Vous êtes sur la page 1sur 228

Turbulence

Turbulence: Physics and Modelling


Bharathram Ganapathisubramani Richard Sandberg
Room 5059, Bldg 13 Room 5007, Bldg 13
Ph: 22305 Ph: 27386
Email: g.bharath@soton.ac.uk Email: r.d.sandberg@soton.ac.uk

Lectures:

Tuesdays at 2PM: Bldg 13, Room 3019


Wednesdays at 10AM: Bldg 7, Room 3021
Fridays at 10AM: Bldg 29, Room 2075
courtesy: S. Lardeau
Turbulence
Focus: Physics of incompressible turbulent flows and
strategies available to address turbulence closure problem
Key outcome:
-Insight in to the nature of turbulence
-Knowledge of tools and concepts used for its analysis
Useful texts:
- Turbulence for the 21st Century, W. K. George
(http://turbulence-online.com)
- Pope, Cambridge University Press
- Tennekes and Lumley, MIT press
- Davidson, Oxford University Press
- D C Wilcox: Turbulence Modeling for CFD, 2nd ed., DCW Industries,
1998
- R. Pletcher, J. Tannehill, D.Anderson, Computational fluid
mechanics and heat transfer, CRC Press, 2012
courtesy: S. Lardeau
House-keeping
Turbulence
70% from exam
15% coursework on processing turbulence data
15% coursework on numerical modelling project

Three lectures a week throughout (36 lectures in all)

First 12 lectures by me on turbulence tools and theory


Next 16 lectures from Richard on numerical modelling
Final 6 lectures from me on miscellaneous/recent advances

courtesy: S. Lardeau
Turbulence
The nature of turbulence
Its origin - stability
Mean and fluctuation quantities
General features and major effects

Tools for studying turbulence


Definitions: stationarity and ergodicity
Statistical tools:
Amplitude statistics: probability density function and moments
Time-domain statistics: correlation and spectral functions.

Equations and scales of motion


Reynolds averaging; momentum and Reynolds stress equations
Turbulence energy equation
Transport, Production and Dissipation of kinetic energy including
Kolmogorov’s hypothesis

courtesy: S. Lardeau
Turbulence
Computational strategies: hierarchy of turbulence closures
Engineering/CFD models: RANS schemes
'Numerical experiments': DNS and LES
Hybrid RANS/LES: DES
Survey of relevance and limitations

Reynolds-Averaged Navier-Stokes (RANS) models


Motivation, philosophy, and classification
Modelling canonical flows: Homogeneous isotropic turbulence, 2D wall layers
Industrial models:
Complex geometry, pressure gradients/separation, surface roughness,
mean three-dimensionality, compressibility
Eddy-viscosity/Boussinesq closures:
Algebraic/zero-, one- and two-equation models
Example: anatomy of the Spalart-Allmaras one-equation scheme
Limitations of eddy-viscosity closures
Guidelines for CFD users

courtesy: S. Lardeau
Turbulence
Direct Numerical Simulations (DNS)
Numerical issues:
Domain size, resolution requirements
Spectral and high-order finite-difference methods
FFTs, aliasing, modified wavenumber analysis
Examples of classical and recent simulations:
plane-channel, 2D and 3D turbulent boundary layers
separation bubbles, shock-boundary layer interaction
Recent developments:
Inflow treatments, and immersed-boundary techniques

Large Eddy Simulations (LES)


Implicit and explicit filtering; subgrid-scale (SGS) modelling
Recent developments: dynamic modelling, approximate deconvolution
Inherent limitations of LES

New frontiers
Unsteady RANS
Hybrid RANS/LES methods

courtesy: S. Lardeau
Turbulence
Equations and scales of motion
Reynolds averaging; momentum and Reynolds stress equations
Turbulence energy equation
Transport, Production and Dissipation of kinetic energy including
Kolmogorov’s hypothesis
Vorticity and enstrophy; vortex stretching
Invariants of velocity gradient tensor

Canonical turbulent flows


Free flows:
Homogeneous isotropic turbulence and homogeneous shear flow
Self-preserving jets, wakes and mixing layers
Wall flows:
2D channel flow
smooth- and rough-wall boundary layers

courtesy: S. Lardeau
Turbulence
Mean-flow similarity
Self-preserving jets, wakes and mixing layers

Fine-scale features
Vorticity and enstrophy; vortex stretching
Invariants of velocity gradient tensor

Revisions (2 lectures)

courtesy: S. Lardeau
Turbulence
Turbulence
“Violent or unsteady movement of air,
“Violent or unsteady movement of air,
water or any other fluid”
water or any other fluid”
-Oxford Dictionary
-Oxford Dictionary

Hokusai ca 1830

The Great Wave off Kanagawa The Dragon of Smoke Escaping from Mt Fuji.
Examplesof
Examples of Turbulence
Turbulence
••Shear-generated
Shear-generatedturbulent
turbulentflow
flow
Fast stream adjacent to a slow stream

Courtesy: S. Lardeau
Examples Turbulence
Examples of Turbulence
•Wall-bounded turbulent flow
The presence
The presence of
of solid
solid wall
wall creates
creates shear
shear
Examples of Turbulence
Examples of Turbulence
• Wake-generated turbulent flow
•Wake-generated turbulent flow

Selkirk Island Guadalupe Island


Selkirk Island Guadalupe Island

Courtesy: G. Fishpool & M. Leschziner


Courtesy: G. Fishpool & M. Leschziner
Examples of Turbulence
•Buoyancy-generated turbulent flow
Examples of Turbulence
Examples of Turbulence
••Atmospheric
Atmospheric turbulence
turbulence

A combination of shear, wall-bounded, buoyant


A combination of shear, wall-bounded, buoyant
and wake turbulence
and wake turbulence
Examples
ExamplesofofTurbulence
Turbulence
•Galactic,
•Galactic,Planetary
PlanetaryororStellar
Stellarturbulence
turbulence
What is Turbulence?

Early View

“Turbulenza” - Leonardo DaVinci

“Observe the motion of the surface of the water,


which resembles that of hair, which has two
motions, of which one is caused by the weight
of the hair, the other by the direction of the
curls; thus the water has eddying motion, one
part of which is due to the principal current,
the other to random and reverse motion”
translated by U. Piomelli

What makes the flow turbulent?


What
Whatmakes
makesthe
theflow
flow turbulent?
turbulent?
Laminar
Laminar Transitional
Transitional Turbulent
Turbulent

Whatdetermines
What determinesthe
thestate
stateofofthe
theflow?
flow?

Forcesacting
Forces actingononthe
thefluid
fluidparticles
particles

•InertialForces
•Inertial Forces •ViscousForces
•Viscous Forces

InertialForces
Inertial Forces UL
ReynoldsNumber
Reynolds Number(Re)
(Re)== ==
ViscousForces
Viscous Forces µ

AsAsReynolds
Reynoldsnumber
number , , the
theflow
flowbecomes
becomesturbulent
turbulent
What makes the flow turbulent?

Laminar Instabilities Turbulent

Low Re Moderate Re High Re

First carried out by O. Reynolds circa 1895

Pipe Flow
Source: Google videos
Properties Turbulence
Properties of Turbulence
“Chaotic” and unsteady with wide range of interacting eddies
However, there is some degree of organisation
“Big whorls have little whorls that feed on their velocity, and little
whorls have lesser whorls and so on to viscosity”
-L.F. Richardson,
-L.F. Richardson, 1922
1922

Courtesy: S.
Courtesy: S. Lardeau
Lardeau
Properties of Turbulence
“Chaotic” and unsteady with wide range of interacting eddies
However, there is some degree of organisation
“Big whorls have little whorls that feed on their velocity, and little
whorls have lesser whorls and so on to viscosity”
-L.F. Richardson, 1922

Reynolds number also indicates


ratio of largest to smallest scales
Higher Re, greater the range of scales

Most processes are highly non-linear and


non-local with multiscale interactions
Turbulence
“Last unsolved problem in classical physics”
A wide variety of engineering applications
• Drag, energy loss fuel consumption, range, economy
• Maximum lift, stall landing, take-off speed, weight
• Aircraft, cars, trucks, trains, ships
• Noise – from aircraft, road vehicle and engine
• Combustion, engine emissions
• Vibrations, buffet, flutter
• Mixing, dispersion
• Erosion
• Heat transfer, heating, cooling
Understanding, Predicting and Controlling
Turbulence is of enormous importance
Turbulence: EngineeringApplications
Turbulence: Engineering Applications
HigherSkin
Higher Turbulence:
SkinFriction
FrictionDrag Engineering
Dragininmost Applications
mosttransport
transport systems
systems
Much higher drag
Higher Skin Friction
Laminar
Drag in most transport systems
Turbulent
U
Much higher drag
Laminar Turbulent
y U

Aircrafts50%
Aircrafts 50% Ships
Ships70%
70% Gas
GasPipelines
Pipelines90%
90%
Aircrafts 50% Ships 70% Gas Pipelines 90%

Source:
Source: Fishpool
Fishpool &&
Source: Leschziner
Leschziner
Fishpool & Leschziner

Wednesday, 23 October 13
Turbulence: Engineering Applications
Separation leads to higher form drag
Turbulence: Engineering Applications
Turbulence and Acoustics

ONERA

ONERA

Ali Mani, Meng Wang and Parviz Moin, CTR, Stanford University

ONERA
Turbulence: Engineering Applications
Turbulence and Mixing

ONERA

ONERA

Kewcharoenwong, Rossi & Vassilicos


Turbulence: lecture
Turbulence: lecture
In this
In this module,
module,we
we limit
limit ourselves
ourselvestotothe
thefollowing
followingconditions
conditions
Weconsider
We consideronly
onlycontinuum
continuumflowsflowsof
ofan
an
incompressible, single-phase, Newtonian
incompressible, single-phase, Newtonian fluid fluid

Continuum: Model fluid from a macroscopic view rather


Continuum: Model fluid from a macroscopic view rather
than the
than the microscopic
microscopic viewpoint
viewpoint

We assume that the fluid flow scales are much larger than
We assume that the fluid flow scales are much larger than
the mean free path of the molecular motions
the mean free path of the molecular motions
Characterised by Knudsen number,

Kn = <<1
L
Wednesday, 23 October 13
Turbulence: lecture

In this module, we limit ourselves to the following conditions


We consider only continuum flows of an
incompressible, single-phase, Newtonian fluid

Incompressible: If entropy is constant along streamline then:


dρ/ρ=-M2du/u (from dp +ρudu =0 and ∂ρ/∂p|s = a2) so that
fractional changes in density are small compared to fractional
changes in velocity, provided M2«1

Relative changes in density is small - Incompressible


Turbulence: lecture

In this module, we limit ourselves to the following conditions


We consider only continuum flows of an
incompressible, single-phase, Newtonian fluid

Two-phase flows are what they seem – mixtures of two


phases (steam, oil & water, &c). Many engineering and
environmental flows are genuinely two-phase, but the
complications are considerable and we avoid them here.

Turbulence in mono-phase flows will be complex enough!


Turbulence: lecture

In this module, we limit ourselves to the following conditions


We consider only continuum flows of an
incompressible, single-phase, Newtonian fluid

Newtonian fluids are those with linear relation


between stress τij and rate of strain Sij, with

Viscous stress Rate of strain


Note: We will try to use tensor notation (also known as
Einstein summation convention most times)
Turbulence: lecture
Turbulence: lecture

source:
source:Turbulence
Turbulence for
for the
the 21st
21st century,
century,W.
W.K.
K.George
George
Thus for incompressible flows, the mass
Turbulence: lecture
Turbulence: lecture conservation

Dρ̃ ∂ ρ̃ ∂ ρ̃
= + ũj =0 (3.3)
Dt ∂t ∂xj
ws that for incompressible flows,
∂ ũj Conservation of mass for
=0 (3.4)
∂xj incompressible flows
In tensor
he stress minusnotation,
the meanyou sum over
normal repeated
stress) indicies,
are represented
(v)
its definition, T̃kk = 0. In many flows of interest, the
@u1 @u2 @u3
nian fluid in which
+ the
+ viscous
= 0 stress can be related to
@x @x @x
titutive relation of the form
1 2 3

! "
(v) @u @v 1 @w
T̃ij = 2µ +s̃ij − +s̃kk δij= 0 (3.5)
@x @y 3 @z
source: Turbulence for the 21st century, W. K. George
rty of the fluid that can be measured in an independent
Turbulence: lecture
Requirements for turbulence
P

As P increases, the deflection in the beam with increase

What happens when P is excessive?


The beam breaks!
Cannot predict this: Because we use linearised equations
Same thing happens repeatedly in fluids
source: Turbulence for the 21st century, W. K. George
frequency (smaller scale) components and thus instabilities ⇒ turbulence. We shall have
Turbulence: lecture
Turbulence: lecture
much more to say on this topic in what follows.

Requirements
Requirements for
for turbulence
turbulence
Example (to show effect of nonlinear terms):
Consider this simple nonlinear equation (1D Burgers equation):
∂u ∂u
+u = 0,
∂t ∂x
and suppose that at t=to, u(x,to)=Acos(kx). For small t-to the solution can be written

∂u
u ( x, t ) = u ( x, t o ) + (t − t o ) + ...
∂t t
o

' ∂u $
= u ( x, t o ) − (t − t o ) %u " + . . .
& ∂x # to
= A cos( kx ) + (t − t o ) A2 k sin(kx ) cos( kx ) + . . .

Notice the smaller spatial scales (i.e. higher wavenumbers) introduced by the nonlinear term.
(Recall sin(kx)cos(kx) = (1/2)sin(2kx).)

Exercise: Repeat this example for a 2D case (i.e. including v∂u/∂y in the equation plus extra
Turbulence:
hout this text, unless explicity lecture
stated otherwise, the density,
Turbulence: lecture
y µ will be assumed constant. With these assumptions, the
Requirements for turbulence
entum equations for aturbulence
Requirements for Newtonian fluid reduce to:
# $
2
∂ ũi ∂ ũi 1 ∂ p̃ ∂ ũi
+ ũj =− +ν 2
∂t ∂xj ρ̃ ∂xi ∂xj
inematic viscosity,this
Non-dimensionalise ν, equation,
has been defined
using as:
a representative U , ⇢, L and P
" #
@ui⇤
@u ⇤ µ@p ⇤
⌫ @ 2 ⇤
ui
⇤ iν ≡
+ uj ⇤ = ρ ⇤ +
@t ⇤ @xj @xi U L @x⇤i 2
ince the density is assumed constant, the
* quantities are non-dimensional tilde is no longer n
mes it will be more
" instructive# and convenient to not explicit
⇤ ⇤ ⇤ 2 ⇤
@u
bility in the stress term,
i ⇤ @u
but
i to @p
refer to1 @
the uincompressible
i m
+ u j = +
@t⇤ @x⇤j @x⇤i Re @x⇤i 2
the following form:
High Reynolds number prevents viscous damping of instabilities
# $ (v)
Wednesday, 23 October 13 ˜
Turbulence: lecture
General features of turbulence

•Chaotic - seemingly random but statistics are reproducible


•Wide range of spatial and time scales. The range increases
with increasing Reynolds numbers.

•It is still a continuum phenomena - The smallest scale of


motion is still much larger than the mean-free path of
molecular motions.

•Nonlinearity ensures that long term evolution is sensitive


to very small changes.
Turbulence:
Turbulence: lecture
lecture
ains a wide range of spatial and time scales – smaller ones ‘living insi
General
And, as wefeatures
General features of
of turbulence
will see, thisturbulence
range increases with Re. Hence it is extrem
ult to•Itcalculate by direct
is rotational application
-i.e. contains of (1.1). Quantify later.
vorticity
otational – i.e. it contains !
vorticity.
=r⇥ u that ω = ∇×U and that taki
Recall
ative ‘amplifies’ the small scale variations. Note that the ‘vorticity tr
We
tion can take
(found the curl
by taking of the
the curl NSN-S
of the equations
eq.) is: and arrive at the
vorticity transport equation

∂ωi ∂ωi ∂ui ∂ ωi2


+ uj = ωj +ν
∂t ∂x j ∂x j ∂x j ∂x j

[1] [2] [3]

Wednesday, 23 October 13
Turbulence:
ion (found by taking theTurbulence: lecture
curl of the N-S eq.) is:
lecture
lecture
ains a wide range of spatial and time scales – smaller ones ‘living insi
General
And, as wefeatures
General features of
of turbulence
will see, thisturbulence
range increases with Re. Hence it is extrem
by ∂ω ∂ωi vorticity
∂uof ∂ ωQuantify
2
ult to•Itcalculate direct
is rotational -i.e. application
contains i (1.1). later.
i
+ uj = ωj +ν i
∂t ∂x ∂x ∂ x ∂x
otational – i.e. it contains vorticity. Recall that jω = ∇×U and that taki
! =r⇥u
j j j

ative ‘amplifies’ the small scale variations. Note that the ‘vorticity tr
We can take the curl [1]the NS[2]equations
of [3] and arrive at the
tion (found by taking the curl of the N-S eq.) is:
[1] - convectionvorticity transport equation
of vorticity
[3] - diffusion of vorticity
e [1] is convection of vorticity, [2] is stretching or tilting of vorticity (ofte
∂ωi ∂ωi ∂ui 2
∂ ωi
[2] -stretching’)
vortex and+ u[3]
Tilting or stretching = ωvorticity
j is of
viscous +ν
diffusion. (We have ignored th
∂t ∂x j ∂x j ∂x j ∂x j
j
Also known as vortex stretching. This happens when
ation term that exists in stratified flow.) Velocity derivatives are do
the rate
of strain has a component along vorticity.
mallest
Thisscales (more
is crucial aslater).
this is[1]
responsible
[2] for [3] production of
enstrophy and hence sustaining turbulence (more later)
Wednesday, 23 October 13
Turbulence: lecture
General features of turbulence

•It dissipates energy - via the cascade process, which


transfers energy from the large- to the small-scales (more
later)

•It is intermittent.
•It is intrinsically 3D. This becomes obvious when we
examine the vorticity transport equation. For 2D flows,
there is no vortex stretching. Therefore, no cascade and
hence no production of enstrophy (exercise - show to self)
Turbulence: lecture
20
Turbulence: lecture
CHAPTER 2. THE ELEMENTS OF STATISTICAL ANALYSIS
Elements of statistical analysis
Elements of statistical analysis
2.2 The Ensemble and Ensemble Averages
Ensemble average
Ensemble
2.2.1 average
The mean (or ensemble) average
The concept of an ensemble average is based upon the existence of independent
statistical events. For example, consider a number of individuals who are simul-
taneously flipping unbiased coins. If a value of one is assigned to a head and the
value of zero to a tail, then the arithmetic average of the numbers generated is
defined as:
1
XN = Σxn (2.1)
N
where our nth flip is denoted as xn and N is the total number of flips.
Now if all the coins are the same, it doesn’t really matter whether we flip
one coin N times, or N coins a single time. The key is that they must all be
Theevents
independent key — is meaning
the events must beof independent
the probability achieving a head or tail in a
given flipThe
mustkey is the events
be completely must
independent be independent
of what happens in all the other flips.
Obviously we can’t just flip one coin once and count it N times; these clearly
would not be independent events.
source: Turbulence for the 21st century, W. K. George
Exercise Carry out an experiment where
Wednesday, 23 October 13 yousource:
flip Turbulence
a coin 100 for thetimes in groups
21st century, of
W. K. George
Exercise Carry out an experiment where you flip a coin 100 times in groups of
Exercise Carry out an experiment where you flip a coin 100 times in groups of
10 flips each. Compare the values Turbulence: lecture
Turbulence:
you get for Xlecture
10 for each of the 10 groups, and
10 flips each. Compare the values you get for X10 for each of the 10 groups, and
note how they differ from the value of X
note how they differ from the value of X .100 .
100
Elements
Unless ofa statistical
very unusual analysis
Unless you had a very unusual experimental
you had experimental result,
result, you
you probably
probably noticed
noticed that
that
the
the value
value of
of the
the X
X1010’s’s was
was alsoalso aa random
random variable
variable and
and differed
differed from
from ensemble
ensemble to to
Ensemble
Ensemble
ensemble.
ensemble. Also
Also theaverage
average
the greater
greater the
the number
number of
of flips
flips in
in the
the ensemble,
ensemble, the
the closer
closer you
you got
got
to
to X
XNN ==1/2.
1/2. Obviously
Obviously the the bigger
bigger N
N,, the
the less
less fluctuation
fluctuation there
there isis in
in XXN ..
N
Now imagine that we are trying to establish the nature of
Now imagine that we are trying to establish the nature of a random variable,a random variable,
x.
x. The
Thenth
nthrealization
realization of of xx isis denoted
denoted as
as xxnn.. The
The ensemble
ensemble average
average of of xx is
is denoted
denoted
as
as X (or ⟨x⟩),
X (or ⟨x⟩), and
and isis defined
defined as as
N
11 !
N
!
X
X== ⟨x⟩
⟨x⟩ ≡
≡ Nlim
lim xxn (2.2)
(2.2)
N →∞ N
→∞ n
N n=1 n=1

Obviously
Obviously itit isis impossible
impossible toto obtain
obtain the
the ensemble
ensemble average
average experimentally,
experimentally, since
since
we
we can
can never
never have have anan infinite
infinite number
number ofof independent
independent realizations.
realizations. The
The most
most wewe
can
can ever
ever obtain
obtain isis the
the arithmetic
arithmetic meanmean for
for the
the number
number of of realizations
realizations we
we have.
have.
For
For this
this reason
reason the the arithmetic
arithmetic mean
mean can
can also
also referred
referred toto as
as the
the estimator
estimator for
for the
the
true
true mean
mean or or ensemble
ensemble average.
average.
Even
Eventhough
thoughthe thetrue
true mean
mean (or(or ensemble
ensemble average)
average) is
is unobtainable,
unobtainable, nonetheless,
nonetheless,
the
the idea
idea isis still
still very
very useful.
useful. Most
Most importantly,
importantly, wewe can
can almost
almost always
always be
be sure
sure the
the
ensemble
ensemble average
average exists,
exists, even
even ifif we
we can
can only
only estimate
estimate what
what itit really
really is.
is. Note
Note that
that
source:Turbulence
source: Turbulence for
for the
the 21st
21st century,
century,W.
W.K.
K.George
George
Turbulence: lecture
Elements of statistical analysis

Ensemble average

Unless stated otherwise, all analyses will utilise the


concept of ensemble average.

This means that we need to be aware of or take in to
account the “statistical differences” between the true
mean and the estimates of mean

source: Turbulence for the 21st century, W. K. George


Turbulence:
2.2. THE ENSEMBLE AND lecture
ENSEMBLE AVERAGES 21
Turbulence: lecture
Elementsunless
Nonetheless, of statistical
stated otherwise,analysis
all of the theoretical deductions in this book
Elements
will of statistical
use this ensemble average; and analysis
therefore are completely general. Obviously
this will mean we have to account for these “statistical differences” between true
Ensemble average
Ensemble
means and average
estimates of means when comparing our theoretical results to actual
measurements or computations.
In general, the xn could be realizations of any random variable. The X defined
by equation 2.2 represents the ensemble average of it. The quantity X is sometimes
referred to as the expected value of the random variable x, or even simply its mean.
For example, the velocity vector at a given point in space and time, ⃗x, t, in a
given turbulent flow can be considered to be a random variable, say ui (⃗x, t). If
(n)
there were a large number of identical experiments so that the ui (⃗x, t) in each of
(n)
them were identically distributed, then the ensemble average of ui (⃗x, t) would
be given by
N
1 ! (n)
⟨ui (⃗x, t)⟩ = Ui (⃗x, t) ≡ lim ui (⃗x, t) (2.3)
N →∞ N
n=1

Note that this ensemble average, Ui (⃗x, t), will, in general, vary with the inde-
pendent variables ⃗x and t. It will be seen later that under certain conditions
the ensemble average is the same as the average which would be generated by
source: Turbulence for the 21st century, W. K. George
averaging in time, or even space. But even when a time (or space) average is
particular
velopmentnature of the flow,equations
of the turbulence or even itsinstatistical
this book character
unless oth
Elements of Turbulence:
thestatistical
equations
the equationsanalysis
derived will be
derived lecture
completely
Turbulence:
will general,
lecture
be completely and and
general, quitequite
ind
Elements ofparticular
2.2.2 statistical
particular nature
Fluctuations
nature analysis
of the
of flow, Turbulence:
or even
theabout
flow, or the
even lecture
its statistical
mean
its character.
statistical character
2.2.2 Fluctuations about the mean
Elements of
Fluctuations
Elements of statistical
statistical
about analysis
the
analysis mean
Elements
Fluctuations
CHAPTER 2. THE ELEMENTS
2.2.2 of
It is often statistical
OF STATISTICAL
It is often
CHAPTER 2. THE ELEMENTS
analysis
importantANALYSIS
about
to know how a random variable is distrib
the
importantANALYSIS
Fluctuations
OF STATISTICAL mean
to know how a random variable is di
about the themean
mean. For example,
2.2.2 Fluctuations Figure 2.1 about
illustrates portions
meanof two rand
Fluctuations
Fluctuations timeabout
mean.
about
which the
For
have mean
example,
the
identical Figure but
Possible
mean
means, 2.1 are
to illustrates
distinguish
obviouslyportions
between
members of of
two diffr
Fluctuations
CHAPTER 2.
CHAPTER
THE It
2. THE
is
ELEMENTS often
time
It OF
ELEMENTSis about
important
STATISTICAL
which
often
OF have the
since the amplitudes of Possible
to know
ANALYSIS
identical
important
STATISTICAL to meanhow
means,
know
ANALYSIS to
a
but
how
their fluctuations distinguish
random
are
a variable
obviously
random betwee
is distrib
members
variable
are not distributed is do
mean. sinceForthe
mean. example,
For these
Figure
amplitudes
example, of two
2.1
their
Figure signals
illustrates
fluctuations
2.1 by
illustrates looking
portions
are of two
not
portions at
rand
distribut
of two
possible to distinguish
timepossible
whichwhichhave these
Possible
between
Possible
identical means, to
totwo
them
but signals
distinguish
by examining
distinguish
are obviously by looking
between
the
between
members
statistic
ofstat
diffo
the fluctuations about thePossible
time to distinguish
have identicalbetween
means,
mean (or simplyto
them
but distinguish
the
by
are examining
obviously betwee
the fluctuations the
members
fluctuations) defi
sincethe thefluctuations
since amplitudes these
about
the amplitudes thetwo
of their
these of two
mean
their signals
fluctuations
signals
(or simply
fluctuations the
by
are
by
not fluctuation
looking
distributed
looking
the
are at
at
fluctuations)
not distribu t
possible to distinguish these
between two
them bysignals
examining by looking
the statistic
possible to distinguish between
the fluctuations
the fluctuations aboutabout
the mean
x′ them
= x′ −
(or simply
the mean (or =
the
by
the
the
x simply
x
fluctuations
X examining
fluctuations
−the
fluctuations)
X
the sta
the fluctuatio
fluctuations)defi
It is easy to see that the average of the
′ fluctuation is zero, i.e.,
x =
It is easy to see that the average ′−X
ofxxthe= fluctuation
x−X is zero, i.
⟨x′ ⟩ = ′0
It is It
easy to seetothat
is easy the average
see that the average of the of fluctuation
= 0 is zero,
⟨x ⟩ fluctuation
the i.e., i.
is zero,
On the other hand, Thethesignals ensemble can averagebe′ distinguished
of the square of th
On the other
not zero. In fact, it is such by
The
hand, thesignals
ensemble⟨x ⟩can

=⟨x be
0⟩ =
average 0 distinguishe
of measure
the square
calculating
an important the
statistical variance we
name,notthe zero.non-zero
withvariance, The
In fact,
The
it
and issignals
such by
signals
represent an canimportant
can
it be
calculating
be
symbolically distinguished
the
statistical
distinguished
by varianc
measure
either var[x
1: A typical random function of time
On
name,ofthe
On other
the
the
mean
hand,hand,
other
withvariance, The
value.
theand thesignals
ensemble
ensemble
represent itcan
average be
of the
average
symbolically distinguishe
of square
the
by of th
square
either v
ure 2.1: A typical random function time non-zero
not zero. In fact, it is by
mean
by
value.
calculating
source: Turbulence
suchcalculating
for the
the 21st
the variance
century, W.
variance
K. George
Wednesday, 23 October 13
efined as:
not
13 name,
zero. In fact, it is such an by
an calculating
important
important
source: statistical
Turbulence the
statistical
for the 21st varianc
measuremeasure
century, W. we
K. Ge
1: A 2.1:
gure typical random
Wednesday, function
23 October
A typical random the
name,
of time
function ofwithvariance,
timethe
non-zero meanand
withvariance,
non-zero represent
and
value.
mean represent
value. it symbolically
it symbolically by either
by var[xv
either
ce is defined as: source: Turbulence for the 21st century, W. K. George
′ 2 2
Turbulence: lecture
Turbulence: lecture
Elements of statistical analysis
Elements of statistical analysis
Figure 2.1: A typical random function of time with non-zero mean value.
Fluctuations about
Fluctuations about the mean the mean
variance is defined as:

var[x] ≡ ⟨(x′ )2 ⟩ = ⟨[x − X]2 ⟩ (2.6)


N
1 !
= lim [xn − X]2 (2.7)
N →∞ N
n=1

Note that the variance, like the ensemble average itself, can never really be mea-
sured, since it would require an infinite number of members of the ensemble.
It is straightforward to show from equation 2.2 that the variance in equation 2.6
can be written as:
var[x] = ⟨x2 ⟩ − X 2 (2.8)
Thus the variance is the second-moment minus the square of the first-moment (or
mean). In this naming convention, the ensemble mean is the first moment.
Exercise Use the definitions of equations 2.2 and 2.7
source: to derive
Turbulence for theequation
21st century,2.8.
W. K. George
Wednesday, 23 October 13
source: Turbulence for the 21st century, W. K. George
sured,
sured, Itsince
It isstraightforward
straightforward
issince itit would
would require toshow
require
to show from equation
an infinite
an infinite
from equation
number
number 2.2
2.2 that
ofthat
of members
members the variance
the variance
of the
of in equation
equation 2.6
the ensemble.
ensemble.
in 2.6
can
canIt be
Itbe
is written as:
isstraightforward
straightforward
written as: to show
to showTurbulence:
Turbulence: lecture
Turbulence: lecture
from equation
from equation 2.2 lecture
2.2 that the
that the variance
variance in in equation
equation 2.6 2.6
2 2
can be
can be written
written as: as: var[x] =
var[x] = ⟨x ⟩ − X⟨x 2 ⟩ − X 2 (2.8)
(2.8)
Elements
Elements
Elements
Thus the variance of
of
of statistical
statistical
statistical
is the analysis
analysis
analysis
var[x]
var[x]
second-moment == ⟨x
⟨x 22 ⟩ − X22

minus
Thus the variance is the second-moment minus the square of the first-moment (or − X the square of the first-moment (2.8)
(2.8)(or
mean).
Thus
mean).
Thus theIn
the In this naming
variance
this
variance naming
isis the convention,
theconvention,
second-moment
second-moment the ensemble
the ensemble
minus the
minus mean
themean
square
square isis of
the
of
the the
the first
first moment. (or
first-moment
moment.
first-moment (or
Fluctuations
Fluctuations
Fluctuations
mean).
mean).
Exercise In
In this
this
Use naming
naming
the
about about
about the
convention,
convention,
definitions of
mean
the
thethe
equations mean
mean
the ensemble
ensemble
2.2 mean
mean
and 2.7 is the
is
to the
derivefirstequation
first moment.2.8.
moment.
Exercise Use the definitions of equations 2.2 and 2.7 to derive equation 2.8.
Exercise
Exercise Use
Use
The variance the
the
variance can definitions
definitions
can also of
of
also referred equations
equations
referred to to as 2.2
2.2
as the and
and
the second 2.7
2.7 to
to
second central derive
derive
central moment equation
equation
moment of 2.8.
2.8.
of x.
x. The
The
The
word
word
The central
Thecentralvariance
variance implies
implies can also
can that referred
also
that the mean
the mean
referred to has
tohasas been
as been
the subtracted
the second
second
subtracted central
central offmoment
off before squaring
moment
before squaring
of x.
of x. Theand
The
and
averaging.
word centralThe
averaging.
word central Theimplies
implies reasons
reasons thatfor
that for
the this
thethismean
mean willhas
will be been
has
be clear subtracted
been
clear below. IfIf two
subtracted
below. two random
offrandom
off variablesand
before squaring
before squaring
variables are
and
are
identically
Figure
averaging.
identically
Figure
averaging. 2.2: The
2.2: distributed,
Two
The randomfor
reasons
distributed,
Two random
reasons then
for
then they
functions
this
they
functions
this will
will must
ofbe
must
of betime
time have
clear
have
clearhavingthe same
below.
the
having
below. same
the mean
theIfIfsame
same
two random
mean
two and variance.
mean
random
and
mean variance.
and
and variance,
variables
variance,
variables are
are
but
but very
identically
very
The
identically different
Thedifferent
variance
distributed,
variance
distributed, higher
isis closely
higher closely moments.
then related
moments.
then they must
related
they to another
must
to another
have the
have the statistical
same mean
statistical
same mean quantity called the
and variance.
quantity
and variance.
called the stan-
stan-
dard
dardThe
The deviation
varianceor
deviation
variance or root
isisroot
closely
closely mean
mean square
related
square
related to (rms)
to(rms)
another
another value
value of the
the random
statistical
of
statistical random
quantity
quantity variable x,which
calledx,which
variable
called the stan-
the stan- isis
dard deviation
dard deviation or or root
root mean mean squaresquare (rms)(rms) value
value of of the
the random
random variable variable x,which
x,which isis
denoted by
denoted by the
the symbol,
symbol, σσxx.. Thus, Thus,
1/2
σ ≡ (var[x])
σxx ≡ (var[x]) 1/2 (2.9)
(2.9)

or σ22 = var[x].
or σxx = var[x].

