behaviour from macro
turbulence point of view taking into account scales between the integral scale L and the
Taylor microscale . As an example we present the range of Kolmogorov length scale
for our cases, it is shown in Figure 6.29.
The Kolmogorov length scale was calculated using the expression
1/ 4
3
=

171
(6.6)
using for the dissipation rate the formula used by Miller and Dimotakis (1991)
= 48(U 03 / d )[( x x0 ) / d ]4
(6.7)
where U0 is the outlet velocity and x0 is the virtual origin. The energy dissipation rate
results are shown in Figure 6.30 for both, the single and two-phase jet case. We should
mention here that the distance from the wall y is not considered in this form.
4E-007
8E-007
6E-007
Normal jet
Kolmogorov scale [cm ]
Kolmogorov scale [cm ]
3E-007
2E-007
1E-007
'Bubble' jet
4E-007
2E-007
0
0
40
80
120
20
40
x/D
60
x/D
Figure 6.29 Kolmogorov scale for normal (left) and bubble (right) jet case as an x/D
10000000000
10000000000
1000000000
1000000000
100000000
100000000
10000000
10000000
D issipation rate [cm 2 /s 3 ]
Dissipation rate [cm 2 /s 3 ]
distance.
Normal jet
Dimotakis
1000000
100000
10000
1000
1000000
100000
10000
1000
100
100
10
10
0.1
0.1
0
40
80
'Bubble' jet
Dimotakis
120
x/D
40
80
120
x/D
Figure 6.30 Dissipation rate for two jet cases as an x/D function.
The integral time scales (Hz) (Figure 6.31) and (s) (Figure 6.32) obtained from the
correlation functions for all jet cases are presented.
172
100
100
ADV frequency limit
ADV frequency limit

10
10
Integral time scale [Hz]
Integral time scale [Hz]
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
1
0.1
0.1
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
0.01
0.01
0
40
80
120
20
x/D
40
60
x/D
Figure 6.31 Integral time scales (Hz) for the single phase (left) and two-phase (right) jet
case for different y distance as an x/D function.
The ADV frequency limit is marked to confirm sufficient frequency range in order
to detect the integral scales ranges (the data below the ADV frequency limit).
16
16
Normal jet
y = 4 cm
y = 8 cm
y = 13 cm
'Bubble' jet
y = 4 cm
y = 8 cm
y = 13 cm
12
Integral time scale [s]
Integral time scale [s]
12
0
0
40
80
120
20
x/D
40
60
x/D
Figure 6.32 Integral time scales (s) for single phase (left) and two-phase (right) jet case
for different y distance as an x/D function.
It is interesting to observe the integral scale peak (especially for the time scale in
seconds case) for the single phase jet for y = 13 cm, for the two phase jet these peaks are
173
observable in all cases however in wall jet case (y = 4cm) for farther downstream
distance x/D.
Let us compare the integral scale Taylor microscale ranges (of our interest) for the
all jet cases, this is shown in Figure 6.33. The calculation of the Taylor microscale is not
entirely straightforward, requiring formation of certain flow correlation function(s), then
expanding in a Taylor series and using the first non-zero term to characterize an
osculating parabola (see Figure 6.28).
1E+012
100000000000
1E+012
100000000000
10000000000
100000000000
10000000000
10000000000
1000000000
100000000
10000000
Integral scale
Taylor microscale
1000000
100000
10000
1000
100
ADV frequency limit
Integral - T aylor scale range [Hz]
1000000000
Integral - T aylor scale range [Hz]
Integ ral - Ta ylo r scale range [H z]
1000000000
100000000
10000000
Normal jet , y = 8 cm
Integral scale
Taylor microscale
1000000
100000
10000
1000
100
10
1
40
80
120
100000
10000
1000
100
ADV frequency limit
0.1
0.1
Integral scale
Taylor microscale
1000000
0.01
10000000
10
ADV frequency limit

10
0.1
100000000
0.01
0
40
80
120
40
x/D
x/D
80
120
x/D
100000000000
10000000000
10000000000
10000000000
1000000000
1000000000
100000000
100000000
10000000
Integral scale
Taylor microscale
1000000
100000
10000
1000
100
ADV frequency limit
10
Integral - Taylor scale range [Hz]
100000000
1000000000
10000000
Integral scale
Taylor microscale
1000000
100000
10000
1000
100
ADV frequency limit

10
10000000
1000000
Integral scale
Taylor microscale
100000
10000
1000
100
ADV frequency limit

10
1
0.1
0.01
0
20
40
x/D
60
0.1
0.1
0
20
40
60
x/D
20
40
60
x/D
Figure 6.33 The integral scale Taylor microscale ranges for the single (top) and two
(bottom) phase jet case as an x/D function.
The results show the macro turbulence to be within range of the ADV device
frequency for some downstream distance range, both for the single phase and two phase
jets, x/D = 40 and x/D = 18 respectively and affirm the correctness of chosen x/D ranges
for the energy spectrum and structure function techniques.
174
CHAPTER 7
SAR AND EXPERIMENTAL IMAGES FRACTAL RESULTS
7.1 Introduction
In this chapter we present fractal results of both, environmental and experimental
jets images in order to compare their fractal features. For the environmental case we use
Synthetic Aperture Radar pictures captured in the northwest region of Mediterranean
Sea. Special attention was focused on the two Catalan river estuaries. Other sea surface
structures were analysed too, to compare fractality of the different cases. For the
experimental case we have done fluorescence jet simulating river estuary into the sea
and capturing sequence images. We assume that a river entering into the sea
surrounding has similar properties to a jet injected by some source but there are not the
same kinds of phenomenon. Final results are presented showing differences for distinct
cases.
7.2 River jets SAR images

We have analyzed two Catalan river (Llobregat and Ter) jets influencing to the
northwest region of Mediterranean Sea. This region is dynamically very active and has
been studied extensively and presents one of the richest biological areas in the
Mediterranean due to the strong upwelling and mixing often caused by the interaction of
bathymetry, the local currents and the dominant winds. These mechanisms may cause
changes local reflectivity in SAR so it is important to distinguish between dynamical
variations at the ocean surface. The basic data (obtained thanks to the Agncia Catalana
de lAigua) of the rivers characteristics and approximate corresponding Reynolds
numbers are shown in Table 7.1.
175
River
Volumetric flow rate
Estuary width
Estuary depth
m3 / s
[m]
[m]
Llobregat
19
260
2,4
71000
Ter
25
240
2,1
102000
Re
Table 7.1 Basic data of the Llobregat and Ter river and corresponding approximate
Reynolds number.
Much larger Reynolds number comparing with the experimental jets permits to
compare different scales effects on the behaviour of the turbulent flows. Described in
chapter 4, ImaCalc program was used to calculate the fractal dimensions using the Box
Counting method.
We have divided every river jet into five parts in order to calculate the fractal
dimension of the each one separately. The different outflow river configurations were
analysed in order to obtain the same wall and free jet phenomenon as used for the
experimental cases, some jets are cuddled to the coast and some are influencing free
to the sea. The distance from the river estuary is marked as x. The sketch examples of
separated zones for different river jets are shown in Figure 7.1. The parts were cut
from original SAR images of high-resolution (Table 7.2) and some examples are
featured in Figure 7.2. We have chosen Barcelona and Girona name group of the
images due to the vicinity of the adequately town.
Name of image
Resolution (pixels)
Barcelona1
2361 x 2422
Barcelona4
1693 x 1625
Girona1
2006 x 2080
Girona2
2007 x 2080
Table 7.2 The resolutions of different original SAR images.
176
A one step of the operation method is shown on Figure 7.3 and the same action was
repeated for every image.
Figure 7.1 The Llobregat (top) and Ter (below) river jets images of SAR presenting the
divisions.
We have analysed also the sequence of the two Barcelonas coast river jets (called
here Barcelona I and Barcelona II) in time, with the object of look up for fractal
dimension behaviour as some period function using the same calculation method. The
time periods for each case are: Barcelona I (26.01.1997 18.10.1997) and Barcelona II
(06.12.1996 15.05.1998). In this case, entire jet zone was taken into account for the
calculation.
Results of the analysis are shown in Figure 7.4. The interesting conclusion is drawn
that the maximum fractal dimension is increasing function as an x distance from the
177
coast. The time dependence is not direct function to affect the fractal aspect of the jet
structures, more dependence is on actual conditions and configuration.
Figure 7.2 SAR images of Barcelona (Barcelona 1 top left and Barcelona 4 top
right) and Girona (Girona 1 below left and Girona 2 below right) coasts.
178
1.6
1.6
Maximum fractal dimension D
Figure 7.3 The operation step for used Box Counting method (for entire jet zone case).
1.2
Barcelona 1
Barcelona 4
Girona 1
Girona 2
0.8
1.2
0.8
Barcelona I
Barcelona II
0.4
0.4
0
0
10
20
Distance [km]
30
40
200
400
Days
600
800
Figure 7.4 Fractal dimension for Barcelona and Girona river jets as a distance x from the
estuary (left) and Barcelona river jets as a time function.
7.3 Sea vortex analysis

In order to compare fractal dimension of different surface structures we analyzed the
sea vortex patterns using SAR images. Two examples of satellite sights with such kind
of phenomenon are shown in Figure 7.5 presenting the northwest part of the
Mediterranean Sea. The selected structures are marked (red boxes with numbers)
containing vorticity occurrences and one convective feature (number 7) apart to
179
compare their fractality. The enlarged areas of interest are shown in Figure 7.6 and the
surface field in pixels is detailed in Table 7.3. Results are presented in Figure 7.7
showing different fractal dimension distribution for the vortex and convective
structures, in first case we have maximum value for grey levels between 180 and 220
and for second case more flat distribution for maximum value is for ranges between 40
and 160, where 0 is for black colour and 255 is white colour. A clear plateau of a
constant value of the maximum fractal dimension indicates that this measure is the same
for the dierent intensity values of the SAR images.
Figure 7.5 Vortex structures in the sea surface (1-6) and a convective pattern (7) in the
northwest Mediterranean Sea SAR images
Number of image
Surface (pixels)
35317
35550
76679
58503
71908
101492
27462
Table 7.3 Image surface (pixels).
180
Figure 7.6 Enlarged vortex (1-6) and convective (7) structures
181
1.6
1.6
1.6
1.2
0.8
Fractal dimension
Fractal dimension
Fractal dimension
1.2
0.8
0.4
0.4
0
50
100
150
Grey level
200
250
300
50
100
150
Grey level
200
250
300
0.8
0.4
0
0
1.2
50
100
150
Grey level
200
250
300
1.6
1.6
1.2
0.8
Fractal dimension
Fractal dimension
Fractal dimension
1.6
1.2
0.8
0.4
0.4
0
100
200
300
0.8
0.4
0
0
1.2
100
Grey level
200
300
100
200
300
Grey level
Grey level
2
Fractal dimension
1.6
1.2
0.8
0.4
0
0
100
200
300
Grey level
Figure 7.7 Fractal dimension for all, under consideration, structures as a grey level
function.
7.4 Experimental jet results

In order to deduce different fractal dimension for environmental and experimental
structures, a basic vertical jet configuration using fluorescence method was performed
and developing sequence of the jet images are shown in Figure 7.8. Not very high outlet
Reynolds number was chosen in order to simulate in approximate way a similar to river
estuary case using laboratory configuration. The 2 second duration sequence is under
consideration (0,67 s between every image). A colour palette of the image was reduced
to get 256 grey levels. The surface of the jet as a number of sequence (or time function)
is presented in Figure 7.9 and reveals almost linear dependence. Results of the fractal
182
dimension as a grey level is demonstrated in Figure 7.10. The truncated grey level range
from 0 to 80 is detected however the maximum fractal dimension number for higher
values (between 80 and 160 depending of the number of sequence) is noticeable. There
was not found direct relation between the time and maximum fractal dimension, the
values are between 1,55 and 1,59. Higher resolution with wider grey level images
experiments are need in order to obtain more detailed distribution of the fractal
dimension as a grey level function. This is done in next point of this chapter.
Figure 7.8 Developing jet sequence.