2.2.3 Higher
2.2.3 Higher moments
moments
source:
source:Turbulence
Turbulencefor
forthe
the21st
21stcentury,
century,W.
W.K.
K.George
George
Figure 2.2: Two random functions of time having the same mean and variance,
Turbulence: lecture
2.2. THE ENSEMBLE AND ENSEMBLE AVERAGES
but very different higher moments.
2

denoted by the symbol, σx . Thus, Turbulence: lecture


Elements of statistical analysis
Turbulence: lecture
Elements ofsymbol,
denoted by the statistical σxanalysis
σ . Thus,
x ≡ (var[x])1/2 (2.9)
2 Elements
Higher moments of statisticalσ analysis 1/2
≡ (var[x])
x (2.9)
or σx = var[x].
Higher moments
or Two
EMBLE AND
signals
σ ENSEMBLE
2
= var[x]. with the same mean and variance
AVERAGES 23
Higher
x
moments
2.2.3 Two
E ENSEMBLE AND signals
Higher
ENSEMBLE with the same mean
moments
AVERAGES 23 and variance
2.2.3 Two signals
Higher with the
moments same mean and variance
Figure 2.2 illustrates two random variables of time which have the same mean
and Figure
also the2.2same
illustrates two random
variances, variables
but clearly theyof are
timestill
which have
quite the sameItmean
different. is useful,
and also
therefore, tothe samehigher
define variances, but clearly
moments they
of the are still quite
distribution todifferent.
assist inItdistinguishing
is useful,
therefore, to define higher moments of the distribution to assist in distinguishing
these differences.
these differences.
TheThe
m-th moment
m-th momentofofthe
therandom variableisisdefined
random variable defined as:as:
Figure 2.2: Two random!Nfunctions
random functions of time having the same mean and
Figure
m
2.2: Two 11 N
! m
of time having the same mean and variance
⟨x
but⟨x
but
m
⟩⟩ =
very
very = lim higher
lim
different
different nxn
m
xmoments.
highermoments. (2.10)(2.10)
→∞N
N→∞ N n=1
n=1

It isItusually
is usually more
more convenient
convenient tobywork
to
denoted work
the with
with
symbol, the
σthe. central moments
central
Thus,
denoted by the symbol, σx . Thus,
x
defined
moments by: by:
defined
!N σ ≡ (var[x])1/2
′ m m 1 N x σ ≡ (var[x])
! m 1/2 (2.9
⟨(x
′ m) ⟩ = ⟨(x − X)m⟩ = Nlim 1 [xn − X] m
x (2.11)
⟨(x ) ⟩ = ⟨(x
or σ−
2 X) ⟩ = lim N n=1
= var[x].
→∞ [xn − X] (2.11)
x
or σx2 = var[x].
N →∞ N n=1
source:
source:Turbulence for for
Turbulence the 21st century,
the 21st W. K. George
century, W. K. George
the random variable about the mean. It is easy to see that the variance is the
Turbulence: lecture
second central moment (i.e.,Turbulence:
m = 2). lecture
Elements
Elements of
of statistical
statistical analysis
analysis
2.3 Probability
Probability
Probability
2.3.1 The histogram and probability density function
The frequency of occurrence of a given amplitude (or value) from a finite num-
ber of realizations of a random variable can be displayed by dividing the range
of possible values of the random variables into a number of slots (or windows).
Since all possible values are covered, each realization fits into only one window.
For every realization a count is entered into the appropriate window. When all
the realizations have been considered, the number of counts in each window is
divided by the total number of realizations. The result is called the histogram
(or frequency of occurrence diagram). From the definition it follows immediately
that the sum of the values of all the windows is exactly one.
The shape of a histogram depends on the statistical distribution of the random
variable, but it also depends on the total number of realizations, N , and the size
of the slots, ∆c. The histogram can be represented symbolically by the function
Hx (c, ∆c, N ) where c ≤ x < c + ∆c, ∆c is the slot width, and N is the number
of realizations of the random variable. Thus the histogram shows the relative
frequency of occurrence of a given value range in a given ensemble. Figure 2.3
source: Turbulence for the 21st century, W. K. George
illustrates a typical histogram. If the size of the sample is increased so thatGeorge
source: Turbulence for the 21st century, W. K. the
of possible values of the random variables into a number of slots (or windows).
of possible
Since valuesvalues
all possible Turbulence:
of theare
random lecture
Turbulence:
variables
Turbulence: lecture
covered, lecture
into a number
each realization of slots
fits into only (or
one windows).
window.
Since all possible values are covered, each realization fits into
For every realization a count is entered into the appropriate window. When allonly one window.
theElements
For
Elements
realizations
Elements of statistical
every realization
ofhave
have a count
statistical
been is analysis
entered
analysis
considered, into
the the appropriate
number of counts window.
in each When isall
window
the realizations of statistical analysis
been considered, the number of counts in each window is
divided by the total number of realizations. The result is called the histogram
divided by the total number of realizations. The result is called the histogram
Probability
(or frequency of occurrence diagram). From the definition it follows immediately
Probability
Probability
(or frequency of occurrence diagram). From the definition it follows immediately
that the sum of the values of all the windows is exactly one.
that the sum of the values of all the windows is exactly one.
The shape of a histogram depends on the statistical distribution of the random
The shape of a histogram depends on the statistical distribution of the random
variable, but it also depends on the total number of realizations, N , and the size
variable, but it also depends on the total number of realizations, N , and the size
of the slots, ∆c. The histogram can be represented symbolically by the function
of the slots, ∆c. The histogram can be represented symbolically by the function
Hx (c, ∆c, N ) where c ≤ x < c + ∆c, ∆c is the slot width, and N is the number
Hx (c, ∆c, N ) where c ≤ x < c + ∆c, ∆c is 25the slot width, and N is the number
of realizations
OBABILITY of the random variable. Thus the histogram shows the relative
of realizations of the random variable. Thus
ROBABILITY 25 the histogram shows the relative
frequency of occurrence of a given value range in a given ensemble. Figure 2.3
frequency of occurrence of a given value range in a given ensemble. Figure 2.3
illustrates a typical histogram. If the size of the sample is increased so that the
illustrates a typical histogram. If the size of the sample is increased so that the
number of realizations in each window increases, the diagram will become less
number of realizations in each window increases, the diagram will become less
erratic and will be more representative of the actual probability of occurrence of
erratic and will be more representative of the actual probability of occurrence of
the amplitudes of the signal itself, as long as the window size is sufficiently small.
the amplitudes of the signal itself, as long as the window size is sufficiently small.
If the number of realizations, N , increases without bound as the window size,
If the number of realizations, N , increases without bound as the window size,
∆c, goes to zero, the histogram divided by the window size goes to a limiting
∆c, goes to zero, the histogram divided by the window size goes to a limiting
curve called the probability density function, Bx (c). That is,
curve called the probability density function, Bx (c). That is,
Bx (c) ≡ lim H(c,source:
∆c, N
source: )/∆c for
Turbulence
Turbulence forthe
the21st
21stcentury,
century,W.
W.(2.12)
K.
K.George
B (c) ≡ lim H(c,source:
∆c, N K.George
)/∆c for the 21st century, W. (2.12)
Turbulence George
Turbulence:
Turbulence: lecture lecture
of realizations of the random variable. Thus the histogram shows the relative
frequency of occurrence of a given value range in a given ensemble. Figure 2.3
illustrates a typical histogram. If the size of the sample is increased so that the
Elements
Elements
number of
of statistical
statistical
of realizations analysis
analysis
in each window increases, the diagram will become less
erratic and will be more representative of the actual probability of occurrence of
Probability
Probability
the amplitudes of the signal itself, as long as the window size is sufficiently small.

If the number of realizations, N , increases without bound as the window size,


∆c, goes to zero, the histogram divided by the window size goes to a limiting
curve called the probability density function, Bx (c). That is,

Bx (c) ≡ lim H(c, ∆c, N )/∆c (2.12)


N →∞
∆c → 0

Note that as the window width goes to zero, so does the number of realizations
which fall into it, N H. Thus it is only when this number (or relative number) is
divided by the slot width that a meaningful limit is achieved.
The probability density function (or pdf) has the following properties:

• Property 1:
Bx (c) > 0 (2.13)
always. source:
source:Turbulence
Turbulence for
for the
the 21st
21st century,
century,W.
W.K.
K.George
George
Bx (c) ≡ lim H(c, ∆c, N )/∆c (2.12)
Turbulence: lecture
Turbulence:
N →∞ lecture
∆c → 0
Elements
Elementsofofstatistical
statisticalanalysis
analysis
Note that as the window width goes to zero, so does the number of realizations
which fall into it, N H. Thus it is only when this number (or relative number) is
Probability
Probability
divided by the slot width that a meaningful limit is achieved.
The probability density function (or pdf) has the following properties:

• Property 1:
Figure 2.3: Histogram, together with Bit’s(c)limiting
> 0 probability density function.
(2.13)
x

always.
• Property 2:
P rob{c < x < c + dc} = Bx (c)dc (2.14)
where P rob{ } is read “the probability that”.

• Property 3: ! x
P rob{c < x} = Bx (c)dc (2.15)
−∞

• Property 4: ! ∞
Bx (x)dx = 1 (2.16)
−∞
source:
source:Turbulence
Turbulencefor
forthe
the21st
21stcentury,
century,W.W.K.K.George
George
−∞

Turbulence: lecture
The condition imposed by property (1) simply states that negative probabilities
Theimpossible,
are condition imposed
are impossible,
realization
while property
takes while
Turbulence: lecture
by property
property
on some value. (4)
(1) simply
(4) assures thatstates
assures(2)that
Property the
gives
that negative
the probability
theprobability
probabilities
is unity
probabilityisofunity
that a
findingthatthea
Elements
realization in aof
takes onstatistical
somearound analysis
value. aProperty (2) gives theproperty
probability
(3) ofprovides
finding the
the
Elements
realization
realization in aof statistical
interval
interval around
probability that the realization is less a analysis
certain value,
certain
than value,
while
while property
a prescribed (3) the
value. Note provides the
necessity
Probability
probability
of that the
distinguishing realization
between the is less than
running a prescribed
variable, x, and thevalue. Note the
integration necessity
variable, c,
Probability
of equations
in distinguishing
2.14 between
and 2.15.the running variable, x, and the integration variable, c,
in equations 2.14 gives
Since Bx (c)dc and 2.15.
the probability of the random variable x assuming a value
SincecBand
between x (c)dc
c +gives the moment
dc, any probability of the
of the random variable
distribution can be xcomputed
assumingby a value
inte-
betweenthe
grating c and c + dc, any
appropriate powermoment
of x of the
over all distribution
possible canThus
values. be computed
the n-th by inte-
moment
26 CHAPTER 2. THE ELEMENTS OF STATISTICAL ANALYSIS
isgrating
given the
by: appropriate power of x over ! ∞
all possible values. Thus the n-th moment
is given by: n ! ∞ cn Bx (c)dc
⟨x ⟩ =
Exercise: Show (by returning nto the definitions) that the value of the moment (2.17)
n
⟨x ⟩
determined in this manner is exactly equal= −∞ c B (c)dc
to xthe ensemble average defined earlier(2.17)
−∞
in equation 2.10. (Hint: use the definition of an integral as a limiting sum.)
If the probability density is given, the moments of all orders can be determined.
For example, the variance can be determined by:
! ∞
2
var{x} = ⟨(x − X) ⟩ = (c − X)2 Bx (c)dc (2.18)
−∞

The central moments give information about the shape of the probability den-
sity function, and vice versa. Figure 2.4 shows three distributions which have the
same mean and standard deviation, but are clearly
source:quite different.
Turbulence Beneath
for the 21st them
century, W. K. George
Turbulence: lecture
2.3. PROBABILITY

Turbulence: lecture
27

Elements
Elementsof
ofstatistical
statisticalanalysis
analysis
Probability
Probability
What are the
distinguishing moments
for these distributions?

Figure 2.4: Relationsource: Turbulence


ofsource: Turbulence
skewness for
to the shape ofthe
for the21st
the 21st
pdf century,
and nature W.
century, K.K.signal.
ofW.
the George
George
function subroutines, but usually Turbulence: lecture
with the independent
Turbulence: lecture
variable normalized by
σx ; i.e., c′ = c/σx . Alternatively we can subtract FxG (c) from unity to obtain the
probability that x ≥ c as 1 − FxG (c). This is related to the complementary error
Elements
Elements
function, erfc(c),of
ofalsostatistical
statistical analysis
analysis
usually easily available.

Skewness
Skewness and Kurtosis
and
2.3.4 Skewness Kurtosis
and kurtosis
Because of their importance in characterizing the shape of the pdf, it is useful
to define scaled (or normalized) versions of third and fourth central moments:
the skewness and kurtosis respectively. The skewness is defined as third central
moment divided by the three-halves power of the second; i.e.,
⟨(x − X)3 ⟩
S= (2.26)
⟨(x − X)2 ⟩3/2
The kurtosis is defined as the fourth central moment divided by the square of the
second; i.e.,
⟨(x − X)4 ⟩
K= 2 2
(2.27)
⟨(x − X) ⟩

source: Turbulence for the 21st century, W. K. George


example,
Turbulence:
Turbulence:
3
⟨(x − X) ⟩ =
!
lecture
(c −lecture

X) B (c)dc 3
x (2.20)
−∞

Elements
Elements ofifstatistical
of
is equal to zero statistical analysis
Bx is an evenanalysis
function.

Probability
Probability distribution
2.3.2 Thedistribution
probability distribution
Sometimes it is convenient to work with the probability distribution instead
of with the probability density function. The probability distribution is defined
as the probability that the random variable has a value less than or equal to a
given value. Thus from equation 2.15, the probability distribution is given by
! c
Fx (c) = P rob{x < c} = Bx (c′ )dc′ (2.21)
−∞

Note that we had to introduce the integration variable, c′ , since c occurred in the
limits.
Equation 2.21 can be inverted by differentiating by c to obtain

dFx
Bx (c) = (2.22)
dc

source:Turbulence
source: Turbulence for
for the
the 21st
21st century,
century,W.
W. K.
K. George
George
Turbulence:lecture
Turbulence:
Turbulence: lecture
lecture
28
28 CHAPTER 2. THE ELEMENTS OF STATISTICAL ANALYSIS
CHAPTER 2. THE ELEMENTS OF STATISTICAL ANALYSIS
Elements
Elements
Elements
28
28
of of
of statistical
statistical
statistical
CHAPTER
CHAPTER 2.
2. THE
THE
analysis
analysis
analysis
ELEMENTSOF
ELEMENTS OFSTATISTICAL
STATISTICALANALYSIS
ANALYSIS
2.3.3 Gaussian (or normal) distributions
2.3.3 Gaussian (or normal) distributions
Normal
Normal
Normal
2.3.3
Oneofofthe
2.3.3
One distribution
distribution
distribution
Gaussian
theGaussian
most
most (orpdf’s
important
(or
important normal)
pdf’s distributions
ininturbulence
normal) turbulenceisisthe
theGaussian
distributions GaussianororNormal
Normaldistri-
distri-
butiondefined
bution definedby by
Oneofofthe
One themostmostimportant
importantpdf’s pdf’sininturbulence
turbulence
1 isisthethe Gaussian
Gaussian or orNormal
Normaldistri-
distri-
B (c) = √ 1 e −(c−X)
−(c−X)
2 /2σ
2 /2σ 2x
2
(2.23)
bution defined
bution defined by by BxG (c) = √ 2πσ e
xG x (2.23)
2πσ1 1 x x −(c−X)2 2 /2σ2 x2 √
B (c)==√√
BxGxG(c) e e−(c−X) /2σx √(2.23)
(2.23)
whereXXisisthe
where themeanmeanand andσσx xisisthe thestandard
standard
2πσ derivation. The
2πσx x derivation. Thefactor
factorofof1/ 1/ 2πσ 2πσx x
insures that the integral of the pdf over all values is unity as required. √√is easy
It
insures
wherethat
where XXisisthe theintegral
the meanand
mean of σ
and the
σx xispdf
isthe
theover all values
standard
standard is unityThe
derivation.
derivation. as factor
The required.
factorofof1/ It is2πσ
1/ easy
2πσ xx
totoinsures
provethat
prove
insures that
thatthis
that
this isthe
theintegral
the
thecase
isintegral case by
ofofthe
bycompleting
the completing
pdf
pdf overall
over
thesquares
the
allvalues
values squares
isisunity
unity inin asthe
asthe
integration
integration
required.
required.
of
ItItisisof
the
the
easy
easy
exponential
exponential
totoprove
provethat (see
(see
that problem
problem
this
this isisthe
2.2).
the2.2).
caseby
case bycompleting
completingthe thesquares
squaresininthe theintegration
integrationofofthe the
TheGaussian
exponential
The
exponential Gaussian distribution
(seedistribution
(see problem2.2).
problem isisunusual
2.2). unusualininthat thatititisiscompletely
completelydetermined determinedby byits
its
firsttwo
first two
The
The moments,
moments,
Gaussian X Xand
Gaussiandistribution andσ.σ.This
distribution This isisnot
isisunusual
unusualnotin typical
in thatitof
typical
that itof ismost
ismost turbulence
completely
turbulence
completely distributions.
determined
determined byits
distributions.
by its
Distribution
Distribution
Distribution
Nonetheless,
first
Nonetheless,
first twomoments,
two moments, completely
Xcompletely
completely
ititisissometimes
sometimes
Xand andσ.σ. useful
This
useful
This determined
to
isdetermined
determined
to
is approximate
not typicalofof
approximate
not typical by byby first
most first
first
turbulence
turbulence
most two two
two
turbulence
turbulence moments
asasbeing moments
moments
being Gaussian,
distributions.
Gaussian,
distributions.
often
often becauseof
Nonetheless,
because
Nonetheless, itof
itisthe
is
the absenceof
sometimes
absence
sometimes of simple
useful
simple
useful alternatives.
totoapproximate
approximate
alternatives. turbulenceasasbeing
turbulence beingGaussian,
Gaussian,
often
often because
ItItisisbecause ofofthe
straightforward
straightforward theabsence
absence
totoshow ofofsimple
show simple
bybyintegratingalternatives.
integrating
alternatives. bybypartspartsthat thatall allthe theeven
evencentral
central
moments
moments ItItisisabove
straightforward
above thesecond
straightforward
the second totoare
show
are
show byintegrating
given
by
given integrating
bybythe thefollowingbyparts
following
by parts thatall
recursive
that
recursive allrelationship,
theeven
evencentral
relationship,
the central
momentsabove
moments abovethe thesecondsecondare aregiven
givenby
n
bythe thefollowing
followingrecursiverecursive
n
relationship,
relationship,
⟨(x−−X)
⟨(x X)n ⟩ ⟩==(n (n−−1)(n 1)(n−−3)...3.1σ
3)...3.1σxnx (2.24)
(2.24)
⟨(x − X) n n ⟩ = (n − 1)(n − 3)...3.1σ nn (2.24)
⟨(x − X) ⟩ = (n4− 1)(n − 3)...3.1σx6x (2.24)
Thusthe
Thus thefourth
fourthcentralcentralmoment momentisis3σ 3σx4 x, ,the
thesixthsixthisis15σ 15σx6 x, ,and
andsosoforth. forth.
4 6
Thusthe
Thus thefourth
fourthcentral
centralmoment momentisis3σ 3σx4 ,x ,the
thesixth
sixthisis15σ 15σx6 ,x ,and
andsosoforth. forth.
Exercise: Prove this. source:
source:Turbulence
Turbulence for
source: Turbulence for the the
for 21st
the century,
21st
21st W.W.
century,
century, K.K.
W. George
K.George
Georg
Turbulence: lecture
Turbulence: lecture
2.4 Multivariate Random Variables
Elements of statistical analysis
2.4.1 Joint pdfs and joint moments
Joint
Joint pdfs
pdfs andandjointjoint moments
moments
Often it is important to consider more than one random variable at a time. For
example, in turbulence the three components of the velocity vector are interre-
It isand
lated usually
must be important in turbulence
considered together. In additiontotoconsider
the marginaljoint
(or single
statistics
variable) statistical moments of flow
already variables
considered, it is necessary to consider the
joint statistical moments.
For example if u and v are two random variables, there are three second-order
moments which can be defined ⟨u2 ⟩, ⟨v 2 ⟩, and ⟨uv⟩. The product moment ⟨uv⟩ is
called the cross-correlation or cross-covariance. The moments ⟨u2 ⟩ and ⟨v 2 ⟩ are
referred to as the covariances, or just simply the variances. Sometimes ⟨uv⟩ is
also referred to as the correlation.
In a manner similar to that used to build-up the probability density function
The joint
from its probability
measurable density
counterpart, function
the histogram, can be
a joint constructed
probability density
using
functionthe(orjoint
jpdf),histogram following
Buv , can be built-up from atheprocedure described
joint histogram. Figure 2.5
illustrates several examples of jpdf’s which have different cross-correlations. For
before, but,
convenience now, wevariables
the fluctuating need uto ′ dov′ this
and can bewith
definedtwo
as variables

u′ = u − Usource:
source:Turbulence
Turbulence for
for the
the 21st
21st century,
century,W. (2.28)
W. K.
K. George
George
alsoIn a manner
referred to as similar to that used to build-up the prob
the correlation.
from
fromIn its its measurable
measurable
a manner similar Turbulence:
to counterpart,
Turbulence:
Turbulence: that usedlecture
counterpart, lecture
lecture
to thethe histogram,
histogram,
build-up a joi
a joint
the probability d
function
from its (or
measurable jpdf),
Turbulence: B
OF counterpart,uv , can
lecture
the be built-up
histogram, a from
joint the
probabjoi
30
30
Elements
function
Elements
CHAPTER

Elements
CHAPTER
2.

ofof of
(or statistical
jpdf),
THE ELEMENTS

statistical
statistical
2. THE ELEMENTS Buv analysis
STATISTICAL
, can be built-up from the joint
ANALYSIS

analysis
analysis
OF STATISTICAL ANALYSIS
illustrates
function
illustrates (or several
jpdf), B examples
, can be of jpdf’s
built-up fromwhich
the have
joint differe
histogra
30
Elements of several
statisticalexamples
analysis of jpdf’s which
uv
CHAPTER 2. THE ELEMENTS OF STATISTICAL ANALYSIS
have
′ different ′
different
illustrates
convenience severalthe examples
fluctuating of jpdf’s which have
variables u and v cross-co
can be d
Joint
Joint pdfs
convenience
Joint pdfs
pdfs and and and
joint
the joint
moments
fluctuating
joint moments
moments variables′ u ′ ′
′and v can be defi
convenience
Joint pdfs and the joint
fluctuating
moments variables u and v can be defined as

= ′u
u − U
u =u u =

− Uu − U

′ ′v
v =v v = = v
−Vv −V − V
u’ and
u’
u’ u’
v’
and
andand
arev’
v’ v’
random
are
areare random
random
random
where
where as as before
before capital
capital lettersletters
are usedare
to used to
represent represent
the mean the
valu
where as before capital ′
letters
′ ′
variables
variables
are about
used
variables
variables

about
the
to
about mean the
represent
about the mean
the
the mean
mean m
fluctuating
fluctuating quantities
quantities u and′ u v and
fluctuating quantities u′ ′and v
are′ random are random
variables variables
with zero me wi
0 0v are random variables with
AApositive value of ⟨u v ⟩ hu ′v i′ >0hu
indicates 0 (positive
0
v 0
that
i > 0u ′
and v
(positive ′
correlation) tend to var
′correlation′
positive
A positive value
value of of ⟨u
⟨uwhen v ⟩
v ⟩ indicates indicates
0 that
′ ′ hu v i > 0 (positive ′correlation)
that u and
u and vtheten ′ v t
negative value indicates that one variable is increasing oth
negative
negative value
value indicates
indicates that
hu v′ i <when
0 0
one
0 0(negative variable
correlation) is increa
decreasing. A zero value ofthat
⟨u v when

0hu0′ v′one
⟩huindicates 0
0variable
i < that (negative
there
v i < 0 (negative correlation)
isiscorrelation
increasin
no corre
decreasing. A zero value
decreasing. A zero value of0 of
hu v⟨u
0 ⟨u
′ ′ v ⟩ indicates that there
i =v0⟩(uncorrelated)
indicates that there is
0 0
0 0 v i = 0 (uncorrelated)
hu
hu v i = 0 (uncorrelated)
source: Turbulence for the 21st century, W. K. George
Figure 2.5: Contours
Wednesday, 23 October of
13 constant probability for four different joint probability
density functions. Try to figure out what the moments would be for each and how source: Turbulence for the 21st century, W. K. Geo
they would differ. source:
source:Turbulence
Turbulencefor
forthe
the21st
21stcentury,
century,W.
W.K.K.George
George
Figure 2.5: Contours
Wednesday, 23 October of
13 constant probability for four different joint probability
also
′ referred ′ to as the correlation.
u and v . As
In a manner similar Turbulence:
will be
Turbulence: seen
to that used lecture
below,
lecture it does not mean
to build-up the probability d
independent.
from
CHAPTERits measurable Turbulence: ANALYSIS lecture
counterpart, the histogram, a joint probab
30
Elements of
Elements
2. THE
of statistical
statistical analysis
ELEMENTS OF
analysis
STATISTICAL

It is (or
function sometimes
jpdf), Buvmore , can be convenient
built-up from to the
deal with
joint val
histogra
30
Elements of statistical analysis
CHAPTER 2. THE ELEMENTS OF STATISTICAL ANALYSIS

illustrates
which have several
been examples
normalized of jpdf’sby which
the have different cross-co
appropriate varian
Joint
Joint pdfs
pdfs
convenience and andjoint
the
jointmoments
fluctuating
moments
variables u ′
and v ′
can be defined as
Joint pdfs and joint moments
coefficient is defined as:
′ ′

u ρ= u≡− U ⟨u v ⟩
uv ′ 2 ′ 2 1/2
v = v − [⟨u

V ⟩⟨v ⟩]
The correlation coefficient u’
is and v’
bounded are
correlation
correlation random
coefficient
by plus
coefficient (-1 to
or
(-1 to 1)
1)
minus
where as before capital letters are used to represent the mean valu
ing perfectquantities
correlation variables about the mean
fluctuating u andand
′ ′
v are the lattervariables
random perfectwith anti-corre
zero me
′ ′ hu0 v 0 i1>: 0perfect correlation
Aspositive
A with the
valuesingle-variable
of ⟨u v ⟩ indicates1 : perfect correlation
pdf, there
(positive
that u are
′correlation)
and v ′
certain
tend to varc
negative
bility value indicates
density function thatmust
when -1:
0 0 -1: oneperfect
variable
hu′ v′ i < 0perfect anti-correlation
is increasing
(negative anti-correlation
satisfy. If B (c
uv 1 2 ,
correlation) c ) the
indicaoth
decreasing. A zero value of ⟨u v ⟩ indicates that there is no corre
variables u and v, then: 0 0
hu v i = 0 (uncorrelated)

• Property 1: source: Turbulence for the 21st century, W. K. George


Figure 2.5: Contours
Wednesday, 23 October of
13 constant probability for four different joint probability
density functions. Try to figure out what the moments would be for each and how
they would differ. source:Turbulence
source: Turbulence for
for the
the 21st
21st century,
century,W.
W.K.
K.George
George
It Itis issometimes
sometimes more
more convenient
convenient totodeal deal with with values
valuesof ofthe
the cross-variances
cross-variances
which
which have
have been
been normalized
normalized Turbulence:
Turbulence:lecture
It is sometimes more convenient to deal with values of the cross-variances
lecture
bybythetheappropriate
appropriate
which have been normalized by the appropriate variances. Thus the correlation
variances.
variances. Thus
Thus thethecorrelation
correlation
coefficient
coefficient is is
defined
defined as:as:
coefficient is defined as:
Elements of
Elements of statistical statistical analysis

ρρuvρuvanalysis

⟨u ′ ′
⟨u
⟨u v2⟩′v ′′⟩v ′ ⟩
(2.30)
(2.30)
uv ≡ [⟨u ′ 2
⟩⟨v
′ 2 ′ ⟩]
′ 21/2 1/2 (2.30)
[⟨u
[⟨u 2
′ ⟩⟨v ⟩⟨v 2
′ ⟩]1/2 ⟩]
Joint
Joint
TheThe pdfs pdfs
correlation
correlationand
correlation andjoint
coefficient
coefficient
coefficient isismoments
joint bounded
is bounded
bounded moments
byby
by plus
plusplus or
or minus
or minus
minus one,
one,
one, the
thetheformer
former
former represent-
represent-
represent-
ingingperfect
perfect
perfect correlation
correlation
correlation andand
and the the
the latter
latter
latter perfect
perfect
perfect anti-correlation.
anti-correlation.
anti-correlation.
AsAs
As withwiththe
with thesingle-variable
the single-variablepdf,
single-variable pdf,there
pdf, thereare
there are arecertain
certain
certain conditions
conditions
conditions the
the thejoint proba-
joint
joint proba-
proba-
bility
bilitydensity
density
density function
function
function must
must
must satisfy.
satisfy.IfIfIf
satisfy. BBuv
Buvuv
(c
(c1(c , c,2)c) 2indicates
1 , 1c2 indicates the
the
) indicates jpdf
jpdf
the ofthe
of
jpdf the
of random
random
the random
variables
variables
variables u uu
and
and
andv,v,v,
then:
then:
then:
• ••Property
Property1:1:1:
Property

BBuv
B (c1,, c1c2,2)c2>
uv(c
uv1(c )>00 (2.31)
(2.31)
(2.31)
always.
always.
always.
• ••Property
Property2:2:2:
Property

PP rob{c1<<uu<<c c1++dc
rob{c dc1, ,cc2 <
< vv <
< cc2 +
+ dc
dc2}}== B
B uv (c
(c 1 ,,cc2 )dc
)dc1 , ,dc
dc (2.32)
2 2 (2.32)
2
P rob{c
1 1 < u <1 c1 + dc
1 12 , c2 < v <2 c2 + dc2 2 } = uv
Buv1(c1 ,2 c2 )dc 1 1 , dc (2.32)

• Property 3:
• •Property 3: 3:
Property !
! ∞ !! ∞ source: Turbulence
source: forfor
Turbulence thethe
21st century,
21st W.W.
century, K. George
K. Georg
always. Turbulence: lecture
Turbulence: lecture
• Property 2:
Elements
Elementsofofstatistical
statisticalanalysis
analysis
Joint
Jointpdfs
P pdfs
rob{c andand
1 < u <joint dcmoments
c1 +joint
1 , c2 <moments
v < c2 + dc2 } = Buv (c1 , c2 )dc1 , dc2 (2.32)

• Property 3: ! !
∞ ∞
Buv (c1 , c2 )dc1 dc2 = 1 (2.33)
−∞ −∞

• Property 4: ! ∞
Buv (c1 , c2 )dc2 = Bu (c1 ) (2.34)
−∞

where Bu is a function of c1 only.

• Property 5: ! ∞
Buv (c1 , c2 )dc1 = Bv (c2 ) (2.35)
−∞

where Bv is a function of c2 only.

The functions Bu and Bv are called the marginal probability density functions,
source:
and they are simply the single variable pdf’s defined Turbulence
earlier. for the
The 21st century,
isW.used
K. George
source: Turbulence forsubscript
the 21st century, W. K. George
Turbulence: lecture
Turbulence: lecture
Elementsofofstatistical
Elements statisticalanalysis
analysis
Figure 2.6: Surface representation of a joint probability density function.
Joint
Joint pdfs
pdfs andand
jointjoint moments
moments
If the joint probability density function is known, the joint moments of all
orders can be determined. Thus the m, n-th joint moment is
! ∞ ! ∞
m n
⟨u v ⟩ = cm n
1 2 Buv (c1 , c2 )dc1 dc2
c (2.36)
−∞ −∞

where m and n can take any value. The corresponding central-moment is:
! ∞ ! ∞
⟨(u − U )m (v − V )n ⟩ = (c1 − U )m (c2 − V )n Buv (c1 , c2 )dc1 dc2 (2.37)
−∞ −∞

In the preceding discussions, only two random variables have been considered.
The definitions, however, can easily be generalized to accommodate any number
of random variables. In addition, the joint statistics of a single random vari-
able at different times or at different points in space could be considered. This
will be discussed later when stationary and homogeneous random processes are
considered.
source:
source: Turbulence
Turbulence forfor the
the 21st
21st century,
century, W.W.
K.K. George
George
(see
(see problem
problem 2.5).
2.5). It
It is
is also
also possible
possible to
to write
write aa multivariate
multivariate Gaussian
Gaussian probability
probability
density
density function
function for
for any Turbulence: lecture
Turbulence:
any number
number ofof random
lecture
random variables.
variables.
Exercise: Prove that equation 2.23 results from integrating out the dependence
Exercise: Prove that equation 2.23 results from integrating out the dependence
Elements
Elementsofofstatistical
of
of either statisticalanalysis
either variable
variable usinganalysis
using equations
equations 2.34
2.34 or
or 2.35.
2.35.

Statistical
Statisticalindependence
independence
2.4.3 Statistical and lack
and
independence oflack
and correlation
lack of correlation
of correlation
Definition:
Definition: Statistical
Statistical Independence
Independence Two Two random
random variables
variables are
are said
said to
to be
be
statistically
statistically independent
independent ifif their
their joint
joint probability
probability density
density is
is equal
equal to
to the
the product
product
of
of their
their marginal
marginal probability
probability density
density functions.
functions. That
That is,
is,

B
Buv (c , c ) = B (c )B (c )
uv (c11 , c22 ) = Buu (c11 )Bvv (c22 )
(2.39)
(2.39)

It
It is
is easy
easy toto see
see that
that statistical
statistical independence
independence implies
implies aa complete
complete lack
lack of
of corre-
corre-
lation;
lation; i.e., uv ≡
i.e., ρρuv ≡ 0.
0. From
From the
the definition
definition of
of the
the cross-correlation,
cross-correlation,
## ∞ ## ∞
∞ ∞
⟨(u
⟨(u −
−U )(v −
U)(v − VV )⟩
)⟩ =
= (c11 −
(c −U )(c22 −
U)(c − VV )B uv(c
)Buv (c11,, cc22)dc
)dc11dc
dc22
−∞ −∞
##−∞
∞ ##−∞

∞ ∞
=
= (c11 −
(c −U )(c22 −
U)(c − VV )B
)Buu(c
(c11)B
)Bvv(c
(c22)dc
)dc11dc
dc22
−∞ −∞
##−∞
∞∞
−∞ ## ∞

=
= (c11 −
(c −UU)B
)Buu(c
(c11)dc
)dc11 (c22 −
(c − VV )B
)Bvv(c
(c22)dc
)dc22
−∞
−∞ −∞
−∞
=
= 00 (2.40)
(2.40)

where
where wewe have
have used
used equation
equation 2.39
2.39 since
since the
the first
first central
central moments
moments are are zero
zero by
by
definition.
definition. source: Turbulence
source:
source: forfor
Turbulence
Turbulence the
for 21st
the
the century,
21st
21st W.W.
century,
century, K.K.
W. George
K. George
George
Turbulence:
Turbulence:lecture
lecture
mean velocities, e.g. U ≡ u ). The total velocity u(x,t) will normally be denoted just by u.
We have to treat the fluctuations statistically (since we normally cannot compute them),
Elements
Elements ofofstatistical
which means having statistical
to analysis
analysis
characterise them in terms of their amplitude and time-domain
features.
Stationarity and
Stationarity and Ergodicity
Ergodicity
2.1.1 Stationarity & Ergodicity
Take two realisations of all possible time histories of some property of a random (or
turbulent) flow:

x1(t)

x2(t)

t=t1
turbulent) flow:
turbulent) flow:
2.1.1 Stationarity & Ergodicity Turbulence: lecture
Turbulence: lecture
Take two realisations of all possible time histories of some property of a random (or
Elements
turbulent) flow:
of statistical analysis
xx11(t)
(t)
Stationarity and
Stationarity and Ergodicity
Ergodicity
x1(t)

xx22(t)
(t)

x2(t)
t=t
t=t11

t=t1
These
These samples
samples could
could be
be produced
produced by
by doing
doing the
the experiment
experiment twice.
twice. The
The collection
collection of
of all
all
possible
possible realisations
realisations
These samples is
could be the
the ‘random
is produced
‘random process’.
process’.
by doing The
The mean
mean
the experiment value
value
twice. at time
time tt11 of
at collection
The all
allallthe
ofof the samples
samples is:
is:
possible realisations is the ‘random process’. The 11value
mean ∞ ∞
at time
x(t ) = lim ∑ x (t
x(t11 ) = lim ) , t1 of all the samples is:
∑ xnn (t11 ) ,
1→∞∞ N
N n=1
NN→∞
n=1
x(t1 ) = lim ∑ x n (t1 ) ,
and
and similarly
similarly for
for powers
powers of
of x(t).
x(t). This
This isis called
N →∞ Nan
called n=1ensemble
an ensemble average.
average.
Turbulence:
Turbulence: lecture
lecture
Turbulence: lecture
Elements of statistical analysis
Elements of statistical analysis
Stationarity
Stationarity and
and Ergodicity
Ergodicity
Stationarity and Ergodicity
If all these ensemble averages do not vary with t1 the process is stationary. In other words, a
If all these ensemble averages do not vary with t1 the process is stationary. In other words, a
stationary process is one in which all possible moments and joint-moments are time-
stationary process is one in which all possible moments and joint-moments are time-
invariant.
invariant.