80000
Surface of jet (pixels)
60000
40000
20000
0
1
3
4
Number of sequence
Figure 7.9 Almost lineal dependence of the surface of the jet on the number of sequence
(or time).
183
1.2
1.2
1
0.8
0.4
0.8
40
80
120
Grey level
160
200
0
0
1.6
40
80
120
Grey level
160
200
1.6
0.4
0.8
0.4
0
40
80
120
Grey level
160
200
160
200
80
120
Grey level
160
200
0.8
0.4
0
0
80
120
Grey level
1.2
Fractal dimension
0.8
40
1.6
1.2
Fractal dimension
1.2
0.8
0.4
0
0
1.2
0.4
Fractal dimension
1.6
Fractal dimension
1.6
Fractal dimension
Fractal dimension
1.6
0
0
40
80
120
Grey level
160
200
40
Figure 7.10 Fractal dimension for jet developing sequence as a grey level function.
Other experiments (images were obtained thanks to prof. Allen Bateman from the
UPC) consist of three different water tank configurations:
A: The water enters the tank through the four screws inclined 30 degrees with
respect to the horizontal ground.
B: The water enters the tank by the short side of the floor and 5 m from the wall.
C: The water enters directly through a pipe of 2 m in diameter, forming a jet.
There was used a chemical blue dye diluted in a concentration 50g used as a
chemical blue dye diluted in a concentration-50g/750 l - 6.7% in volume. The volume
flux is 15 l/min. In Figure 7.11 we present some examples of sequence images for the
three different cases. The main objective of these experiments is to see which of these
configurations get better mixing efficiency in the tank.
184
C
Figure 7.11 Sequence images for the three cases of the experimental setup (A, B and C).
Results of the multifractal analysis are presented in Figure 7.12. A number of every
fractal dimension curve is related to the sequence number, for example the number 0040
is for 40th sequence, etc. It is interesting to observe the fractal dimension curve
maximum displacement with the number of sequence. For initial time of the jet
developing a maximum fractal dimension has peaks for the higher grey levels and with
time it tends to smaller grey level values. It is perceptible for all configurations with
different tempo. There is noticeable other peak for high grey levels (about 220) but
almost not discernible for the configuration C. For this case the mixing efficiency factor
has larger influence what is perfectly seen on the last sequence image. For the other two
cases decreasing of the peak with time is detected confirming the influence of the
mixedness. It is important conclusion allowing relating the mixing efficiency with
multifractal dimension so main objective of these experiments was achieved.
Future work may include investigation of the interaction between various parallel
jets or plumes relating the fractal dimension and mixing efficiency. Gonzalez-Nieto et
al. (2009) found that for this kind of configuration the mixing efficiency is increasing.
185
1.6
1.6
Fractal dimension D
Fractal dimension D
1.2
0.8
Configuration A
0040
0090
0140
0190
0240
0.4
1.2
0.8
Configuration B
0100
0132
0170
0240
0300
0.4
0
0
50
100
150
Grey level
200
250
50
100
150
Grey level
200
250
1.6
Fractal dimension D
1.2
0.8
Configuration C
0030
0070
0120
0170
0210
0.4
0
0
50
100
150
Grey level
200
250
Figure 7.12 Fractal dimension D as a grey level function for the three
configurations: A, B and C.
Let us analyse the maximum fractal dimension for every sequence for the three
configurations; this is shown in Figure 7.13. Similar, but with different values,
maximum fractal dimension curves distribution as a number of sequence functions are
observable for the configurations B and C. Note that there are horizontally (the jet is
parallel to the largest tank side) positioned jet cases. The configuration A has distinct
maximum fractal dimension curve distribution and in the three cases we can notice a
peak for some time value. More detailed analysis (larger number of analysed images)
could reveal more information about fractality behaviour of the different configurations.
The range of the maximum fractal dimension value is wider (D = 1,48 1,65)
comparing with lower resolution fluorescent jet analyzed before, where D = 1,55 and
1,59, confirming the importance of the image quality.
186
1.68
Configuration A
Configuration B
Configuration C
1.64
1.6
1.56
1.52
1.48
1
3
Number of sequence
Figure 7.13 Maximum fractal dimensions as a number of sequence function for the
three configurations (A, B and C).
The effect of the wall presence is other parameter taken under consideration in this
work. We analysed different images of the configurations divided our jets into two
categories: a free and wall jet and results are presented in Figure 7.14. The number of
sequence was not the object of our interest in this case; we tried to look for some
relation between these two kinds of jets from the fractal point of view in order to
investigate effect of the wall. We can distinguish some differences in fractal dimension
distribution between free and wall jet cases, especially for the configuration A it is more
accented, where more typical wall jet occurs on the perpendicular to the jets axis tank
side (impinging jet case). Nevertheless, we should emphasize here that the set of our
configurations is more related to the free jet set up, typical wall jet cases should resolve
any doubts. The maximum fractal dimension seems to be higher for the wall jet case
testifying different mixing conditions. The fractal dimension analysis is valuable tool to
investigate the boundary layer effect on the jet behaviour.
187
1.6
1.6
Fractal dimension D
Fractal dimension D
1.2
0.8
Configuration A
Free jet cases
Wall jet cases
0.4
1.2
0.8
Configuration B
Free jet cases
Wall jet cases
0.4
0
0
50
100
150
Grey level
200
250
50
100
150
Grey level
200
250
Fractal dimension D
1.6
1.2
0.8
0.4
Configuration C
Free jet cases
Wall jet cases
0
0
50
100
150
Grey level
200
250
Figure 7.14 Fractal dimension as a grey level function comparing a free and wall jet
cases.
Flohr and Olivari (1994) described distribution of a passive scalar in a turbulent jet
using fractal geometry. They analysed images of cross-sections of a circular turbulent
jet at moderate Reynolds numbers and show fractal scaling of iso-concentration lines
and that the fractal dimension varies for different concentration levels. Suggestion that
Reynolds number and downstream position are not affecting the multiftactal structure
was supposed but they were wrong. Catrakis and Dimotakis (1996) found a fractal
measures to be functions of both scalar threshold and Reynolds number. The coverage
of level sets of jet-fluid concentration in the two-dimensional images was found to
possess
a scale-dependent-fractal dimension
that
increases
continuously with
increasing scale, from near unity, at the smallest scales, to 2, at the largest scales.
In the turbulent mixing of a passive scalar, surfaces of constant mixed-fluid
concentration, or isoscalar surfaces, are highly convoluted in turbulent-jet flows, as well
188
as other high-Reynolds-number turbulent flows. The jet Reynolds number was varied in
this investigation in the range from 4.5 x l03 to 18 x l03.
Here we present some examples of different Reynolds numbers jets in order to
refute theory of Flohr and Olivari (1994) and reaffirm that of Catrakis and Dimotakis
(1996). In Figure 7.15 we show examples of analysed from the multifractal point of
view images and in Figure 7.16 results of fractal dimension analysis of the jet with
different Reynolds numbers: 1000, 2500 and 10000. More number of images was
processed to obtain relationship between the multifractal dimension and the Reynolds
number. Clearly, the fractal dimension distribution dependence on the Reynolds number
is detectable and some similarity of the curves is shown. The maximum fractal
dimension is decreasing with increase of the Reynolds number.
Figure 7.15 Examples of the analysed jets images (Dimotakis (2000, 2001).
1.6
Fractal dimension D
1.2
0.8
Re = 1 000
Re = 2 500
Re = 10 000
0.4
0
0
50
100
150
Grey level
200
250
Figure 7.16 Fractal dimension D as a grey level function for three different Reynolds
numbers: 1000, 2500 and 10000.
189
7.5 Fractal dimension analysis conclusions
Differences in the multifractal dimension distributions are easily observable

comparing the different sea surface structures and much smaller Reynolds number flows
using laboratory experiments with LIF (Laser Induced Fluorescence) and capturing
images of the evolution of a single jet demonstrating that multifractal analysis can be
used to distinguish among various turbulent structures both in environmental and
laboratory flows.
On the basis of the fractal dimension analysis of the SAR images we can notice that
fractal dimension value is increasing with distance from the source of river jets. The
outermost parts of the jet have had more time to acquire the turbulent structure. In most
cases the maximum fractal dimension of the outer parts of the jets tends to have a
maximum of about 1.4. It is apparent that the vorticity originated by the wall (or
boundary layer) increases large-scale turbulence, and probably also mixing at the region
of interaction between the jet and the vortices produced by it. This seems to occur both
in the jets and plumes experiments indicating an increase in the maximum fractal
dimension of the interface centre between D =1.3 and 1.4.
In most cases there is a monotonic increase of the maximum fractal dimension as
the jet moves away from the source (rivers in most cases). In some cases, however, the
starting values are higher (1.5-1.7) which could possibly indicate a non equilibrium high
Reynolds number initial turbulence forcing.
In the case of the SAR images region of the jet close to an estuary is characterised
by the black hue (smooth surface of the sea). Further from the estuary the environmental
conditions cause that the sea is chopper and we can observe whiter levels of reflectivity
which indicates a rougher sea surface (the jet is also more complex because of the
environmental dispersion). The black colour of the jets close to the coast represent
fresh water normally mixed with tensioactive pollutants flowing to the sea.
In the case of the jet experiment, the turbulence is produced by the momentum and
local shear produced at the jets source. In the SAR detected river induced jets, most of
the turbulence is produced by environment. In the case of experimental jet we have to
take into account the 3 dimensionality of the flow and also factors like the range of
scales or the different number of image pixels considered.
190
The jets that attach themselves to the coast, behave more like wall jets and possibly
the influence of the boundary layer produces the observed smaller increasing of the
fractal dimension as a downstream function but with higher values as was shown for the
experimental jets also.
We can observe also that grey level for maximum fractal dimension value is
increasing with distance from the source. It is not shown here due to the space limitation
but deduced by detailed, part by part, analysis results that are available.
Probably, because there is no ambient external turbulence for experimental jet case,
the level of fluorescence concentration at which the fractal dimension is maximal does
not vary much. Furthermore, we should mention that the calculation method was
different, for the environmental jets we divided every one into five parts, for the
experimental case we analysed the whole image of every sequence.
The vortical structures exhibit a slightly higher fractal value (1.6) for the higher
SAR reflectivity (white) and a linear increase from the darker features (lines that
indicate the eddy structure, mostly due to Langmuir convergence lines). It is obvious
that the dark features, being elongated and smoother have a smaller fractal dimension
than the background area between the spiral structures. But the appearance of a linear
increase is not clear. The fact that convective structures take place in all images and the
structures are marked both by darker (meaning a smooth surface) and white (rougher
sea surface areas) zones explains that at a wider range the maximum fractal dimension
(D) is about the same as seen in Figure 7.7 for convective cells case. The integral scales
which seem dominant range between 15 and 20 km showing a limit at the Rossby
deformation radius.
191
CHAPTER 8
CONCLUSIONS
The main aim of this thesis is to describe and demonstrate key aspects of the
structure of the turbulent jets and plumes from the experimental and environmental
point of view and their interaction with coherent structures and the effects connected
with them. Most practical flows occurring in nature and in engineering applications
involve non-homogeneous turbulent flows that are affected by boundaries or body
forces. Great progress has been made in the last years on the structure and theory of
homogeneous and isotropic turbulence, but non-homogeneous or boundary affected
flows lack a comprehensive theory. The mechanics of the turbulent jets, although
studied during the last decades, still is a paradigm of flow behaviour, together with
wakes and boundary layers and of great interest to researchers.
Moreover, there are still some important problems unsolved in the behaviour of the
turbulent cascades and their structure because of the past limits of measurements
methods. In recent years, thanks to improved remote sensing we can observe
concentration the research work on the environment. It was observed by Craft and
Launder (2001), since the three-dimensional wall jet is an acutely sensitive flow for
assessing turbulence models; it is necessary to establish definitive experimental or,
possibly, LES results for the fully developed limit. In our work special attention was
focused on this region.
The present work is based on laboratory experiments and environmental

observations discussing different kind of common phenomenon related to these flows
taking into account a diverse range of Reynolds numbers Re. The series of detailed
experiments were performed in the laboratory of Fluid Dynamics of the UPC consisting
on visualizations and turbulent measurements of the velocity fields in order to obtain a
comprehension on the interaction that lead to mixing and mass transport between
boundaries and jets or free shear layers. A strong motivation has been provided for a
careful re-analysis of the older experiments with implementation of the new advances.
The used configurations are composition of the different basic cases. Elevating an
axisymmetric jet from a wall and adding a back plane is complex due to the numerous
192
interactions that determine the behaviour of the flow. This situation seems to be most
common in practical jet applications. Because most practical applications involve the
near field of wall jets, it is important to study the distribution of velocities and turbulent
stresses in these regions.
Different downstream ranges were considered to investigate the jet structures
behaviour. Law and Herlina (2002) performed measurements up to 50 nozzle diameters,
while Lane-Serff (1993) observed in the far field at distances 250d to 550d of the flow.
In this thesis we consider intermediate ranges up to x/D 110 in order to complement
previous studies.
Most of the existing works are based on the mean velocity results (Tachie et al.
2002 or Kwon and Seo 2005), here we focus on the turbulent velocity fluctuations
components.
The obtained data in the fully developed range was found to be reasonably
consistent with similarity as shown also by Eriksson et al. (1998).
We proven the existence of the different size scales interacting, for example the
interaction of large turbulence scales in the outer layer of the jet with smaller scales in
the inner layer creates a complicated flow field and determines the development of the
wall jet. The energy spectrum and structure function techniques were useful to
determine these effects.
The higher Reynolds number (5000 for the lower Reynolds number case and 20000
and 13000 for higher Reynolds number cases) comparing with previous works (LaneSerff 1993 or Webster et al. 2001) were considered in this study in order to investigate
the flows with higher turbulence intensities.
Different experimental methods and techniques were used, such as Acoustic

Doppler Velocimeter, Laser Induced Fluorescense or image capturing to get detailed
results. The Acoustic Doppler Velocimetry (ADV) was used, focusing special attention
on large sample data in order to obtain more detailed statistical, spectral and structure
functions information. The direct ADV measurements showed that spreading in the
spanwise direction is increased by the approach of the wall and accentuated in the near
wall region and the interaction of the wall roll-up breaks the symmetry of the jet, the
wall-jet data confirmed axisymmetry (Knowles and Myszko 1998). The strong lateral
spreading arises from the creation of streamwise vorticity as was shown by Craft and
Launder (2001) and recirculation zone is revealed within the flow that carries material
193
from the periphery of the wall jet back to its initial regions (Fairweather and Hargrave
2002). It is clear that the production of vortices is the key element in initial jet dilution. We
proven that, because a free jet entrains fluid from both sides, it spreads faster, and,
therefore, it centreline velocity decays faster than that for the wall jet in the flow
development region near the nozzle exit.
The experiments show how external velocity fluctuations change significantly the
energy of fluctuations in the mixing outer layer, causing a rapid growth of the wall jet.
Spectral estimations of the flow components agree with theoretical spectral estimations
based on Kolmogorovs model (Voulgaris and Trowbridge 1997), especially for the
fully development flow region. In other, less stable regions we should compare results
with other experimental techniques and use more advanced post-processing filters.
We aimed to understand the behaviour of turbulent jets in a deeper and more

detailed way incorporating the recent advances in non-homogeneous turbulence, for
example, statistics, structure function analysis, multifractal techniques and extended
self-similarity, between others.
The statistical turbulence magnitudes, such as the standard deviation, skewness, and
kurtosis are useful tools to investigate different jet structures phenomenon, for example
the wall, buoyancy or mixing effects. The energy spectrums allow us to analyse energy
transfer for different scales and we fix our attention on the macro turbulence where the
larger scales and higher energy play dominant role. The different additional techniques,
for example line fits or adimensional numbers were used to facilitate detailed
information. More detailed techniques are mentioned also as possible future work on
this topic. The wall effect is perceptible on the energy spectrums modifying the
turbulence cascade.
The structure function analysis and extended self similarity were performed to
manifest no homogeneity of the jets as occur in real life. The series of velocity fields
were measured with an ADV velocimetry during long steady periods in order to
improve the convergence of higher-order structure functions but due to the
configurations complexity and non-homogeneity high deviations from the existing
estimations were observed. We proved that the energy spectrum is modified by
attributing less energy to small scales (in some jet regions) in comparison with that
predicted by the standard Kolmogorov distribution and it enhances diffusion with
respect to the usual Richardson law of non-intermittent turbulence (Jou 1997). The
194
higher orders of the structure functions are characterized by very irregular scaling
exponent, contrasting the theoretical results for homogeneous, isotropic and fully
developed turbulence. They suffer from the experimental conditions being affected by
recirculation large scale structures present in the flow for farther downstream distances
especially. More detailed analyses of the intermittency must focus on the exponents of
the scaling laws for the different moments of the velocity.
The importance of this study intensifies the fact that a numerical resolution of
existing problems is possible only if experimental results are available for the studied
configuration. The originality of the used experimental setups gives detailed results as a
base for comparisons.
Other part of this work is related to the multi-fractal analysis of SAR (Synthetic
Aperture Radar) and experimental jet (and plume) images. A comparison of the multifractal structure of these two types of techniques allows us to determine the spectral
structure of the turbulence cascade and helps to determine mixing properties and
relationships between them. Other aims of the investigation under way were to
determine the structure of ocean surface detected jets (SAR) and compare wall and
boundary effects on the structure of jets including baroclinic vorticity production. An
innovative technique used to investigate the turbulent interactions within the inverse and
direct cascades near jets was to measure the spectral and fractal structure of the non
homogeneous jets and develop multi-fractal techniques which we believe will be useful
for environmental and industrial monitoring.
Turbulence is sometimes described as objects consisting of a hierarchy of scales.
Such a description leads to the expectation that the theory of fractals must be applicable
to turbulence (Jaw and Chen 1999); this is confirmed in this work.
Both, the environmental (SAR) and experimental images were analysed to reveal the
fractal structure of non homogeneous jets affected by different levels of turbulence.
Other kind of the sea structures (vortices, convective structures) were tested also. The
structure of turbulent jets with boundary layer effect for laboratory and field
observations is described by Sekula and Redondo (2008).
The image analysis was able to help in the study of the structure of the turbulence, if
both velocity structure functions and multifractal measures can be compared both at the
large (or external) scales, of the order of the integral length scale and at the small (or
195
internal) scales where mixing takes place. It was interesting to relate the scaling
exponent of the structure functions or the spectral slope to the fractal dimension as was
shown in Chapter 3.
The results of the fractal analysis are only a geometrical tool but it was interesting to
compare changes in the fractal dimension as an indication of the self-similarity
transition to fully developed turbulence between the dierent experimental and
environmental set-ups. Information about the mixing can be extracted from the
evolution of the fractal dimension measurements.
We have conrmed that multifractal analysis can be used to distinguish among
dierent turbulent structures, both in the laboratory and in the environment field.
On the basis of the fractal dimension analysis of the SAR images we noticed that the
fractal dimension value was increasing with distance from the source of jets. Even in the
cases when just a single image was available, the outermost parts of the jet have had
more time to acquire the turbulent structure.
A new type of experiment showing the evolution of a jet in a turbulent environment
of the type shown by Redondo et al. (1995) will be able to investigate both types of
behaviour detected in the SAR image multifractal analysis:
1. When the initial level of jet induced turbulence is lower than the ambient one,
then there is a growth in the fractal dimension maximum as well as some changes in the
level of grey (roughness) at which this maximum occurs.
2. When the initial turbulent forcing (mostly due to the shear structure) comes from
the outlet of the jet, and this level is higher than the environmental average turbulence,
then the maximum fractal dimension decreases until a level of 1.4.
The multifractal dimension analysis confirmed different fractality characteristics of