But it is also possible to describe the properties of x(t) by computing time averages over
But it is also possible to describe the properties of x(t) by computing time averages over
specific samples (realisations). Consider the n’th sample, xn(t). It has a mean value given by:
specific samples (realisations). Consider the n’th sample, xn(t). It has a mean value given by:

T
T 1
xn = lim 1 ∫ xn (t )dt.
xn = Tlim
→∞ T

0
xn (t )dt.
T →∞ T
If x(t) is stationary AND x n does not depend on0 n, the process is ergodic. So, for ergodic
If x(t) is stationary AND x n does not depend on n, the process is ergodic. So, for ergodic
processes, the time-averaged mean values are equal to the corresponding ensemble averages,
processes, the time-averaged mean values are equal to the corresponding ensemble averages,
i.e. x(t) = xn . Note that only stationary processes can be ergodic. Fortunately, in practice,
Turbulence:
Turbulence: lecture
lecture
Elements of statistical analysis
Elements of statistical analysis
Stationarity
Stationarity and and Ergodicity
Ergodicity
For
For ergodic
ergodic processes,
processes, the
the time
time averages
averages areare equal
equal toto the
the
corresponding
corresponding ensemble
ensemble averages
averages

x(t) = xn
Only
Only stationary
stationary processes
processes areare ergodic
ergodic
In practice, random data representing physical phenomen
In practice, random data representing physical phenomena
areare ergodic.
ergodic. Therefore,
Therefore, wewe
cancan analyse
analyse them
them based
based onon
single observed time history, provided that the
single observed time history, provided that the data data
record is long enough for the quantity of interest
record is long enough for the quantity of interest
Wednesday, 23 October 13
Turbulence:
Turbulence: lecture
lecture
Elements of statistical analysis
Elements of statistical analysis
Stationarity
Stationarity and and Ergodicity
Ergodicity
For
For ergodic
ergodic processes,
processes, the
the time
time averages
averages areare equal
equal toto the
the
corresponding
corresponding ensemble
ensemble averages
averages

x(t) = xn
Only
Only stationary
stationary processes
processes areare ergodic
ergodic
In practice, random data representing physical phenomen
In practice, random data representing physical phenomena
areare ergodic.
ergodic. Therefore,
Therefore, wewe
cancan analyse
analyse them
them based
based onon
single observed time history, provided that the
single observed time history, provided that the data data
record is long enough for the quantity of interest
record is long enough for the quantity of interest
Wednesday, 23 October 13
n=1

Turbulence:
Turbulence:
=
= N
11
⟨xlecture
lecture
⟨x ⟩⟩
N
!
N
!
n
n
(2.47)
(2.47)
N n=1
n=1
Elements
Elements of
of statistical
statistical analysis
analysis
=
= NN
11
NXX==X
X (2.48)
(2.48)
N
Estimation
Estimation
(Note
(Note that
that the from
from
the expected
expected finite
finite
value
value of each number
of number
each xxnn is
is just
just of the
since
since realisations
of XXrealisations
the xxnn are
are assumed
assumed
identically distributed). Thus x is, in fact, an unbiased estimator for the mean.
identically distributed). Thus xNN is, in fact, an unbiased estimator for the mean.
The
The question
question ofof convergence
convergence of
of the
the estimator
estimator can
can be be addressed
addressed by
by defining
defining the
the
22
square of variability of the estimator, say
square of variability of the estimator, say ϵX ϵ N
,, to
to be:
be:
XN

22
22 var{X
var{X N }
} ⟨(X
⟨(X N −
− X)
X) ⟩⟩
ϵϵXN ≡≡ 2
N =
=
N
2
(2.49)
(2.49)
2.5. XN X 2 X 2
2.5. ESTIMATION
ESTIMATION FROM
FROM A A FINITE
X
FINITE NUMBER
NUMBER X OFOF REALIZATIONS
REALIZATIONS 37
37

Now
Now we we want
want to
to examine
examine what
what happens
happens toto ϵϵX N
XN
as
as the
the number
number of
of realizations
realizations
increases.
increases. For
For the
the estimator
estimator toto converge
converge it
it is
is clear
clear that
that ϵϵxx should
should decrease
decrease as
as the
the
number
number of of samples
samples increases.
increases. Obviously,
Obviously, we
we need
need to
to examine
examine the
the variance
variance of
of X
XNN
first.
first. It
It is
is given
given by:
by:
22
var{XNN } = ⟨(XNN − X) ⟩⟩
var{X } = ⟨(X − X)
!! N
##22
11 "
N
"
= ⟨⟨ N
= xxnn − X) ⟩⟩
− X) (2.50)
(2.50)
N n=1 n=1
!! N N
##22
"
1 "
N source:
1 Turbulence
source:
source: Turbulence
"N
Turbulence for
forthe
for the21st
the 21stcentury,
21st century,W.
century, W.K.
W. K.George
K. George
George
"
Turbulence: lecture
Turbulence: lecture
Elements of statistical analysis
2.5. ESTIMATION FROM A FINITE NUMBER OF REALIZATIONS 37

Estimation
Estimation to from
Now we wantfrom finite
finite
examine number
whatnumber
happens to ϵof ofnumber
realisations
X realisations
as the of realizations
N
increases. For the estimator to converge it is clear that ϵx should decrease as the
We examine
number of samplesthe variance
increases. of X
Obviously, weN,need to examine the variance of XN
first. It is given by:
2
var{XN } = ⟨(XN − X) ⟩
! N
#2
1 "
= ⟨ xn − X) ⟩ (2.50)
N n=1
! N N
#2
1 " 1 "
= ⟨ xn − X) ⟩ (2.51)
N n=1 N n=1
! N
#2
1 "
= ⟨ (xn − X) ⟩ (2.52)
N n=1
since ⟨XN ⟩ = X from equation 2.46. Using the fact that the operations of av-
eraging and summation commute, the squared summation can be expanded as
follows:
! N
#2 N "
N
" 1 " source:
source:Turbulence
Turbulence for
for the
the 21st
21st century,
century,W.
W. K.
K. George
George
⟨ (x − X) ⟩ = ⟨(x − X)(x − X)⟩
n
N
Turbulence:
Turbulence:
! N
lecture
n=1
lecture
N
#2
1 " 1 "
= ⟨ xn − X) ⟩ (2.51)
Elements
Elements of statistical analysis
of statistical analysis
!
N n=1 N n=1
#2
N
1 "
= ⟨ (xn − X) ⟩ (2.52)
Estimation
Estimation from finite number
N n=1 of realisations
from finite number of realisations
since ⟨XN ⟩ = X from equation 2.46. Using the fact that the operations of av-
Using
eraging the
Usingandthe fact
fact that
summationthat summation
summation
commute, and averaging
andsummation
the squared averagingcan commute,
commute,
be expanded as
follows:
! N
#2 N "N
" 1 "
⟨ (xn − X) ⟩ = 2
⟨(xn − X)(xm − X)⟩
n=1 N n=1 m=1
N
1 " 2
= 2
⟨(x n − X) ⟩
N n=1
1
= var{x}, (2.53)
N
where the next to last step follows from the fact that the xn are assumed to
be statistically independent samples (and hence uncorrelated), and the last step
from the definition of the variance. It follows immediately by substitution into
equation 2.49 that the square of the variability of the estimator, XN , is given by:
1 var{x}
ϵ2X = source: Turbulence for the 21st century, W. K. George
source: Turbulence for the 21st century, W. K. George
N
Turbulence:
=
1 "
2
lecture
Turbulence: lecture
⟨(x n − X) 2

N n=1
Elements
Elementsofofstatistical
statistical=analysis
1
analysis
var{x}, (2.53)
N
where the next to last step follows from the fact that the xn are assumed to
Estimation
Estimation fromfromfinite number
finite numberof realisations
of realisations
be statistically independent samples (and hence uncorrelated), and the last step
from the definition of the variance. It follows immediately by substitution into
The
Thesquare
equation 2.49 thatof
square ofthe
the thevariability
variability
square ofofXXofNNthe
of the variability isisgiven
givenby
byXN , is given by:
estimator,
1 var{x}
ϵ2XN
=
N X2
$ %2
1 σx
= (2.54)
N X
Thus the variability of the estimator depends inversely on the number of in-
dependent realizations, N , and linearly on the relative fluctuation level of the
random variable itself, σx /X. Obviously if the relative fluctuation level is zero
(either because there the quantity being measured is constant and there are no
measurement errors), then a single measurement will suffice. On the other hand,
as soon as there is any fluctuation in the x itself, the greater the fluctuation (rel-
ative to the mean of x, ⟨x⟩ = X), then the more independent samples it will take
to achieve a specified accuracy.
source:
Example: In a given ensemble the relative Turbulence
Turbulencefor
fluctuation
source: forthe 21st
21stcentury,
level
the is 12%
century,W.W.K.K.(i.e.,
George
George
Turbulence:lecture
Turbulence: lecture
Elementsofofstatistical
Elements
36
statisticalanalysis
analysis
CHAPTER 2. THE ELEMENTS OF STATISTICAL ANALYSIS

Estimation
Estimation from
from finitefinite number
number of realisations
of realisations
Convergencetototrue
Convergence truemean
mean

Convergencetotowrong
Convergence wrongmean
mean

Noconvergence
No convergence

source:
source: Turbulence
Turbulence forfor
thethe 21st
21st century,
century, W.W. K. Georg
K. George
Turbulence: lecture
Turbulence: lecture
Time/space domain analysis
Time/space domain analysis
So far, we have looked at amplitude statistics. These
So far, we have looked at amplitude statistics. These do not do not
tell
tell you
you anything
anything about
about temporal/spatial
temporal/spatial structures
structures
Amplitude and frequency modulation in wall turbulence
Turbulent flow
Turbulent flow Synthetic signal
Synthetic signal
10 10
Original signal Synthetic signal
5 5
0 0
–5 –5
–10 –10
5 10 15 20 5 10 15 20

0 1
We
We
2
need
3 need
4
additional
5 additional
6 7
tools
(c) to
8 tools to
tell
0.04tell
these
these
0.06
apart
0.08 apart
0.10 0.12 0.14
00 100
104 104
can depend only on the time difference t′ − t. And ⟨u(t)v(t′ )w(t′′ )⟩ can depend
only on the two time differences Turbulence: lecture
Turbulence:
t′ − t and t′′ lecture
− t (or t′′ − t′ ) and not t, t′ or t′′
directly.
Time/space domain analysis
Autocorrelation
8.2 The autocorrelation
Autocorrelation
One of the most useful statistical moments in the study of stationary random pro-
cesses (and turbulence, in particular) is the autocorrelation defined as the aver-
age of the product of the random variable evaluated at two times, i.e. ⟨u(t)u(t′ )⟩.
Since the process is assumed stationary, this product can depend only on the time
difference τ = t′ − t. Therefore the autocorrelation can be written as:

C(τ ) ≡ ⟨u(t)u(t + τ )⟩ (8.1)

The importance of the autocorrelation lies in the fact that it indicates the
“memory” of the Measure ofis,memory
process; that ofwhich
the time over the aprocessprocess is correlated with
itself. Contrast the two autocorrelations shown in Figure 8.1. The autocorrelation
of a deterministic sine wave is simply a cosine as can be easily proven. Note that
You
there iscan replace
no time beyondtime
whichwith
it can space and itto will
be guaranteed tell you
be arbitrarily about
small since
longevity of the process
it always “remembers” when it began, and thus always remains correlated with
itself. By contrast, a stationary random process like the one illustrated in the
figure will eventually lose all correlation and go to zero. In other words it has a
“finite memory” and “forgets” how it was. Note source:
thatTurbulence
source: Turbulence
one must for
for the
be21st
the 21st century, W.
W. K.
century,to
careful K. George
George
make
Turbulence:lecture
Turbulence: lecture
Time/spacedomain
Time/space domainanalysis
analysis
Autocorrelation
Autocorrelation
8.2. THE AUTOCORRELATION 151

Figure 8.1: Autocorrelations for two random processes and a periodic one.

source:
source: Turbulenceforforthe
Turbulence the21st
21stcentury,
century,
W.W.
K.K. George
George
roughness with the most recent bumps having the most influence and with dis-
Turbulence: lecture
tant bumps eventually forgotten. On the other hand if the car is traveling down a
railroad track, the periodic crossing of the railroad ties represents a deterministic
Time/space domain analysis
input and the motion will remain correlated with itself indefinitely. This can be
a very bad thing if the tie crossing rate corresponds to a natural resonance of the
suspension system of the vehicle.
Autocorrelation
Since a random process can never be more than perfectly correlated, it can
never achieve a correlation greater than is value at the origin. Thus

|C(τ )| ≤ C(0) (8.2)

An important consequence of stationarity is that the autocorrelation is sym-


metric in the time difference, τ = t′ − t. To see this simply shift the origin in time
backwards by an amount τ and note that independence of origin implies:

⟨u(t)u(t + τ )⟩ = ⟨u(t − τ )u(t)⟩ = ⟨u(t)u(t − τ )⟩ (8.3)

Since the right hand side is simply C(−τ ), it follows immediately that:

C(τ ) = C(−τ ) (8.4)

source: Turbulence for the 21st century, W. K. George


Turbulence: lecture
Turbulence: lecture
Time/space
152 domain analysis
CHAPTER 8. STATIONARITY RANDOM PROCESSES

8.3 The autocorrelation coefficient


Autocorrelation
Autocorrelation
It is convenient to define the autocorrelation coefficient as:

C(τ ) ⟨u(t)u(t + τ )⟩
Sometimes, R = ρ(τ ) ≡
C(0)
=
⟨u2 ⟩
(8.5)

where

⟨u2 ⟩ = ⟨u(t)u(t)⟩ = C(0) = var[u] (8.6)


Since the autocorrelation is symmetric, so is its coefficient, i.e.,

ρ(τ ) = ρ(−τ ) (8.7)

It is also obvious from the fact that the autocorrelation is maximal at the origin
that the autocorrelation coefficient must also be maximal there. In fact from the
definition it follows that
ρ(0) = 1 (8.8)
and
ρ(τ ) ≤ 1 (8.9)
for all values of τ . source: Turbulence for the 21st century,
century,W.
W. K.
K. George
George
and
Turbulence:
Turbulence:
ρ(τ ) ≤ 1 lecture
lecture (8.9)

Time/space domain analysis


Time/space
for all values of τ .

Integral
Integral
8.4 The length
length scale
scale
integral scale
One of the most useful measures of the length of time a process is correlated with
itself is the integral scale defined by
! ∞
Tint ≡ ρ(τ )dτ (8.10)
0

It is easy to see why this works by looking at Figure 8.2. In effect we have replaced
the area under the8.5.correlation
8.5. coefficient
THE TEMPORAL
THE TEMPORAL by a rectangle of height153
TAYLOR MICROSCALE
TAYLOR MICROSCALE unity and width
153

Tint .

Spatial
8.5 The equivalant:
temporalL Taylor microscale
The autocorrelation can be expanded about the origin in a MacClaurin series; i.e.,
" " "
dC "" 1 2 d C ""
2
1 3 d C ""
3
C(τ ) = C(0) + τ " + τ " + τ " (8.11)
dτ " 2 dτ "
2 3! dt "
τ =0 source:Turbulence
source: Turbulence for
τ =0 for the
theτ21st
21st century,
=0 century,W.
W.K.
K.George
George
Figure 8.2: The autocorrelation coefficient showing relation of the integral scale
Turbulence: lecture
Turbulence: lecture
It is easy to see why this works by looking at Figure 8.2. In effect we have replaced
the area under the correlation coefficient by a rectangle of height unity and width
Time/space
Time/space domain
Tint .
domain analysis
analysis
Taylor
Taylor microscale
8.5 Themicroscale
temporal Taylor microscale
The autocorrelation can be expanded about the origin in a MacClaurin series; i.e.,
" " "
dC "" 1 2 d C ""
2
1 3 d C ""
3
C(τ ) = C(0) + τ " + τ "+ τ (8.11)
"
dτ " 2 dτ "
2 3! dt "
Figure 8.2: The autocorrelation coefficient showing relation of the
τ =0 τ =0 integral scale
τ =0

toBut
thewe
area under
know thethe autocorrelation
autocorrelation is coefficient
symmetric curve.
in τ , hence the odd terms in τ
must be identically zero (i.e., dC/dτ |τ =0 = 0, d3 /dτ 3 |τ =0 , etc.). Therefore the
expansion of the autocorrelation near the origin reduces to:
!
1 2 d C !!
2
C(τ ) = C(0) + τ ! + ··· (8.12)
2 dτ !
2
τ =0

Similarly, the autocorrelation coefficient near the origin can be expanded as:
!
1 d ρ !!
2
ρ(τ ) = 1 + ! τ2 + · · · (8.13)
2 dτ !
2
τ =0

where we have used the fact that ρ(0) = 1. If we define ′ = d/dτ we can write
source:
source:Turbulence
Turbulencefor
forthe
the21st
21stcentury,
century,W.
W.K.
K.George
George
Turbulence: lecture
Similarly, the autocorrelation coefficient near the origin can be expanded as:
Turbulence:! lecture
!
1 d2 ρ !
Time/space domain analysis
2
ρ(τ ) = 1 + τ + ··· (8.13)
Time/space domain analysis 2
!
!
2 dτ τ =0
where we have used the fact that ρ(0) = 1. If we define ′ = d/dτ we can write
Taylor
Taylor microscale
microscale
this compactly as:

1 ′′
ρ(τ ) = 1 + ρ (0)τ 2 + · · · (8.14)
2
Since ρ(τ ) has its maximum at the origin, obviously ρ′′ (0) must be negative.
We can use the correlation and its second derivative at the origin to define a
special time scale, λτ (called the Taylor microscale 1 ) by:

2
λ2τ ≡ − ′′ (8.15)
ρ (0)
Using this in equation 8.14 yields the expansion for the correlation coefficient
near the origin as:
τ2
ρ(τ ) = 1 − 2 + · · · (8.16)
λτ
1
The Taylor microscale is named after the famous English scientist G.I. Taylor who invented
it in the 1930’s. Among his many other accomplishments he designed the CQR anchor which is
still found on many boats today. source:
source:Turbulence
Turbulence for
for the
the 21st
21st century,
century,W.
W. K.
K. George
George
Turbulence:
Turbulence: lecture
lecture
Turbulence: lecture
Time/space
Figure
Time/space
Figure
domain
8.3: The autocorrelation
Time/space domain
domain analysis
8.3: The autocorrelation
coefficient for positive time lags together with its
analysis
analysis
coefficient for positive time lags together with its
oscullating parabola showing the Taylor microscale.
oscullating parabola showing the Taylor microscale.
Taylor microscale
Taylormicroscale
Taylor microscale
Thus very near the origin the correlation coefficient (and the autocorrelation as
Thus very near the origin the correlation coefficient (and the autocorrelation as
well) simply rolls off parabolically; i.e.,
well) simply rolls off parabolically; i.e.,
τ22
ρ(τ ) ≈ 1 − τ2 (8.17)
ρ(τ ) ≈ 1 −λτ 2 (8.17)
λτ
This parabolic curve is shown in Figure 8.3 as the osculating (or ‘kissing’) parabola
This parabolic curve is shown in Figure 8.3 as the osculating (or ‘kissing’) parabola
which
CHAPTER approaches zero exactly
8. STATIONARITY RANDOM asPROCESSES
the autocorrelation coefficient does. The inter-
which approaches zero exactly asPROCESSES
the autocorrelation coefficient does. The inter-
cept
CHAPTER of this 8. osculating
STATIONARITYparabola
RANDOM with the τ -axis is the Taylor microscale, λτ .
cept of this osculating parabola with the τ -axis is the Taylor microscale, λτ .
The Taylor microscale is significant for a number of reasons. First, for many
The Taylor microscale is significant for a number of reasons. First, for many
random processes (e.g., Gaussian), the Taylor microscale can be proven to be
random processes (e.g.,
the average distance For For
For many
many random
Gaussian),
betweenmany random processes,
the Taylor
random
zero-crossing
microscaleTaylor
processes,
ofprocesses,
a random variable
can be microscale
Taylor
Taylor microscale
proven
in microscale
time. This
to be
is approximately true
is approximately true
can
for be
can
can be shown
shown
turbulence
for be
as to
shownasthe
turbulence tobe
well.
to be
be
well.
equal
Thus
equal
Thus
one to
the average distance between zero-crossing of a random variable in time. This
equal to
can
one to
the
the
quickly
canthe
average
quickly average
estimate the
average
estimate the
Taylor microscale bydistance between
simply observing zero-crossings
zero-crossings using an oscilloscope
Taylor microscale bydistance
trace. distance between
simply observing the zero-crossings
between zero-crossings
zero-crossings using an oscilloscope
trace.
The Taylor microscale also has a special relationship to the mean square time
The Taylor microscale also has source: Turbulence
2 a special relationship to for
thethemean
21st century,
square W. K. George
time
derivative of the signal, ⟨[du/dt] ⟩.
Wednesday, 23 October 13 2 This is easiestsource:to
source: derivefor
Turbulence
Turbulence ifthe
for we
the consider
21st
21st W.two
century,W.
century, K.K.George
George
dt dt
cept dt

dtosculating dξparabola withdτ dτ
of this
Turbulence:
Turbulence:lecture
the τ -axis is the Taylor microscale, λτ .
lecture
The Taylor microscale is significant for a number of reasons. First, for many
unction
function
random only
only ofof τ(e.g.,
processes τ, , the
the derivative
derivative
Gaussian), the Taylor ofofmicroscale
itit with
with canrespect
be proven to
respect tobeξξ isis
to
hus we
hus we Time/space
Time/spacedomain
theare
are domainanalysis
average
left
left analysis
distance
with:
with: between zero-crossing of a random variable in time. This
is approximately true for turbulence as well. Thus one can quickly estimate the
Taylor microscale
Taylor microscale by simply observing the zero-crossings using an oscilloscope
Taylor microscale
trace. du ′′
du du du d 22
d
isisrelated ⟨⟨to the ⟩ ⟩ =
mean = −− C(τ
squared C(τ ) )
of the derivative of the (8.23)
(8.23)
signal
related
The
dtto
Taylor
dt′ ′ the
dt
dt mean
microscale also
dτsquared
has
dτ 2
2
2 a special
of the derivative of the signal
relationship to the mean
derivative of the signal, ⟨[du/dt] ⟩. This is easiest to derive if we consider two
square time

′ ′′
need the
needatthe second
stationary random
second trick.
signals,
trick. Let’s
say u
Let’s and evaluate
u , we obtainboth
evaluate′ ′
sides
by evaluating
both sides at
the
at t
same
t= t
signal
= t (or
(or
two different times, say u = u(t) and u = u(t ). The first is only a function
emean
mean of t,square
square derivativeis only aas:
derivative
and the second as: of t′. The derivative of the first signal is
function
du/dt and the second du′ /dt′ . Now lets multiply these together and rewrite them
!! ""2 ##
as:
du 2
dd 22 ##
⟨⟨ du ⟩⟩==−− 2 ′C(τ
##
))# #2 (8.24)
dt dτ C(τ
du2du #d# (8.24)
dt dτ = τ =0u(t)u′(t′) τ ′=0 (8.18)
dt′ dt dtdt
22
finition
efinition
where ofof
thethe
the Taylor
Taylor
right-hand microscale
sidemicroscale and
and
follows from our the
thefacts
that uthat
facts
assumption is notC(0)
that aC(0) ==⟨u
function of⟨u⟩⟩
′ ′
), t
this nor
is u a functionthe
exactly of t. same as:
τ ), this is exactly the same as:
Using
Usingstandard
standard
!! ""2
calculus
calculusand
andsome
somemanipulation,
manipulation,
2 22
du
du ⟨u
⟨u ⟩⟩
⟨⟨ ⟩⟩==22 2 (8.25)
(8.25)
dtdt λλτ 2
τ source: Turbulence for the 21st century, W. K. George
source: Turbulence for the 21st century, W. K. George
Turbulence: lecture
Turbulence: lecture
Time/space
Time/space
152 domain
domain analysis
analysis
CHAPTER 8. STATIONARITY RANDOM PROCESSES

Cross correlation
8.3 The autocorrelation coefficient
Cross correlation
It is convenient to define the autocorrelation coefficient as:

C(τ ) ⟨u(t)u(t + τ )⟩
ρ(τ ) ≡ = (8.5)
C(0) ⟨u2 ⟩

where
What if the second ⟨usignal is not
2
“u”, but
⟩ = ⟨u(t)u(t)⟩ = C(0) = var[u]
another variable?
(8.6)

⇢⇢AB
hA(t)B(t +
Since the autocorrelation
hA(t)B(t
(⌧ ) = p p+ ⌧ )i Cross
⌧ )i correlation
is symmetric, so is its coefficient, i.e.,
AB (⌧ ) = p p ρ(τ
hA i hB 2ii ) = ρ(−τ ) between A & B (8.7)
hA 2
2 i hB 2

It is also obvious from the fact that the autocorrelation is maximal at the origin
that the autocorrelation coefficient must also be maximal there. In fact from the
Gives you information about how one signal is related to
Gives you itinformation
definition follows that aboutρ(0)how one signal is related to
=1 (8.8)
another - We will come back to this in Reynolds shear stress
and
ρ(τ ) ≤ 1 (8.9)
for all values of τ . source:
source:Turbulence
Turbulence for
for the
the 21st
21st century,
century, W.
W. K.
K. George
George
Turbulence: lecture
Time/space domain analysis
Matlab implementation

>>[R,lags] = xcorr(A,B,‘unbiased’);
>> R = R./(std(A).*std(B));
>> plot(lags.*dt,R);

If A = B, then we have autocorrelation


`Unbiased’ is crucial since we need to account for the limited
number of samples as we go farther towards the edge of the signals

Ideally, we want to compute correlations within the central part of


the signal
source: Turbulence for the 21st century, W. K. George
Turbulence: lecture
Turbulence: lecture
Time/space domain
Time/space domain analysis
analysis
Fourieranalysis
Fourier analysis
This provides
This provides aa frequency
318
frequency analysis
analysis of
of aa signal
signal -- amplitude
amplitude or
APPENDIX C. FOURIER ANALYSIS OF TIME VARYING SIGNALS
or
energy
C.1 energy
Fourier content
content in different
series in different frequencies
frequencies
If we have a periodic signal (i.e., a signal that repeats itself every time interval T )
it can be developed in a Fourier series as:

! " # ∞
! " #
t t
u(t) = A0 + An cos 2πn + Bn sin 2πn (C.1)
n=1 T n=1 T

The frequencies present in this decomposition, fn = n/T , are harmonics (or in-
teger multiples) of the fundamental frequency 1/T . The Fourier coefficients, An
and Bn , are given by:
" #
1 $ T /2 t
An = u(t) cos 2πn dt (C.2)
T −T /2 T
$ " #
1 T /2 t
Bn = u(t) sin 2πn dt (C.3)
T −T /2 T
source:Turbulence
source: Turbulencefor
forthe
the21st
21stcentury,
century,W.
W.K.K.George
George
∞ " #
44 ! 1 t
Turbulence: lecture

! 1 sin 2πn t #
! 1 " t
Turbulence: lecture

u(t) =
u(t) =
=π 4 sin 2πn
2πnT (C.4)
(C.4)
u(t) π n=1,3,5... n
n sin T (C.4)
π n=1,3,5...
n=1,3,5... n T
Time/space domain
Time/space
Figure C.1domain analysis
analysis
illustrates that for nn = 11 (dashed curve) the result isis just
Figure C.1 illustrates that for n = 1 (dashed curve) the result is just a sine curve.
Figure C.1 illustrates that for = (dashed curve) the result just aa sine
sine curve.
curve.
As the Fourier components for increasing values of nn are added, however, the
As the Fourier components for increasing values of n are added, however, the
As the Fourier components for increasing values of are added, however, the
Fourieranalysis
Fourier analysis
resulting
resulting curve
curve approaches
approaches the
the square
square wave
wave
resulting curve approaches the square wave signal.
signal.
signal.

ItIt isis sometimes convenient to use aa complex notation in the formulation of


It is sometimes convenient to use a complex notation in the formulation of
sometimes convenient to use complex notation in the formulation of
Fourier series. We can define aa complex coefficient as C nn = A nn −− iB nn and rewrite
Fourier series. We can define a complex coefficient as Cn = An − iBn and rewrite
Fourier series. We can define complex coefficient as C = A iB and rewrite
equations
equations (C.2–C.3)
(C.2–C.3) as:
as:
equations (C.2–C.3) as:
11 $$$ TTT/2
/2 −i2πnt/T
CCnn == 1 /2 u(t)e
u(t)e −i2πnt/Tdt
−i2πnt/T dt (C.5)
(C.5)
Cn = TT −T −T /2
/2
u(t)e dt (C.5)
T −T /2
The ratio between the real and imaginary values can provide phase information
The ratio between the real and imaginary values can provide phase information
The ratio between the real and imaginary values can provide phase information
about
about the
the signal.
signal.
about the signal.
ItIt isis also convenient to introduce negative values of n. This corresponds to
It is also convenient to introduce negative values of n. This corresponds to
also convenient to introduce negative values of n. This corresponds to
negative frequencies (which can be thought of as waves going backwards in time).
negative frequencies (which can be thought of as waves going backwards in time).
negative frequencies (which can be thought of as waves going backwards in time).
Then the values are symmetric in the sense that A(−n) = A(n) and B(−n) =
Then the values are symmetric in the sense that A(−n) = A(n) and B(−n) =
Then the values are symmetric in the sense that A(−n) = A(n) and B(−n) =
B(n). The use of negative values of nn means that the reconstruction formula now
B(n). The use of negative values of n means that the reconstruction formula now
B(n). The use of negative values of means that the reconstruction formula now
can
can be
be written
written very
very compactly
compactly as
as
can be written very compactly as

!

!

! Cn e+i2πnt/T
u(t)
u(t) =
= C e +i2πnt/T
+i2πnt/T (C.6)
(C.6)
u(t) = C e
n
n=−∞ n (C.6)
n=−∞
n=−∞
source:Turbulence
source: Turbulence for
for the
the 21st
21st century,
century,W.
W. K.
K. George
George
C.2 Turbulence: lecture
Fourier transform
TheTime/space domaininanalysis
Fourier series discussed the previous section only applies for a periodic
deterministic signal; i.e., a signal that is exactly the same from period to period.
Fourier analysis
But sometimes we must deal with a single pulse or even a random processes. We
therefore need to be able to decompose a signal that is not repeatable. We can
Fourier series is only valid for time-periodic signals.
overcome this by using the Fourier transform. This can be viewed as a Fourier
series in the limit for which the period, T becomes infinite, and the correspond-
ing values of n/T become a continuous range of frequencies f , meaning that all
We need to deal with
frequencies are now possible.
signals that result from random processes
We define then, the Fourier transform of the function u(t) as:
! ∞
û(f ) = e−i2πf t u(t)dt (C.7)
−∞

These are really the continuous counterpart to the Fourier series coefficients of a
periodic signal, and can similarly be used to reconstruct the original signal. We
call this reconstruction the inverse Fourier transform and define it as:
! ∞
u(t) = e+i2πf t û(f )df (C.8)
−∞

As in theWe assume
complex Fourierthat
series,these integrals
we use negative values converge!
of the frequency f .
An implicit assumption is that the integralssource:
converge. This
Turbulence is,21st
for the of course,
century, W.never
K. George
F [F(f ) ⊗ F(g)] = f g (C.13)
Turbulence: lecture
Turbulence: lecture
In words, this means that the convolution of two functions in the time space
Time/space
Time/space
corresponds domain
domain
to the product analysis
analysis
of the transformed functions in the frequency space –
and vice versa.
Fourier analysis
Fourier analysis
But,
But, we
C.4 The we cannot
cannot
finite collect data
data for
for infinite
collecttransform
Fourier infinite time
time
In any application of Fourier analysis we are always limited by the length of the
time record, T . This means that the most we can expect to be able to transform
is the finite time transform given by:
! T /2
ûiT (f ) = e−i2πf t u(t)dt (C.14)
−T /2

where for convenience we have written it over the symmetric interval in time
(−T /2, T /2).
Now with a little thought, it is clear that we are actually taking the Fourier
The
The length
transform
length of
of the
of the product
the oftime-record becomes
two functions, the
time-record correlation really
becomes (the partimportant
really we want) plus
important
the window function; i.e.,
! ∞
ûiT (f ) = F[u(t)wT (t)] = source:
e−i2πfTurbulence
t
u(t)wTfor the 21st century, W. (C.15)
(t)dt K. George
source: Turbulence for the 21st century, W. K. George
Turbulence: lecture
Turbulence: lecture
is the finite time transform given by:
! T /2
Time/space domain analysis
Time/space domainiT analysis
û (f
−T /2
) = e −i2πf t
u(t)dt (C.14)

Fourier analysis
where for convenience
Fourier we have written it over the symmetric interval in time
(−T /2, Tanalysis
/2).
Now with a little thought, it is clear that we are actually taking the Fourier
transform of the product of two functions, the correlation (the part we want) plus
the window function; i.e.,
! ∞
ûiT (f ) = F[u(t)wT (t)] = e−i2πf t u(t)wT (t)dt (C.15)
−∞

where wT (τ ) is defined by:

1, −T /2 ≤ τ ≤ T /2
wT (τ ) = (C.16)
0, |τ | > T /2

From the results of the preceding section we immediately recognize that the
Whatwedoes
What
Fourier transform does this
seek is this windowof function
window
the convolution function do?transform with
do?
the true Fourier
the Fourier transform of the window function; i.e,,
source:Turbulence
source: Turbulence for
for the
the 21st
21st century,
century,W.
W.K.
K.George
George
Turbulence: lecture
Time/space domain analysis
Fourier analysis
The window contaminates the amplitude at a given
frequency by leaking information from other frequencies
Most people use a top-hat window without actually
realising that they are using a top-hat window
Need to ensure the record length is much much larger
than the largest scale to be considered
Conservatively, keep the record length at least one order
of magnitude larger than the integral scale
source: Turbulence for the 21st century, W. K. George
Turbulence:lecture
Turbulence: lecture
Time/spacedomain
Time/space domainanalysis
analysis
Fourier
Fourier analysis
analysis
Usually,we
Usually, wesample
samplesignals
signalsdigitally
digitallyfor
fora alength
lengthofoftime
timeatata a
givenfrequency
given frequencyofofacquisition
acquisition
D.1. ALIASING OF PERIODICALLY SAMPLED DATA 3

(a) Fourier transform of true signal

Thefrequency
The frequencyofofacquisition
acquisitionisis |û( f )|

veryimportant
very important
-2fs -fs 0 fs 2fs

Wewill
We willincur
incurALIASING
ALIASINGif if (b) Fourier transform of properly sampled signal

samplingfrequency
sampling frequencyisistoo
toolow
low |û( f )|

Satisfy“Nyquist
Satisfy “Nyquistcriterion”
criterion”
-2fs -fs 0 fs 2fs

(c) Fourier transform of aliased signal


|û( f )|

Sampleatatleast
Sample leasttwice
twicethe
thefrequency
frequency
asasthe
themaximum
maximumfrequency
frequency -4fs -3fs -2fs -fs 0 fs 2fs 3fs 4fs

source:
source: Turbulence
Turbulence forfor
thethe 21st
21st century,
century, W.W.
K. K. Georg
George
2 Turbulence:
The Discrete Fourier lecture
transform
Time/space
section domain
B.1, the problem is bothanalysis
that the signal is only known at discrete
s and that we only will have data for a limited time period. Now we will look
w Fourier analysiscan be made. First we have to change eqs. (C.7) to
a discrete formulation
te time estimate, Continuous fourier transform
T !
D.2. THE DISCRETE FOURIER TRANSFORM
−i2πf t 329
uˆT (f ) = e u(t)dt (D.14)
0
times. We really have
What
that wedeliberately
we have have is discrete
shifted thedata: Discrete
time axis by +T /2. fourier transform
This introduces
+iπf T
ear phase shift equivalent to multiplying
un = u(n∆t) the Fourier coefficients
n = 0, 1, 2, . . . , N − 1 by e (D.15)
ared to the symmetric finite transform of the previous chapter. Generally
will not present
Using a problem,
the basic but
definitions of should
integral not be ignored.
calculus, we can discretize the integral of
his isequation (D.14)
not exactly as: we want, since we have sampled the velocity at discrete
what
N
! −1
uˆT (f ) = e−i2πf n∆t un ∆t (D.16)
n=0

N is the total number of samples and T = N ∆t is the total sample time. The time
between samples, ∆t, is given by the sampling frequency, ∆t = 1/fs = T /N . This
is, of course, an approximation which becomes exact in the limit as the number
of points, N , becomes infinite and as the interval between them, ∆t, goes to zero.
Now since we only have N data points, we can only
source: calculate
Turbulence for theN21st
independent
century, W. K. George
equation (D.14) as:
Turbulence:
Turbulence:
uˆ (fT)= e lecture
lecture
N
! −1
u ∆t −i2πf n∆t
n (D.16)
n=0

Time/space
Time/space domain
domain
N is the total number analysis
analysis
of samples and T = N ∆t is the total sample time. The time
between samples, ∆t, is given by the sampling frequency, ∆t = 1/fs = T /N . This
is, of course, an approximation which becomes exact in the limit as the number
Fourieranalysis
Fourier analysis
of points, N , becomes infinite and as the interval between them, ∆t, goes to zero.
Now since we only have N data points, we can only calculate N independent
Fourier coefficients. In fact, since we are in the complex domain, we can only
calculate N/2, since the real and imaginary parts are independent of each other.
So we might as well pick the frequencies for which we will evaluate the sum of
equation D.16 for maximum convenience. For almost all applications this turns
out to be integer multiples of the inverse record length; i.e.,
m m
fm = = m = 0, 1, 2, . . . , N − 1 (D.17)
T N ∆t
Substituting this into equation D.16 yields our discretized Fourier transform as:
" N −1
#
1 !
ûT (fm ) = T e−i2πmn/N un m = 0, 1, 2, . . . , N − 1 (D.18)
N n=0

This equation can be evaluated numerically for each of the frequencies fm defined
in eq. (D.17). The Fourier coefficients, ûT (fm ), found from eq. (D.18) are complex
numbers. For future reference note they negative frequencies are mapped into
fm = N − 1, N − 2, N − 3, N − m instead of at negative values of m, as illustrated
in Figure D.1.
source:
It is easy to show the original time series datasource:
points Turbulence
Turbulence
can for21st
for the
be recovered theby
21st century,
century,
using W. K.W. K. George
George
N n=0
Turbulence: lecture
Turbulence: lecture
This equation can be evaluated numerically for each of the frequencies f defined m
in eq. (D.17). The Fourier coefficients, ûT (fm ), found from eq. (D.18) are complex
Time/space
Time/space
numbers.
domain
For futuredomain
analysis
analysis
reference note they negative frequencies are mapped into
fm = N − 1, N − 2, N − 3, N − m instead of at negative values of m, as illustrated
Fourier
Fourier
in Figure D.1. analysis
analysis
It is easy to show the original time series data points can be recovered by using
the inverse discrete Fourier transform:
"N −1 #
1 !
un = e+i2πmn/N uˆT (fm ) (D.19)
T m=0

Evaluating
Evaluating these discrete
The real and these discrete
imaginary coefficients
value of atcoefficients isis numerically
numerically
frequency fm corresponds taxing
taxing
to the coefficients
of the cosine and sine parts respectively, in a reconstruction of the signal. In fact if
Fast
Fast Fourier
Fourier Transform
Transform (FFT)
(FFT) allows
allows inexpensive
inexpensive computation
we divide uˆT (fm ) by T, then it is exactly the Fourier series coefficient that would
computation
represent a periodic signal of period T , but equal to our piece of the record over
theMatlab
interval 0, implementation
Matlab implementation
T . Thus the algorithms for treating a discretized periodic signal and
a discretized piece of an infinitely long signal are the same. Because of this the
>> uh
>> uh == fft(A);
fft(A);

source:Turbulence
source: Turbulencefor
forthe
the21st
21stcentury,
century,W.
W.K.K.George
George
Turbulence: lecture
Time/space domain analysis
Fourier analysis a = sin(2*pi*t/5);
T = 101
dt = 1
uh = fft(a);
ah = abs(uh);
plot(ah); Peak at m = 21
You only need to
consider till N/2

f = (m-1)/T In matlab, zero index does not exist


So, m=1 is in fact zero frequency
f = 0.2 (= 1/5 that we put in)
Turbulence:lecture
Turbulence:
Turbulence: lecture
lecture
Time/space domain
Time/spacedomain
Time/space analysis
domainanalysis
analysis
Power
Power
Power spectral
spectral
spectral density function
density
density function
function
Now
Now
Now that
that
that we
we we know
know
know how
how
how toto
to compute
compute
compute fourier
fourier
fourier transforms,
transforms,
transforms, we
wewe can
can
can use
use
use
this
this
this toto
to see
see
see the
the
the energy/power
energy/power
energy/power content
content
content within
within
within signals
signals
signals

Define: S
Define: SABAB(f ) = Â (f(f
(f ) = Â

) )B̂(f
B̂(f
) )
RRR(in
(in(indimensional
dimensional
dimensional form)
form)
form) andSSform
and
and Sform
form perfect
perfect
perfect fourier
fourier
fourier transform
transform
transform pairs
pairs
pairs
Different
Different
Different people
people
people use
use
use different
different
different symbols
symbols
symbols ofofof
SSS
Obviously,
Obviously,
Obviously, the
thethesymbol
symbol
symbol or
or orletter
letter
letter does
does
does not
not
not matter
matter
matter
Z Z1 1
i2⇡f ⌧
SABSAB(f(f
)=)= e e i2⇡f ⌧
RABRAB
(f(f
)d⌧)d⌧
11
ZZZ111
i2⇡f
i2⇡f ⌧⌧
RRAB
RAB(⌧(⌧(⌧
AB ))=)== eee SSAB
i2⇡f ⌧ SAB(f(f(f
AB )df
)df)df
11
1
Turbulence:
Turbulence: lecture
lecture
Time/space
Time/space domain
domain analysis
analysis
Power
Power spectral
spectral density
density function
function
Z 1
Z 1
i2⇡f
i2⇡f ⌧ ⌧
RABRAB
(⌧ )(⌧=) = e e SAB SAB (f )df
(f )df
1 1

ForFor
⌧ =⌧ 0,
= if
0, we
if we evaluate
evaluate thisthis integral,
integral,
Z 1
RAB (0) = SAB (f )df
1

This
This tells
tells us us
howhow
thethe cross
cross correlation
correlation at at zero
zero time
time laglag
comes
comes from
from thethe entire
entire distribution
distribution of of frequencies
frequencies

source:
source: Turbulence
Turbulence for21st
for the the century,
21st century,
W. K.W. K. Geo
George
Turbulence: lecture
Turbulence: lecture
Time/space domain analysis
Time/space domain analysis
Power spectral density function
Power spectral density function
If A = B = u
Z1
1
Ruu
uu (0) = Suu
uu (f )df
1
1
Experimentally,negative
Experimentally, negativefrequencies
frequenciesdo
donot
notmake
makesense
sense
SSuuuuisisan
aneven
evenfunction
functionand
andhence
Z
hencewe
wecan
canwrite
writethis
thisas,
as,
1
1
Ruu
uu (0) = 2 Suu
uu (f )df
00
Z 1
1
Ruu
uu (0) = Euu
uu (f )df
00
source:Turbulence
source: Turbulence for
for the
the 21st
21st century,
century,W.
W. K.
K.George
George
source: Turbulence for the 21st century, W. K. George
Turbulence:
Turbulence: lecture
lecture
Time/space
Time/space domain
domain analysis
analysis
Power spectral
Power densitydensity
spectral functionfunction
Z 1
2
hu i = Ruu (0) = Euu (f )df
0
2 2
RuuR
(0)
uu (0)
has has units
units of of energy
energy per per
unitunit
massmass
(u )(u )

df has
df has units
units of frequency
of frequency

EuuE ) (f
(fuu has) has units
units of energy
of energy per per
unitunit
massmass
per per frequency
frequency

source:source: Turbulence
Turbulence for the for
21stthe 21st century,
century, W. K. G
W. K. George
Turbulence: lecture
Time/space domain analysis
Matlab implementation: Power spectrum
>> u = load(‘hwdata.dat’); % load data
>> N = length(u) % number of samples
>> uh = fft(u); %Do FFT
>> f = [0:N-1]./(N*dt)
“dt is the time separation between successive time
records”

>> eh = (conj(uh).*uh)/(N*dt);
% calculate spectrum. Dividing by N*dt ensure %that
the units are correct

>>plot(f(2:N/2),eh(2:N/2));
% The area under the curve here should be equal to
the variance of the signal
source: Turbulence for the 21st century, W. K. George
Turbulence: lecture
Turbulence: lecture
Time/space domain analysis
Matlab implementation:
Matlab implementation:Power spectrum
Power spectrum
First record
8
10 128 records

7
10

6
10
S(f)

5
10
Take lots of blocks of
data and average
4
10

3
10

2
10

2 3 4
10 10 10
f [Hz] source: Turbulence for the 21st century, W. K. George
Turbulence: lecture
Turbulence: lecture
Time/space domain analysis
Time/space domain analysis
Taylor’s hypothesis
Taylor’s hypothesis
Most of the time, we obtained temporal data
Most of the time, we obtained temporal data
Almost always, the theories in turbulence are related to space
Almost always, the theories in turbulence are related to space
Need a way to convert time to space
Need a way to convert time to space

Proc. R. Soc. Lond. A February 18,


1938, Vol 164 No. 919 pp 476-490;
doi:10.1098/rspa.1938.0032

x= U t or l = f /U
Turbulence: lecture
Time/space domain analysis
Taylor’s hypothesis

At a first glance, it looks pretty frozen


But, if you start peeling the layers, then not really frozen
There is an entire range of convection velocities
It is not really correct, but, we are limited at this time
Turbulence: lecture
3.2 Equations for the Average Velocity
Equations
Turbulence forchaotic
is that average
state ofvelocities
motion characteristic of solutions to the equa-
tions of motion at high Reynolds number. Although laminar solutions to the
It is convenient
equations to are
often exist that analyse turbulent
consistent flows conditions,
with the boundary by decomposing
perturba-
it in to(sometimes
tions to these solutions the mean even and the fluctuation
infinitesimal) can cause them to become
turbulent. To see how this can happen, it is convenient to analyze the flow in two
parts, a mean (or average) component and a fluctuating component. Thus the
Thus, the instantaneous
instantaneous velocity andvelocities andbestresses
stresses can can be written as,
written as:

ũi = Ui + ui
p̃ = P + p
(v) (v) (v)
T̃ij = Tij + τij (3.11)
(v)
where Ui , p, and Tij represent the mean motion, and ui , p, and τij the fluctuating
motions. This technique for decomposing the instantaneous motion is referred to
as the Reynolds decomposition. Note that if the averages are defined as ensemble
means, they are, in general, time-dependent. For the remainder of this book,
unless otherwise stated, the density will be assumed constant so ρ̃ ≡ ρ and its
fluctuation is zero.
Substitution of equations 3.11 into equationssource:
3.10Turbulence
yields for the 21st century, W. K. George
as the Reynolds decomposition. Turbulence:
T̃(v)
ν≡
p̃(v)
Turbulence:
ij p̃
= Note
P+
(v) =ρ P
T̃(v) T
ij = =T (v)
(v)pthat
+
ij
T(v) +
(v)p
ij+ +
τ (v) lecture
(v) if the averages are defined
τijlecture
ij
(v)
(3.9)
(3.11)
(3.11)
as e
means, they are, in general, T̃ T̃ij time-dependent.
ij = Tijij + τijijτ (v)
For the remainder (3.11)
(3.11) of th
that
unless Equations
wheresince
U , theand
p,
otherwise
where i U , p, and T for
density
(v)
T
(v)
stated, average
is assumed
represent
representthe thethe velocities
constant,
mean
density
mean motion, theand
will
motion,
tilde
be
and u
u ,isp,no
assumed
longer
and τ the
i , p, and τ ij constant
the
necessary.
fluctuating
fluctuatingso ρ̃ ≡ ρ
Equations for average the velocities
i ij
(v) i ij
ometimes
where U
motions.
where , it
p,
i U This will
and Tbe more
ij
represent
(v)
technique instructive
the
forfordecomposingmean and convenient
motion,
the and u
instantaneousto
, p, not
and explicitly
τ
and uii , p, andmotion the is include
fluctuating
referredtoto
motions. i , p, and
This ijTij represent
technique decomposing mean motion,
the instantaneous τijij theis
motion fluctuating
referred
fluctuation
mpressibility
motions.
asmotions.
the This
Reynolds
istechnique
in zero.
the stressfor
decomposition. term, but
fordecomposing
Note to that
refer
theififto the incompressible
instantaneous motion ismomentum
referred to
This
as the Reynolds technique
decomposition. Notethat
decomposing the the averages
averages are
instantaneous
the defined
motion
are defined is as ensemble
referred
as to
ensemble
tion Substitution
asmeans,
the Reynolds
inthe
as the
means, following
Reynolds
they they are, of equations
decomposition.
form:
decomposition.
are, in in general, NoteNote 3.11
general,time-dependent. thatinto
that
time-dependent. theequations
ififthe averages
averages
For
For thethe are are3.10
defined
defined
remainder
remainder yields
as this
as
of
of ensemble
ensemble
this book,
book,
means,
means,
unless
unless theythey
otherwise are,
otherwise are,in ingeneral,
stated,
# general,
stated, the the time-dependent.
time-dependent.
density
density $ will will bebe For
For the
assumed
assumed the remainder
remainder
constant
constant so of
of
ρ̃ ≡
ρ̃ this
this
≡ ρ and
ρ book,
book,
and its
its
! " (v)
unless
unless otherwise
fluctuation otherwise
is stated,
zero. stated,∂ ũthe
i thedensity ∂ ũi will
density willbe be∂ p̃assumed
assumed∂ T̃ij constant
constant so ρ̃ρ̃ ≡ ≡ ρρ and and (v)its (v)
its
fluctuation
∂(U i + isu zero.
i ) ρ + ũj ∂(U = i−+ ui )+ ∂(P + p) ∂(T(3.10)
ij + τij )
fluctuation
ρ Substitution
fluctuation is is
Substitutionzero.
zero. +equations
of of (Uj + 3.11
∂t
equations u∂xj )jinto
3.11 ∂xi
intoequations
equations =∂x
3.10
3.10 j−yields
yields +
Substitution
Substitution∂t of ofequations equations3.11 3.11into ∂x
intoequations
equations
j 3.10
3.10 yields yields∂xi ∂xj
form has ! !the advantage that it is easier""to keep track of the(v) exact (v)role
(v) of the
∂(U ∂(U + +
u u
)
! ρ! i i i i + (U + u ) i i) ∂(U∂(U + + u u ) )
i"i" ∂(P
∂(P ++ p)
p) ∂(T ij ++ ττ ij
ij ))
ρ
us stresses.
∂(U ∂(U + +
u u
) )+ (U + u )
∂(U ∂(U + + u u ) ) = = −−∂(P
∂(P + + p)p) +
+ ∂(T
(v)
++ τ
τ
(v)
(v)
)) (3.12)
(3.12)
This
ρ ρ equation
i ∂t ∂t
i i i
++ can
(U(U j
j
+
j
j
+ now
u u
j
j
)j )
j
bei ∂xi ∂xaveraged
j
ij i
= = − − to ∂xyield
∂x ii + + an equation
ij ∂x jj
ij
ij expressing
(3.12)
(3.12) mo
∂t
∂t for the averaged ∂x ∂x j j motion. ∂x ∂xii ∂xjj
conservation This equation can nowbebeaveraged
averagedtotoyield yieldan
Note
anequation
that
equation expressing
the
expressing momentum
operations of a
This equation can now momentum
and differentiation
This equation can commute;
now beaveraged
averaged
conservation for the averaged motion. Note that the operations of
conservation
This equation for
can the
now averaged
be i.e.,
motion. to the
toyield
yield
Note average
anthat
an equation
equation the of a
expressing
operations
expressingderivative momentum
averaging
momentum
of is the sam
averaging
conservation
derivative
and of
differentiationforthe
the the averagedmotion.
average.
commute; motion.
Also,
i.e., the Noteaverage
the
average that
of a the operations
of
derivative a fluctuating
is the of averaging
same as quantity
conservation for averaged
and differentiation commute; i.e., the average of a derivative is the same Note that the operations of as the
averaging the
and and differentiation
derivative
differentiation of the commute;i.e.,
average.
commute; i.e.,the
Also, the
the average
average
average ofofaaaderivative
of derivative
fluctuating is
is the
quantity
the same
same isas the
zero.
as the
2
2
Thus the
derivative
derivative
Thus the
equation
of
of
the
the
equation
for
average.
average. the
for theAlso,
Also,
Also, averaged
averaged
the
the
average
average motion
motion reduces
of
of a
a reduces
fluctuating
fluctuating
to:
to:
quantity
quantity is
is zero.
zero. 2
2
derivative
Thus of the average. the average of a fluctuating quantity is zero.
Thusthe theequation
equationfor for the
the averaged
averaged motion motion reduces reducesto: to:
Thus the equation for !! the averaged "motion " reduces(v)to: (v)
!! ∂U ∂U i i
"
∂U"i ∂Ui ∂P ∂P
∂T (v)
ij ∂T ∂ui ∂u
ρ ∂U + U ∂U = − ∂P + ∂T (v) − ρ⟨u ij
∂u ⟩ i
(3.13)
! ρ ∂Uii
ρρ∂Ui ∂t+ + UU
+
∂U
j
∂U
j U
∂x
"i j = − ∂P
i
= − ∂P ∂x= +
+
− ∂T∂x
∂T
ij
(v)
ij j −

+ ρ⟨u
ρ⟨u
∂u
j
∂u
j ∂x i ⟩− ρ⟨uj

i ⟩
(3.13)
(3.13)
ρ ∂t
∂t +∂t U j ∂x
j ∂xj
i
j
j
=∂x − j ∂x
∂xi +
i
i
∂x
∂x
ij
∂xj j i− ρ⟨u ∂x
j ∂x
j ∂xj ⟩
ji j
j ∂x(3.13)
j
where the remaining ∂t fluctuating ∂xj product ∂xi term∂xhas j been moved ∂xj to the right-hand
where
where the remaining
thethe remaining fluctuating
fluctuating product
product term
term has
lasthas been
been moved
moved totothe the right-hand
right-hand
where
where
sidethe
the
of remaining
remaining
equation. fluctuating
Whether
fluctuating
or
product
not product
this
term has
termterm
source:
been has
Turbulence
is zero
moved
likebeen
to
for
the
the
the moved
21st
other to
century,
fluctu-
right-hand
W. the rig
K. George
side
side ofof
ating
Wednesday, 23the
theequation.
terms
October
dependsWhether
equation.
13 Whether or not this
on the correlation this last
of last
terms term
term source:
in isiszero
the zero likefor
Turbulence
like
product. the
the
In the other
21st
other
general, fluctu-
century,
fluctu- W. K. George
these
This
This equation
equation can
can nownow bebeaveraged
averaged totoyield
yield ananequation
equation expressing
expressing momentum
momentum
This equation can now be averaged to yield an equation expressing momentum
This equation
conservation
conservation
conservation
conservation
for
for for
can now
the
the the averaged
averaged
averaged
for thecommute;
averagedi.e.,
Turbulence: lecture
be averaged
Turbulence:
motion.
motion.
motion.
motion.
to Note
yield an
lecture
NoteNote
Note of
that
that
equation
that
that
the
the the
the
expressing
operations
operations
operations
operations of
momentum
ofof averaging
averaging
averaging
ofsame
averaging
and
and differentiation
differentiation commute; i.e., the
the average
average of a aderivative
derivative isisthethe same asasthe
the
and
and differentiation
differentiation commute;
commute; i.e.,
i.e., the
the average
average ofof a a derivative
derivative is is the
the same
same asasthe 2
the
derivative of the average. Also, the average of a fluctuating quantity is zero.
Equations
Equationsfor
derivative
derivative
Thus
Thus foraverage
derivative
the
theaveragevelocities
of
ofofthe the
the
equation
equation
average.
velocities
average.
average.
for
for thethe
Also,
Also,
Also,
averaged
averaged
the the
the
average
average
average
motion
motion reduces
of
reduces
a fluctuating
ofofa afluctuating
fluctuating
to:to:
quantity
quantity
quantity
is 2 2
zero.
isiszero.
zero. 2
Thus the equation for the
Thus the equation !for the averaged averaged motion reduces
motion reduces to: to:
"
! ! " (v)
! ∂U∂U i ∂U∂U "
i" ∂P ∂P ∂T∂T(v)
ij
(v) ∂u∂u i
ρ ρ∂U∂Ui +
i
+U∂U
Uj
∂U i
i = = − ∂P
− ∂P + ∂T
+ ∂T ij
ij
(v) − − ρ⟨u
ρ⟨u j∂u∂ui ⟩i
⟩ (3.13)
(3.13)
ρ ρ ∂t∂t i+ Uj ∂x j ji = − ∂xi+ ∂x ij − ρ⟨u ∂x
j j⟩i (3.13)
j ∂x ∂xi+ ∂x∂xj − ρ⟨u∂x j ∂x
j j
∂t∂t + U∂x j = − ∂x j⟩ (3.13)
j
∂xj ∂xii ∂xjj j
∂xj
where
where the
the remaining
remaining fluctuating
fluctuating product
product term
term hashas been
been moved
moved totothe
the right-hand
right-hand
where
where
side the
the
ofoftheremaining
remaining
equation. fluctuating
fluctuating
Whether product
product
oror not thisterm
term
last has
has
term been
been moved
moved
isiszero like to tothe
the the right-hand
right-hand
other fluctu-
side side
of the the equation.
equation. Whether
Whether or not not thisthis lastlast
term term is zerozero likelike thethe other
other fluctu-
fluctu-
ating
side
atingofterms
the
terms depends
dependson
equation. onthe
Whether thecorrelation
or not this
correlation ofofterms
last
terms ininthe
term is product.
the zero
product.like In Ingeneral,
the these
other fluctu-
general, these
ating terms
correlations depends
are on the correlation of terms in the product. In general, these
ating
correlations arenot
terms depends notzero.
on
zero.the correlation of terms in the product. In general, these
correlations
The mass are not zero.
conservation
zero. equation
correlations are not
The mass conservation equationcan canbebesimilarly
similarlydecomposed.
decomposed.InInincompress- incompress-
The
ible The mass
form,mass conservation
substitution
conservation of equation
equations
equation can
3.11
can bebe
intosimilarly
equation
similarly decomposed.
3.4 yields:
decomposed. InInincompress-
incompress-
ible form, substitution of equations 3.11 into equation 3.4 yields:
ible
ibleform,
form,substitution
substitutionofofequationsequations3.11 3.11into intoequation
equation3.4 3.4yields:
yields:
∂(U
∂(U j j++ uju)j )
∂(U + u ) ==0 0 (3.14)
(3.14)
∂(U∂x
j +j uj )= 0
j ∂x
j
(3.14)
44 CHAPTER∂x 3. j =
REYNOLDS 0 AVERAGED EQUATIONS (3.14)
2 ∂xj
j
These
2 are
These areeasily
easilyproven
proven from
fromthe definitions
the ofofboth.
definitions both.
2
These
2
Theseare
areeasily
easilyproven
provenfrom
fromthe
thedefinitions
definitionsofofboth.
both.
of which the average is:
∂Uj
=0 (3.15)
∂xj
It is clear from equation 3.15 that the averaged motion satisfies the same form
of the mass conservation equation as does the instantaneous motion, at least
source: Turbulence
source: Turbulenceforforthe
the21st
21stcentury, W.W.
century, K.K.
George
George
It is clear from equation 3.15 that the averaged motion satisfies the same form
Turbulence:
of the mass conservation equation lecture
as does the instantaneous motion, at least
for incompressible flows. How much simpler the turbulence problem would be
Equations
if the same werefortrueaverage velocitiesUnfortunately, as is easily seen from
for the momentum!
equation 3.13, such is not the case.
Equation 3.15 can be subtracted from equation 3.14 to yield an equation for
the instantaneous motion alone; i.e.,
∂uj
=0 (3.16)
∂xj
Again, like the mean, the form of the original instantaneous equation is seen to
be preserved. The reason, of course, is obvious: the continuity equation is linear.
The momentum equation, on the other hand, is not; hence the difference.
Equation 3.16 can be used to rewrite the last term in equation 3.13 for the
The entire problem of turbulence would be so much easier
mean momentum. Multiplying equation 3.16 by ui and averaging yields:

if the same thing were⟨utrue


i
∂uj
∂x
for
⟩= 0 momentum equation
(3.17)
j

This can be added to ⟨uj ∂ui /∂xj ⟩ to obtain:


∂ui ∂ui ∂uj ∂
⟨uj ⟩ + 0 = ⟨uj ⟩ + ⟨ui ⟩= ⟨ui uj ⟩ (3.18)
∂xj ∂xj ∂xj ∂xj
source:operations
where again the fact that arithmetic and averaging Turbulence for commute
the 21st century,
has W. been
K. George
∂uj
=0 (3.16)
Turbulence: lecture
∂xj
Again, like the mean, the form of the original instantaneous equation is seen to
Equations for reason,
be preserved. The average velocities
of course, is obvious: the continuity equation is linear.
The momentum equation, on the other hand, is not; hence the difference.
Equation 3.16 can be used to rewrite the last term in equation 3.13 for the
mean momentum. Multiplying equation 3.16 by ui and averaging yields:
∂uj
⟨ui ⟩=0 (3.17)
∂xj
This can be added to ⟨uj ∂ui /∂xj ⟩ to obtain:
∂ui ∂ui ∂uj ∂
⟨uj ⟩ + 0 = ⟨uj ⟩ + ⟨ui ⟩= ⟨ui uj ⟩ (3.18)
∂xj ∂xj ∂xj ∂xj
where again the fact that arithmetic and averaging operations commute has been
used.
The equation for the averaged momentum, equation 3.13 can now be rewritten
as: ! " (v)
∂Ui ∂Ui ∂P ∂Tij ∂
ρ + Uj =− + − ρ⟨ui uj ⟩ (3.19)
∂t ∂xj ∂xi ∂xj ∂xj
The last two terms on the right-hand side are both divergence terms and can be
combined; the result is:
! "
∂U ∂U ∂P ∂source:
# Turbulence for the
$ 21st century, W. K. George
used. Turbulence:
Turbulence: lecture
lecture
where again the fact that arithmetic and averaging operations commute has been
where
The again
used.
equation the for
factthe
The equation for the averaged
used.
as:
that averaged
Turbulence:
arithmetic
momentum, lecture
and averaging
momentum, equation operations
3.13 cancommute has been
now be rewritten
equation 3.13 can now be rewritten
Equations
as: The equation
Equations
The equation for
for
for
!
foraverage
∂Uaverage
the
i averaged
velocities
! the averaged" momentum, equation
∂Ui velocities
" momentum,
∂P ∂T
(v)
equation
(v) ∂
3.13 can now be rewritten
3.13 can now be rewritten
Equations
as:
as: for
ρ !
ρ !∂U average
∂U i + Uj ∂Uivelocities
∂t i + Uj∂x " =
ji " = −∂P

∂xi + ∂T
ij
∂P + ∂Tij − ∂ ρ⟨ui uj ⟩
∂x(v)j − ∂x
(v)
j ρ⟨ui uj ⟩
(3.19)
(3.19)
∂t i + U ∂x ∂U ∂xi + ∂T ∂xij j ∂
∂x
ρ ∂U j ∂U ji = − ∂P
ij − ∂ jρ⟨ui uj ⟩ (3.19)
The last two terms on ρ ∂t + U
the right-hand
j∂xj = − side
∂x +
are∂x both − ρ⟨u
divergence
∂x u ⟩
i j terms and can (3.19)be
The last two terms on ∂t the right-hand
∂xj side
∂xii are both
∂xj j divergence
∂xjj terms and can be
combined;
The last twotheterms
resulton is:the right-hand side are both divergence terms and can be
combined;
The last two theterms
resulton is:the right-hand side are both divergence terms and can be
! "
combined; the result !∂Uis: ∂U " ∂P ∂ # $
combined; the result ρ ! ∂Uis:
i i
i + Uj ∂Ui" = − ∂P + ∂ Tij(v) − ρ⟨ui uj ⟩
# (v) $
(3.20)
ρ !∂U ∂t i + U j∂x
∂U ji " = − ∂xi ∂x
∂P + ∂ T
j ## (v) ij − ρ⟨u u
i j$⟩ (3.20)
ρ ∂U ∂t i + U ∂x ∂U ji = − ∂P ∂xi + ∂x ∂ j T (v) − ρ⟨u u ⟩ $ (3.20)
j ij − ρ⟨ui uj ⟩
Now the terms in square ρ +
∂t bracketsU j∂xj =
on −
the +
∂xiright∂xhave
j T the
ij i j
dimensions of (3.20)
stress. The
∂t ∂x ∂x ∂x
Now the terms in square brackets on the right have the dimensions of stress. The
j i j
first
Nowterm is, in fact,
the terms in the viscous
square brackets stress. The
on the second
right haveterm, on the other
the dimensions hand, isThe
of stress. not
first
Now term
the is,
termsin fact,
in the
square viscous
brackets stress.
on The
the second
right have term,
the on the
dimensions other
of hand,
stress. is not
The
afirst
stress
termat is,
all,inbutfact,simply
the a re-worked
viscous stress. version
The of theterm,
second fluctuating
on the contribution
other hand, tonot
is the
afirst
stress at is,
term all,inbut
fact,simply
the a re-worked
viscous stress.version
The of theterm,
second fluctuating
on the contribution
other hand, tonot
is the
non-linear
a stress at acceleration
all, but simply terms.
a The fact
re-worked that of
version it the
canfluctuating
be writtencontribution
this way, however,
to the
non-linear
a stress at acceleration
all,atbut simply terms. The
a re-worked fact that
version it can
of the be written
fluctuating this way,
contribution however,
to theit
indicates that
non-linearthat least
acceleration as far as the mean motion is concerned, it acts as though
indicates
non-linear at least asterms.
acceleration far asThe
terms. Thethe fact
mean
fact
that
that
it can
motion
it can is be
be
written this
concerned,
written this
way,
it acts as however,
way, though it
however,
were a stress
indicates that —athence
least its far
as name, as themean
the Reynolds
motion stress.
is In the it
concerned, succeeding
acts as sections
though it
were a
indicatesstress —
that at of hence
least its name,
asdifference the
far as the will Reynolds
mean motion stress. In the succeeding
is concerned, it acts as though it sections
the consequences
wereconsequences
a stress — hence this its difference
name, the will be
Reynolds examined.
stress. In the succeeding sections
the of this be examined.
were a stress — hence its name, the Reynolds stress. In the succeeding sections
the consequences of this difference will be examined.
the consequences of this difference will be examined.