the wall and free jet cases. The jets that attach themselves to the coast, behave more like
wall jets and possibly the influence of the boundary layer produces the observed smaller
increasing of the fractal dimension as a downstream function. We found the grey level
for maximum fractal dimension function as a downstream distance.
For the experimental jet case there was not variation of the maximum fractal
dimension as a grey level function, because there is no ambient external turbulence and
196
the calculating method was different. The wall effect has influence on fractal analysis
results changing (increasing) the maximum fractal dimension value for the wall jet
configuration.
It is interesting to note how mixed tensioactive products in the ocean surface
produce more complex fractal structures at intermediate levels that would indicate
probably the values with highest gradients.
The relation between fractal analysis and spectral analysis can be very useful to
determine the evolution of scales, even in the eld with video recordings, or as
presented here with SAR images. The use of routine satellite information by SAR or
other sensor types may be of great interest to build a seasonal database of the dynamic
conditions of the mesoscale turbulence in the ocean, after several years of observations
the dominant patterns and the causes for dierent topological characterisations might be
understood better (Redondo 2002). The dierent multifractal formalisms can be used to
discriminate between dierent physical processes that, despite being similar, have
dierent transport mechanisms for the dierent scales, or time. From the comparison of
the multifractal plots of the well dened SAR detected vortices with those of convective
cells, vortices show a maximum complexity for the low reectivity values, while
convection, probably because the basic instability happens everywhere at the same time,
exhibits almost the same fractal dimension for a wide range of intermediate SAR
reectivity (Platonov et al. 2008).
It is possible to relate fractal dimension D, scaling exponent q and the order of the
structure function q as was presented in chapter 3, by Jou (1997) and by Sekula and
Redondo (2008).
8.1 Future works
Further work might consist on more advanced turbulent jet configurations with
application of the experimental techniques (for example, LIF, PIV) with more powerful
and precise instrumentation (cameras, lens, lasers, etc). As an example, a jet affected by
grid stirred turbulence is shown on Figure 8.1.
197
Figure 8.1 Experimental setup of a horizontal jet affected by a grid stirred turbulence.
In the same way future work should include the study of the structure functions of
higher orders, for this reason we need larger measurements data by increasing the
capturing time and employing the higher frequencies mensuration apparatus. Fractal
analysis would be also useful in the study of the different image structures.
The fractal tool used in SAR ocean images may also be used for further fractal
analysis of environmental flows.
198
REFERENCES
ABRAHAMSSON H., JOHANSSON B., LOFDAHL L. (1997) The turbulence field of a fully
developed three-dimensional wall jet. Internal report 97/1, Dept. Thermo- and Fluid Dynamics,
Chalmers University of Technology, Goteborg, Sweden
AGRAWAL A. (2005) Measurement of spectrum with particle image velocimetry, Experiments
in Fluids 39: 836 840
AGRAWAL A., PRASAD A. K. (2003) Integral Solution for the Mean Flow Profiles of Turbulent
Jets, Plumes and Wakes, Journal of fluids engineering, Vol. 125
ALLISON S. W., PARTRIDGE W. P. Laser-Induced Fluorescence Imaging
ANSELMET F., ANTONIA R. A., DANAILA L. (2001) Turbulent flows and intermittency in
laboratory experiments, Planetary and Space Science 49: 1177-1191
ANTONIA R. A., ZHOU T., ROMANO G. P. (1997) Second- and third-order longitudinal
velocity structure functions in a fully developed turbulent channel flow, Phys. Fluids Vol. 9, No.
11
BABIANO A., DUBRULLE B., FRICK P. (1995) Scaling properties of two dimensional
turbulence Phys. Rev, E. 52, 3719
BABIANO A., DUBRULLE B., FRICK F. (1997) Some properties of two dimensional
turbulence, Phys. Rev. E. 55, 2693
BATCHELOR G. K. (1954) Heat convection and buoyancy effects in fluids, Quarterly Journal of
the Royal Meteorological Society, Volume 80 Issue 345, Pages 339 358
BENAISSA A., FLECK B. A., POLLARD A. Wall Effects on Large Scale Structure Evolution in
an Axisymmetric Jet
BENZI R., CILIBERTO S., TRIPICCIONE R., BAUDET C., MASSAIOLI F., SUCCI S.(1993)
Extended self-similarity in turbulent flor, Physical Review E, Volume 48, Number 1
BEZERRA M. O., DIEZ M., MEDEIROS C., RODRIGUEZ A., BAHIA E., SANCHEZ-ARCILLA
A., REDONDO J. M. (1997) Study on the Influence of Waves on Coastal Diffusion Using Image
Analysis, Applied Scientific Research, Volume 59, Numbers 2-3 / June, 1997
BORATAV O. N., PELZ R. B. (1997) Structures and structures functions in the inertial range of
turbulence, Phys. Fluids 9 (5)
CANTALAPIEDRA I. R., REDONDO J. M. (1995) Mixing in zero-mean flow turbulence
Mixing in geophysical flows Eds. J.M. Redondo and O. Metais CIMNE, Barcelona
199
CAO N., CHEN S., SHE Z. S. (1996) Scaling and Relative Scalings in the Navier Stokes
Turbulence, Physical Review Letters, Volume 76, Number 20
CATRAKIS H. J., DIMOTAKIS P. E. (1996) Mixing in turbulent jets: scalar measures and
isosurface geometry, J. Fluid Mech. (1996), vol. 317, pp. 369 - 406
CAVALLETTI A. , P. A. DAVIES P. A. (2003) Impact of Vertical, Turbulent, Planar, negatively
Buoyant Jet with Rigid Horizontal Bottom Boundary, Journal of Hydraulic Engineering, vol. 129,
No 1, pp 54 62
CHENG N. S., LAW A. W. K. (2001) Measurements of Turbulence Generated by Oscillating
Grid, Journal of Hydraulic Engineering
CRAFT T. J., LAUNDER B. E. (2001) On the spreading mechanism of the three-dimensional
turbulent wall jet, J. Fluid Mech., vol. 435, pp. 305-326
DALZIEL S. S., REDONDO J. M. (2007) New visualization methods and self-similar analysis in
experimental turbulence studies, Models, Experiments and Computation in Turbulence, R.
Castilla, E. Oate and J. M. Redondo (Eds.), CIMNE, Barcelona, Spain
DE SILVA I. P. D., FERNANDO H. J. S. (1998) Experiments on collapsing turbulent regions in
stratified fluids, J. Fluid Mech. Vol. 358, pp. 29 60.
DIMOTAKIS P. E. (2000) The mixing transition in turbulent ows, J. Fluid Mech. (2000), vol.
409, pp. 69-98.
DIMOTAKIS P. E. (2001) Recent advances in turbulent mixing, Mechanics for a New
Millennium, 327344.
EAD S. A., RAJARATNAM N. (2004) Plane Turbulent Wall Jets on Rough Boundaries with
Limited Tailwater, Journal of Engineering Mechanics
EAD S. A., RAJARATNAM N. (2002) Plane Turbulent Wall Jets in Shallow Tailwater, Journal
of Engineering Mechanics
ERIKSSON J. G., KARLSSON R. I., PERSSON J. (1998) An experimental study of a twodimensional plane turbulent wall jet, Experiments in Fluids 25, 50-60
FAIRWEATHER M., HARGRAVE G.K. (2002) Experimental investigation of an axisymmetric,
impinging turbulent jet.1. Velocity field, Experiments in Fluids 33, 464 471
FISCHER H. B. (1979) Mixing in inland and coastal waters, Academic Press, Inc. Ltd, London
FLOHR P., OLIVARI D. (1994) Fractal and multifractal characteristics of a scalar dispersed in a
turbulent jet, Physica D: Nonlinear Phenomena, Volume 76, Issues 1-3, Pages 278-290
FORTHMANN E., (1936) Turbulent Jet Expansion, English Translation, NACA Technical
Memorandum TM-789.
200
FRAUNIE P., BERREBA S., CHASHECHKIN Y. D., VELASCO D., REDONDO J. M. (2008)
Large eddy simulation and laboratory experiments on the decay of grid wakes in strongly stratified
flows, IL NUOVO CIMENTO C 31, 909-930
FRISCH U. (1995) Turbulence. The legacy of A. N. Kolmogorov, Cambridge University Press
GEORGE W. K., ABRAHAMSSON H., ERIKSSON J., KARLSSON R. I., LOFDAHL L.,
WOSNIK M. (2000) A similarity theory for the turbulent plane wall jet without external stream J.
Fluid Mech., vol. 425, pp. 367-411
GIOSTRA U., CAVA D., SCHIPA S. (2002) Structure functions in a wall-turbulent shear flow,
Boundary Layer Meteorology 103: 337 359
GONZALEZ-NIETO P. L., CANO J. L., REDONDO J. M. (2007) Buoyant mixing processes
generated in turbulent plume arrays, Fisica de la Tierra, 19 205-217
GONZALEZ-NIETO P. L., CANO J. L., REDONDO J. M. An experimental model of mixing
processes generated by an array of top-heavy turbulent plumes IL NUOVO CIMENTO, DOI
10.1393/ncc/i2009-10330-y
GRAU J. (2006) Introduction to multifractal analysis in the environment, Lectures in
environmental turbulence, CIMNE, Barcelona
HINZE J. O. (1975) Turbulence 2nd edition, New York, McGraw-Hill
HSIAO F. B., SHEU (1994) Double row vortical structures in the near field region of a plane wall
jet, Experiments in Fluids, Volume 17, Number 5, pp. 291-301
JAW S. Y., CHEN C. J. (1999) Near-Wall Turbulence Modelling Using Fractal Dimensions,
Journal of Engineering Mechanics
JIRKA G. H. (2004) Integral Model for Turbulent Buoyant Jets in Unbounded Stratified Flows.
Part I: Single Round Jet, Environmental Fluid Mechanics 4, pp. 1-56
JOU D. (1997) Intermittent turbulence: a short introduction, Sci. Mar., 61 (Supl.1): 57-62,
Lectures on plankton and turbulence, C. Marrase, E. Saiz and J. M. Redondo (eds.)
KIT E. L. G., STRANG E. J., FERNANDO H. J. S. (1997) Measurement of turbulence near shearfree density interfaces, J. Fluid Mech. Vol. 334, pp. 293-314
KNOWLES K., MYSZKO M. (1998) Turbulence measurements in radial wall-jets, Experimental
Thermal and Fluid Science 17, pp. 71-78
KOLMOGOROV A. N. (1941) Local structure of turbulence in an incompressible fluid for very
large Reynolds numbers Comptes rendus (Doklady) de lAcademie des Sciences de lU.R.S.S., 31:
301-305
KOLMOGOROV A. N. (1962) A refinement of previous hypotheses concerning the local structure
of turbulence in a viscous incompressible fluid at high Reynolds number, J. Fluid. Mech. 13, pp 82-85
201
KWON S. J., SEO I. W. (2005) Reynolds number effects on the behaviour of a non-buoyant round
jet, Experiments in Fluids 38, 801 812
KUANG C. P., LEE J. H. W. (2006) Stability and Mixing of a Vertical Axisymmetric Buoyant Jet
in Shallow Water, Environmental Fluid Mechanics 6: 153-180
LANE-SERFF G. F. (1993) Investigation of the fractal structure of jets and plumes, J. Fluid
Mech., vol. 249, pp. 521-534
LAUNDER B. E., RODI W. (1983) The turbulent wall jet measurements and modelling,
Annual review of fluid mechanics, vol. 15, pp. 429-459
LAW A. W. K., HERLINA (2002) An Experimental Study on Turbulent Circular Wall Jets,
Journal of Hydraulic Engineering
LESIEUR M. (1997) Turbulence in Fluids. Third Revised and Enlarged Edition, Kluwer
Academic
Publishers
LIST E. J. (1982) Turbulent jets and plumes Annual review of fluid mechanics, vol. 14, pp.
189212
LUMLEY J. L., YAGLOM A. M. (2001) A century of Turbulence, Flow, Turbulence and
Combustion, 66: 241-286
MAHJOUB O. B. (2000) Non-local dynamics and intermittency in non-homogenous flows
Doctoral thesis, Universitat Politecnica de Catalunya
MAHJOUB O. B., REDONDO J. M., BABIANO A. (1998) Structure Functions in Complex
Flows, Applied Scientific Research 59: 299-313
MAHJOUB O. B., GRANATA T., REDONDO J. M. (2001) Scaling Laws in Geophysical Flows,
Phys. Chem. Earth (B), Vol. 26, No. 4:281-285
MANDELBROT B. (1983) The Fractal Geometry of Nature, New York: Freeman
MATULKA A. (2009) Turbulent Structure in Environmental Flows: Effects of Stratification and
Rotation, PhD Thesis, Universitat Politecnica de Catalunya, Barcelona
MEDINA P., SANCHEZ M. A., REDONDO J. M. (2001) Grid Stirred Turbulence: Applications
to the Initiation of Sediment Motion and Lift-Off Studies, Phys. Chem. Earth (B), Vol. 26, No. 4,
pp. 299 304
MI J., KALT P., NATHAN G. J., WONG C. Y. (2007) PIV measurements of a turbulent jet
issuing from round sharp-edged plate, Exp. Fluids 42: 625 637
MICHALLET H., MORY M. (2004) Modelling of sediment suspensions in oscillating grid
turbulence, Fluid Dynamic Research 35: 87-106
MILLER P. L., DIMOTAKIS P. E. (1991) Reynolds number dependence of scalar fluctuations in
a high Schmidt number turbulent jet, Phys. Fluids A 3(5)
202
NIKORA V. I., GORING D. G. (1998) ADV Measurements of Turbulence: Can We Improve

Their Interpretation? Journal of Hydraulic Engineering
ORLINS J. J., GULLIVER J. S. (2000) Measurements of Free Surface Turbulence
ORLINS J. J., GULLIVER J. S. (2003) Turbulence quantification and sediment resuspension in an
oscillating grid chamber, Experiments in Fluids 34: 662-677
PANTZLAFF L., LUEPTOW R.M. (1999) Transient positively and negatively buoyant turbulent
round jets, Experiments in Fluids 27, 117 125
PERRIER E., TARQUIS A. M., DATHE A. (2006) A program for fractal and multifractal analysis
of two-dimensional binary images: Computer algorithms versus mathematical theory, Geoderma
134: 284 294
PLATONOV A. (2002) Aplicacin de imgenes de satlite SAR en los estudios de contaminacin
marina y de dinmica de las aguas en el Mediterrneo Noroccidental, Doctoral thesis, Universitat
Politcnica de Catalunya
PLATONOV A., TARQUIS A., SEKULA E., REDONDO J. M. (2007) SAR observations of
vortical structures and turbulence in the Ocean, MODELS, EXPERIMENTS AND
COMPUTATION IN TURBULENCE, R. Castilla, E. Oate and J.M. Redondo (Eds.), CIMNE,
Barcelona, Spain
PLATONOV, A., CARRILLO A., MATULKA A., SEKULA E., GRAU J., REDONDO J. M.,
A.M. TARQUIS A. M. (2008) Multifractal observations of eddies, oils spills and natural slicks in
the ocean surface, IL NUOVO CIMENTO C, 31, 861-880.
RAJARATNAM N. (1976) Turbulent jets, Amsterdam (etc), Elsevier
REDONDO, J.M. (1990), The structure of density interfaces, PhD Thesis, Cambridge University.
REDONDO J. M. (2002) Mixing eciency of dierent kinds of turbulent processes and
instabilities, Applications to the environment,in Turbulent Mixing in Geophysical Flows, edited by
Linden P. F. and Redondo J. M. (CIMNE, Barcelona), pp. 131-157.
REDONDO J. M., CANTALAPIEDRA I. R. (1993) Mixing in horizontally heterogenous flows,
Applied Scientific Research 51: 217-222.
REDONDO J.M., DIAZ, BEZERRA (2009), Sintethic Turbulence. ERCOFTAC SERIES, In Press.
REDONDO, J. M., METAIS, O. (1995) Cloud entrainment by internal or external turbulence,
Mixing in geophysical flows, p379-392
REDONDO J. M., X. DURRIEU DE MADRON, P. MEDINA, M. A, SANCHEZ, E. SCAF (2001)
Comparison of sediment resuspensin measurements in sheared and zero-mean turbulent flows,
Continental Shelf Research 21, 2095 - 2103
203
REDONDO J. M., GRAU J., PLATONOV A., GARZN G. (2008) Analisis multifractal de
procesos autosimilares: imagenes de satelite e inestabilidades baroclinas, Revista Internacional de
Mtodos Numricos para Clculo y Diseo en Ingeniera, Vol. 24, 1, 25-48
REDONDO J. M., LINDEN P. F. (1998) Report on Turbulence and Mixing in Geophysical Flows
II, Applied Scientific Research 59: 89-110
REDONDO J. M., PLATONOV A. K. (2001) Aplicacin de las imgenes SAR en el estudio de la
dinmica de las aguas y de la polucin del mar Mediterrneo cerca de Barcelona Ing. Agua 8 15-23
REDONDO J. M., SANCHEZ M. A., CANTALAPIEDRA I. R. (1996) Turbulent mechanisms in
stratified fluids, Dynamics of Atmosphers and Oceans 24, 107 115
REDONDO J. M., PLATONOV A. (2009) Self-similar distribution of oil spills in European
coastal waters, Res. Lett. 4 014008
ROBERTS P. J. W., MAILE K., DAVIERO G. (2001) Mixing in Stratified Jets, Journal of
Hydraulic Engineering
RODRIGUEZ A., SNCHEZ-ARCILLA A., REDONDO J. M. AND Msso C. (1999)
Macroturbulence measurements with electromagnetic and ultrasonic sensors: a comparison under
high-turbulent flows, Experiments in Fluids, Volume 27, Number 1 / June, 1999, pp. 31-42
SANCHEZ M. A., REDONDO J. M. (1998) Observations from Grid Stirred Turbulence, Applied
Scientific Research 59: 243-254
SCASE M. M., CAULFIELD C. P, DALZIEL S. B., HUNT J. C. R. (2006) Time-dependent
plumes and jets with decreasing source strengths, J. Fluid Mech., vol. 563, pp. 443-461
SCASE M., CAULFIELD C., DALZIEL S., HUNT J. (2006) Plumes and jets with time-dependent
sources in stratified and unstratified environments, 6th International Conference on Stratified
Flows, Perth, Western Australia
SCHERTZER D., LOVEJOY S., SCHMITT F., CHIGIRINSKAYA Y., MARSAN D. Multifractal
Cascade Dynamics and Turbulent Intermittency
SCHNEIDER M. E., GOLDSTEIN R. J. (1994) Laser Doppler measurement of turbulence
parameters in a two-dimensional plane wall jet, Physics of Fluids, Volume 6 / Issue 9
SEKULA E., REDONDO J.M (2006) Turbulent structure of axisymetric wall jets, Lectures in
Environmental Turbulence, Eds. Babiano A., Fraunie P., Redondo J.M. and Vassilicos J.C, Ed.
C.U.M. CD, Barcelona.
SEKULA E., REDONDO J. M. (2008) The structure of turbulent jets, vortices and boundary layer:
Laboratory and eld observations, IL NUOVO CIMENTO Vol. 31 C, N. 5-6
SEURONT L., SCHMITT F., LAGADEUC Y. (2001) Turbulence intermittency, small-scale
phytoplankton patchiness and encounter rates in plankton: where do we go from here?, Deep-Sea
Research I 48 pp. 1199-1215
204
SFORZA P.M., HERBST G. (1970) A study of three-dimensional, incompressible, turbulent wall

jets AIAA J 8: 276-283
SHE Z. S., CHEN S., DOOLEN G., KRAICHNAN R. H., ORSZAG S. A. (1993) Reynolds
Number Dependence of Isotropic Navier Stokes Turbulence, Physical Review Letters, Volume
70, Number 21
SOCOLOFSKY S. A., ADAMS E. E. (2002) Multi-phase plumes in uniform and stratified
crossflow, Journal of Hydraulic Research, Vol. 40, No. 6
SOUS D., BONNETON N., SOMMERIA J. (2005) Transition from deep to shallow water layer:
formation of vortex diploes, European Journal of Mechanics B/Fluids 24: 19-32
SREENIVASAN K. R. (1999) Fluid turbulence, Reviews of Modern Physics, Vol. 71, No. 2
SREENIVASAN K. R., ANTONIA R. A. (1997) The phenomenology of small-scale Turbulence,
Annu. Rev. Fluid Mech. 29: 435-72
TACHIE M. F., BALACHANDAR R., BERGSTROM D. J. (2002) Scaling the inner region of
turbulent plane wall jets, Experiments in Fluids 33, 351 354
TACHIE M. F., BALACHANDAR R., BERGSTROM D. J. (2004) Roughness effects on
turbulent plane wall jets in an open channel , Experiments in Fluids 37, 281 292
TAILLEUX R. (2009) On the energetics of
stratified turbulent mixing, irreversible
thermodynamics, Boussinesq models, and the ocean heat engine controversy, J. Fluid Mech,
Accepted.
TARQUIS A. M., MORATO M. C., CASTELLANOS M. T., PERDIGONES A. (2008)
Comparison of structure function and detrended uctuation analysis of wind time series, Il Nuovo
Cimento, Vol. 31 C, N. 5-6
TENNEKES H., LUMLEY J. L. (1972) A first course in turbulence, The MIT Press
TSAI Y. S., HUNT J. C. R., NIEUWSTADT F. T. M., WESTERWEEL J., GUNASEKARAN B. P.
N. (2007) Effect of Strong External Turbulence on a Wall Jet Boundary Layer, Flow Turbulence
Combust
TSINOBER A. (2001) An Informal Introduction to Turbulence, Kluwer Academic Publishers
TURCOTTE D. L. (1992) Fractals and chaos in geology and geophysics, Cambridge University
Press
TURNER J. S. (1973) Buoyancy effects in fluids, Cambridge University Press
WAHL T. L. (2000) Analyzing ADV data using WinADV, 2000 Joint Conference on Water
Resources Engineering and Water Resources Planning & Management
WANG X. K., TAN S. K. (2006) Experimental investigation of the interaction between a plane
wall jet and a parallel offset jet, Exp Fluids
205
WATER VAN DE W., HERWEIJER J. A. (1999) High-order structure functions of turbulence, J.

Fluid mech., vol. 387, pp. 3-37
WEBSTER D. R., ROBERTS P. J. W., RAAD L. (2001) Simultaneous DPTV/PLIF measurements
of a turbulent jet, Experiments in Fluids 30: 65-72
VENAS B., ABRAHAMSSON H., KROGSTAD P. A., LOFDAHL L. (1999) Pulsed hot-wire
measurements in two- and three- dimensional wall jets, Experiments in Fluids 27: 210-218
VINDEL J. M., C. YAGE, REDONDO J. M. (2008) Structure function analysis and
intermittency in the atmospheric boundary layer, Nonlin. Processes Geophys., 15, 915929
VOULGARIS G., TROWBRIDGE J. H. (1997) Evaluation of the Acoustic Doppler Velocimeter
(ADV) for Turbulence Measurements, Journal of Atmospheric and Oceanic Technology, Volume
15
YAGE, C., VIANA S., MAQUEDA G., REDONDO J. M. (2006). Influence of stability on the
flux-profile relationships for wind speed, m, and temperature, h, for the stable atmospheric
boundary layer. Nonlinear Processes in Geophysics, 13, 185203.
ZHANG J. B., CHU V. H. (2003) Shallow Turbulent Flows by Video Imaging Method, Journal
of Engineering Mechanics
ZHOU M. D., WYGNANSKI I. (1993) Parameters governing the turbulent wall jet in an external
stream, AIAA journal, vol. 31, no5, pp. 848-853
206
APPENDIX 1
In Chapter 1 of this thesis we introduced basic information about turbulence and
mentioned state of the art situation on the turbulent flows. Here we will continue to
describe the actual work done on this subject.
Ap.1.1 Turbulence
As was mentioned in introduction chapter of this thesis (Chapter 1) there are many
books precisely describing problems of turbulence and turbulent phenomenon. We mention
here some more complete and used during the evolution of this study in chronological
order:
P. A. Davidson (2004) Turbulence: an introduction for scientists and engineers,
A. Tsinober (2001) An Informal Introduction to Turbulence,
M. Lesieur (1997) Turbulence in Fluids. Third Revised and Enlarged Edition,
U. Frish (1995) Turbulence. The legacy of A. N. Kolmogorov,
N. Rajaratnam (1976) Turbulent jets,
J. O. Hinze (1975) Turbulence,
J. S. Turner (1973) Buoyancy effects in fluids,
H. Tennekes, J. L. Lumley (1972) A first course in turbulence.
The first one is based on a course by the author at the University of Cambridge and it is
a comprehensive text on turbulence and fluid dynamics aimed at undergraduates and
graduates in applied mathematics, physics, and engineering, and provides an ideal reference
for industry professionals and researchers. It bridges the gap between elementary accounts
of turbulence found in undergraduate texts and more rigorous accounts given in
monographs on the subject. Containing many examples, the author combines the maximum
of physical insight with the minimum of mathematical detail where possible. The text is
divided into three parts: Part I consist of a traditional introduction to the classical aspects of
turbulence, the nature of turbulence, and the equations of fluid mechanics. Mathematics is
207
kept to a minimum, presupposing only an elementary knowledge of fluid mechanics and

statistics. Part II tackles the problem of homogeneous turbulence with a focus on numerical
methods. Part III covers certain special topics rarely discussed in introductory texts. Many
geophysical and astrophysical flows are dominated by the effects of body forces, such as
buoyancy, Coriolis and Lorentz forces. Moreover, certain large-scale flows are
approximately two-dimensional and this has led to a concerted investigation of twodimensional turbulence over the last few years. Both the influence of body forces and twodimensional turbulence are discussed.
The book An Informal Introduction to Turbulence by A. Tsinober is an informal
introduction to the turbulence of fluids and it addresses the unresolved issues,
misconceptions, controversies, and major problems of the turbulence of fluids rather than
the conventional formalistic elements and models. The scope of the book is focused on the
purely basic aspects of the turbulent flows of incompressible fluids. The book is
intentionally written to appeal to a broad readership with the aim of making the turbulence
of fluids interesting and comprehensible to the interested engineer.
The book Turbulence in Fluids. Third Revised and Enlarged Edition by M. Lesieur is
an exhaustive monograph on turbulence in fluids in its theoretical and applied aspects, with
many advanced developments using mathematical spectral methods, direct-numerical
simulations and large-eddy simulations. It is still of great actuality on a topic of the utmost
importance for engineering and environmental applications, and presents a very detailed
presentation of the field. This book is an attempt to reconcile the theory of turbulence, too
often presented in a formal, isolated mathematical context, with the general theory of fluid
dynamics. It reviews, in a unifying manner, the main characteristics and general theorems
of rotational fluids (liquids or gases), with applications to aerodynamics and geophysical
fluid dynamics.
The book Turbulence. The legacy of A. N. Kolmogorov by U. Frish attempts to
predict the properties of flow and presents a modern account of turbulence. Elementary
presentations of dynamical systems ideas, probabilistic methods (including the theory of
large deviations) and fractal geometry make this a self-contained textbook. This is the first
book on turbulence to use modern ideas from chaos and symmetry breaking.
208
Turbulent jets of N. Rajaratnam is one of the first books about different configurations
of the turbulent jets. Some examples are plane turbulent free jets, circular confined jets or
cylindrical wall jet and this book focus of the theory.
The phenomena treated in the book Buoyancy effects in fluids by J. S. Turner all
depend on the action of gravity on small density differences in a non-rotating fluid. The
author gives a connected account of the various motions which can be driven or influenced
by buoyancy forces in a stratified fluid, including internal waves, turbulent shear flows and
buoyant convection. It is an excellent introduction to a rapidly developing field, first
published in 1973.
A First Course in Turbulence by H. Tennekes, J. L. Lumley is the first book
specifically designed to offer the student a smooth transitionary course between elementary
fluid dynamics and the professional literature on turbulent flow, where an advanced
viewpoint is assumed. In particular, dimensional analysis is used extensively in dealing
with turbulence problems whose exact solution is mathematically elusive. Dimensional
reasoning, scale arguments, and similarity rules are introduced at the beginning and are
applied throughout. A discussion of Reynolds stress and the kinetic theory of gases provide
the contrast needed to put mixing-length theory into proper perspective: the authors present
a thorough comparison between the mixing-length models and dimensional analysis of
shear flows. This is followed by an extensive treatment of vorticity dynamics, including
vortex stretching and vorticity budgets. Two chapters are devoted to boundary-free shear
flows and well-bounded turbulent shear flows. The examples presented include wakes, jets,
shear layers, thermal plumes, atmospheric boundary layers, pipe and channel flow and
boundary layers in pressure gradients. The spatial structure of turbulent flow has been the
subject of analysis in the book up to this point, at which a compact but thorough
introduction to statistical methods is given. This prepares the reader to understand the
stochastic and spectral structure of turbulence. The remainder of the book consists of
applications of the statistical approach to the study of turbulent transport (including
diffusion and mixing) and turbulent spectra.
Large number of works has been done related to the turbulence phenomenon and we
present some describing shortly different research aspects.
209
She et al. (1993) explored Reynolds Number dependence of turbulence energy spectra
and higher-order moments of velocity differences by numerical integrations of the
incompressive Navier-Stokes equations. The simulations have spatial resolutions up to 5123
and cover 15 Re 200, where Re is the Taylor microscale Reynolds number.
Mixing in zero-mean-flow turbulence was investigated by Cantalapiedra and Redondo
(1995). They investigated the local characteristics of horizontally distributed mixing and
presented a series of experiments on the mixing across a density stratified interface when
the turbulence is produced by a distribution of air bubbles rising through a section of an
initially sharp interface. The turbulence that causes the mixing can be related to the
characteristics of the bubbles. Conductivity probes were used to find point density
fluctuations at the turbulent density interface. The experiments suggest that there is a limit
to the amount of kinetic energy input rate which is converted into potential energy due to
geometrical considerations. In mixing produced by rising bubbles across a density
interface, mixing efficiencies can be measured for a range of Richardson number Ri.
A good introduction describing new visualization methods and self-similar analysis in
experimental turbulence studies was done by Dalziel and Redondo (2007). They present
results on several laboratory experiments where a sharp density interface generated by
either salt concentration or heat, advances due to grid stirred turbulence or to buoyancy.
The fractal and multifractal measures were used as a function of the position and time of
the advance of the mixing fronts as topological indicators of the multi-scale of these
complex flows.
Other configuration and effects to investigate related to turbulence is zero-mean