source: Turbulence for the 21st century, W. K. George


source: Turbulence for the 21st century, W. K. George
Wednesday, 23 October 13
Wednesday, 23 October 13 source: Turbulence for the 21st century, W. K. George
! "
Turbulence: lecture
(v)
∂Tij
= − Turbulence:
− lecture
∂Ui ∂Ui ∂P ∂
ρ + Uj + ρ⟨ui uj ⟩ (3.19)
∂t ∂xj ∂xi ∂xj ∂xj
stThe problem
twoThe
terms with turbulence
problem with turbulence
on the right-hand side are both divergence terms and can be
ed; the result is:
! "
∂Ui ∂Ui ∂P ∂ # (v) $
ρ + Uj =− + Tij − ρ⟨ui uj ⟩ (3.20)
∂t ∂xj ∂xi ∂xj
e terms in square brackets on the right have the dimensions of stress. The
Appearance
Appearance of Reynolds
of Reynolds stress
stressis the
is problem
the problem
m is, in fact, the viscous stress. The second term, on the other hand, is not
at all, but simply a re-worked version of the fluctuating contribution to the
They
ear They areare
accelerationcreated
created
terms. byfact
The thethat
by flow
the canthat
itflow wewe
bethat
writtenwant tohowever,
thiswant
way, study/
to study/
understand/predict
es in
that at least as far as the meanthe first
motion
understand/predict in the first place! place!
is concerned, it acts as though it
stress — hence its name, the Reynolds stress. In the succeeding sections

•The
sequences of this difference will be examined.
• equations
The are
equations not
are closed
not closed
•Simple
• models
Simple modelsdo not
do work
not (more
work (morelater)
later)
•Compiling tables/handbooks carry substantial risk
•Compiling tables/handbooks carry substantial risk

source: Turbulence for the 21st century, W. K. George


source: Turbulence for the 21st century, W. K. George
Turbulence: lecture
(v)
∂Ui ∂Ui ∂P ∂Tij ∂
ρ
∂t
+ Uj
∂xj
Turbulence:
=−
∂xi
+
∂xj
− lecture
∂xj
ρ⟨ui uj ⟩ (3.19)

Closure
ast two terms onand
Closure and eddy
eddy viscosity
the right-hand
viscosity
side are both divergence terms and can be
ned; the result is:
! "
∂Ui ∂Ui ∂P ∂ # (v) $
ρ + Uj =− + Tij − ρ⟨ui uj ⟩ (3.20)
∂t ∂xj ∂xi ∂xj
the terms in square brackets on the right have the dimensions of stress. The
There
erm is, in fact, are 6stress.
the viscous unknowns
The second forterm,
theon3 the
equations
other hand, is not
ss at all, but simply a2re-worked
2 version2 of the fluctuating contribution to the
near acceleration hu 1 i, hu
terms. 2 i,fact
The hu3that
i, huit1can
u2 i,behu 1 u3 i&hu
written this 2 u3 ihowever,
way,
ates that at least as far as the mean motion is concerned, it acts as though it
These
a These
stress —have to be related to the mean
have to be related to the mean motion in order to
hence its name, the Reynolds stress. In motion
the in
succeeding order
sections to
onsequences be able
of this to solve
difference will befor the mean motions
examined.
be able to solve for the mean motions
The fact that we do not have these relations is referred to
The fact that we do not have these relations is referred to
asasthe closure problem
the closure problem

source: Turbulence for the 21st century, W. K. George


dnesday, 23 October 13
source: Turbulence for the 21st century, W. K. George
equation reduces to:
Turbulence: ∂ ρ̃lecture
Turbulence:
Dρ̃
Dt
=
∂t
∂ ρ̃
∂x
lecture
+ ũ =0
j
j
(3.3)
Closure
equation and
3.2 it eddy
Closure and eddy viscosity
From follows viscosity
that for incompressible flows,
∂ ũj
=0 (3.4)
∂xj
Analogy
Analogy with
with (the stress minus the mean normal stress) are represented
The viscous stresses
viscous
viscous
by the
stress
tensor T̃stress
(v) (v)
. From its definition, T̃ = 0. In many flows of interest, the
ij kk
fluid behaves as a Newtonian fluid in which the viscous stress can be related to
theWe wrote
motion down
by a a constitutive
We wrote down a constitutive relationship,
fluid constitutive relation of relationship,
the form
! "
(v) 1
T̃ij = 2µ s̃ij − s̃kk δij (3.5)
3
The viscosity, µ, is a property of the fluid that can be measured in an independent
experiment. s̃ij is the instantaneous strain rate tensor defined by
# $
1 ∂ ũi ∂ ũj
s̃ij ≡ + (3.6)
2 ∂xj ∂xi

The
Fromviscosity
The only
its definition,
viscosity
s̃kkdepends
= ∂ ũk /∂xkon
only depends
. If the fluid
the flow and not on
is incompressible,its
s̃kk motions
on thetofluid and not on its motions
= 0 and the
Newtonian constitutive equation reduces
(v) source: Turbulence for the 21st century, W. K. George
T̃ = 2µs̃ (3.7)
pressure. Someone somewhere Turbulence: lecture
Turbulence: lecture
else has already compiled this information from
independent experiments.
It is tempting to try such an approach for the turbulence Reynolds stresses
Closure and eddy viscosity
(even though we know the underlying requirements of scale separation are not
satisfied). For example, a Newtonian type closure for the Reynolds stresses, often
We can
referred to aswrite a similar
an “eddy” equations
or “turbulent” viscosityfor thelooks
model, Reynolds
like: stress
! "
1 1
−ρ⟨ui uj ⟩ + ⟨ui ui ⟩ = µt Sij − Skk δij (3.23)
3 3
where µt is the turbulence “viscosity” (also called the eddy viscosity), and Sij is
the mean strain rate defined by:
# $
1 ∂Ui ∂Uj
Sij = + (3.24)
2 ∂xj ∂xi

The second term vanishes identically for incompressible flow. For the simple
The second
case of term onshear
a two-dimensional theflow,
right hand3.23side
equation is zero
for the (continuity)
Reynolds shear stress
reduces to
∂U1
−ρ⟨u1 u2 ⟩ = µt (3.25)
∂x2
Note this “model” is the direct analogy to the Newtonian model for viscous
(v)
stress in a fluid. The Reynolds stresses, ⟨−ui ujsource:
⟩ replaces theforviscous
source:Turbulence
Turbulence for the
the 21st
stress,
21st century,
century,W.
τijGeorge
W.K.
.
K.George
Turbulence:
Turbulence:
#
1 ∂U lecture
lecture
∂U i
$
j
Sij = + (3.24)
2 ∂xj ∂xi
Closure
Closure and
and eddy
eddy viscosity
viscosity
The second term vanishes identically for incompressible flow. For the simple
case For
For
of aa simple
simple case,
case,
a two-dimensional 2D
2D
shear shear
shear
flow, flow
flow
equation 3.23 for the Reynolds shear stress
reduces to
∂U1
−ρ⟨u1 u2 ⟩ = µt (3.25)
∂x2
Note this “model” is the direct analogy to the Newtonian model for viscous
(v)
stress in a fluid. The Reynolds stresses, ⟨−ui uj ⟩ replaces the viscous stress, τij .
The counterpart to the mechanical pressure is the mean normal Reynolds stress,
Direct
⟨ui ui ⟩/3. Andanalogy
Direct like it’s to
analogy to the
fluidthe Newtonian
Newtonian
counterpart model
model
it, the for
for viscous
Reynolds viscous
stress canstress
stress
depend only
on the mean strain rate at a single instant and single location in the flow, so has
The
The counterpart
counterpart to
to the
the mechanical
mechanical pressure
pressure isis the
the the
the normal
normal stress
stress
Reynolds
Reynolds stress
stress depends
depends on
on mean
mean strain
strain at
at aa single
single location
location /instant
/instant
in
in the
the flow
flow and
and has
has no
no history
history or
or non-local
non-local dependence
dependence
This
This isis fatal!!!
fatal!!!
source:
source:Turbulence
Turbulencefor
forthe
the21st
21stcentury,
century,W.
W.K.
K.George
George
Turbulence: lecture
Turbulence: lecture
Closure
Closure and
Turbulence:
and eddy
eddy viscosity
viscosity
lecture
Closure and eddy viscosity
52
Example,
Example, buoyant
buoyant plume
plume 3.6. THE EDDY VISCOSITY
CHAPTER 3. REYNOLDS AVERAGED EQUATIONS
Example, buoyant plume 3.6. THE EDDY VISCOSITY
52 Want
Want
CHAPTER to
to“predict”
3. REYNOLDS AVERAGED the
“predict” the mean
mean flow
EQUATIONS flow at
at
Want to “predict”
different
different the mean
downstream
downstream flow at
locations
locations
different downstream locations

Figure 3.1: Schematic of axisymmetric plume


source:
source:Turbulence
Figure 3.1: Schematic of axisymmetric plume Turbulencefor
forthe
the21st
21stcentury,
century,W.
W.K.
K.George
George
Closure and eddy viscosity
Turbulence:3.6.lecture
52
Example, buoyant plume THE EDDY VISCOSITY
CHAPTER 3. REYNOLDS AVERAGED EQUATIONS
Closure and eddy viscosity
Want to “predict” the mean flow a
Example, buoyant plume different downstream locations
Model generated from
measurements of Reynolds
stresses
Only good for this
particular set of conditions

As good as predicting Figure 3.1: Schematic of axisymmetric plume


yesterday’s weather! source: Turbulence for the 21st century, W.
no history or non-local dependence. This absence will turnFigure
out to be3.2: Mean
fatal velocity profile
in most
Wednesday, 23 October 13
applications. Moreover, unlike like the viscosity, µ, which depends only on the
fluid and not the motion itself, the “turbulence viscosity”, µt , depends entirely on
the motion.
source: Turbulence for the 21st century, W. K. George
Turbulence: lecture
Turbulence: lecture
Closure and eddy viscosity
3.6. THE
3.6. THE EDDY
EDDY VISCOSITY
VISCOSITY 55
55

Example, buoyant plume Figure 3.4: Reynolds shear stress profiles for axisym
Does not really work for other quantities of interest
such as vertical components of Reynolds stress

source:
source: Turbulence
Turbulence
Figure for the
for
3.5: Radial the 21st century,
21st
turbulentcentury, W.K.
W.
heat flux K.George
for George
axisymm
direction. In other words, the turbulence for which a constitutive equation is being
Turbulence:
written is not an isotropic “medium”. In fact, inlecture
this specific example the problem
is that the vertical component of the heat flux is produced more by the interaction
Closure
of buoyancy and and eddy viscosity
the turbulence, than it is by the working of turbulence against
mean gradients in the flow. We will discuss this in more detail in the next chapter
Example, buoyant plume
when we consider the turbulence energy balances, but note for now that simple
gradient closure models never work unless gradient production dominates. This
This
rules means
out many flowsthat turbulence
involving is not
buoyancy, and an isotropic
also many medium
involving recirculations
or separation where the local turbulence is convected in from somewhere else.
A more general form of constitutive equation which would allow for the non-
isotropic nature of the “medium” (in this case the turbulence itself) would be
! "
1 1
−ρ⟨ui uj ⟩ + ⟨uk uk ⟩δij = µijkl Skl − Smm δkl (3.26)
3 3
This closure relation allows each component of the Reynolds stress to have its own
unique We
value have traded
of the eddy 6 unknowns
viscosity. It is easy to seefor
that81 it isunknowns!!!
unlikely this will solve
the closure problem since the original six unknowns, the ⟨ui uj ⟩, have been traded
for eighty-one new ones, Closure problem
µijkl . Even if some remains!
can be removed by symmetries,
the remaining number is still formidable. More important than the number of
There are
unknowns, however, nothere
is that general
is no solutions
independentto or this problem
general means for selecting
Flow
them is turbulent
without consideringanda the turbulence
particular depends
flow. This is because onturbulence
circumstances
is indeed
a property of the flow, not of the fluid.
source: Turbulence for the 21st century, W. K. George
3.7 Turbulence:
The Reynolds lecture
Stress Equations
Reynolds
It is clear from thestress equations
preceding section that the simple idea of an eddy viscosity
might not be the best way to approach the problem of relating the Reynolds
It is
stress clear
to the that
mean eddy
motion. viscosity
An alternative approach
approach is is
to flawed.
try to derive dynamical
equations for the Reynolds stresses from the equations governing the fluctuations
themselves. Such an approach recognizes that the Reynolds stress is really a
So, we 6need
functional of the dynamical
velocity; that equations
is, the stress atfor the depends
a point Reynolds stresses
on the velocity
that recognises
everywhere and for allthat
past these
times, notstresses depend
just at the point inon velocities
question and at a
particular instant in time.
everywhere now with
The analysis begins andthe in equation
past timesfor the instantaneous fluctuating ve-
locity, equation 3.21. This can be rewritten for a Newtonian fluid with constant
Equation
viscosity as: for instantaneous fluctuating velocity,
# $ (v) # $ % &
∂ui ∂ui ∂p ∂τij
∂Ui ∂ui ∂ui
ρ + Uj =− + − ρ uj − ρ uj − ⟨uj ⟩ (3.27)
∂t ∂xj ∂xi ∂xj ∂xj ∂xj ∂xj

Note that the free index in this equation is i. Also, since we are now talking about
turbulence again, the capital letters represent mean or averaged quantities.
6
Take ANS equation is
functional anda subtract
functionmean momentum equation
of a function

source: Turbulence for the 21st century, W. K. George


Turbulence: lecture
. THE REYNOLDS STRESS EQUATIONS 57
7. THE Reynolds
REYNOLDSstress equations
STRESS EQUATIONS
3.7. THE REYNOLDS STRESS EQUATIONS
3.7. THE REYNOLDS STRESS EQUATIONS 57
57
57
Multiplying
Multiplying equation 3.27 by uukk and averaging yields:
Multiplying equation 3.27 by uk and averaging yields:
equation 3.27 by and averaging yields:
Multiplying equation 3.27 by uk and averaging yields:
!! "" (v)
!
∂u
! ∂u "" ∂p ∂τ (v)
(v)
ρρ ⟨u ∂ui i⟩∂u
+
∂u U ⟨u ∂ui∂u i⟩ = − ⟨u ∂p
∂p ⟩ + ⟨u ∂τ ij
∂τ
(v)
⟩⟩⟩ (3.28)
⟨uρkρk ∂t
⟨u ⟩ +i

i U+ Ukk⟨u ∂u
j j ⟨u ⟩ i

i = = − − ⟨u
⟨u
kk ∂p ⟩⟩ ++ ⟨u
⟨ukk ij
ij (3.28)
(3.28)
⟨uk ⟩ + Uj ∂x⟨u
k ⟩ = − ∂x
⟨uk ⟩ + k ∂x
⟨u ⟩ (3.28)
∂t ∂t
k
∂t
j ∂xj∂x
k
j∂xj j ! ∂x
k∂x
i
∂xiii "
k ∂x
∂x
j
#jjj $$
! !! """ #
## $$
∂U∂U
∂U i ii ∂u
∂u
∂ui
−ρ
−ρ ⟨uk⟨u ∂U
u j⟩uj⟩⟩⟩ i −− ρρρρ ⟨u ∂ui i
uuujjj i⟩⟩⟩⟩
kkku
−ρ −ρ ⟨u⟨u
k kujku ∂x
j ∂x
−− ⟨u
⟨u
⟨u k j∂xj
∂xj jj
∂x j ∂xj j

ow Now
Nowsince
w since
since bothboth
both
since iiand
both
and iiand
kand
k kkare
are
arefree
arefree
freefreeindices
indices
indices
indices they
theythey
they can
can be
can
can be
bebe interchanged
interchanged
interchanged
interchanged to
to
to
toyield
yield
yield
yieldaaaasecond
second
second
second
77by 7:7
equation
ation givengiven
by
equation
uation by : :by :
givengiven
!! !! "" "" (v)
∂u ∂u ∂uk ∂p ∂τ ∂τ (v)
∂τ
(v)
(v)
kik ∂uk⟩ + Uj∂u ∂u ∂p ∂p
k kj
ρ ⟨u
∂u ⟨u
∂u k ⟩
k = − ⟨u ∂p ⟩ + ⟨u ∂τ kj kj ⟩ (3.29)
ρρ ⟨uiρi ∂t⟨u⟩i⟩∂t
⟨u ++UU⟩j+
⟨u U j ⟨ui∂x
⟩ = ⟩ =− −⟨u ⟨ui∂x⟩k⟩++
⟩+⟨u ⟨u i ∂x ⟩⟩⟩ (3.29)
(3.29)
i k i i kj
∂t j ⟨u
i i∂x ∂xj ⟩ j = − ⟨ui i∂x∂xk ⟨ui i ∂x ∂x j (3.29)
∂t ∂xj j !∂xkk " ∂x j j j# $$
! ! ! " " ##
∂U k" # ∂u k $
$
−− ρρ ⟨u u ∂U
⟩ ∂U k − ρ ⟨u u ∂u∂u k⟩
i⟩uj∂U ∂uk⟩⟩
⟨u k
⟩∂xkj −−ρ ρ ⟨u⟨u k
iu
i j i j
−− ρρ ⟨u u
⟨ui uj ⟩ ∂x − ρ ⟨ui uj ∂xj ⟩
i j i u j j∂x
∂x
∂xj
j
j ∂x
∂xj
j
j

Equations
Equations 3.28
3.28 and
and 3.29
3.29 can
can be
be added
added together
together to
to yield
yield an
an equation
equation for
for the
the
source: Turbulence for the 21st century, W. K. George
Equations 3.28 and 3.29 can be added together to yield an equation for the
ρ ⟨ui ⟩ + Uj ⟨ui ⟩ = − ⟨ui ⟩ + ⟨ui ⟩ (3.29)
∂U ∂uk
∂t ∂xj
Turbulence:
− ρ ⟨ui uj!⟩ lecture
∂x k k
− "ρ ⟨u#i uj
∂xj
⟩ $
∂xj ∂Uk ∂x∂u
jk
− ρ ⟨ui uj ⟩ − ρ ⟨ui uj ⟩
Reynolds stress equations ∂xj ∂xj
ons 3.28 and 3.29 can be added together to yield an equation for the
stress,Equations 3.28 and 3.29 can be added together to yield an equation for the
Reynolds stress,
! !
" "
∂⟨ui uk ⟩ ∂⟨ui∂⟨u
uk ⟩ i uk ⟩ ∂⟨ui uk ⟩ 1 1 ∂p ∂p ∂p ∂p
+ Uj + Uj = − = ⟨u −i ⟨ui⟩ + ⟨u ⟩ +k ⟨uk ⟩ ⟩
∂t ∂t ∂xj ∂xj ρ ρ∂xk ∂xk ∂xi∂xi
! ! ""
∂u k
∂uk ∂u ∂u
i i
− ⟨u−i uj ⟨ui uj⟩∂x +j⟨u⟩ +u ⟨uk uj ⟩ ⟩
k j ∂xj
(3.30)
(3.30)
∂x
⎡j ∂xj ⎤
⎡ 1 ∂τ
(v)
∂τ
(v)⎤
⎣ (v) kj (v)ij ⎦
1 ⎣+ ρ∂τ⟨u kj i ∂x ⟩ + ⟨u ∂τkij∂x ⟩
+ ⟨ui ⟩ +j ⟨uk ⟩j⎦
ρ ! ∂xj ∂xj "
! − ∂Uk ∂U i
⟨ui uj ⟩ + ⟨uk uj ⟩ "
∂Uk ∂xj ∂U∂xi j
− ⟨ui uj ⟩ + ⟨uk uj ⟩
∂x j ∂x j
It is customary to rearrange the first term on the right hand side in the fol-
lowing way:
ustomary to rearrange the first" term on !the right hand side in the fol-
y:
Let’s
! look
∂p
at this
∂p
equation term-by-term
∂ui ∂uk
"
⟨u ⟩ + ⟨u
i ⟩ = k ⟨p + ⟩ (3.31)
! ∂xk " ∂xi ! ∂xk ∂x
" i for the 21st century, W. K. George
source: Turbulence
ρ ⟨ui ⟩ + Uj ⟨ui ⟩ = − ⟨ui ⟩ + ⟨ui ⟩ (3.29)
∂U ∂uk
∂t ∂xj
Turbulence:
− ρ ⟨ui uj!⟩ lecture
∂x k k
− "ρ ⟨u#i uj
∂xj
⟩ $
∂xj ∂Uk ∂x∂u
jk
− ρ ⟨ui uj ⟩ − ρ ⟨ui uj ⟩
Reynolds stress equations ∂xj ∂xj
ons 3.28 and 3.29 can be added together to yield an equation for the
stress,Equations 3.28 and 3.29 can be added together to yield an equation for the
Reynolds stress,
! !
" "
∂⟨ui uk ⟩ ∂⟨ui∂⟨u
uk ⟩ i uk ⟩ ∂⟨ui uk ⟩ 1 1 ∂p ∂p ∂p ∂p
+ Uj + Uj = − = ⟨u −i ⟨ui⟩ + ⟨u ⟩ +k ⟨uk ⟩ ⟩
∂t ∂t ∂xj ∂xj ρ ρ∂xk ∂xk ∂xi∂xi
! ! ""
∂u k
∂uk ∂u ∂u
i i
− ⟨u−i uj ⟨ui uj⟩∂x +j⟨u⟩ +u ⟨uk uj ⟩ ⟩
k j ∂xj
(3.30)
(3.30)
∂x
⎡j ∂xj ⎤
⎡ 1 ∂τ
(v)
∂τ
(v)⎤
⎣ (v) kj (v)ij ⎦
1 ⎣+ ρ∂τ⟨u kj i ∂x ⟩ + ⟨u ∂τkij∂x ⟩
+ ⟨ui ⟩ +j ⟨uk ⟩j⎦
ρ ! ∂xj ∂xj "
! − ∂Uk ∂U i
⟨ui uj ⟩ + ⟨uk uj ⟩ "
∂Uk ∂xj ∂U∂xi j
− ⟨ui uj ⟩ + ⟨uk uj ⟩
∂x j ∂x j
It is customary to rearrange the first term on the right hand side in the fol-
lowing way:
ustomary to rearrange the first" term on !the right hand side in the fol-
y:
Let’s
! look
∂p
at this
∂p
equation term-by-term
∂ui ∂uk
"
⟨u ⟩ + ⟨u
i ⟩ = k ⟨p + ⟩ (3.31)
! ∂xk " ∂xi ! ∂xk ∂x
" i for the 21st century, W. K. George
source: Turbulence
+ ⎣⟨ui ⟩ + ⟨uk ⟩⎦
Turbulence:
Turbulence:
ρ
!
lecture
lecture
∂x j ∂xj
"
∂Uk ∂Ui
Reynolds stress equations
− ⟨ui uj ⟩
Reynolds stress equations ∂xj
+ ⟨uk uj ⟩
∂xj

It is customary to rearrange the first term on the right hand side in the fol-
lowing way:
! " ! "
∂p ∂p ∂ui ∂uk
⟨ui ⟩ + ⟨uk ⟩ = ⟨p + ⟩ (3.31)
∂xk ∂xi ∂xk ∂xi

+ [⟨pui ⟩δkj + ⟨puk ⟩δij ]
∂xj

The The first


first term term
on the righton the right
is generally is
referred pressure-strain
to term
as the pressure strain-rate term.
The first
The second term
term is onas the
written right term,
a divergence is pressure-strain termto as
and is generally referred
the pressure diffusion term. We shall see later that divergence terms can never
create The
The second
second
nor destroy term,
term,
anything; written
written
they in
can simply divergence
in move
divergence form,
form,
it around from one is
is to
place
another. pressure-diffusion term
pressure-diffusion term
7
Alternatively equation 3.21 can be rewritten with free index k, then multiplied by ui and
averaged

source: Turbulence for the 21st century, W. K. George


source: Turbulence for the 21st century, W. K. George
Wednesday, 23 October 13
ρ ⟨ui ⟩ + Uj ⟨ui ⟩ = − ⟨ui ⟩ + ⟨ui ⟩ (3.29)
∂U ∂uk
∂t ∂xj
Turbulence:
− ρ ⟨ui uj!⟩ lecture
∂x k k
− "ρ ⟨u#i uj
∂xj
⟩ $
∂xj ∂Uk ∂x∂u
jk
− ρ ⟨ui uj ⟩ − ρ ⟨ui uj ⟩
Reynolds stress equations ∂xj ∂xj
ons 3.28 and 3.29 can be added together to yield an equation for the
stress,Equations 3.28 and 3.29 can be added together to yield an equation for the
Reynolds stress,
! !
" "
∂⟨ui uk ⟩ ∂⟨ui∂⟨u
uk ⟩ i uk ⟩ ∂⟨ui uk ⟩ 1 1 ∂p ∂p ∂p ∂p
+ Uj + Uj = − = ⟨u −i ⟨ui⟩ + ⟨u ⟩ +k ⟨uk ⟩ ⟩
∂t ∂t ∂xj ∂xj ρ ρ∂xk ∂xk ∂xi∂xi
! ! ""
∂u k
∂uk ∂u ∂u
i i
− ⟨u−i uj ⟨ui uj⟩∂x +j⟨u⟩ +u ⟨uk uj ⟩ ⟩
k j ∂xj
(3.30)
(3.30)
∂x
⎡j ∂xj ⎤
⎡ 1 ∂τ
(v)
∂τ
(v)⎤
⎣ (v) kj (v)ij ⎦
1 ⎣+ ρ∂τ⟨u kj i ∂x ⟩ + ⟨u ∂τkij∂x ⟩
+ ⟨ui ⟩ +j ⟨uk ⟩j⎦
ρ ! ∂xj ∂xj "
! − ∂Uk ∂U i
⟨ui uj ⟩ + ⟨uk uj ⟩ "
∂Uk ∂xj ∂U∂xi j
− ⟨ui uj ⟩ + ⟨uk uj ⟩
∂x j ∂x j
It is customary to rearrange the first term on the right hand side in the fol-
lowing way:
ustomary to rearrange the first" term on !the right hand side in the fol-
y:
Let’s
! look
∂p
at this
∂p
equation term-by-term
∂ui ∂uk
"
⟨u ⟩ + ⟨u
i ⟩ = k ⟨p + ⟩ (3.31)
! ∂xk " ∂xi ! ∂xk ∂x
" i for the 21st century, W. K. George
source: Turbulence
Turbulence:
Turbulence: lecture
lecture
Turbulence: lecture
58 CHAPTER 3. REYNOLDS AVERAGED EQUATIONS
58 Reynolds stress
Reynolds CHAPTERequations
Reynolds stress equations
CHAPTER 3.
stress equations
CHAPTER 3. REYNOLDS
REYNOLDS AVERAGED
AVERAGED EQUATIONS
EQUATIONS
3. REYNOLDS AVERAGED EQUATIONS
The third term on the right-hand side of equation 3.30 can similarly be re-
The third term on the right-hand side of equation 3.30 can similarly be re-
hewritten
third term
as: on the right-hand side of equation 3.30 can similarly be re-
hewritten
third term
as: on the right-hand side of equation 3.30 can similarly be re-
en as: ⎡ ⎤ % &
en as: ⎡ ∂τ (v)
(v)
(v)
∂τ⎤ij(v) ⎤ % (v) ∂uk ∂u &
⎡ ⎣(v)
⟨u kj
∂τkj ⟩ + ⟨u(v) ∂τ⎤ij ⟩⎦ = −
% ⟨τ ∂u ⟩ + ⟨τ
(v)
∂u i
&⟩ (3.32)
⎡ i k (v) k (v) i
∂τ⎣(v)
⟨ui ∂x ⟩ +
j ∂τ⟨u(v)
k ∂x ⟩ = −
j
⎦ % ij
⟨τ∂u ∂x ⟩ + ⟨τ ∂u∂x &⟩
kj
(v) ij k j (v)kj i j (3.32)
⎣⟨ui ∂τkj ∂x ∂τij ∂x⎦ ∂u ∂x (v) ∂u ∂x
⎣⟨ui ∂xkj ⟩⟩ + ⟨u
j k ij ⟩ ⎦ j= −
+ ⟨uk ∂x ⟩ = −
⟨τij(v) k ⟩ +
⟨τij∂ ∂xj ⟩ (v)
j ⟨τkj
+ ⟨τkj ∂xj(v)⟩
i ⟩j (3.32)
(3.32)
∂xjj ∂xjj + ∂ ∂x
[⟨uji τkj(v)⟩ + ⟨u∂xk τijj(v)⟩]
+∂ ∂xj [⟨ui τkj ⟩ + ⟨uk τij ⟩]
+ ∂ ∂x (v)
[⟨ui τkj(v)⟩ + ⟨uk τij(v)⟩]
j (v)

The first of these The The second


second term
term + is
is aa
∂x divergence
[⟨uai τNewtonian
divergence term
⟩ + ⟨u
term ij ⟩] the last is the
k τfluid,
The first of these The is alsosecond termterm. is ∂x
a divergence term
is also a divergence term. For
j kj
a divergence jFor a Newtonian fluid, the last is the
so-called “dissipation
so-called
first of these“dissipation
is alsoThe
The
The a first
of
first
first
of term
Reynolds
term
term
Reynolds
divergence on
stress”
on
on
stress”
term. the
the
the
byright
byright
For right
the
a hand
the turbulence
hand
hand
turbulence
Newtonian side
side
side is,
viscous
is,
is,
viscous
fluid,
stresses. This
thestresses.
last is This
the
is
first easily seen
of “Dissipation
these by
is by
alsosubstituting
aofdivergence the Newtonian
term. Forby constitutive
aconstitutive
Newtonian relation
fluid, to to obtain:
theobtain:
last is the
is “Dissipation
easily seen
“Dissipation
led “dissipation of
of
of Reynolds Reynolds
Reynolds
substituting
Reynolds the stress”
stress”
by the%by
Newtonian
stress” stress” by turbulent
turbulent
turbulent
turbulence viscous
viscous
relation
viscous
viscous stress
stress
stress
stresses. This
lled “dissipation % &
of Reynolds stress”& by the% turbulence viscous& stresses. This &
%
ily seen byIfIfsubstituting
we
1 substitute
we
ily seen byIfsubstituting
we
1 ⟨τ
(v) ∂uk the the
substitute
substitute
ij(v) ∂u k the the
⟩ + ⟨τ
(v) ∂u
the ∂uNewtonian
Newtonian
Newtonian
i
(v) Newtonian
Newtonian
kj i ⟩ = 2ν ⟨s constitutive
∂uk relation
constitutive
constitutive
ijconstitutive
∂u
constitutive k ⟩ + ⟨s
relation
kj
∂urelation,
∂u i to obtain:
relation,
relation,
i ⟩ to obtain:(3.33)
% ρ ⟨τij ∂xj ⟩ + ⟨τkj ∂x & j ⟩ =%2ν ⟨sij ∂xj ⟩ + ⟨skj ∂xj&⟩ (3.33)
1 % (v)ρ ∂uk ∂xj (v) ∂ui ∂x & j % ∂uk∂xj ∂u∂x i
j&
1 ⟨τatij(v)all∂uobvious
It is not k ⟩ + ⟨τkj (v) ∂uthis
what i⟩ = has 2νto ⟨sdo ∂uk ⟩dissipation,
ij with + ⟨skj ∂ui ⟩but it will become (3.33)
It is ρnot ⟨τatij all j ⟩ + ⟨τkj
∂xobvious what∂xjthis⟩ = has 2νto ⟨s doij ∂x
withj ⟩dissipation,
+ ⟨skj ∂xj ⟩but it will become (3.33)
clear ρlater on ∂x whenj ∂xj the trace of the
we consider ∂xjReynolds∂x stress
j equation, which
clear later on when we consider the trace of the Reynolds stress equation, which
notis
is
theall
at
the
kinetic
obvious
kinetic
energy
energy whatequation
this
equation
for the
has
for to
the
turbulence.
do with source:
turbulence. dissipation,
source:Turbulence
Turbulence forbut
for the
the
it century,
21st
21st
will W.
century,
become
W. K.
K. George
not at all obvious what this has to do with source: dissipation,
Turbulence forbut it century,
the 21st will W. K. George
becomeGeorge
ρ ⟨ui ⟩ + Uj ⟨ui ⟩ = − ⟨ui ⟩ + ⟨ui ⟩ (3.29)
∂U ∂uk
∂t ∂xj
Turbulence:
− ρ ⟨ui uj!⟩ lecture
∂x k k
− "ρ ⟨u#i uj
∂xj
⟩ $
∂xj ∂Uk ∂x∂u
jk
− ρ ⟨ui uj ⟩ − ρ ⟨ui uj ⟩
Reynolds stress equations ∂xj ∂xj
ons 3.28 and 3.29 can be added together to yield an equation for the
stress,Equations 3.28 and 3.29 can be added together to yield an equation for the
Reynolds stress,
! !
" "
∂⟨ui uk ⟩ ∂⟨ui∂⟨u
uk ⟩ i uk ⟩ ∂⟨ui uk ⟩ 1 1 ∂p ∂p ∂p ∂p
+ Uj + Uj = − = ⟨u −i ⟨ui⟩ + ⟨u ⟩ +k ⟨uk ⟩ ⟩
∂t ∂t ∂xj ∂xj ρ ρ∂xk ∂xk ∂xi∂xi
! ! ""
∂u k
∂uk ∂u ∂u
i i
− ⟨u−i uj ⟨ui uj⟩∂x +j⟨u⟩ +u ⟨uk uj ⟩ ⟩
k j ∂xj
(3.30)
(3.30)
∂x
⎡j ∂xj ⎤
⎡ 1 ∂τ
(v)
∂τ
(v)⎤
⎣ (v) kj (v)ij ⎦
1 ⎣+ ρ∂τ⟨u kj i ∂x ⟩ + ⟨u ∂τkij∂x ⟩
+ ⟨ui ⟩ +j ⟨uk ⟩j⎦
ρ ! ∂xj ∂xj "
! − ∂Uk ∂U i
⟨ui uj ⟩ + ⟨uk uj ⟩ "
∂Uk ∂xj ∂U∂xi j
− ⟨ui uj ⟩ + ⟨uk uj ⟩
∂x j ∂x j
It is customary to rearrange the first term on the right hand side in the fol-
lowing way:
ustomary to rearrange the first" term on !the right hand side in the fol-
y:
Let’s
! look
∂p
at this
∂p
equation term-by-term
∂ui ∂uk
"
⟨u ⟩ + ⟨u
i ⟩ = k ⟨p + ⟩ (3.31)
! ∂xk " ∂xi ! ∂xk ∂x
" i for the 21st century, W. K. George
source: Turbulence
(v) k (v) i k i
⟨τij ⟩+ ⟨τkj ⟩ = 2ν ⟨sij ⟩ + ⟨skj ⟩ (3.33)
ρ ∂xj Turbulence:
∂xj lecture
∂xj ∂xj

It is not at all obvious what this has to do with dissipation, but it will become
Reynolds
clear later on whenstress equations
we consider the trace of the Reynolds stress equation, which
is the kinetic energy equation for the turbulence.
Now if we use the same trick from before using the continuity equation, we
can rewrite the second term on the right-hand side of equation 3.30 to obtain:
% &
∂uk ∂ui ∂
⟨ui uj ⟩ + ⟨uk uj ⟩ = ⟨ui uk uj ⟩ (3.34)
∂xj ∂xj ∂xj

This is also a divergence term.


Note
We that
can use inthe
all of each ofwethe
pieces havethree terms
developed aboveon the RHS
to rewrite of 3.30
equation
as:
equation 3.30, we have now identified a divergence term

Now, lets group terms that has this divergence form and
rewrite equation 3.30.

source: Turbulence for the 21st century, W. K. George


Turbulence: lecture
Reynolds
REYNOLDS stressEQUATIONS
STRESS equations 59
Rate of change of Reynolds
stress following mean motion !
Pressure-strain
"
term
∂ ∂ p ∂ui ∂uk
⟨ui uk ⟩ + Uj ⟨ui uk ⟩ = ⟨ + ⟩
∂t ∂xj ρ ∂xk ∂xi
#
∂ 1
+ − [⟨puk ⟩δij + ⟨pui ⟩δkj ] − ⟨ui uk uj ⟩
∂xj ρ Transport
+2ν[⟨sij uk ⟩ + ⟨skj ui ⟩]}
! "
∂Uk ∂Ui
− ⟨ui uj ⟩
∂xj
+ ⟨uk uj ⟩
∂xj
production
! "
∂uk ∂ui
− 2ν ⟨sij ⟩ + ⟨skj ⟩ (3.35)
∂xj ∂xj dissipation
e so-called Reynolds Stress Equation which has been the primary
much of the turbulence modeling efforts of the
source: past forfew
Turbulence decades.
the 21st century, W. K. George
2. the turbulence transport (or divergence) term
Turbulence:
Turbulence: lecture
lecture
3. the “production” term, and
Reynolds
Reynolds stress equations
stressterm.
4. the “dissipation” equations
Obviously these equations do not involve only Ui and ⟨ui uj ⟩, but depend on many
more new unknowns.
It is clear that, contrary to our hopes, we have not derived a single equation
relating the Reynolds stress to the mean motion. Instead, our Reynolds stress
transport equation is exceedingly complex. Whereas the process of averaging the
equation for the mean motion introduced only six new independent unknowns,
the Reynolds stress, ⟨ui uj ⟩, the search for a transport equation which will relate
these to the mean motion has produced many more unknowns. They are:

⟨pui ⟩ − 3 unknowns (3.36)


⟨ui sjk ⟩ − 27 (3.37)
⟨sij sjk ⟩ − 9 (3.38)
⟨ui uk uj ⟩ − 27 (3.39)
∂ui
⟨p ⟩ − 9 (3.40)
∂xj
T OT AL − 75 (3.41)

source:
source:Turbulence
Turbulencefor
forthe
the21st
21stcentury,
century,W.
W.K.K.George
George
Turbulence: lecture
Reynolds stress equations
Closure problem cannot be solved

By deriving new dynamical equations,

or by using “eddy” viscosity models

We need to understand how the turbulence behaves

This may allow us to develop new constitutive models for


turbulence and also understands its limitations
Turbulence is a topic that we are still studying!
So, this module by definition is incomplete
source: Turbulence for the 21st century, W. K. George
Turbulence: lecture
Kinetic energy of turbulent fluctuations
REYNOLDS STRESS EQUATIONS 59
We can start with the Reynolds stress equations and
contract the free indices (i.e. i = k)
! "
∂ ∂ p ∂ui ∂uk
⟨ui uk ⟩ + Uj ⟨ui uk ⟩ = ⟨ + ⟩
∂t ∂xj ρ ∂xk ∂xi
#
∂ 1
+ − [⟨puk ⟩δij + ⟨pui ⟩δkj ] − ⟨ui uk uj ⟩
∂xj ρ
+2ν[⟨sij uk ⟩ + ⟨skj ui ⟩]}
! "
∂Uk ∂Ui
− ⟨ui uj ⟩ + ⟨uk uj ⟩
∂xj ∂xj
! "
∂uk ∂ui
− 2ν ⟨sij ⟩ + ⟨skj ⟩ (3.35)
∂xj ∂xj
e so-called Reynolds Stress Equation source:
which has been the primary
Turbulence for the 21st century, W. K. George
Turbulence: lecture
Kinetic energy of turbulent fluctuations
REYNOLDS STRESS EQUATIONS 59
We can start with the Reynolds stress equations and
contract the free indices
! "
∂ ∂ p ∂ui ∂uk
⟨ui uk ⟩ + Uj ⟨ui uk ⟩ = ⟨ + ⟩
∂t ∂xj ρ ∂xk ∂xi
#
∂ 1
+ − [⟨puk ⟩δij + ⟨pui ⟩δkj ] − ⟨ui uk uj ⟩
∂xj ρ
+2ν[⟨sij uk ⟩ + ⟨skj ui ⟩]}
! "
∂Uk ∂Ui
− ⟨ui uj ⟩ + ⟨uk uj ⟩
∂xj ∂xj
! "
∂uk ∂ui
− 2ν ⟨sij ⟩ + ⟨skj ⟩ (3.35)
∂xj ∂xj
e so-called Reynolds Stress Equation source:
which has been the primary
Turbulence for the 21st century, W. K. George
Turbulence: Turbulence: lecture
it. Our hope is that by understanding more about turbulence itself, we will gain
lecture that will work, at least
insight into how we might make closure approximations
Kinetic Hopefully,
sometimes. energy of turbulent
we will also gain fluctuations
an understanding of when and why they
Kinetic
will energy of turbulent fluctuations
not work.
We can start
An equation with
for the the Reynolds
fluctuating stress
kinetic energy equations
for constant density and
flow can be
We can
obtained start
directly fromwith the Reynolds
the Reynolds stress
stress equation equations
derived and 3.35,
earlier, equation
contract
bycontract
the
contractingthe
free indices
free indices
the free indices. The result is:
! "
∂ ∂
⟨ui ui ⟩ + Uj ⟨ui ui ⟩
∂t ∂xj
# $
∂ 2
= − ⟨pui ⟩δij − ⟨q 2 uj ⟩ + 4ν⟨sij ui ⟩
∂xj ρ
∂Ui ∂ui
−2⟨ui uj ⟩ − 4ν⟨sij ⟩ (4.1)
∂xj ∂xj

where the incompressibility condition (∂uj /∂xj = 0) has been used to eliminate
the pressure-strain rate term, and q 2 ≡ ui ui .
The last term can be simplified by recalling that the velocity deformation rate
We can simplify this equation further
We can simplify this equation further
tensor, ∂ui /∂x j , can be decomposed into symmetric and anti-symmetric parts;

source: Turbulence for the 21st century, W. K. George


Wednesday, 23 October 13
61 source: Turbulence for the 21st century, W. K. George
Turbulence: lecture
Turbulence: lecture
62 CHAPTER 4. THE TURBULENCE KINETIC ENERGY
Kinetic
62 energyCHAPTER
of turbulent
4. THEfluctuations
TURBULENCE KINETIC ENERGY
i.e.,
i.e., ∂ui
∂u = sij + ωij (4.2)
∂xij = sij + ωij
i
(4.2)
ij ij
∂xjj
where the symmetric part is the strain-rate tensor, sij , and the anti-symmetric
where
part isthe
the symmetric
rotation-ratepart is theωstrain-rate
tensor, ij , defined by:
tensor, sijij, and the anti-symmetric
part is the rotation-rate tensor, ωijij, defined ! by: "
1 ! ∂ui ∂uj"
ωij = 1 ∂ui − ∂uj (4.3)
ωijij = 2 ∂xij − ∂xji (4.3)
2 ∂xjj ∂xii
Since the double contraction of a symmetric tensor with an anti-symmetric tensor
Since the double
is identically contraction
zero, it follows of a symmetric
immediately tensor with an anti-symmetric tensor
that:
is identically zero, it follows immediately that:
∂ui
⟨sij ∂ui ⟩ = ⟨sij sij ⟩ + ⟨sij ωij ⟩
⟨sijij ∂xij⟩ = ⟨sijijsijij⟩ + ⟨sijijωijij⟩
∂xjj = ⟨s s ⟩ (4.4)
ij ij
= ⟨sijijsijij⟩ (4.4)
Now it is customary to define a new variable k, the average fluctuating kinetic
Nowper
energy it isunit
customary
mass, by:to define a new variable k, the average fluctuating kinetic
energy per unit mass, by: source:
source:Turbulence
Turbulence for
for the
the 21st
21st century,
century,W.
W.K.
K.George
George
Turbulence:
Turbulence: lecture
lecture
Turbulence: lecture
∂u
∂u
∂u ii
== ⟨s
⟨s ⟨s+ ∂u
⟨sijij i ⟩⟩⟩
⟨sijijsssijij⟩⟩⟩+ ⟨sijijωωωijij⟩⟩⟩
+⟨s
⟨s i
= ⟨s
⟨sijij ∂x
= ⟨sijijsijij⟩ + ⟨sijijωijij⟩
∂x ⟩j
Kinetic energy
Kinetic energy of
of turbulent
turbulent fluctuations
fluctuations ∂xjj
∂x j

Kinetic energy of turbulent


=
== ⟨s
⟨sfluctuations
⟨s ijsssijij⟩⟩⟩
ij
= ⟨sijijsijij⟩
(4.4)
(4.4)
(4.4)
(4.4)
Nowitititisisiscustomary
Now
Now customaryto
customary todefine
to defineaaanew
define newvariable
new variablek,
variable k,the
k, theaverage
the averagefluctuating
average fluctuatingkinetic
fluctuating kinetic
kinetic
Now it is customary to define a new variable k, the average fluctuating kinetic
energyper
energy
energy perunit
per unitmass,
unit mass,by:
mass, by:
by:
energy per unit mass, by:
1
11 1
11 22 111 22 22 ⟩ + ⟨u 22 ⟩]
kk ≡
≡ 1 ⟨u⟨u u u ⟩⟩ =
= 1 ⟨q
⟨q 2 ⟩⟩ =
= 1 [⟨u
[⟨u 2 ⟩⟩ +
+ ⟨u
⟨u 2 ⟩ + ⟨u 2 ⟩] (4.5)
(4.5)
≡ 22⟨u
kk ≡ ⟨uiiuiui ii⟩i⟩i =
= 22⟨q⟨q2 ⟩⟩ = [⟨u2111⟩1⟩ +
= 22[⟨u +⟨u⟨u2222⟩2⟩ +
+⟨u⟨u2333⟩]
3⟩] (4.5)
(4.5)
22 22 22
Bydividing
By
By dividingequation
dividing equation4.1
equation 4.1by
4.1 by222and
by andinserting
and insertingthis
inserting thisdefinition,
this definition,the
definition, theequation
the equationfor
equation forthe
for the
the
By dividing equation 4.1 by 2 and inserting this definition, the equation for the
averagekinetic
average
average kineticenergy
kinetic energyper
energy perunit
per unitmass
unit massof
mass ofthe
of thefluctuating
the fluctuatingmotion
fluctuating motioncan
motion canbe
can bebere-written
re-written
re-written
average kinetic energy per unit mass of the fluctuating motion can be re-written
as:
as:
as:
as:
!!
! """ ###
$
$
$
! ∂∂ ∂∂ " ∂
∂ # 11 11 $
∂∂ + + U
U j
∂ kk =
∂ = ∂∂ −
− 11 ⟨pu⟨pu ⟩δ
i ⟩δ ij −−11 ⟨q⟨q
22
2
2 u ⟩ + 2ν⟨s u ⟩
ujj⟩j⟩⟩+ 2ν⟨sijijijuuui⟩i⟩i⟩
+2ν⟨s
∂t ++ UU
∂t j ∂x jj
∂xjj
∂x k
k == ∂xjj
∂x −− ⟨pu
⟨pu
ρ i ij
ρ ii⟩δij
⟩δ ij−− 2 ⟨q
⟨q
2 j uu + 2ν⟨s ij i
∂t
∂t ∂xjj ∂x
∂xjj ρρ 22
∂Uii
∂U
∂U
−⟨u
−⟨u iuu j⟩∂U
⟩ ii − −2ν⟨s
2ν⟨sijijsssijij⟩⟩⟩
2ν⟨s (4.6)
(4.6)
−⟨u u
i
−⟨uiiujj⟩ ∂xj⟩ ∂x j

− 2ν⟨sijijsijij⟩ (4.6)
(4.6)
∂xjj
∂x j

Another
Another
Therole
The
The roleof
role of way
way
ofeach
eachof
each ofto
to
of these
these arrive
arrive
these terms
terms at
at
termswill
will
will this
this
be
be beexamined equation
equation
examined
examined
The role of each of these terms will be examined in detail later. First note that
indetail
in
in detail
detail is
is by
by
later.
later.
later. taking
taking
Firstnote
First
First dot
dot
note
note that
that
that
anAnother
an alternativeform
alternative way
form
form ofto
of thisarrive
this equationat
equation canthis
can bederived
be equation
derived
derived byleaving
by leaving
leaving is theby
the taking
theviscous
viscous
viscous dot
stress
stress in
in
product
product
an
an alternative
alternative of
of formmomentum
momentum
of
of this
this equation
equation equation
equation
can
can be
be derived with
with by
by velocity
velocity
leaving the
fluctuation
fluctuation
viscous stress
stress in
in
product
terms
terms
terms
terms
the
ofof
of the
the
the
of the strain
thefluctuating
of momentum
strain
strain
strain
fluctuatingmomentum
fluctuating
rate.
rate.
rate.
momentumfrom
momentum
We
We
We can
can
can
rate. We can obtain equation
obtain
obtain
obtain
fromequation
from equation3.21
equation
the
the
the with
appropriate
appropriate
appropriate
the appropriate
3.21by
3.21 by
velocity
bysubstituting
form of thefluctuation
substituting
substituting
form
form
form of
of
of the
the
the
the
the
equationfor
equation
equation
equation for
incompressible
incompressible
for
for
the
the fluctuating momentum from equation 3.21 by substituting
source:
source: Turbulence for
Turbulence for the the
the21st
the incompressible
incompressible
21st century,W.
century, W.K.K.George
George
Newtonian
Newtonian
Newtonian constitutiveequation
13 constitutive
constitutive equationinto
equation intoitititto
into toobtain:
to obtain:
obtain: source: Turbulence for the 21st century, W. K. George
Newtonian
Wednesday,
Wednesday, 23October
23 October13
constitutive equation into it to obtain:
2 i i i i
∂ ∂
++UUj zero, uitui =
entically − 1
follows ∂p + ν ∂
immediately u i − that:
u ∂U i − uj ∂u i − ⟨uj ∂u i⟩ (4.7)
∂t
∂t
j∂xj
∂xj
i = −ρ ∂xi + Turbulence:
ν∂xj 2 − uj∂xjlecture
ρ ∂xi Turbulence:
2
∂xj
j − uj∂xj − ⟨uj∂xj ⟩
∂xjlecture ∂xj ∂xj
(4.7)
∂ui
we Kinetic
e take the
Kinetic
take the energy
scalar
energy
scalar productof
⟨s
product turbulent
of ⟩
this
of turbulent
ij of
∂x this
j
with
=with ⟨sthe
ij
the fluctuations
s ij ⟩ + ⟨s
fluctuating
fluctuations
fluctuatingij ω ij ⟩
velocity
velocity itself
itself and
and average,
average,
ollows (after some rearrangement) that:
ollows (after some rearrangement)
!! "" = that:
⟨s
##ij sij ⟩ Used
Used for for modelling
modelling
$$ (4.4)
∂∂ ∂∂ ∂∂ 11 11 2 ∂∂
Now it∂tis+customary
+UUj j∂xj kkto define == a new variable −− ⟨pu ⟨pui ⟩δ
k, −− average
ijthe
⟩δ ⟨q⟨q 2uuj ⟩⟩+ νν kk kinetic
fluctuating
+
∂t ∂x ∂x
∂x j ρ ρ
i ij 22
j ∂x
∂xj
gy per unit mass, by: j j j
∂U
∂U i ∂u∂u i ∂u∂u i
−− ⟨u ⟨ui u uj ⟩⟩
i −
− ν⟨ν⟨
i i⟩
⟩ (4.8)
(4.8)
1 1 2 i j ∂x 1
∂x jj 2 ∂x∂x ∂x
j2j ∂xj j 2
k ≡ ⟨ui ui ⟩ = ⟨q ⟩ = [⟨u1 ⟩ + ⟨u2 ⟩ + ⟨u3 ⟩] (4.5)
2 2 2

BothBoth
By dividing equations
equations
equation represent
4.1 by 2 and inserting thisthe
represent the same
same
definition, thing
the thingfor the
equation
age kinetic energy per unit mass of the fluctuating motion can be re-written
Used
Used for
for understanding
understanding
! " # $
∂ ∂ ∂ 1 1 2
+ Uj k = − ⟨pui ⟩δij − ⟨q uj ⟩ + 2ν⟨sij ui ⟩
∂t ∂xj ∂xj ρ 2
∂Ui
−⟨ui uj ⟩ − 2ν⟨sij sij ⟩ (4.6)
∂xj
The role of each of these terms will be examined in detail later. First note that
source: Turbulence for the 21st century, W. K. George
source: Turbulence for the 21st century, W. K. George
Turbulence:
Turbulence:lecture
now that it never occurs in nature. Nonetheless, many flows can be assumed
lecture
to be homogeneous at the scales of turbulence which are important to this term,
so-called local homogeniety.
Each term in the equation for the kinetic energy of the turbulence has a distinct
Kinetic
Kineticenergy
energyof
ofturbulent
turbulentfluctuations
fluctuations
role to play in the overall kinetic energy balance. Briefly these are:

• Rate of change of kinetic energy per unit mass due to non-stationarity; i.e.,
time dependence of the mean:
∂k
(4.9)
∂t

• Rate of change of kinetic energy per unit mass due to convection (or advec-
tion) by the mean flow through an inhomogenous field :

∂k
Uj (4.10)
∂xj

• Transport of kinetic energy in an inhomogeneous field due respectively to


the pressure fluctuations, the turbulence itself, and the viscous stresses:
! "
∂ 1 1
− ⟨pui ⟩δij − ⟨q 2 uj ⟩ + 2ν⟨sij ui ⟩ (4.11)
∂xj ρ 2

• Rate of production of turbulence kinetic energy from the mean flow (gradi-
ent):
∂Ui
−⟨ui uj ⟩ (4.12)
∂xj

• Rate of dissipation of turbulence kinetic energy per unit mass due to viscous
stresses:
ϵ ≡ 2ν⟨sij sij ⟩ (4.13)
source: Turbulence
source: for the
Turbulence 21st century, W. K. George
These terms will be discussed in detail in the succeeding sections, and for
the the
role21st
of century, W. K. George
Turbulence:
Turbulence:
Turbulence: lecture
lecture
lecture
1
1 2
1
1 2 Turbulence:
2 2
ui ui ⟩ = 1 ⟨q2 ⟩ = 1 [⟨u21 ⟩ + ⟨u22 ⟩ + ⟨u23 ⟩] lecture
(4.5)
uui uui ⟩KE= transport
⟨q 2⟩ = 2[⟨u12⟩ + ⟨u22⟩ + ⟨u32⟩] (4.5)
i KE transport
⟩ = 2⟨q ⟩ = [⟨u ⟩ + ⟨u ⟩ + ⟨u ⟩] (4.5)
i
KE transport
2
2 22 1 2 3
4.1 by
.1 by This
This 2
2 and
and transport
inserting this
transport
inserting term
term
this is
definition, best
isismotion
definition, best
the
the understood
equation for the
understood
equation for by
by
the considering
considering aaa
4.1 This
by
r unit
rr unit
2mass
andtransport
inserting term
this
of the fluctuating
turbulent
definition,
flow best
confined understood
the equation
can
70 in a by
for the
be re-written
control considering
CHAPTERvolume
unit mass of
of the
the fluctuating
massturbulent motion
flow confined
fluctuating motion can be re-written
70 in a control
can be CHAPTERvolume
re-written 4. THE TURBULENCE KINE
turbulent
Figure 4.1: Turbulence flow
# Turbulence
confinedconfined 70 in a control
by rigid walls. CHAPTERvolume
$ by rigid walls.
4. THE TURBULENCE KINET
4. THE TURBULENCE KINE
Figure
Figure 4.1: Figure
4.1: Figure 4.1:
Turbulence Turbulence
confined
confined by rigid confined
walls.
by rigid walls.
∂ # 1
# Figure 4.1: Turbulence
1
4.1: Turbulence
$
confined
$ by
confined by rigid
rigid walls.
walls.
= ∂∂ −11 ⟨pui ⟩δij −11 ⟨q22 uj ⟩ + 2ν⟨sij ui ⟩
2
=
= ∂x ∂xj − − ρ⟨pu
⟨pu i ⟩δ
⟩δ ij − − 2⟨q
⟨q uuj ⟩⟩ + + 2ν⟨s
2ν⟨s ij uu i ⟩⟩
tion 4.6∂x j ρρ
forj the fluctuating
i ij 2
kinetic2
j ij i
energy can be integrated over the
on 4.6 for
eld: Integrate
the fluctuating ∂U i kinetic energy can be integrated over the
ion
ld:
of
4.6 for
Integrate
equation
the
−⟨u
−⟨u
uthis
fluctuating
this
4.6 ⟩
i j for
u ⟩
∂U
∂U within
i
within
i
kinetic

the

2ν⟨s
2ν⟨s this
this
energy
fluctuating s control

control
ij ij
s ⟩
can be
volume
integrated
volume
kinetic energy
over
(4.6)
(4.6)
the
can be integrated ov
eld:
of
of Integrate
equation
equation
! ! !
−⟨u 4.6
i
i this
4.6 u j
" ⟩
j for
∂x
∂x
for within
the
j
the− 2ν⟨s this
fluctuatingij control
ij
fluctuating s ij
ij ⟩ kineticvolume
kinetic # energyenergy (4.6) can
can be
be integrated
integrated ove
ov
me to
! ! yield:
! ∂ "" 1 ∂x
1 j
j 1 ##
me
mese to
to !
terms yield:
! !
yield: will ∂∂ be −
− 1 ⟨pu
examined
⟨pu
⟩δ
i ij −
in11 ⟨q
detail
2
2
u ⟩ +
2uj ⟩ +later.
j ν⟨s u
ij i ⟩
First dV
note that
Vo ∂x ρ i ⟩δ ij − 2⟨q ν⟨s ij u i⟩ dV
se
se terms
terms will
will
V"o ∂x
∂x
be
bej
j examined
− ρ ⟨pu
examined
ρ
i ij in2detail
⟩δ −
in 2
⟨q
"
detail u j later.
⟩ + ν⟨s
later. # ij First
u ⟩
First
i dV note
note that
that #
sequation
equation! ! ! ! !
"can be!derived
derived by ""leaving the
## viscous stress in ##
V
!! !! "can
o
1bej 1 by
sWe equation
= −11 ⟨pu
can !
be !
! i!derived
⟩δij!−11 ⟨q∂ ∂
2 leaving
2by uj ⟩leaving
+−ν⟨s 11 the
ij uthe
viscous
i ⟩ nviscous
j dS 1
1 stress
stress
2 (4.31) in
in
= Apply
can obtain
− divergence
ρ⟨pu the
⟩δ appropriate
− 2⟨q ∂
theorem,
u ⟩ + ν⟨s 1
j ⟩ + ν⟨sijconvert
form ⟨pu
u ⟩ of
i ⟩δ
i ⟩ nj dS this to
n the
dS
ij − 1 ⟨q
equation 2 u ⟩
surface+
for
(4.31)
j ν⟨s ij u
integration
i ⟩ dV
j −form ⟨pu ⟩δ j −equation ⟨q ⟩ ⟩
2
We =
Apply
can So −
obtain ρ ⟨pu
divergencethei ⟩δij − ⟨q
i appropriate
ij 2 theorem,
∂x u − ρ u
convert
ij i of
⟨pu i ⟩δ the
ij −this 2 to
⟨q 2uu surface
j ⟩ +
(4.31)
for
+ ν⟨s
ν⟨s ij u
integration
u i ⟩ dV
dV
We
m fromApply
can So divergence
Sobtain
o
equation ρ the 3.21appropriate
V
V
by
V"
2 theorem,
o
∂x
o ∂x
j
substituting
jj
form
ρρ convert of
ithe the
ij this
equation
2
incompressible
2 to surface
j
Figure
for
4.1: # integration
Turbulence ij i
confined by rigid walls.
m
m vefrom
used equation
from equation
the divergence !3.21
3.21 by
! theorem"
by o substituting
substituting
— again! the
the incompressible
incompressible Figure 4.1: #
Turbulence confined by rigid walls.
quation
e used the into it
divergence !!to !! obtain:
theorem "
1 1— again! 1 1 2
Figure 4.1: #
Turbulence confined by rigid walls.
ve
med used
uation
quation the
into
our enclosure
into = =
divergence
itit to
to
to havetheorem
obtain:
obtain: − − —
⟨pu⟨pu again!
rigid 1walls; therefore i
⟩δ⟩δ ij − − 1the
⟨q ⟨q u
2normal
2 u j
⟩ ⟩ ++ ν⟨s
component
ν⟨s iju u i
⟩ ⟩ n n j
dSdS
kinetic energy can be inte(
meded ourour enclosure
enclosure =
velocity (un =2 uj nj )!must
to
to have
!
have S rigid
rigid
o −
be"zero"
ρ ρ
walls;⟨pu
walls; #
# therefore
⟩δ
itherefore
ij
on itheijsurface − 2 2
the
thesidenormal
⟨q
side
of
u
normal
since ⟩
equation
jjthere
ofofequation
volume to $+
$component
ν⟨s
4.6 forij
component
4.6
yield: can
for
the
u ⟩
ij fluctuating
be
the i
no n
fluctuating
i j dS
jkinetic energy can be integ
1 ∂p
elocity
1velocity
∂p (u
(u ∂
= 2 u
n∂ uij −
u n
i ) !
must SS o
∂U∂U
be "
zero
i ρ
n =2 uj nj ) must be zero on the surfacevolume
o on# the ∂u
∂usurface
i 2 side
since
volume
since ∂u∂uequation
tothere
i $
yield:
tothere
4.6
can
can
for
bethe
be
fluctuating
no
no "
kinetic energy can be inte
it
1it ∂p (the
(the +
+ νν
kinematic

kinematic u 2i − u
boundaryu
j boundary
j ∂U ii − u −
condition).
condition). u j ∂u This
This
i − ⟨u
immediately
i −immediately
⟨u j ∂u i ⟩
yield: !
i ⟩ eliminates !(4.7)
eliminates
! !(4.7)
!
∂" 1 1
#

ρreρit
tions
∂x∂x
(the
we to
i +have
theν ∂x
kinematic used
surface
∂x 2 j −
2 integral
u
boundary
j
the
integral
j ∂x∂x divergence
from
j −
condition).
the u j
⟨pu
j ∂x
∂x
This −
j theorem
n j
j ⟩ and ⟨u
immediately
j
⟨q
j ∂x
2
2∂x u— ⟩
source:
j j again!
nj ⟩ ! !(4.7)
eliminates
Turbulence
terms. !
! But ∂∂x
for" −the
11ρ⟨pu
21st
⟨pu
i ⟩δ
ij century,
−1
2
⟨q
j W.
u ⟩ K.
+ George
ν⟨s
ij i u
#
⟩#
ρ ∂x ∂x ∂x ∂x ∂x source: Turbulence ∂ −
for the 21st −− 12⟨q
century,
2
W.⟩ K. George
e wei have used the divergence
ions to
i the surface
j
Wednesday, 23 October 13 from
jj the ⟨pu j theorem
n jjj⟩ and ⟨q 2u — j jj again!
n
j ⟩
source: terms. Vo
Turbulence
! !Vo
But
"∂x
j
j
for the
−ρ
i ⟩δ
21st
⟨pu ij
⟩δ
i ij century, u +K.ν⟨s
2 ⟨q uW.⟩ +
2 j
j # uu⟩⟩ d
George
ij i
ν⟨s
ij i
Vo ∂xj
ρ 2
# " #
=
! ! "

!
equation can1 be derived
⟨pu
!
⟩δ
!

1 ∂
⟨q
by
2
u ⟩
Turbulence:
leaving
+ 1
ν⟨s u ⟩
#
the viscous
n dS
lecture
1 stress in
2 (4.31)
+ ν⟨s
We Apply
can Su ⟩ ρ nthe
j dSappropriate
i
o i divergence
ijobtain
ij j
2theorem, (4.31)
−form
⟨pu
ij i
convert i ⟩δthe
of ij −
j
this ⟨qtousurface
equation j ⟩ for ij ui ⟩ dV
+ ν⟨sintegration
Vo ∂xj ρ 2 Figure 4.1: Turbulence confined by rigid walls.
KE transport
from equation !
3.21
!
by
" substituting the incompressible #
used the divergence theorem1— again! 1 2
uation into it to obtain:
d our enclosure = to have rigid − walls;
⟨putherefore
i ⟩δij − the ⟨q u
normalj ⟩ + ν⟨s
component ij u i ⟩ n j dS (
gain! ! S " ρ #
locity (un =2 uj nj ) must be zero on the surfacevolume 2 side of equation
sincetothere
70 $ 4.6 for the fluctuatingCHAPTER kinetic 4. THEenergy can be integr
TURBULENCE KINE
∂uiyield: can be no
o
∂p ∂ ui ∂Ui ∂ui
t therefore ν 2 the
(the+kinematic unormal
−boundary
j −component
condition).uj This−immediately
⟨uj ⟩ eliminates
! !(4.7)
!

"
1 1 2
#
∂x
onswe
the itohave
the Now
surface
∂x j consider
used
surface the
integral
since
∂x j thethe
divergence
from
there boundary
can
∂x njj ⟩ andconditions
⟨puj theorem
be no
2
⟨q u—∂x source:
jn
Turbulence
j j ⟩again!
terms. VBut
"
∂xj
for
o
the

ρ
21st
⟨pu i ⟩δ century,
ij −
2
⟨q uW.
j⟩ +K.ν⟨s
George
#
ij ui ⟩ dV

Wednesday, 23 October 13 ! !
is zero on the surface also. This can be seen in two ways: either
Normal velocity component is zero
1 1 2
et
on).assumed
of this
he no-slip This with our
condition the enclosure
immediatelyfluctuating
which together
to have
velocity
eliminates rigid
itself walls;
and
with the kinematic boundary
= therefore
average, S

ρ
⟨pu
o
i the
⟩δ ij −
2 normal
⟨q uj ⟩ + ν⟨sij uicompo
⟩ nj dS

mean that velocity


angement)
res ui isthat: zero (uon2nthe = boundary,
uj nj ) must or by benoting
zerowe from on used the thesurface
Cauchy’s since there can b
⟨pu j n j ⟩ and ⟨q u j n j ⟩ terms. are
But eliminated
where have divergence theorem — again!
hrough
νsij nj is the it (the viscous
# kinematic
contribution boundary
to the condition).
normal
of
Wecontact
the
assumed our
mean $
velocity (u
This
enclosure
force
=
per
u n )
immediately
to have rigid walls; therefore the norm
must be zero on the surface
elimin
since th
can
e surfacebe
ntributions seen

(i.e.,
Either tn )in
(v)
to 1
whose
no-slip
the two (u
surface
i
ways:
scalar
= 1
product
0) or
integral
2
either
with flow
slip (sui must
fromij = ∂ 0)
the
through be
⟨pu
it (the
n
identically
n ⟩
j j

and
kinematic boundary condition).
⟨q 2
u n ⟩ terms.
This immediate
=
s zero. Therefore the entire − ⟨pu ⟩δ
i ij − ⟨q u ⟩ +
integral is identically
j ν
the k
zero and its net
contributions to j
the j
surface integral from thej ⟨puj j j and ⟨q uj n
n ⟩ 2

er
st with
term
o the rate
∂xthe
is
Therefore, jzero kinematic
of change of the
ρon the surface
kinetic
boundary
2
lastenergy
termis is also.
zero.
the ∂x
also This
zerothe no-slip condition which together with the kinemae
last term
j iscan
zero onbe seen
the surface in
also. two
This can beways:seen in two
by invoking
ary,
voking
nly effector by
the
of the noting
∂Ui condition
no-slip
turbulence from∂u Cauchy’s
i ∂u
transport i which
terms (in together
a fixed
condition insuresvolume with
that is at
ui Figure the
zero kinematic
onTurbulence
4.1: the boundary,confinedor by boun
byrigid
notingwalls.fr

to Role
− ⟨uof i uj ⟩ this transport
− ν⟨ term
⟩ is to
theorem move
that νsij njenergy the viscousfrom
(4.8)
is contribution one to the normal cont
tomove
tion the
insures energy
normal from
that uone
∂xj contacti
place
is ∂x
zero
j to
∂x janother,
force
on per
the neither
boundary,
unit area on creating
the surface or nor
(i.e., by
t (v)
n ) whose noting scalar productfrom with uCau i must
n the process. This is, of course,
em that νs n is the place
viscous
why theyto another
are
contribution
collectively
zero since uside
n isofzero.
to to
equation
the
called
Therefore
4.6 for the
normal
thefluctuating
entire integral
of kinetic contact
kineticisenergy
identically
can be zer
energy is zero. force
int

oduct with ij uj must be


terms. This spatiali transport of kinetic energy Thus identically contributionvolume
is accomplished
the only
thetorate
yield:
effect
of change
of by
the ! !turbulence " transport terms (in a fix#
The transport happens (v)
by acceleration of adjacent fluid elements
!
rea
ral on
on ofisadjacent the
identicallysurface (i.e.,
fluid due zero t
to pressure
n )
and andwhose scalar
its viscous
netleast) stressesproduct
can be to move (theenergy with
first from∂x one

u
Vo
− must
1
⟨pu ⟩δ −
1
be
i ρplace to another,
j
i ij
2
⟨q u ⟩identi
2
j neither
+ ν⟨s ij iu⟩

by upressure
n is zero.orTherefore
viscous stress or by integral
the fluctuation itself zero and it destroying it in the process.
! !This" is, of course, why they are # colle
ince the entire is identically 1 1 2
rgy is zero. the transport terms. This
= spatial −
So
transport
ρ
⟨pui ⟩δij −of kinetic
2
⟨q uj ⟩ + energy
ν⟨sij ui ⟩ isnjacc
the acceleration of adjacent fluid due to pressure and viscous stres
dS

bution to the rate of change of kinetic energy isused


where we have zero.
the divergence theorem — again!
port terms (in a fixed volume at source: Turbulence
We assumed for theto21st
our enclosure havecentury, W.therefore
rigid walls; K. George
the no
Turbulence: lecture
Turbulence: lecture
KE transport
4.4. THE TRANSPORT (OR DIVERGENCE) TERMS 71
The fact that this term only transports energy is exploited
in turbulence modelling
and last terms respectively), and by the physical transport of fluctuating kinetic
energy by the turbulence itself (the middle term).
The rationale:
This This
role of these termtransport
turbulence will transport energy
terms in moving from
kinetic regions
energy around
is often exploited by turbulence modellers. It is argued, that on the average, these
of high energy to regions of low energy
terms will only act to move energy from regions of higher kinetic energy to lower.
Thus a plausible first-order hypothesis is that this “diffusion” of kinetic energy
should be proportioned to gradients of the kinetic energy itself. That is,