turbulence produced by grid stirring.
Sanchez and Redondo (1998) presented some experimental techniques used to study
sediments extracted from the sea. The characterization of sediment behaviour is very
important in coastal dynamics, deep sea topography or global models of circulation. The
way to obtain enough information in order to characterize the sediment in natural state is
not a solvable problem, even in the easier, well-controlled, world of the laboratory. Nonhomogeneity and non-stationarity of the experiments introduce important difficulties.
210
In other paper, together with Cantalapiedra (1996) they studied probability distribution
of basic instabilities appearing in stratified flow and point density fluctuations. Various
parameters of the mixing process have been changed in the experiments, to investigate
mixing. From the density fluctuation measurements or conversely from the mixedness
values in the oscillating grid experiments, we see that a region of intermediate Ri
(Richardson number) allows higher direct transport of unmixed fluid across the interface.
This mechanism should provide higher mixing efficiencies at this number showing
consistency with previous works.
Measurements of turbulence near shear-free density interfaces were done by Kit et al.
(1997). The experimental configuration consisted of a two-layer fluid medium in which the
lower layer was maintained in a turbulent state by an oscillating grid. Special attention was
given to the study of how a shear-free turbulent flow is distorted by the introduction of an
interface with the hope of applying results to the modelling of turbulence near density
interfaces. The introduction of a density interface into a mean shear-free turbulent flow
distorts the structure of turbulence around it. The effects of the interface are felt mainly at
large scales of turbulence. The rate of turbulent kinetic energy dissipation near the
interface shows a sharp decrease.
De Silva and Fernando (1998) investigated the properties of a collapsing turbulent patch
generated within a linearly stratified fluid by a sustained energy source and its long-time
evolution in the presence of lateral boundaries. An oscillating grid spanning the width of
the experimental tank was used as the turbulence source. Interest of this paper was to
calculate the eddy diffusivity of buoyancy associated with turbulent mixing with the patch.
During the initial spreading regime, ambient fluid enters to patch via return currents is
subjected to mixing and then is fed into the intrusion.
Orlins and Gulliver (2000) researched free-surface turbulence. It is important to the

transfer of heat and mass across an air-liquid interface. In the absence of wind shear and
surface waves, turbulence generated below the water surface is one of the controlling
factors in air-water mass transfer. Measurements were made using Particle Image
Velocimetry.
211
Redondo et al. (2001) presented in their paper two laboratory experiments with the
results of sediment lift off under sheared and zero-mean flow turbulence. They compared
seabed sediments extracted from the Gulf of Lyon. The conclusion is that video digitizing
of field events seems to be a promising technique for extended quantitative measurements
of lift off of sediments. A new approach to study sediment behaviour was used, while grid
stirring turbulence opens up new possibilities.
Medina et al. (2001) presented a laboratory experimental set-up for studying the
behaviour of sediment in presence of a turbulent field with zero mean flow. Particular
interest is shown on the initiation of sediment motion and in the sediment lift-off. A new
approach to study sediment behaviour was used showing that it is interesting to compare in
further detail the sediment transport mechanism in sheared flow and in zero-mean flow.
Cheng and Law (2001) investigated turbulence generated by a vertically oscillating grid
in a water tank using the Digital Particle Image Velocimetry technique. Their statistical
turbulence characteristics computed based on the experimental data agree well with the
previous results. The study quantitatively reveals the flow structure in the region near the
grid. It was found that the structure is closely related to the grid geometry.
In other work, the same authors (2003) used an oscillating grid chamber to study
sediment suspension, desorption of compounds from the resuspended sediment and airwater mass transfer. The chamber was designed to allow researchers to study desorption of
contaminants from cohesive sediments and the flux of those contaminants to the vapour
phase. They characterized the relationship between turbulence and sediment entrainment in
the device and developed similitude relationships between the turbulence in the device and
that found in a boundary layer flow. The turbulent flow field in the oscillating grid chamber
described was fairly uniform at sufficient distances away from the grid, regardless of the
method used to attach the grid to the drive shaft. The oscillating grid mixing chamber can
be used as an analogue to open-channel flow systems by setting operational parameters of
the grid (stroke, frequency, etc.) such that the total kinetic energy of the turbulence matches
that expected either at the bed or free surface for an open channel flow.
Modeling of this kind of processes was done by Michallet and Mory (2004). They used a
k- model to describe the steady of fine-grained sediments maintained in suspension by
purely diffusive turbulence, as generated in oscillating grid turbulence experiments. The
212
behaviour was shown to depend both on the bulk Rouse number and the product of the bulk
Rouse number and the bulk Richardson number, built on oscillating grid parameters. The
paper focuses on concentrated sediment suspensions.
Ap.1.2 Turbulent jets and plumes

Large existing research work on turbulent jets is available and here we concentrate in
general on experimental (but not only) part. To give some more clear idea of the enormity
of research problem we will list in several groups (different research aspects) studies done
till now. Special attention was directed to last years publications that include recent, more
complete and advanced results.
The most popular, on the turbulence and turbulent jets point of view, studies have taken
into consideration in this thesis are edited by the next publishers:
Experiments in Fluids
Journal of Fluid Mechanics
Journal of Hydraulic Engineering
Journal of Engineering Mechanics
Applied Scientific Research
There are many possibilities of jets configurations from the basic ones to much more
complicated. Here we will mention some examples of existing works.
One of the first and good works on the turbulent jets is the publication of List (1982). He
made an introduction for turbulent jets plumes and tried to provide current references and
suggest approaches for future research. Overall conclusion from a review of the literature
on jets and plumes is that some areas of the flow field received attention quite out of
proportion to other areas. Worthy work would be paper that contributes substantially to the
understanding of jet flows in the region of flow within ten jet diameters of the discharge
orifice. The author mentioned some existing gaps at that moment which deserved research
attention and there is actually developing research work on these subjects:
213
A comparison of turbulence intensity and turbulent flux measurements in jets,

where the data is obtained using new experimental methods.
Measurements of heat and momentum fluxes in buoyancy-dominated flows to help

establish the basic nature of such flows.
An assessment of the appropriateness of the self-similar hypothesis for jets and

plumes in density-stratified and uniform flow environments.
An evaluation of entrainment and entrainment hypotheses for jets and plumes in

density-stratified and uniform flow environments.
Establishment of mixing criteria at neutral density levels for jets and plumes in
stratified environments.
Determination of the influence of ambient turbulence levels on dilution within