1 1 2 ∂k
− ⟨puj ⟩ − ⟨q uj ⟩ + ν⟨sij ui ⟩ = νke (4.32)
ρ 2 ∂xj

where νke is an effective diffusivity like the eddy viscosity discussed earlier. If we
use the alternative form of the kinetic energy equation (equation 4.8), there is no
We
need to took
model the 9 unknowns
viscous andit lumped
term (since involves only them together
k itself). Therefore our
model might be:
1 1 2 ∂k
− ⟨puj ⟩ − ⟨q uj ⟩ =source:
νkealtTurbulence for the 21st century, W. K.(4.33)
George
ρ 2 ∂x
Turbulence: lecture
KE transport
We replaced complex physical mechanisms by a simple
gradient diffusion process
Surprisingly this works pretty well
In fact, this is at the root of almost all engineering
turbulence modelling
But, by definition we do not allow counter-gradient
diffusion
Although maybe true on average, there is no reason for
gradient diffusion to be true locally

source: Turbulence for the 21st century, W. K. George


Turbulence:
Turbulence:lecture
now that it never occurs in nature. Nonetheless, many flows can be assumed
lecture
to be homogeneous at the scales of turbulence which are important to this term,
so-called local homogeniety.
Each term in the equation for the kinetic energy of the turbulence has a distinct
Kinetic
Kineticenergy
energyofofturbulent
turbulentfluctuations
fluctuations
role to play in the overall kinetic energy balance. Briefly these are:

• Rate of change of kinetic energy per unit mass due to non-stationarity; i.e.,
time dependence of the mean:
∂k
(4.9)
∂t

• Rate of change of kinetic energy per unit mass due to convection (or advec-
tion) by the mean flow through an inhomogenous field :

∂k
Uj (4.10)
∂xj

• Transport of kinetic energy in an inhomogeneous field due respectively to


the pressure fluctuations, the turbulence itself, and the viscous stresses:
! "
∂ 1 1
− ⟨pui ⟩δij − ⟨q 2 uj ⟩ + 2ν⟨sij ui ⟩ (4.11)
∂xj ρ 2

• Rate of production of turbulence kinetic energy from the mean flow (gradi-
ent):
∂Ui
−⟨ui uj ⟩ (4.12)
∂xj

• Rate of dissipation of turbulence kinetic energy per unit mass due to viscous
stresses:
ϵ ≡ 2ν⟨sij sij ⟩ (4.13)
source: Turbulence
source: forfor
Turbulence thethe
21st century,
21st W.W.
century, K.K.
George
George
These terms will be discussed in detail in the succeeding sections, and the role of
4.3. THE
THE PRODUCTION
PRODUCTION 67
4.3.
4.3 The
4.3 The Kinetic Turbulence:
KineticTurbulence:
Turbulence:
Energy of
Energy of thelecture
lecture
lecture
the Mean Motion
Mean Motion and and
67

4.3 The
KE production
4.3 production
The
the Kinetic
Kinetic
“Production” Energy
Energy of
of
of the
the Mean
Mean
Turbulence Motion
Motion and
and
KE the “Production” of Turbulence
the “Production”
the “Production” of of Turbulence
Turbulence
An equation
An equation for
for the
the kinetic
kinetic energy
energy of
of the
the mean
mean motion
motion can
can be
be derived
derived by
by aa pro-
pro-
cedure
cedure exactly
exactly
An equation
equation analogous
foranalogous
the kinetic
kinetic toenergy
to that applied
that applied
of the to the
to
the mean
mean themotion fluctuating
fluctuatingcan bebe motion.
motion.
derivedThe The
by aameanmean
pro-
An for the energy of motion can derived by pro-
motion
motion was shown
was
cedure exactly
exactlyshown in equation
in
analogous equation
to that 3.19applied
3.19
that to be
to be given
giventheby:
to the by:fluctuating motion. motion. The The mean mean
cedure analogous to applied to fluctuating
motion was
motion was shown
shown !! inin equation
equation 3.19 3.19
"" to to bebe given
given by: by:
(v)
∂U
∂U i ∂U∂U i ∂P
∂P ∂T
∂T
(v)
ij ∂∂
ρρ!! ∂U
i
++UUjj i
"" ==− − ++ (v) ij
(v) − − (ρ⟨uiiuujj⟩)
(ρ⟨u ⟩) (4.21)
(4.21)
∂tii
∂t
∂U ∂x
∂U∂xjiji
∂U ∂xii ∂T
∂P
∂x
∂P ∂T∂xijjj
∂x ∂x
∂x
∂∂ jj
ρρ ∂t + +UUjj ==− − +
+ ij
−− ⟩)
(ρ⟨uiiuujj⟩)
(ρ⟨u (4.21)
(4.21)
∂t ∂xjj
∂x ∂xii
∂x ∂xjj
∂x ∂xjj
∂x
By taking
By taking the
the scalar
scalar product
product of of this
this equation
equation with with the the meanmean velocity,
velocity, UUii,, we we can
can
obtain
By taking
obtain
By an equation
taking
an equation
the scalar
the scalarfor for the kinetic
product
the
product kinetic
of this
of thisenergy
equation
energy
equation of the
of the
with
with mean
mean motion
themotion
the as:
mean velocity,
mean velocity,
as: UUii,, we
we can
can
obtain an
obtain an equation
equation for the kinetic energy of of the
the meanmean(v)motion
motion as: as:
!! for the kinetic "" energy
∂∂ ∂∂ UU i ∂P
∂P UU i ∂T
∂T
(v)
ij ∂⟨uiiuujj⟩⟩
∂⟨u
UUii!! ∂t ++UUjj "" U Uii ==− − i +
+
i ij
(v) − −UUii (4.22)
(4.22)

∂t∂ ∂x
∂x∂∂jj UU
ρ ∂P
∂xii UU
ρii ∂P
∂x ρ ∂T
ρii ∂T ∂x
∂x
(v)
ijjj ∂⟨u
∂⟨u ∂x
∂x ujjjj⟩⟩
iu
UUii ∂t + +UUjj Uii =
U =− − ++ ij
−−UUii i (4.22)
(4.22)
∂t ∂xjj
∂x ρρ ∂x ∂xii ρρ ∂x ∂xjj ∂xjj
∂x
Unlike the
Unlike the fluctuating
fluctuating equations,
equations, there there isis no no need
need to to average
average here,here, since since all all the
the
terms
terms
Unlike are
Unlikearethe already
thealready
fluctuating
fluctuating averages.
equations, there
averages.
equations, there isis no no need
need to to average
average here,here, since since all all the
the
terms
terms In are
In exactly
are already
exactly
already theaverages.
the same manner
averages.
same manner that that we we rearranged
rearranged the the terms
terms in in thethe equation
equation
for In
for theexactly
In
the kineticthe
exactly
kinetic energy
the
energy same
same ofmanner
of the fluctuations,
manner
the fluctuations,
that we
that we can
can rearrange
we rearranged
rearranged
we rearrange
the terms
the the in
terms
the equation
in
equation for the
the equation
the equation
for the
kinetic
for the
kinetic
for energy
the energy
kineticof
kinetic of the mean
energy
the
energy mean
of the
of the flow
flow to obtain:
to obtain:we
fluctuations,
fluctuations, we can
can rearrange
rearrange the the equation
equation for for the
the
kinetic energy
kinetic energy!!of of thethe mean
mean flow flow"" toto obtain:
obtain: source: Turbulence for the 21st century, W. K. George
source: Turbulence for the 21st century, W. K. George
∂ ∂ Ui ∂P Ui ∂Tij ∂⟨ui uj ⟩
Ui
∂t
+ Uj
Turbulence:
Turbulence: lecture
∂xj
Ui = −
lecture
ρ ∂xi
+
ρ ∂xj
− Ui
∂xj
(4.22)

KE
KE production
production
Unlike the fluctuating equations, there is no need to average here, since all the
terms are already averages.
In exactly the same manner that we rearranged the terms in the equation
for the kinetic energy of the fluctuations, we can rearrange the equation for the
kinetic energy of the mean flow to obtain:
! "
∂ ∂
+ Uj K=
∂t ∂xj
# $
∂ 1 1
− ⟨P Ui ⟩δij − ⟨ui uj ⟩Ui + 2ν⟨Sij Ui ⟩
∂xj ρ 2
∂Ui
+⟨ui uj ⟩ − 2ν⟨Sij Sij ⟩ (4.23)
∂xj

where
1 2 1
K ≡ Q = Ui Ui (4.24)
2 2
The role of all of the terms can immediately be recognized since each term has its
counterpart in the equation for the average fluctuating kinetic energy.
Comparison of equations 4.23 and 4.6 reveals that the term −⟨ui uj ⟩∂Ui /∂xj
appears in the equations for the kinetic energy of BOTH the mean and the fluc-
tuations. There is, however, one VERY important difference.
source:
source:Turbulence forThis
Turbulencefor the “production”
the21st
21st century,
century,W.
W.K.
K.George
George
e already averages.
actly the
actly the same
same manner
manner
⟨s ij
∂u Turbulence:
that
i Turbulence:
that
∂ui ⟩Turbulence:
= ⟨s
we
we rearranged
rearranged
⟨sij ssij ⟩⟩ ++ ⟨s lecture
lecture
⟨slecture
the
the
ωij ⟩⟩
ij ω
terms
terms in
in the
the equation
equation
inetic energy
energy of
of the⟨s ∂x ⟩
the ijfluctuations,
fluctuations, = weij can
ij can rearrange
rearrange
ij ij the equation
equation for
for the
the
inetic ∂xj j we the
KE
nergyKE ofproduction
nergy ofproduction
the mean
the mean flowflow to to obtain:
obtain:
= ⟨s ⟨sij ssij ⟩⟩ (4.4)
= ij ij (4.4)
! "
!∂ ∂ "
ow it is customary

ow it is customary to
∂ define
to define a new variable k, the average
a new variable k, the average fluctuating kinetic fluctuating kinetic
+ U
+ Uby: j K =
y per unit
∂t mass, j ∂xj K =
y per unit
∂t mass, by: ∂x#j $
∂ # 1 1 $
1
∂1 ⟨u − 1⟩ ⟨P 1 1
2 − 11 ⟨u2u ⟩U +
k ≡ i−u i =
⟨P 1
U
U ⟨q
i ⟩δ
⟩δ2 ⟩
ij =
− [⟨u
⟨u 2i ⟩
u +
j ⟩U ⟨u
i
2 2ν⟨S2 U ⟩
+
2 ⟩ + ⟨u
2ν⟨S ⟩]
23ij U i⟩ (4.5)
k ≡∂x2j⟨ui ui ρ ⟩ = 2 ⟨q i ⟩ ij = 22[⟨u1i ⟩ j+ ⟨u
1 i 2 ⟩ + ⟨u3ij
2 ⟩] i (4.5)
∂x2j ρ 2 22
∂U
y Positive
dividing
Positive
y dividing here
equation
here
Positive
equation 4.1
+⟨u
4.1
+⟨u
by
by i u
u
2
2j ⟩

i jmass
and
∂U
and
∂x
iinserting this definition, the equation for the
−− 2ν⟨S
iinserting
2ν⟨S S
ijthis
S
ij ij

ij ⟩ definition, the equation (4.23)
for the
(4.23)
ge kinetic energy per unit
ge kinetic energy per unit mass j of the fluctuating
∂xj of the fluctuating motion motion can can be
be re-written
re-written

! " # $
!∂ ∂ " ∂ 1
# 21 1 1 $
∂ + Uj ∂ k = K K≡ ≡ Q Q
−∂ 1 1
2 =
⟨pu 1 U
⟩δi U −
i 1 ⟨q 2
u ⟩ + 2ν⟨s u ⟩ (4.24)
∂t + Uj ∂xj k = ∂xj 2 − ρ= ⟨pu2iU
i ⟩δiijU−
ij 2 j
i 2 ⟨q uj ⟩ + 2ν⟨sij ui ⟩ ij i (4.24)
∂t ∂xj ∂xj 2 ρ 2 2
of all of the terms can immediately be
∂U recognized since each term has its
of all of the
art in the Negative
terms can
Negative
equation here
immediately
here
for the
−⟨u i uj ⟩
be irecognized
∂U i
−⟨ui uj ⟩fluctuating
average ∂x
− 2ν⟨s ij sij ⟩
− 2ν⟨sijkinetic
sij ⟩
since each term has
energy.
its
(4.6)
(4.6)
Negative
art in the equation for the average fluctuating ∂xj j kinetic energy.
parison
parison of
of equations
equations 4.23
4.23 and
and 4.6
4.6 reveals
reveals that
that the
the term
term −⟨u
−⟨u i u
u j ⟩∂U
⟩∂U i /∂x
/∂x j
he
he role
n role of
the of each
equations of these terms
for the
each of these the will
kinetic
terms be examined
energy
will be examined in
of BOTH
BOTH detail
in detail later.
thelater. First
meanFirst
i j notei thatj
andnote the that
fluc-
n the
ernativeequations
form of for
this kinetic
equation can energy
be derived of by leaving the mean
the and
viscous the
stress fluc-
in
There
ernative
There is, is,
formhowever,
however, one VERY
of this equation
one VERY important
canimportant
be derivedsource: difference.
by leaving
source:Turbulence
difference. This
Turbulencethe
for the
for the
This “production”
viscous stress
21st century,
21st century,
“production” W. K.
W. in
K. George
George
CHAPTERTurbulence:
4. THE TURBULENCE
Turbulence: KINETIC EN
lecture KINETIC EN
lecture
CHAPTERTurbulence:
4. THE TURBULENCE
lecture
KE
KE production
production
KE
this
this production
68 “production ” term provides
CHAPTER the only means
4. THE TURBULENCE KINETIC by which
ENERGY ene
68 “production CHAPTER ” term provides the only means
4. THE TURBULENCE KINETIC by which
ENERGY ene
hangedThisbetween
This term
term the
isis mean
the
the only
only flow way
way and by
by the fluctuations.
which
which energy
energy can
can be
be
anged between
This
Therefore term
this the
is mean
the
“production only
” flowway
term andbythe
provides the fluctuations.
which
only energy
means can
by which be canm
energy
standing
Therefore
the manner
transferred
transferred
this manner
“production
in
from
from which
” which
term the
the this
mean
mean
provides
energy
to
to the
the exchange
fluctuations
fluctuations between
standing the in thisandthe only
energy means by which
exchange energy canme
between
g motions transferred
be interchanged
is
be interchanged
between
accomplished
between from
the
the mean the
mean
mean
flow
represents
flow and
Understanding the manner in which this energy exchange
to
the
the onethe offluctuations
fluctuations.
the most
fluctuations. betweenchallengin
mean and
g motions is
Overall,
Overall,
Understandingaccomplished
this
this
the exchange
exchange
manner inrepresents
can
can
which be
be
this one of
understood
understood
energy the
exchange most by
by
between challengin
mean and
urbulence.
Overall, The
fluctuating motions overall
this exchange exchange
is accomplished can
represents
can be be
one ofunderstood
understood by
the most challenging
by exploi
prob-
urbulence.
fluctuating
lems in The
treating
treating
motions
turbulence. overall
the
the
is exchange
Reynolds
Reynolds
accomplished
The overall can
stress
stress
represents
exchange can be
as
as
one
be aa
ofunderstood
“stress”
“stress”
the most
understood by
challenging
by exploit
prob-
exploiting the
which treats −ρ⟨u u ⟩ as a stress, the Reynolds stress. The term
lems
hich in
analogy treating
treatsturbulence.
which−ρ⟨u the
i The
⟩ Reynolds
i uj−ρ⟨u
treats asi uaj ⟩ as
stress, stress
a stress,the as a “stress”
j overall exchange can be understood by exploiting the
theReynolds
Reynolds stress.
stress. The term:
The term:
analogy which treats −ρ⟨ui uj ⟩ as a stress, the Reynolds stress. The term:
−ρ⟨u i uji⟩∂U
−ρ⟨u uj ⟩∂Ui /∂x
i /∂xjj (4.25)
−ρ⟨u i uji⟩∂U
−ρ⟨u uj ⟩∂Ui /∂x
i /∂xjj (4.25)
can be thought of as the working of the Reynolds stress against the mean ve-
hought
can be of as
thought
locityofgradient
theof working
as
of the the workingof
flow, exactly
the
of the
as the
Reynolds
Reynolds stress
stress
viscous stresses against
resist
against
the mean
deformation
the ve-by
m
ought
dient as the
of the flow, working
exactly of the
as the Reynolds stress against the m
locity gradient
the instantaneous of the flow,
velocity exactly
gradients.as Thisviscous
the viscous stresses
stresses
energy expended resist
against resist
deformation deforma
the Reynolds by
dient
the
ntaneous of the
instantaneous
stress during flow, exactly
velocity
deformation
velocity by
gradients. as
gradients.
the meanthe
ThisThis viscous
energy
motion
energy ends stresses
expended
up in
expended theagainst resist
fluctuating
against deforma
the Reynolds
motions,
the R
stress
taneous during
however, deformation
velocity
while by the
thatgradients.
expended mean
This
against motion
energy
viscous ends up goes
in thedirectly
expended
stresses fluctuating
against motions,
to internal the en-R
inghowever,
deformation
while that by the
expended mean
against motion
viscous ends
stresses up
goes in the
directly
ergy. As we have already seen, the viscous deformation work from the fluctuating fluctuating
to internal en- m
ng
whiledeformation
that(orexpended by the mean motion ends up in the fluctuating m
ergy. As
motions we dissipation) against
have already seen, the viscous
viscous
will eventually stresses
deformation
send this goeskinetic
work
fluctuating from directly
the energytooninte
fluctuating to
while that
motions
internal
we internal (orexpended
dissipation)
energy
have already as well.
seen, against
the viscous
will eventually
viscous send thisstresses
deformation goes
fluctuating directly
kinetic
work energytooninter
from the to
fluc
we have Now, energy
just inascase
already well.youthe
seen, viscous
are not all thatdeformation
clear source: Turbulence
exactly
source: how
Turbulence for
work
forthe the21st
the 21stcentury,
fromcentury,W.
dissipation W.K.K.
the George
fluc
terms
George
internal energy asaswell.
internal
urbulence.
Now,
energy
The
just in case
well.
overall
you are not Turbulence: lecture
Turbulence:
exchange
all that lecture
canexactly
clear
be understood
how the dissipation
Now, just in case you are not all that clear exactly how the dissipation terms
by terms
exploi
which treats
really −ρ⟨u
accomplish i ufor
this ⟩ the
as ainstantaneous
stress, the
theinstantaneous Reynolds
motion, ititmight stress.
mightbe usefultotoThe
useful term
examine
KEKE production
really
exactly
accomplish
production
exactly how
howthe
this
theabove
j for
aboveworks.
works. WeWebegin
beginby
motion,
bydecomposing
decomposingthe
be
themean
examine
meandeformation
deformation
rate
ratetensor
tensor∂U ∂Ui /∂x
i /∂xj jinto
intoits
itssymmetric
−ρ⟨u u ⟩∂U
symmetric and
and antisymmetric
/∂x
antisymmetric parts,
parts, exactly
exactlyasaswe
we
i j i j
did
didfor
forthe
theinstantaneous
instantaneousdeformation
deformationrateratetensor
tensorininChapter
Chapter3;3;i.e.,
i.e.,
hought of as the working of ∂U∂U ithe
i= S Reynolds
+ Ω stress against the
(4.26) m
∂x = Sijij + Ωijij (4.26)
adient of the flow, exactly as∂xjthe viscous stresses resist deforma
j

where
wherethe
ntaneous themeanmeanstrain
velocity rate
rateSS
gradients.
strain ijijisisdefined
definedby
This energy
by expended against the R
!! ""
ing deformation by the mean SSijij==
1motion
1 ∂U∂U i i ∂U ends
++ ∂Uj
j up in the fluctuating (4.27)
m
2 ∂x ∂x (4.27)
while that expended against viscous 2 ∂xj j stresses∂xi i goes directly to inte
and
we haveandthe mean
meanrotation
already
the seen,rate
rotation isisdefined
the
rate definedby
viscous by deformation work from the flu
!! ""
or dissipation) will eventually ΩΩ ijij==
11send∂U∂U i i this
−−
∂U∂U j j fluctuating kinetic ener
(4.28)
2 ∂x ∂x (4.28)
nergy as well. 2 ∂xj j ∂xi i
justSince
in
Sincecase
ijijisisyou
ΩΩ are notand
antisymmetric
antisymmetric all−⟨u
and that
−⟨u iu iu clear
j ⟩j ⟩is exactly
issymmetric,
symmetric, theirhow
their the isdissipatio
contraction
contraction iszero
zerososo
ititfollows
followsthat:
omplish thisthat: for the instantaneous ∂U∂U ii motion, it might be useful to
−⟨u
−⟨u iu j ⟩j ⟩
iu
==−⟨u −⟨uiu j ⟩S
iu j ⟩S
ijij (4.29)
(4.29)
ow the above works. We begin ∂x∂xj j by decomposing the mean defo
or ∂UiEquation
/∂x
Equation
j
4.29
into isisan
4.29its analog
analogtotothe
symmetric
an theandmean viscous
viscousdissipation
meanantisymmetric dissipationtermparts,
term given
given for
exactl
for
source: Turbulence for the 21st century, W. K. George
incompressible
incompressibleflow
flowby:
by: source: Turbulence for the 21st century, W. K. George
of energy by the mean flow is usually negligible compared to the energy lost to
Turbulence: lecture
Turbulence: lecture
the turbulence through the Reynolds stress terms. (Remember, there is a term
ntisymmetric
exactly like this inand −⟨uenergy
ii ujj ⟩equation
is symmetric,
ntisymmetric and −⟨u u ⟩ is symmetric, their contracti
the kinetic for the fluctuating their motion,contracti
but in-
KE production
volving
t: for almostonly fluctuating quantities; namely, 2µ⟨sij sij ⟩.) We shall show later that
t: always in turbulent flow, ⟨sij sij ⟩ >> Sij Sij . What this means is that
∂U ∂U iflow is almost entirely due to the turbulence.
i
the energy dissipation
−⟨u u ⟩ −⟨uii ujj ⟩
in a
= −⟨u u ⟩Sturbulent = −⟨uii ujj ⟩Sijij
There is a very important difference
∂x ∂xjj between equations 4.29 and 4.30. Whereas
the effect of the viscous stress working against the deformation (in a Newtonian
The
4.29 product
fluid)is to of
an analog
is always twoenergy
remove toterms
the from here
mean
the
4.29 is an analog to the mean viscous dissipation ter flowcan beSij positive
viscous
(since Sij > dissipation
0 always),or the
negative
effectter
of the Reynolds stress working against the mean gradient can be of either sign, at
le flow
least inby:
le flow by: How do
principle. Thatweis, itknow
can either that thisenergy
transfer is production?
from the mean motion to
the fluctuating motion, or vice versa.
Almost always (and(v)
(v)
∂U
especially
∂U
i in situations
(v)
i = Therefore,
(v) S −⟨u
of engineering importance), −⟨ui uj ⟩
and Sij have theTopposite
ijT =T
sign. T ij ij = i u2µS
S = 2µS S j ⟩Sij > ij 0Salmost
ij always, so ki-
ij ∂x ij ij ij ij
netic energy is removed from ∂x
jj
the mean motion and added to the fluctuations.
Since the term −⟨ui uj ⟩∂Ui /∂xj usually acts to increase the turbulence kinetic
energy, it is usually referred to as the “rate of turbulence energy production”, or
show that this term transfers
how that this term transfers (or dissipates) the mean ki
simply the “production”.
(or dissipates) the mean ki
ternalNow energy, since exactly how thethe
ternal energy, since exactly the same term appears with
that we have identified same
averaged termaccount
equations appears with
for the ‘pro-
duction’ of turbulence energy from the mean motion, it is tempting to think we
have understood the problem. In fact, ‘labeling’ phenomena is not the same as
‘understanding’ them. The manner in which the turbulence motions cause this
exchange of kinetic energy between the mean and fluctuating motions varies from
source:Turbulence
source: Turbulence for
for the
the 21st
21st century,
century,W.
W.K.
K.George
George
flow to flow, and is really very poorly understood. Saying that it is the Reynolds
Turbulence: lecture
The role of the pressure strain rate terms can best be illustrated by looking at a
Turbulence: lecture
simple example. Consider a simple homogeneous shear flow in which Ui = U (x2 )δ1i
and in which the turbulence is homogeneous. For this flow, the assumption of
KE
KE production
production
homogeneity insures that all terms involving gradients of average quantities vanish
(except for dU1 /dx2 ). This leaves only the pressure-strain rate, production and
Simple
Simple homogeneous
homogeneous
dissipation terms; thereforeshear
shear flow
flow
equations (jet,
4.37mixing-layer,
(jet,
4.36, and 4.40 reduce pipe,
mixing-layer, pipe,channel)
to: channel)
1-component:
! "
∂⟨u21 ⟩ ∂u2 ∂u3 ∂U1
= − ⟨p ⟩ + ⟨p ⟩ − ⟨u1 u2 ⟩ − ϵ1 (4.41)
∂t ∂x2 ∂x3 ∂x2
2-component:

∂⟨u22 ⟩ ∂u2
= +⟨p ⟩ − ϵ2 (4.42)
∂t ∂x2
3-component:
∂⟨u23 ⟩ ∂u3
= +⟨p ⟩ − ϵ3 (4.43)
∂t ∂x3
where

ϵ1 ≡ 2ν⟨s1j s1j ⟩ (4.44)


ϵ2 ≡ 2ν⟨s2j s2j ⟩ (4.45)
ϵ3 ≡ 2ν⟨s3j s3j ⟩ (4.46)

It is immediately apparent that only ⟨u21 ⟩ can source:


directly
source: receive
Turbulence
Turbulence forenergy
for the
the21st from
21stcentury, the
century,W.
W.K.K.George
George
Turbulence:lecture
lecture
The role of the pressure strain rate terms can best be illustrated by looking at a
Turbulence:
simple example. Consider a simple homogeneous shear flow in which Ui = U (x2 )δ1i
and in which the turbulence is homogeneous. For this flow, the assumption of
KEproduction
KE production
homogeneity insures that all terms involving gradients of average quantities vanish
(except for dU1 /dx2 ). This leaves only the pressure-strain rate, production and
Simple
Simple homogeneous
homogeneous
dissipation shearflow
terms; thereforeshear
equationsflow (jet,
4.36,(jet, mixing-layer,
4.37 mixing-layer,
and pipe,channel)
pipe,
4.40 reduce to: channel)
1-component:
! "
∂⟨u21 ⟩ ∂u2 ∂u3 ∂U1
= − ⟨p ⟩ + ⟨p ⟩ − ⟨u1 u2 ⟩ − ϵ1 (4.41)
∂t ∂x2 ∂x3 ∂x2
2-component:

∂⟨u22 ⟩ ∂u2
= +⟨p ⟩ − ϵ2 (4.42)
∂t ∂x2
3-component:
∂⟨u23 ⟩ ∂u3
= +⟨p ⟩ − ϵ3 (4.43)
∂t ∂x3
where
Productiononly
Production onlyhappens
happensininstreamwise
streamwisecomponent
component
ϵ1 ≡ 2ν⟨s1j s1j ⟩ (4.44)
Pressureredistributes
Pressure redistributesenergy
energybetween
betweenthe
thecomponents
components
ϵ2 ≡ 2ν⟨s2j s2j ⟩ (4.45)
ϵ3 ≡ 2ν⟨s3j s3j ⟩ (4.46)

It is immediately apparent that only ⟨u21 ⟩ can source:


directly
source: receiveforfor
Turbulence
Turbulence energy 21stfrom
the21st
the the
century,
century, W.W.
K.K. George
George
ence is homogeneous. For this flow, the assumption of
Turbulence:
t all terms involving gradients of averagelecture
quantities vanish
his leaves only the pressure-strain rate, production and
KE
s shear production
flow
4.36,(jet,
4.37 mixing-layer, pipe, channel)

#
ore equations and 4.40 reduce to:
Simple homogeneous shear flow (jet, mixing-layer, pipe, channel)
! "
∂u2 has ∂uto3 be ∂U1 positive
− ⟨p ⟩ + ⟨p ⟩ − ⟨u1 u2 ⟩ − ϵ1 (4.41)
∂x2 negative

1 12
∂x3 ∂x2

That means, on average, when u1 is negative u2 is positive


∂u2 and vice-versa
+⟨p ⟩ − ϵ2 (4.42)

⟨pui⟩δij − ⟨q uj⟩+ν⟨sijui⟩ dV
Recall∂xfrom
2 signal analysis, Contribution by fluctuations
Z 1 across different frequencies to
hu1 u2 i = û⇤1 (f ) û2 (f ) df the mean shear stress
∂u3 1
+⟨p ⟩ − ϵ3 (4.43)
∂x3
Without mean shear, we have no production
y happens in streamwise component source: Turbulence for the 21st century, W. K. George
ϵ ≡ 2ν⟨s s ⟩ (4.44)
e already averages.
actly the same manner Turbulence:
Turbulence:
∂uithat we rearranged lecture
lecture
the terms in the equation
⟨sij ⟩ = ⟨sij sij ⟩ + ⟨sij ωij ⟩
inetic energy of the fluctuations,
∂xj we can rearrange the equation for the
KE dissipation
KEofproduction
nergy the mean flow to obtain:
= ⟨sij sij ⟩ (4.4)
! "
ow it is ∂customary to∂ define a new variable k, the average fluctuating kinetic
+ Uj K=
y per unit
∂t mass, by:∂xj
# $
∂1 1 1 2 11 2
k ≡ ⟨ui− ui ⟩ ⟨P
= U⟨qi ⟩δ⟩ − [⟨u
ij = ⟨u1i u⟩ j+⟩U⟨u
i+
2 2ν⟨S2 U ⟩
2 ⟩ + ⟨u3ij⟩] i (4.5)
∂x2j ρ 2 22
∂Uiinserting this definition, the equation for the
Positive here +⟨ui uj ⟩ ∂x − 2ν⟨Sij Sij ⟩
y dividing equation 4.1 by 2 and (4.23)
ge kinetic energy per unit massj of the fluctuating motion can be re-written
The
! viscous" dissipation
# in the mean is usually very $ small
∂ ∂ ∂ 1 21 1 1
compared
+ Uj to the
k = Kviscous
≡ Q =dissipation
U U 2 in the fluctuations
− ⟨pui ⟩δij − ⟨q uj ⟩ + 2ν⟨sij ui ⟩
i i (4.24)
∂t ∂xj ∂xj 2 ρ 2 2
hsij−⟨u
sij iu >>
of all of the terms can immediately hS S
be irecognized
∂U ij ij i
since each term has its
Negative here i j⟩
art in the equation for the average fluctuating
∂xj
− 2ν⟨sij sij ⟩
kinetic energy.
(4.6)
Almost
parison all of the 4.23
of equations dissipation
and 4.6 in turbulence
reveals that theis termvia the −⟨ufluctuations
i uj ⟩∂Ui /∂xj
he role of each of these terms will be examined in detail later.
n the equations for the kinetic energy of BOTH the mean and the fluc- First note that
ernative
There is,form of this equation
however, one VERY canimportant
be derivedsource:
by leaving
Turbulence the
difference. for theviscous
This 21st stress
century,
“production” in
W. K. George
Turbulence: lecture
now that it never occurs in nature. Nonetheless, many flows can be assumed
Turbulence: lecture
to be homogeneous at the scales of turbulence which are important to this term,
so-called local homogeniety.
Each term in the equation for the kinetic energy of the turbulence has a distinct
Kinetic energy of turbulent fluctuations
role to play in the overall kinetic energy balance. Briefly these are:

• Rate of change of kinetic energy per unit mass due to non-stationarity; i.e.,
time dependence of the mean:
∂k
(4.9)
∂t

• Rate of change of kinetic energy per unit mass due to convection (or advec-
tion) by the mean flow through an inhomogenous field :

∂k
Uj (4.10)
∂xj

• Transport of kinetic energy in an inhomogeneous field due respectively to


the pressure fluctuations, the turbulence itself, and the viscous stresses:
! "
∂ 1 1 2
− ⟨pui ⟩δij − ⟨q uj ⟩ + 2ν⟨sij ui ⟩ (4.11)
∂xj ρ 2

• Rate of production of turbulence kinetic energy from the mean flow (gradi-
ent):
∂Ui
−⟨ui uj ⟩ (4.12)
∂xj

• Rate of dissipation of turbulence kinetic energy per unit mass due to viscous
stresses:
ϵ ≡ 2ν⟨sij sij ⟩ (4.13)
source: Turbulence
source: Turbulence for
for the
the 21st
21st century,
century,W.
W. K.
K. George
George
These terms will be discussed in detail in the succeeding sections, and the role of
Turbulence:
Turbulence:
4.2 The Rate of Dissipationlecture
lecture
of the Turbulence
KE
KE dissipation
dissipation
Kinetic Energy.
The last term in the equation for the kinetic energy of the turbulence has been
identified as the rate of dissipation of the turbulence energy per unit mass; i.e.,
! "
∂ui ∂ui ∂ui ∂uj
ϵ = 2ν⟨sij sij ⟩ = ν ⟨ ⟩+⟨ ⟩ (4.14)
∂xj ∂xj ∂xj ∂xi

It is easy to see that ϵ ≥ 0 always, since it is a sum of the average of squared


quantities onlyThis (i.e., term
This ⟨sij sij ≥is
term is0).always
always greater
greater
Also, since it occursthan
than
on thezero
zero
right hand side of
the kinetic energy equation for the fluctuating motions preceded by a minus sign,
Since
Since it
it is
is preceded
preceded by
by negative
negative sign,
sign, it
it is
is always
it is clear that it can act only to reduce the kinetic energy of the flow. Therefore
always
it causes a negative rate of change of kinetic energy; hence the name dissipation.
aa sink
sink
Physically, energy is dissipated because of the work done by the fluctuating
Energy
viscous stresses is
Energy in dissipated
is dissipated
resisting because
because
deformation of the thefluidturbulent
the material by viscous
turbulent viscous
the fluctuating
stress
strain rates; is
i.e.,doing work to resist the deformation of the
stress is doing work toϵ =resist (v) the deformation of the
⟨τij sij ⟩ (4.15)
fluid
fluid element
element by
by turbulent
turbulent strain
strain rate
rate
This reduces to equation 4.14 only for a Newtonian fluid. In non-Newtonian
fluids, portions of this product may not be negative implying that it may not all
represent an irrecoverable loss of fluctuating kinetic energy.
source: Turbulence for the 21st century, W. K. George
source: Turbulence for the 21st century, W. K. George
Recall
Recall (Section
(Section 4.1)
4.1) that
that the
the dissipation
dissipation rate
rate scales
scales with
with velocity
bevelocity –and
and length
–length scales
scales
ofof
ofof
vorticity.
vorticity.
vorticity.Note
vorticity.
Note
Note
Note
that
that
that
that
the
the
the
the
former
former
former
former Turbulence:
must
must
must
must lecture
therefore
therefore
Turbulence: lecture
Turbulence: lecture
therefore
therefore be
be
be
positive
positive
positive
positive –
vortex
vortex
– vortex
vortex
stretching
stretching
stretching
stretching
appropriate
appropriate to the
to turbulent
amplifies
amplifies the large motions
large motions
turbulent –inεin
vorticity
vorticity – ε
the ∝
∝the 3 3
. as
Kolmogorov
L/ .LKolmogorov
u'/ mean,
u'mean, as (1941)
(1941)
expected.
expected. – K41
– K41 – proposed
– proposed
amplifies turbulent vorticity in the mean, as expected.
KE
KE
that:
hat: KE dissipation
amplifies turbulent
dissipation
dissipation mostly
vorticity
mostly
mostly
in in
the
in insmall
mean,
small
small scales
as expected.
scales
scales
1. 1.Hypothesis
Hypothesis ofof local
local isotropy:AtAt
isotropy: sufficiently
sufficiently high
high Re,
Re, the
the small-scale
small-scale motions
motions are
are
DEFINITIONS
DEFINITIONS ofof Isotropy
Isotropy and
and homogeneity.
homogeneity.
DEFINITIONS
DEFINITIONS ofof Isotropy
Isotropy and
and homogeneity.
homogeneity.
statistically isotropic (say for length scales
statistically isotropic (say for length scales l<ll<lEI).
EI).