buoyant jets.
Establishment of criteria for turbulent buoyant jet bifurcation.
Research contributions in all of these areas provide much clearer guidelines and better
insight to engineers and geophysicists concerned with buoyancy-driven and jet like motions
in the ocean and atmosphere.
An experimental study was carried out into a singular jet impinging onto a flat ground
board by Knowles and Myszko (1998). The jet was running at a fixed nozzle pressure ratio
of Reynolds number 90 000. Wall-jet data confirmed axisymmetry and showed the mean
velocity profiles to reach self-similarity within 2.5 nozzles diameter. Nozzle height is also
shown to affect wall-jet growth with increasing height generally causing a reduction in
wall-jet thickness.
Eriksson et al. (1998) made Laser-Doppler measurements in a plane turbulent wall jet at
a Reynolds number based on inlet velocity, Re = 9600. The initial development as well as
the fully developed flow was studied. The experiment showed momentum conserving to the
streamwise position x/b = 150. The data in the range 40 x/b 150 was found to be
reasonably consistent with similarity. The near wall region of the turbulent wall jet was
resolved for the first time, including the inner, negative, peak in the shear stress.
214
Pantzlaff and Lueptow (1999) measured the transient character of the jet issuing from
an upward nozzle centred at the bottom of a vertical cylindrical tank into bulk liquid of a
different density using flow visualization and PIV for varying densimetric Froude numbers
by varying the jet Reynolds numbers and the ratios of fluid densities. A key result of this
work is relation between the penetration depth and time scale with buoyancy flux F and the
jet momentum flux M for both positively and negatively buoyant jets.
A new theory for the turbulent plane wall jet without external stream was proposed by
George et al. (2000) based on a similarity analysis of the governing equations. It is
hypothesized that the inner part of the wall jet and the inner part of the zero-pressuregradient boundary layer are the same. A new theory has been set forth in which the outer
wall jet is governed by different scaling parameters than commonly believed. A strong
motivation has been provided for a careful re-analysis of the older experiments, and
perhaps a new generation of experiments over the entire range of Reynolds numbers.
The paper by Craft and Launder (2001) explored, using different levels of turbulence
closure, the computed behaviour of the three-dimensional turbulent wall jet in order to
determine the cause of the remarkably high lateral rates of spread observed in experiments.
Their computations confirm that the strong lateral spreading arises from the creation of
streamwise vorticity, rather than from anisotropic diffusion. The driving vorticity source is
created by the anisotropy of the Reynolds stresses in the plane perpendicular to the jet axis
rather than to the bending of mean vortex lines. It may be observed too, since the threedimensional wall jet is an acutely sensitive flow for assessing turbulence models; it would
be desirable to establish definitive experimental or, possibly, LES results for the fully
developed limit.
Roberts et al. (2001) made experiments that were performed to measure the mixing
characteristics of turbulent momentum jets discharged horizontally into a linearly stratified,
stationary environment. These flows can occur when sewage is discharged into water
bodies such a lakes. Special attention was focus on the minimum dilution and gross flow.
The centreline dilution was found to follow the results for an unstratified jet up to a point
where the jet begins to collapse under the influence of the stratification.
Fairweather and Hargrave (2002) obtained mean and fluctuating velocities and shear
stresses in an air jet impinging on a flat surface by Particle Image Velocimetry. A
215
recirculation zone is revealed within the flow that carries material from the periphery of the
wall jet back to its initial regions.
Some part of this thesis is related to, call here, bubble jet or more adequate, the twophase jet. Multi-phase plumes in uniform and stratified crossflow were studied by
Socolofsky and Adams (2002). In uniform crossflow, multi-phase plumes behave as mixed
single-phase plumes up to critical height, where the entrained fluid separates from the
dominant dispersed phase. From the experimental results, an empirical relationship was
calibrated. First, crossflow cause separation when their horizontal motion overcomes the
restoring force of entrainment and advects the entrained fluid away from the rising
dispersed phases. Second, stratification causes separation when the entrained fluid becomes
too dense to be carried by the dispersed phase and it intrudes horizontally into the ambient
fluid. Further experiments in stratifies crossflow confirmed the assumption that plumes
could be classified as either crossflow or stratification dominated. The modelling algorithm
developed to predict the fate of the separated fluid above the separation height allows
single-phase models to be applied to multi-phase plumes in crossflow.
The results of a laboratory investigation on the characteristics of a circular threedimensional turbulent wall jet were presented by Law and Herlina (2002). Measurements
were performed up to 50 nozzle diameters using a combined Particle Image Velocimetry
and Planar Laser Induced Fluorescence approach that captured simultaneously the planar
velocity and concentration fields. Results show that the velocity profiles exhibited
similarity in both the streamwise and spanwise directions after 20d and 25d respectively.
On the concentration characteristics, the streamwise profile showed a vertical gradient at
the wall that is consistent with the no-flux boundary condition.
The study of Tachie et al. (2002) reported refines laser Doppler anemometry
measurements in a turbulent plane wall jet. The mean velocity profiles in outer coordinates
(Um, y1/2) are shown.
Cavalletti and Davies (2003) showed the results of a series of laboratory modelling
experiments for the case of a vertical, turbulent, plane, negatively buoyant jet impinging on
a horizontal solid surface placed at some distance below the jet source. The results show
that the impingement results in the generation of a complex two-dimensional disturbance
216
field at the site of the impact and the generation of a buoyancy-driven boundary current
carrying away fluid from the impingement zone.
Integral methods were used to derive similarity solutions for several quantities of
interest including the cross-stream velocity, Reynolds stress, the dominant turbulent kinetic
energy production term and eddy diffusivities of momentum and heat for axisymmetric and
planar turbulent jets, plumes and wakes by Agrawal and Prasad (2003). A universal
constant is evaluated for axisymmetric and planar plumes. Expressions for cross-stream
velocity indicate outflow for jets and axisymmetric plumes near the axis, while inflow
occurs in the far field. Expressions for the entrainment coefficients of planar jets and
plumes are developed along the lines of their axisymmetric counterparts. Unlike plumes
and jets, the normalized cross-stream velocity and Reynolds stresses for wakes are
functions of downstream distance.
Zhang and Chu (2003) studied shallow turbulent flows produced in a tank of small
thickness investigating the friction effects on large-scale turbulent motion of small depth.
The experiments in the double-tank apparatus have provided data for space-and-time
variations of the tracer concentration in the shallow turbulent jets.
Tachie et al. (2004) reported the effects of the surface roughness on the mean flow
characteristics for a turbulent plane wall jet created in an open channel. They summarized
that the inner region of the mean velocity profiles for the smooth-wall turbulent wall jet and
low Reynolds number turbulent boundary layer collapse reasonably well.
The paper of Ead and Rajaratnam (2004) attempted to describe effects of boundary
roughness and tailwater depth on the characteristics of plane turbulent wall jets on rough
beds, which are important in the field of hydraulic engineering. The time-averaged axial
velocity profiles at different sections in the wall jet were found to be similar, with some
difference from the profile of the classical plane wall jet. The normalized boundary layer
thickness /b, where b is the length scale of the velocity profile, was equal to 0.35 for wall
jets on rough boundaries compared to 0.16 for the classical wall jet.
Kwon and Seo (2005) investigated the behaviour of a non-buoyant circular water jet
discharged from a contraction nozzle, for Reynolds numbers ranged from 177 to 5142
based on the experimental results of the mean velocity field obtained by particle image
velocimetry. The mean axial velocity profile near the nozzle for a Reynolds number of 177
217
did not show a top-hat profile, whereas the other cases with a Reynolds number higher than
437 showed almost top-hat distributions.
Sous et al. (2005) studied in their paper the transition from a deep to a shallow water
layer and the formation of quasi-two-dimensional vortex dynamics. Vortices are
experimentally generated by a circular horizontal turbulent pulsed jet. Qualitative and
quantitative experimental investigations have been shown on the jet development.
The dual-jet flow generated by a plane wall jet and parallel offset jet at an offset ratio of
d/w = 1.0 (d jets nozzle diameter, w distance between jets axis) was investigated using
Particle Image Velocimetry (PIV) by Wang and Tan (2006). The particle images are
captured, processed and subsequently used to characterize the flow in terms of the 2D
velocity and vorticity distributions. Statistical characteristics of the flow, including the
mean velocities and the Reynolds stresses, have been obtained. While the free shear layers
of a single jet (either wall or offset) are standard Kelvin-Helmholtz roll-ups, the near field
of the dual-jet flow is characterized by the periodic Karman-like vortex street in the inner
shear layer regions.
Kuang and Lee (2006) studied the stability, mixing and effect of downstream control on
axisymmetric turbulent buoyant jets discharging vertically into shallow stagnant water
using 3D Reynolds-averaged Navier-Stokes equations (RANS) combined with a buoyancyextended k- model. The steady axisymmetric turbulent flow, temperature and turbulence
fields are computed using the finite volume method on a high-resolution grid. The
computations provide the unavailable to that moment, detailed flow and temperature fields.
An example of a shallow thermal discharge jet is provided by the cooling water discharge
site of the Sizewall Power Stations on the Thames River estuary.
The classical bulk model for isolated jets and plumes due to Morton, Taylor & Turner
was generalized to allow for time-dependence in the various fluxes driving the flow by
Scase et al. (2006). Separable time-dependent solutions for plumes and jets were found.
Their experiments demonstrated the robust nature of turbulent plumes in unstratified
ambient fluids. When the background ambient fluid is linearly density stratified with
constant buoyancy frequency, N, they found that a reduction in source buoyancy leads to
the establishment of a new lower buoyancy plume with lower rise height than the initial
plume.
218
The initial stage of the development of a wall jet under the influence of strong external
turbulence was studied in a novel shear-flow mixing-box experiment by Tsai et al. (2007).
An experimental study at the inlet Reynolds number of 2200 was carried out using LIF
flow visualization and LDA measurements. The experiment shows how external velocity
fluctuations change significantly the velocity profile and the energy of fluctuations in the
mixing outer layer, causing a rapid growth of the wall jet.
Ap.1.3 Fractal methods

Lane-Serff (1993) in his paper described an experimental study of round, turbulent jets
and plumes, investigating the effects of buoyancy on the fractal structure. For both jets and
plumes, the Reynolds number at the source was in the range 800 to 1800 and the flow was
observed in the far field at distances 250d to 550d. The measured dimensions of the
contours were found to be the same for jets and plumes despite the presence of buoyancy
forces in the latter case. It should be noted that presuming the flows to be Reynolds-number
independent leads to the same value of the fractal dimension regardless of the flow type.
To introduce the intermittency effects of the energy cascade process into turbulence
modeling, a new fractal turbulence scale base on k, , and the fractal dimension of
turbulence dissipation D was introduced by Jaw and Chen (1999). A fractal scale and low
Reynolds number turbulence model is proposed. The proposed turbulence model is first
examined in detail by predicting a two-dimensional channel flow and then applied to
predict a backward-facing step flow. Turbulence is sometimes described as objects
consisting of a hierarchy of scales. Such a description leads to the expectation that the
theory of fractals must be applicable to turbulence. Fractal dimensions of various turbulent
flows were determined, e.g., regions of active dissipation, turbulent/nonturbulent interface
of boundary layer flow, axisymmetric jet, plane wake, mixing layer and so on. Among
them, the energy cascade, dissipation process is a common characteristic of all of the
turbulent flows. Numerical results are compared with the DNS budgets, experimental data
and the other models.
Perrier et al. (2006) presented a tool to carry out the multifractal analysis of binary,
two-dimensional images through the calculation of the Rnyi dimensions and associated
statistical regressions. The estimation of a (mono) fractal dimension corresponds to the
219
special case. When we work with images from the real world instead of mathematical
models, we cannot rely on any theoretically based calculation of the fractal, non-fractal or
multifractal behaviour. The difficulties associated with real images are that they have a
finite size, their scale boundaries are defined by the image resolution and we use to use
numerical log/log fits (which tend to straighten any curve). Generalized dimensions can be
obtained with the method of moments for any image and box size range. The software
depicted in their paper provides a tool for comparing the result for different box size ranges.
This paper is a contribution in the wide scientific research field of the comparison of
discrete computer algorithms versus mathematical theory.
Ap.1.4 Experimental methods

Voulgaris and Trowbridge (1997) evaluated the ADV for turbulence measurements.
Simultaneous measurements of open-channel flow were undertaken in a 17-m flume using
and ADV and a laser Doppler velocimeter. Flow velocity records obtained by both
instruments were used for estimating the true (ground truth) flow characteristics and the
noise variances encountered during the experimental runs. Mean flows measured with the
ADV agree within 1% with those measured by the LDV. ADV-measured Reynolds stress
values were underestimated by only 1%, even without any correction for noise. Turbulence
intensity of the vertical component can be resolved accurately by the sensor, while the
intensity of the downstream components suffers from a high noise term that is an
inescapable feature of the geometry of the ADV. Good agreement was found between
bottom shear velocities calculated using the Reynolds stress and logarithmic profile
methods. Spectral estimations of the horizontal and vertical flow components agree with
theoretical spectral estimations based on Kolmogorovs model. The noise variance along
the measuring beam of the ADV signal consists of an electronic-related component and a
flow-related component.
Interesting discussion about the improvement of ADV measurement interpretation was
done by Nikora and Goring (1998). They developed a simple technique to reduce the
Doppler noise influence on measurements of turbulence using the Acoustic Doppler
Velocimeter (ADV) and derived basic relationships to estimate turbulence characteristics
using information about the Doppler noise. The procedure involves, firstly, measurement of
220
instantaneous velocities in the flow under consideration; secondly, measurement of the

Doppler noise components in still water taken from the flow investigated; and thirdly,
estimation of turbulence characteristics using both sets of measurements and relationships
developed in the paper. This technique was tested for three typical environments and it was
shown that the Doppler noise can change the true turbulence characteristics significantly,
even for high level turbulence flows.
Simultaneous measurements of instantaneous velocity and concentration fields were
performed using digital particle tracking velocimetry and planar laser-induced fluorescence
for a turbulent jet at a Reynolds number Re of 3000 by Webster et al. (2001). The
measurements of mean velocity, turbulence stresses, mean concentration and turbulent flux
of tracer all collapse onto self-similar profiles in the far field of the jet. In each case, the
profile shape and magnitude agreed well with the previous velocity and concentration
measurements. The intention was to use the measurement system to study complex
environmental flows.
Agrawal (2005) showed that the measurement of turbulent spectrum using whole field
velocity techniques such as Particle Image Velocimetry (PIV) is possible. Data from the
axial plane of a self-similar turbulent axisymmetric jet, at a Reynolds number, based on
Taylor microscale of 30 were analysed. The method proposed is distinct because the
vortices are directly identified by author and an estimate of the number of vortices is also
possible. On the other hand, in conventional methods Fourier transform are used on time
series of a signal and the vortices roughly correspond to a Fourier mode such that their
energy is the square of the amplitude of that mode.
Mi et al. (2007) presented an experimental investigation of a turbulent jet issuing from a
round sharp-edged orifice plate into effectively unbounded surroundings at Re = 72,000.
Planar measurements of velocity were conducted using Particle Image Velocimetry (PIV)
in the near and transition regions. Primary coherent structures occur in the near field of both
the smooth contraction jet and the sharp-edged orifice plate jet, which can be revealed by
the instantaneous streamlines relative to a frame of reference translating near to the mean
vortex convection velocity. The flow structure of the sharp-edged orifice plate jet in the
near field is more three-dimensional than that of the smooth contraction jet. The exit
221
turbulence intensity of the sharp-edged orifice plate jet depends upon the upstream
conditions and these influences propagate downstream.
Ap.1.5 Structure functions and intermittency

Extended self-similarity in turbulent flows was studied by Benzi et al. (1993). They
reported undetected form of self-similarity not known until that time. ESS holds at high as
well as at low Reynolds number and it is characterized by the same scaling exponents of the
velocity differences of fully developed turbulence. Their claims are supported by a number
of experimental and numerical results described elsewhere and an attention is confined to a
set of experimental data obtained by a hot-wire measurements of the velocity field in a
wind tunnel.
More particular case was done by Cao et al. (1996). They carried out high-resolution
direct numerical simulations of 3-D Navier Stokes turbulence with normal viscosity and
hyperviscosity. It is found that the inertial-range statistics, both the scaling and the
probability density functions are independent of the dissipation mechanism, while the neardissipation-range fluctuations show significant structural differences. It was examined the
behaviour of the energy spectrum in the Fourier space too.
Boratav and Pelz (1997) investigated the deviations from the Kolmogorov 1941 laws of
inertial range of turbulence using the results from the direct numerical simulations of an
unforced flow starting from a high-symmetry initial condition. The resolution is 3003 points
and the Taylor scale Reynolds number is in the order of 100. The most exciting result in
this work is that the deviation from the Kolmogorov scaling of the lateral derivatives of the
structure functions is much different (and more) than those associated with the longitudinal
structure functions. They also investigated what kinds of flow structures are responsible for
different intermittency deviations along the longitudinal and lateral directions. A large
portion of this work is devoted to investigating whether the extended self similarity can be
used in turbulent flows.
Antonia et al. (1997) measured second and third-order moments of temporal increments
of the longitudinal velocity fluctuation using hot wires in a fully developed turbulent
channel flow and analyzed using values of the mean energy dissipation rate inferred from
direct numerical simulation data for the same flow and Reynolds numbers. The results
222
indicate that dissipative range scales approach isotropy relatively rapidly when the mean
shear is negligible.
Mahjoub et al. (1998) in their work proposed a classification to determine the
intermittency and mixing ability. The variation of the structure functions and the scaling
exponent in decaying non-homogeneous turbulence produced by a grid and by a jet was
measured with a sonic velocimeter. Their investigation showed in most cases that 3
(absolute scaling exponent) depends strongly on the separation distance, except near the
source of turbulence at high Re. The BDF model (after Babiano, Dubrulle and Frick) was
used to check their experimental results and relate the values of the relative scaling
exponents. The two flows studied show different distributions of intermittency.
It is difficult to calculate high-order structure functions due to a large calculating error
nevertheless there are works investigating it, for example by Van de Water and Herweijer
(1999). They studied the scaling behaviour of high-order structure functions in a variety of
laboratory turbulent flows. The statistical accuracy of the structure function benefits from
novel instrumentation for its real-time measurement. Structure functions are studied in two
different kinematical situations. The (standard) longitudinal structure functions are
measured using Taylors hypothesis. In the transverse case an array of probes is used and
no recourse to Taylors hypothesis is needed. The experimental results are discussed in the
light of the multifractal model that explains intermittency in a geometrical framework. In
this paper they gave firm evidence for the anomaly of scaling exponents and gave initial
results for the transverse scaling exponent. The transverse scaling exponent deviates more
strongly from the Kolmogorov (1941) result than the longitudinal scaling exponent. An
intriguing question is whether new physics will emerge when using ever longer integration
times that allow the measurement of ever larger-order moments. The authors believe that
experiments on the small-scale structure of turbulence should exploit different geometric
arrangements for measuring velocity statistics. They did not touch upon the scaling with
Reynolds number.
Structure functions in a wall-turbulent shear flow were done by Giostra et al. (2002).
Wavelet and quadrant analyses were applied to turbulent velocity data in order to
investigate the transition from the anisotropy of energy-containing eddies to the isotropy of
the inertial subrange scales. The quadrant analysis of the wavelet coefficients of
223
longitudinal and vertical velocity components allows the evaluation of the velocity
structure functions and the momentum cospectrum as a function of the separation distance
and of the quadrants. The analysis was performed on turbulent velocity data relative to
near-neutral atmospheric conditions. A conditioning procedure was applied in order to
eliminate the small-scale intermittency. The different characteristics of the dominant
structures (ejections and sweeps) and of the secondary ones (reflections and deflections)
were stressed by the quadrant analysis of the structure functions. The proposed technique
seems to represent a powerful tool for investigating the interaction between the anisotropic
energy-containing scales and the quasi-isotropic scales.
224
APPENDIX 2
Babiano, Dubrulle and Frick (BDF) method
There are many conflicting theories that may be used to relate intermittency, scaling
exponents of the structure functions, and sometimes a geometrical characterization of the
space occupied by different levels of dissipation in a turbulent flow, we can thus
characterize some of the different turbulent situations from the parameters of the BDF
model explained in Mahjoub (2000) which originated from Babiano et al. (1995, 1997).
The advantage of this method is that it may also be used in highly non-homogeneous
turbulent cascade situations.
In the inertial range, the q-th order velocity structure functions, Sq, follow a power
relation with the scale; that is to say, in case of a temporal scale, :
Sq
(A2.1)
where the q are the absolute scaling exponents of the velocity structure functions.
Moreover, there is also a power relation between Sq and the third order structure function,
S3, throughout the inertial range:
S q S3
(A2.2)
where q is the so-called relative scaling exponent.

According to this last power relation is more evident, in general, than the first one.
In 3D turbulence, in homogeneous and isotropic conditions, the Kolmogorov theory
establishes that the q, present a linear dependency with the order q of the corresponding
velocity structure function; specifically:
q = q/3
225
(A2.3)
According to this theory, the PDFs of the velocity increments present a Gaussian form.
Later, a deviation with regard to this linear form has been observed; this deviation is the
so-called intermittency. Moreover, when there is intermittency, the tails of the PDFs of the
velocity increments depart from the Gaussian form and can be approached by stretched
exponentials. Therefore, the scaling exponents can be used to characterize the turbulence
intermittency.
There are many turbulence models that try to explain the phenomenon of the
intermittency. The BDF model defines the following structures:
q+1
q =
(A2.4)
where is the energy transfer due to non-lineal terms of the Navier-Stokes equation with
an approached value that can be drawn from the expression:
v 3 (t + ) v 3 (t )
(A2.5)
The averages of the expression correspond to the usual decomposition:
= +
(A2.6)
The dependence of the fluctuations with the scale shows the degree of intermittency,
while the dependence of the averages
with this scale refers to a lack of stationarity.
Now,
0 = lim q
q 0
= lim q
q
226
(A2.7)
Therefore:
0 =
(A2.8)
Assuming a power-law of the structure with the scale:
(A2.9)
The scaling exponent of the 0 order structure, 0, shows the relative contribution to the
energy transfer of the structures the least intermittent. Thus, this exponent is not
representative of the degree of intermittency, but of the degree of dependence with the scale
showing the average value of the energy transfer, that is to say, the degree of stationarity
(the greater is 0, the greater is the lack of stationarity). Conversely, the scaling exponent
represents the contribution of the most intermittent structures and thus, this parameter can
be used as an intermittency parameter (more precisely, the greater the difference between
and 0; the greater the intermittency; when the difference is zero, we face an absence of
intermittency).
Assuming that > 0 and that q can be expressed as:
q = + ( 0 )h(q )
(A2.10)
being h(q) a monotone decreasing positive function of q, less or equal to 1.

According to the above mentioned expressions, h(q) must tend to 1 when q tends to 0 and
h(q) must tend to 0 when q tends to the infinite.

The value of h(q) that accomplishes these conditions is:
h(q ) = exp( aq )
a being the slope of h when q = 0:
227
(A2.11)
a = h' (0 ) =
q' (0 )
0
(A2.12)
A relation can be established between the structures of one order and those of the
immediately previous order:
lq+1
l lq
So
lq
q1
l l
h (q )
h (q1)
(A2.13)
h(q )
corresponds to the parameter of the She-Leveque model.
h(q 1)
It follows that
( )
h(q )
= exp( a ) = q
( q1 )
h(q 1)
(A2.14)
Broadly speaking, this parameter depends on q, although in homogeneity and isotropy

conditions, as we have already shown, it has a universal value equal to 2/3.
On the other hand, we can come to the following proportionality through an iterative
q 1
process and representing the summation
h(q ) through I(q):

q =0
lq
l
l ( 0 )[I (q )q ]
(A2.15)
ul3
l
Assuming that the quotient
presents the same scaling properties than
, we
l
ul3
can obtain an estimate for the relative exponents:
228
q =
q q
q
+ I
3
3 3
being =
0
3
(A2.16)
q
If we give expression to I as a function of the parameter , then we have:
3
q =
1 q / 3 q
q
+

3
3
1
(A2.17)
Therefore, this model presents two characteristic parameters: and . The first one is
indicative of the degree of intermittency and the second one is dependent of the order of the
structure function. must be higher or equal to 0, and must be included in the interval
(0,1). According to that, it is possible to establish the relative scaling exponents when the
former parameters are modified.
For a certain order of the structure function and value of the parameter , the parameter
also present a clear relation with the scaling exponent (and thus, with the intermittency).
The close study of intermittency can explain sometimes the complex behaviour of
stratified and rotating non/homogeneous flows or when geometrical forcing changes the
topology of the turbulent cascade from a 2D to a 3D situation (Mahjoub et al. 1998, 2000),
this structure function based procedure has also been used in the atmosphere by Yague et
al. (2006), Vindel et al. (2008).
It is possible to find some pairs of values , , in order to represent the different

relative scaling exponents which characterize a certain atmospheric situation. However, this
fit is not possible in some special cases, corresponding to very stable turbulent atmospheric
situations, or where internal waves and turbulence are strongly interacting.
229

Thesis Final Julio 15 Emil

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Thesis Final Julio 15 Emil

Uploaded by

Copyright:

Available Formats

UNIVERSITAT POLITCNICA DE CATALUNYA

Departament de Fsica Aplicada

THE STRUCTURE OF TURBULENT JETS:

Director: Prof. Jos Manuel Redondo Apraiz

Dedicated to my girlfriend, Amaia and to my parents, Zbigniew and Lucyna.

To Departament de Fsica Aplicada de la Universidad Politcnica de Catalunya for facilities

State of the art

CHAPTER 2. BACKGROUND AND THEORETICAL CONSIDERATIONS

Turbulence and turbulent flows

Extended Self Similarity (ESS)

Plane turbulent free jet

Circular turbulent jet

Plane turbulent wall jets

Other jet and plume configurations

The turbulent boundary layers

CHAPTER 3. THEORY SAR AND FRACTALS

SAR (Synthetic Aperture Radar)

Self-affine fractals. Relationship between turbulence spectra and

The spectral method

Fractal characteristics of intermittent turbulence

CHAPTER 4. METHODOLOGY AND EXPERIMENTAL SETUP

Acoustic Doppler Velocimeter (ADV)

Laser Induced Fluorescence (LIF) and Planar Laser Induced

Particle Image Velocimetry (PIV)

Particle Tracking Velocimetry

CHAPTER 5. EXPERIMENTAL RESULTS

Lower Reynolds number jet experiments

Standard deviation (r.m.s. turbulence)

High Reynolds number wall jets experiments

Turbulence and mean velocity parameters

High Reynolds number, two-phase Bubble jet

Correlation and Covariance

Comparison of the three cases of the jet configurations

CHAPTER 6. SPECTRAL MEASUREMENTS, STRUCTURE FUNCTION

Energy spectrum Fast Fourier Transform (FFT)

Normal jet, y = 4 cm case

Normal jet, y = 8 cm case

Normal jet, y = 13 cm case

Bubble jet, y = 4 cm case

Bubble jet, y = 8 cm case

Bubble jet, y = 13 cm case

Energy spectrum results discussion

Structure function analysis

Normal jet, y = 4 cm case

Normal jet, y = 8 cm case

Normal jet, y = 13 cm case

Bubble jet, y = 4 cm case

Bubble jet, y = 8 cm case

Bubble jet, y = 13 cm case

Extended Self Similarity (ESS)

Correlation and scales

CHAPTER 7. SAR AND EXPERIMENTAL IMAGES FRACTAL RESULTS

River jets SAR images

Sea vortex analysis

Experimental jet results

Fractal dimension analysis conclusions

translational invariance. Obviously in non-homogeneity these properties depend on the

The study of turbulence clearly is an interdisciplinary activity. It is important to know

Irregularity (or randomness): the turbulent flow is unpredictable.

Mentioned before high Reynolds number

Three-dimensional vorticity fluctuations: turbulence is rotational and threedimensional.

Dissipation: turbulent flows are always dissipative.

Continuum: turbulence is a continuum phenomenon, governed by equations of fluid

Diffusivity is the diffusion coefficient. It is proportionality constant between the molar