1.
2. 2.1st1. 1.Similarity
AA turbulent
turbulent
1Similarity
st flow
flow is is
hypothesis:
hypothesis: deemed
deemed
In In every
every isotropic
isotropic if if
turbulent
turbulent allall
the
flow
flow the
atat statistics
statistics ofof
sufficiently
sufficiently the
the
high
high fluctuating
fluctuating
Re,
Re, thethe statistics
statistics
1.AA turbulent
turbulent flow
flow is is deemed
deemed isotropic
isotropic if if
allall
thethe statistics
statistics ofof the
the fluctuating
fluctuating
of of
thethevariables
variables (e.g.
(e.g.
small-scale
small-scale ui’)
ui’)
motions
motions areare
have
have independent
independent
a a universal
universal of
ofform,
form, coordinate
coordinate
uniquely
uniquely system
system rotations
rotations
determined
determined by ν
by ν oror
and
and ε. ε.
variables
variables (e.g.
(e.g. ui’)
ui’) areare independent
independent ofof coordinate
coordinate system
system rotations
rotations oror
reflections
reflections – i.e.
– i.e. there
there η η
≡is(≡is
νno
(3no
ν/ ε/)preferred
3preferred
ε
1 / 41 / 4
) direction.
direction.
reflections
reflections – i.e.
– i.e. there
there is isnono preferred
preferred direction.
direction.
Thus:
Thus: v≡ v (≡εν(εν) ) 1/ 41/ 4
asas given
given inin Section
Section 4.14.1
2. 2. AA turbulent
turbulent flow
flow isτisτ=said
said =
(ν to

/ to
εbe
/)ε be homogeneous
homogeneous
1 / 21 / 2
) if if
allall
the the fluctuating
fluctuating statistics
statistics are
are
2. 2. AA turbulent
turbulent flow
flow is is said
said
η η toto bebe homogeneous
homogeneous if if
allall
the the fluctuating
fluctuating statistics
statistics areare
The
The independent
independent
region
region l<l
l<l is is of spatial
ofreferred
spatial
referred totoasposition
position
asthe the (and,
(and,
universal
universal strictly,
strictly, allall
equilibrium
equilibrium joint
joint probability
probability
range.
range. density
density
independent
independentEI EI ofof spatial
spatial position
position (and,
(and, strictly,
strictly, allall joint
joint probability
probability density
density
functions
functions remain
remain unchanged
unchanged if if distances
distances are
aredisplaced
displaced byby a constant
a constant
Recall
Recall that the
that
functions
functionsReynolds
the Reynolds
remain
remain number
number
unchanged
unchanged basedbased
if if onon
distancesthese
these
distances scales
are
arescales
displacedis
displaced unity
is unity
byby and and note
a constant
a constantnote thatthat
∂∂
velocity
velocity gradients
gradients
amount).
amount). So, of
for
So, the
of
forthedissipative
dissipative
example,
example, homogeneityeddies
eddies
homogeneity are
areO(O(
implies
implies
1/1/τ ητthat
).η that
).Recall
Recall
∂ ∂(u also
also
i 'u
(u the =ratios
')the
j ')j =
i 'u 0. ofof
0.ratios
amount).
amount).So, forfor
So, example,
example, homogeneity
homogeneity implies
implies thatthat∂x∂j x(u (u'u'u') =') 0.
j i ij j
= 0.
∂x∂j xj
smallest
smallest toto
largest
largest scales
scales (eqs.
(eqs. (4.3)
(4.3) and and (4.4)).
(4.4)).

3. 3.Since, atat
Since, sufficiently large
sufficiently Re,
large there
Re, must
there bebe
must asource:
a range
range
source:Turbulence
of
source: scales
of scales
Turbulence for
Turbulencefor the
which
which21st
thethe
for are
21st century,
21staremuch W.
much
century, K.
K.George
W.W.
century, George
K. Georg
! ! Recall
Recall " " 11
(Section
22
(Section 4.1)
3 3!
4.1)
! that
that "
the "dissipation
the dissipation
! ! rate " "scales
rate scales with
with velocity
velocity andand length
length scales
scales
!Recall "
Recall "(Section
(Section
2 4.1) !
4.1) !
that
that "
the "
the dissipation !
dissipation
2 ! "
rate "
ratescales!
scales
2 ! with
with" "
velocity
velocity and
and length
lengthscales
scales
∂u

∂u ∂uof
1
2
1
2
2vorticity. Note
1 1of vorticity.
⟩ = 1 1 1∂u
Note ⟨
∂u that
∂u
1that
2
1
2
1 the⟩former
1 1
= Turbulence:
Turbulence:
Turbulence:
2the former
1 1
Turbulence:
∂u⟨
∂u
∂u
∂u lecture
lecture
must
∂u
must
2
lecture2
2
!
2therefore
2 23 3 therefore
2
⟩ lecture
"be2 2positive
2 – vortex stretching
be positive – vortex stretching
∂u ∂u∂u ∂u ∂u
vative
ative
tive
appropriate ⟩
appropriate ⟩
appropriate
moments
appropriate =
moments=
moments
toto
the
to to⟨
the
the

the in
large
in
large
large
in
large ⟩
isotropic
motions
isotropic
motions
motions
⟩==
isotropic
motions–⟨ –


ε ε
ε∂u∝ε 3 3u'
turbulence
∝ ∝ u' / ⟩
.L ⟩
turbulence

–turbulence . Kolmogorov
/..Kolmogorov
LKolmogorov
Kolmogorov
1 1 1 (1941)
(1941)
(1941)
(1941) – – – K41
K41
K41
– K41 – – –1proposed
1
proposed
1
proposed
– proposed 2∂u
2 2
∂xamplifies 2
turbulent ∂x vorticity in2∂x
the ∂x
mean,
⟨⟨ ⟨ ∂x ⟩⟩ ⟩===−⟨
as expected. −⟨
u'u'/ L
/ L
∂x ∂x
KE1 1
KE KE 2
dissipation
1 1amplifies turbulent
dissipation
2
dissipation∂x ∂x 2 2
mostly
mostly
2
mostly
2
in
2 2vorticity in the 1mean, ∂x
in in
small
1
1 1
small
small asscales scales
expected.
scales −⟨ ⟩ ⟩
powerful
that:
werful
hat: KE
that: dissipation
that: influence
werful
werful influence
influence
influenceonon1the !
on !
!on
1soall
!
thethe
the∂u
∂u
" mostly
"
" "
velocity
velocity
velocity2
velocity
22 2
1
! !in !
derivative
derivative
derivative
1 ∂u
small
derivative
!
∂u" "
"
2 22
" scales
moments.
moments.
2
moments.
moments. ∂x
1 ∂x
1 In
1 In
In Infact,
fact,
fact,
fact, 2 2 2 1∂x
1∂x
1 ∂x∂x ∂x ∂x
mssindependent,
is independent,
independent, = 1 so ∂uall
⟨ of
1 of
1the
1
180 180 ⟩ 1
1the others
others
= ∂u
cancan
⟨ be
3 3be
3 3 expressed in terms of
CHAPTER
expressed
CHAPTER⟩ in terms
9. 9. HOMOGENEOUS
of
HOMOGENEOUS RANDO
RANDOM
independent,
1.Hypothesis =
Hypothesis
1.ifHypothesis
Hypothesis= so so⟨ ⟨
all
all
of ofofofthe
180the
local
local 180 ⟩ ⟩
others=
others
isotropy:
isotropy:= ⟨ ⟨
cancan
At be
Atbe ⟩
expressed
expressed
CHAPTER
sufficiently
CHAPTER

sufficiently ininterms
9.
high
high terms
9.
Re, Re, of
the of
HOMOGENEOUS
the small-scale
HOMOGENEOUS
small-scale RANDOM
motions
RANDOM
motions areare
1.,f1.
fwe we
wechoose
we choose
choose
DEFINITIONS
choose
of
toexpress
toto
local
of local
2∂x
2 express
2toexpress
express ∂x
of
isotropy:
∂x
the
isotropy:
the the
3Isotropy
the
3 2
others
3others
others 2
othersAt
and
in 2At
∂x
in
in
sufficiently
sufficiently
∂x
in ∂x
term
term
termterm
1 of of
homogeneity.
1of of
⟨[∂u
⟨[∂u
⟨[∂u
high
⟨[∂uhigh
/∂x
1 /∂x
/∂x
Re,
1 /∂x
]1
Re,
]
2 the
2 ]21⟩,
⟩, ⟩,
2 the
thesmall-scale
⟩,the
]the small-scale
the motions
motions
2
are
are
∂u
2 2 33 3 ∂u ∂u∂u ∂u ∂u
DEFINITIONS
statistically
statistically
statistically
of
isotropic
!
isotropic
statisticallyisotropic ! !
isotropic
3
! Isotropy
" "
"(say
(say
"(say
2
2 for
2(say
2 for
for
and
forlength
!
length!
length
1 1
homogeneity.
! "scales
length
! "
"
scales
2 2 l<l
scales
2
1
l<l
1
"2 l<ll<l).EI).
scales ).
EI).
1 1
===−⟨ −⟨ −⟨ ⟩ ⟩
! = =
" 1 1 1
1 ∂u 2

∂u⟨
"2"22 ⟨! ! "2"22 ⟩ =
∂u
!
∂u
2 2 "
2
⟩ =⟩ 1
=
1
1 ∂u 3
1
!
! ! ⟨ "2"22 ⟨
∂u
∂u⟨∂u" 3 3
3
EIEI
⟩! ⟩! " " ! ! (9.76)
(9.76)
" " ∂x∂x ∂x
!3 ! 3∂x ∂x
3"
∂x
"121 1
!! =
!
∂u
" 2
2
!
∂x
∂u
" 2 !
∂u2
" 2
∂x ⟩! !
∂u
"
∂u "
2 22 2
1
!
1!(9.76)
∂u
"
∂u "
2 22 2
1
!
1!
∂u
"
∂u "
2 22
2. 1.st1
∂u∂u
A Similarity
st turbulent
2. 1 ⟨11 ⟨Similarity
Similarity
∂u 1 2 2 ∂x
∂u ∂x
∂u
hypothesis:
flow
hypothesis:
∂u 2is3 deemed ∂u2 ∂u2
In
Inevery
∂u every ∂x3∂x
every
isotropic turbulent
turbulent
2 ∂u
if ∂u
all
2 ⟨ ⟨⟨if ⟨all flow
the
flow
1 1
at sufficiently
1
statistics1
atsufficiently
sufficiently
⟩sufficiently
=⟨(9.75)∂u∂u
of the
1
⟨ high
1
high
high 1 Re,
1
fluctuating
Re,
⟩= ∂u ∂u
thethe
∂u∂u
∂u
⟨ statistics
2
∂u
statistics
2
⟩ 3
2.2. st st
A
1
1.⟨ Similarity
⟨ turbulent
1 1
∂x ⟩ ⟩
=
⟩ =
=⟩⟨hypothesis:
=⟨
⟨ ⟨flow
2 2
∂x
3
hypothesis:
32 3⟩ ⟩==
is
⟩ ⟩⟨In
=
deemed
= ⟨
⟨ In ⟨ every
3 3 turbulent
3
∂x ⟩,2⟩,
isotropic
⟩, ⟩,
2
turbulent flow
∂x
1
flow
1
the
∂x ⟩at⟩⟩at= = (9.75)
statistics
=(9.75)
(9.75)
2

⟨2 of
1

2=
∂x
1
high
the
∂x ⟩
2 −⟨

=Re,
=
−⟨ ∂u
Re,the
⟨ the

fluctuating
⟩ = 2 ⟨2 2∂x
2
∂u
statistics
statistics
2∂x
2
2 ⟩ ⟩

1 33
= "2 ⟩ ⟩
∂x ∂x ∂x ∂x 1 2 ∂x 2 2 ∂x
ofof
∂x
of
ofthe
∂x
the
the
1 11
variables
the
1
small-scale
small-scale
small-scale
∂x
(e.g.
small-scale ∂x2u
motions
2
2
motions
i
(e.g.motions
2
’) are
motionshave
u ’) arehavehave
∂x
a
∂x 3 33
independent
have a a
3
universal
universal
universal
a universalofform, of
form,
form,
∂x1 11
coordinate
form, uniquely
uniquely
uniquely
uniquelysystem
2
system = ∂x
determined
!
determined
determined
! ! "
2
−⟨
rotations
determined
! "
"by
"
by by
2ν ν
2 2 by ν and
2
andν
or
and
! !ε
∂x
!
∂x
and!

1 1
ε∂u.
1
ε .
"
. 2 32∂x
" "
variables i independent coordinate
= 1
1 1 ∂u

1
⟨∂u
∂urotations
∂u
1 11 1
2
⟩ ⟩1∂x
=1∂x
1or

1∂u
3
⟨∂u
∂u 3
3
3∂x 3
3∂x
3
2
⟩2 2

2 2
reflections – i.e. there η η≡isη
≡ no
ν ≡
3 33ν ν
ε ε ε
preferred
3 1 / 41 / 4
( / ) direction. =
== ⟨ ⟨2 ∂x⟩3 ⟩== ⟨ ⟨2 ∂x⟩1 ⟩ =
reflections – i.e. thereηis≡ no (ν preferred
( / ε)
/ )
1 / 14 / 4
( / ) direction. 2 22 ∂x∂x ∂x
3 33 2 22 ∂x∂x∂x 1 1 1 "
! " ! " !
! It
!
ItItThus:
!
Thus:
∂u∂u ∂ufollows
"
follows
Thus:
follows
Thus:
1
"
"
2
1
22 2 immediately
immediately
immediately
∂uv∂u v ≡(εν
v≡∂u
≡v
1∂u( ≡ εν
εν (
1/ 1/
(∂u
1 ) εν
41/ 44
∂u) 2
)2
1/ 4
) thethe the dissipation
dissipation
dissipation
asas as
asgiven
=
given
given
1 1
1
!
!given
!
1∂u∂u
⟨in
∂u
2
in
∂u
in
"
"2in
"
2 for
Section
2
2
2
2
Sectionfor
⟩1for
2 Section
Section
=14.1
1
!
4.1
∂u
⟨∂u
∂u 3
isotrop
"2"
isotropi
! ! 4.1
isotropic
14.1 ∂u3
3
" 2
2
3
2

⟨ 11
⟨ A turbulent ⟩
⟩ ==flow = −⟨
−⟨isτ said11 22 ⟩
⟩ = == ⟨ ⟨
⟨ ⟩ ⟩
=
⟩ =
= ⟨ ⟨
⟨ ⟩ ⟩

simply⟨
2. ⟩
expressed −⟨ τ = τ
as:
= ν =
ν
to ε⟩ν
be
( ε / ε 1 /2
homogeneous
)
1 / 121 // 22
if all the 2 2∂x∂x ∂x
fluctuating
∂x 2∂x∂x
statistics
2 ∂x are
simply
imply 2. ∂x ∂x
A ∂xexpressed
expressed
1turbulent
11
1 flow is ∂xas:
∂x as:
τ ∂x
said
η =
2 22∂x
η ηη ( ∂x(
(ν/∂x/
/
2to1 be )ε )
) 2 2
11 homogeneous if all the fluctuating statistics are
1 3 3
3
3 2 2 ∂x 2 2
2
2

The independent
The region
region l<l
l<l isof
is spatial
and
and
∂u
referred
andand
∂u
referred position
to∂u ∂u
to
as as
thethe(and, strictly,
universal
universal all joint
equilibrium
equilibrium probability
range.
range. density
The region l<l
The independent
region l<l is
EI referred
EI
EIEI isof
= ∂u
referred
spatial
−⟨ to
2 2to
2∂uas
2 asthe universal
3 3 the⟩ (and,
position3 3 universal equilibrium
equilibrium
strictly, range.
all jointrange. probability ! density
"
functions
= = −⟨
remain
−⟨
unchanged
∂x ∂x⟩ ⟩
if distances are displaced by ! ! a constant " "2 2
2 2
Recall
Recall
Recall
Recall that
that
that
that
functions thethe
the
the Reynolds
Reynolds
Reynolds
Reynolds
remain ∂x
∂x ∂u
number
number
∂u
number
3number
unchanged ∂x
33∂x1 11based
∂u3
if based
based
1based
distanceson
onthese
onon these
these
these
are
!!
scales
scales
scales
! !"
scales
displaced
∂u
"
2"22is
" is
2
is
unity
is unity
unity
unity
by and
a and
and andnote
constant
∂u1
note
note
note
1
1 1
∂u
that that
that
that
velocity
velocity
velocity gradients
gradients
gradients
amount).
velocity So,
gradients =ofofϵϵ=
of
the
for
−⟨
of
ϵ==
∂u
the
−⟨
the
15ν⟨
the
2
∂u
15ν⟨
15ν⟨
∂u 2∂u∂u
2 ⟩⟩ ⟩
∂u
dissipative
dissipative
dissipative
2
example, 3
dissipative
3
3
3
⟩ eddies
eddies
eddies
homogeneity
⟩ eddies are
are⟨ ⟨
O(
∂u
are
⟨ O(
implies
are O(
∂u
∂u
⟨ O(
1
τ
1/
1
1
τ
1/
τ⟩
1
). τ

).

that
). =⟩).Recall
=
Recall
= =
∂Recall
−⟨
Recall −⟨
∂u
−⟨
also
∂u
∂u1
also
(9.77)
−⟨ also
∂u
1
also
1
the
∂u
1
∂u2
the
the ⟩
2
the
2 ⟩
2
⟩ratios
ratios

ratios
ratios ofof
of of
amount). =
= −⟨ example,
So, for ∂x
∂x ∂x 3
∂x
∂x
∂x
∂x∂x
3 ∂x⟩homogeneity implies
2 2
1/
1/
∂x
∂x∂x ∂x
1 1
η η
η
1 that
1
η
∂ x
(9.77)
(9.77)

j
(u
(u ∂x 'u
∂x
i ∂x
i 'u j
')
∂x
2j ∂x
')
2
21
=
∂x
=
0.
∂x
21 1 1
∂x
1
1 0.1
1
smallest
smallest to to largest
largest scales
scales
33 22
(eqs.(eqs. (4.3)
(4.3) andand (4.4)).
(4.4)). ∂xj ∂u∂u ∂u ∂u
∂u 23 ∂u
∂u
smallest
smallesttotolargest
largestscales
scales(eqs.
(eqs.(4.3)(4.3)and and(4.4)).
(4.4)). = =
= −⟨ −⟨
−⟨ ∂x ∂x 2 2
2 ∂u

3
3
3
⟩⟩ ⟩
But
But
But
ely any
iately
any
any ofofof
telythethethe
thethe
the other
other
other derivatives
dissipation
derivatives
derivatives
dissipation
dissipation could
forfor
could
could
for have
have
have been
been
isotropic
isotropic use
isotropic
been used
used turbulence
turbulence
turbulence can
can can bebebe= −⟨ quite
quite
quite ∂x ∂x
∂x3 ∂x ∂x
3 ∂x
3 31 1 1
1
3. Since, at sufficiently large Re, there must be a range of scales ∂u
which ∂u ∂u ∂u
3.3. Since,
Since,atatsufficiently
sufficientlylargelargeRe, Re,there
theremust source:
mustbebea arange
range Turbulence
source:
source:of
source: ofscales
scales
Turbulence
Turbulence for
Turbulencefor the
∂u
for
which
which
for 21st
∂u
the
the 21st
2 ∂u
the areare
century,
21st
∂u
2are
21st
2 3 much
23century,
much
3
century, much
3 W.
W.K.
century,
W. George
K.W.
K. K. Geo
Georg
George
Turbulence:
Turbulence: lecture
lecture
Turbulence: lecture
KEKE dissipation
dissipation
4. Scales of Motion in Turbulent Flow
Scales
4.1 Generalof turbulent
Scales of turbulent motion
Comments motion
4. Scales of Motion in Turbulent Flow
4.1 General Comments
We have noted the wide range of time scales that exist in typical turbulent flows.
We have noted the wide range of time scales that exist in typical turbulent flows.
Consider a turbulent shear flow:
Consider a turbulent shear flow:
U
U λ
Remember the
Remember the
λ

δ ∼ Lx
power spectrum
power spectrum
δ ∼ Lx

U=0
U=0
x
Supposed to give
Supposed to give us us
x
frequency by frequency
frequency by frequency
The following length and velocity scales apply:
The following length and velocity scales apply:
distribution of energy
distribution of energy
x, U global length and velocity scales content (or scale-by-scale)
content (or scale-by-scale)
x, U global length and velocity scales
Lx, u’ length and velocity scales of the large motions
Lx, u’ length and velocity scales of the large motions
λ, u’ length and velocity scales for dissipation process
λ, u’ length and velocity scales for dissipation process
η, v length and velocity scales for the smallest (viscous) motions – Kolmogorov
η, v length and velocity scales for the smallest (viscous) motions – Kolmogorov
scales – see later
scales – see later
ξ, a length and velocity scales of the molecular motions (mean free path and
ξ, a source:
length and velocity scales of the molecular motions (mean free path and Turbulence for the 21st century, W. K. George
speed
Wednesday, 23 October 13 of sound)
source: Turbulence for the 21st century, W. K. George
Turbulence:lecture
Turbulence: lecture
KE dissipation
KE dissipation
Scales of turbulent motion
Scales of turbulent motion A lot of the energy
A lot of the energy is is
in the large-scales
in the large-scales
ZZ 1
2 1
hu
2 i=
hu i = E uu (f
E (f )df )df
0 uu
0

Apply Taylor’s hypothesis


Apply Taylor’s hypothesis
Z 1
2
hu i = Euu (k)dk
0

This log-log representation


This log-log representation
shows that energy distrbiution
shows that energy distrbiution
source: Turbulence for the 21st century, W. K. George
source: Turbulence for the 21st century, W. K. George
Wednesday, 23 October 13
Turbulence:
Turbulence: lecture
lecture
Turbulence: lecture
KE
KE dissipation
dissipation
KE dissipation
Scales
Scales of
Scales of turbulent
of turbulent motion
turbulent motion
motion
Under
Under homogeneous
Under homogeneous isotropic
homogeneous isotropic assumptions
isotropic assumptions
assumptions
ZZ 1
Z 1 ZZ 1
Z 1
1 1
KE
KE
KE==
= E
E
Euu
uu
uu
dk
dk
dk =
=
= k
kk E
E
E d(ln(k))
d(ln(k))
uu d(ln(k))
uu
uu
00
0 00
0
*
* ✓✓
*✓ ◆ +
++ *
**✓ ◆ +
+
◆◆2 ✓✓ ◆◆2 +
@u
@u
@u11
1
22 @u
@u
@u11
1
22
Dissipation
Dissipation
Dissipation== 2⌫hs
=2⌫hs hs
2⌫hshsij
ij
hs s
ijs
ij s
ijsij
ij
sij
ij i
i = 15
iiii==15
15 = 7.5
==7.5
7.5
ij ij
ij @x
@x
@x11
1 @x
@x
@x222
In
In fourier
Infourier space,
fourierspace, @u
space, @u /@x
@u111/@x isis just
/@x111 is just premultiplication
just premultiplication with
premultiplication with wavenumber
with wavenumber
wavenumber
*
* ✓✓
*✓ ◆ +
++ Z Z
@u ◆◆2
22 ZZ 1
11 ZZ 1
11
15⌫ @u
@u 1
11 = 15⌫ k 2
22Euu (k)dk = 15⌫ k 33
3 EEuu (k)d(ln(k))
15⌫
15⌫ @x == 15⌫
15⌫ kk E
E uu
uu(k)dk
(k)dk =
= 15⌫
15⌫ kk E uu
uu(k)d(ln(k))
(k)d(ln(k))
@x
@x111 00
0 00
0

Let’s
Let’s see
Let’s see where
see where the
where the kinetic
the kinetic energy
kinetic energy and
energy and dissipation
and dissipation are
dissipation are
are
distrbuted
distrbuted across
distrbuted across the
across the scales
the scales
scales
Wednesday, 23 October 13
Turbulence: lecture
KE dissipation
Scales of turbulent motion
spread increases with
Reynolds number

kE kE
3

ln(k)
This shows that the dissipation is in the small-scales
How does it get there?
Turbulence:lecture
Turbulence: lecture
Turbulence: lecture
Turbulence: lecture
KEdissipation
KE dissipation
KE dissipation
KE dissipation
Scalesofofturbulent
Scales turbulentmotion
motion
Scales
Scales of turbulent
of turbulent motion
motion
consider the continuity
consider the continuity equation, rr
equation, · u· =
u= 0 and
0 and take
take thethe r ·r!·=
curl,
curl, ! 0=
consider the continuity equation, r · u = 0 and take the curl, r · ! = 0
Integratethis
Integrate thisover
overthe
the volume,
volume, and
and apply
apply divergence
divergence theorem
theorem,
Integrate this
Integrate this over
over the
the volume,
volume, and
and apply
apply divergence
divergence theorem,
theorem,
!A
!A == constant
constant
!A == constant
constant
t=t =t1t1
!A t =t =
t t>2 >
t t1
2 1
tt == tt11 tt == tt22 >> tt11

!1!, 1A
, 1A1 !2!>
2 >
!1!, 1A, 2A<
2 <
A1A1
! 1 , A1 !2 > !1 , A2 < A1
Thefluctuating
The fluctuatingfield
fieldstretches
stretchesthe theintialintial vortex
vortex and
and throu
through
The fluctuating
The fluctuating field
field
thisprocess stretches
stretches
processtransfers the
the
transfersenergy intial
intial
energytoto vortex
vortex
the and
and
smaller through
through
scales
this the smaller scales
this process
this process transfers
transfers energy
energy to to the the smaller
smaller scales
scales
Thisprocess
This processcontinues
continuesuntil
untilit itis islimited
limited
byby viscosity
viscosity
This process
This process continues
Wednesday,
Wednesday,
continues until
23 October
23 October 13 13
until itit isis limited
limited byby viscosity
viscosity
Turbulence: lecture
KE dissipation
Scales of turbulent motion
“Big whorls have little whorls
that feed on their velocity, and
little whorls have lesser
whorls and so on to viscosity”
-L.F. Richardson, 1922
This describes the cascade process
"Great fleas have little fleas
upon their backs to bite 'em,
And little fleas have lesser
fleas, and so ad infinitum."
- Jonathan Swift, 1915
1 ∂U1i ∂U∂U
i j ∂Uj
Sij =Sij = + + Turbulence: lecture (4.27)(4.27)
2 ∂x2j ∂xTurbulence:
∂x
j i ∂xi lecture
enE isKE isdissipation
dissipation
defined
rate by by
defined
! ! " "
It turns
turns out
∂U1i that
out1 that the
∂U∂U the dissipation
i dissipation
j ∂U j rate rate is determined
is determined by the
by the
Ωij =Ωij = − − (4.28)(4.28)
large-scales
arge-scales2 ∂x and
2j ∂xand
not not
j i the
∂x the small-scales
∂xismall-scales eveneven
thoughthough
the the
and −⟨u
etric
dissipation
dissipation is
⟩ isusymmetric,
and u−⟨u
is happening
happening at
their their
⟩ is symmetric,
at
the the small-scales
small-scales
contraction is zero
contraction is so
zero so
i j i j

03 03
∂Ui ∂Ui
scales
u
as
u (4.29)(4.29)
−⟨ui u−⟨u
j ⟩ i uj ⟩= −⟨u
∂xj ∂xj
scales
j ⟩Siijuj ⟩S
=i u−⟨u ij as
Lx Lx
nalog to the
an analog to mean viscous
the mean dissipation
viscous term term
dissipation givengiven
for for
Then,
y: Then, if dissipation
if dissipation rate rate is controlled
is controlled by process,
by this this process,
dimensionally,
mensionally,
∂U dissipation
dissipation
i ∂U (v)
rate rate should
should also also
scalescale
the the
same same
way way
(v) (v) i (v)
Tij Tij = Tij=STijij=S2µS Sij ij Sij
ij =ij2µS (4.30)(4.30)
03
∂xj ∂xj 03 u
u
t term transfers
2⌫hs ij s ij
(or dissipates)
i = ✏ ⇠
the mean
⇡ A
this term transfers (or dissipates) Lkinetic
the mean
x
energy
kinetic L
energy
x
since since
ergy, exactly the same
exactly term term
the same appears with with
appears the opposite
the opposite
Cornerstone
Cornerstone of turbulence
of turbulence theory
theory
Turbulence:
Turbulence:
Turbulence: lecture
lecture
lecture
Turbulence: lecture
KEKE
KE dissipation
dissipation
dissipation
KE dissipation
Lets
Lets
Lets look
look
look at
Letsat
at the
thethe scaling
scaling for dissipation
theforscaling
scaling
look at dissipation itself itself
itself
for dissipation
03 u03
u u03
Turbulence: lecture
⌫hs⌫hs ijisij
ij sij i ⇠ij sij i ⇠
⇠⌫hs
Lx Lx Lx
n Lets
Lets
saysay
that
Letsthat
saythe
the velocity
velocity
that thescalescale isscale
is the
velocity the same,
same, but,but,
is the thethe
same, length
length
but, the
scale What
is other
different, scaling
(sort law
of can
obvious we think
from of?
the spectrum)
scale is different,
scale is (sort
different,of obvious
(sort
ook at the scaling for dissipation itself of from the
obvious spectrum)
from the spec
*✓ *✓ ◆ +◆* +
2✓ ◆
02u02 03
030302u 03 2
@u1@u1
u uuu u ✏ ⇠ ✏⌫ ⇠ ⌫ ✏ ⇠ ⌫
@u 1
⌫hs⌫ij s⌫ij i ⇠

2

⌫ ⇠ @x1@x1 @x1
2 LLx xL
2x Lx
Recall,
Recall, fromfrom
the the signal
signal
Recall, analysis
analysis
from the signa
the velocity
: Taylorscale
: Taylor ismicroscale
: Taylorthe same,
microscalebut, the length
microscale *✓ *✓ ◆ +◆* +
2 ✓ 2 ◆22 +
2
ent,
Not the(sort of obvious
smallest scale from the spectrum)
of motion. @u1 @u1 hu@ui1hu i
⇠ ⇠2 2 ⇠
Not
Not the the
Notsmallest
smallest
the scale
scale of
smallest of motion.
motion.
scale of motion. @x @x1 @x1
Characteristic
03 scale for dissipation
* ✓ ◆2 + 1

Charactersistic
Charactersistic
u scale
scale forfor dissipation
dissipation
Charactersistic scale
✏⇠⌫ for dissipation
@u 1 : : Taylor
Taylor microscale
microscale
: Taylor mi
⇠ @x1
Turbulence: lecture
Turbulence:lecture
Turbulence: lecture
KE dissipation
KEdissipation
KE dissipation
spread increases
spreadincreases
spread with
increaseswith Reynolds
withReynolds
Reynolds
0202 0303 number
number
number
uu uu
⌫⌫ 22 ⇠⇠
LLxx
EEE DD
D
IfIfIfwe
we rearrange
werearrange this
rearrangethis equation,
thisequation,
equation,
1/2
1/2
==Re
ReLLxx ln(k)
ln(k)
ln(k)
LLxx
This shows
Thisshows
This that
showsthat the
thatthe ratio
theratio between
ratiobetween the
betweenthe characteristic
thecharacteristic length
characteristiclength
length
scale ofofdissipation
scaleof
scale dissipation
dissipationandand the
andthe characteristic
thecharacteristic length
characteristiclength scale
lengthscale of
scaleofof
energy containing
energycontaining
energy motions
containingmotions increases
motionsincreases with
increaseswith Reynolds
withReynolds number
Reynoldsnumber
number
What
Whatare
What are the
arethe smallest
thesmallest scales
smallestscales of
scalesof motion?
ofmotion?
motion?
Turbulence: lecture
Turbulence: lecture
KE dissipation
KE dissipation
Smallest
Smallest scales
scales of
of motion
motion Length
Length scale:
scale: ⌘⌘
IfIf we scale dissipation rate with these
we scale dissipation rate with these Velocity
Velocity scale:
scale: uu⌘⌘
scales,
scales, 22
Time
Time scale:
scale: tt⌘⌘
u
u⌘⌘
2
✏✏ ⇠ ⌫ ⌘
⇠ ⌫ ⌘222

But, at this scale the energy is immediately dissipated
But, at this scale the energy is immediately dissipated by by
viscosity and converted to internal energy
viscosity and converted to internal energy
The local Reynolds number has to be close to 1
The local Reynolds number has to be close to 1 (whatever(whatever
comes
comes from
from large
large scales
scales isis immediately
immediately lost)
lost)
u⌘⌘⌘
Re
Re⌘⌘ = ⇡1

es
es where
where
seswhere the
the
wherethe Reynolds
Reynolds
theReynolds number
Reynoldsnumber based
numberbased on
basedon local
onlocal velocity
localvelocity and
velocityand
and
Turbulence:
Turbulence:
Turbulence:
Turbulence: lecture
lecture
lecture
lecture
rrunity.
er unity.
unity. Thus
Thus
unity.Thus we
Thuswe introduce
introduceaaanew
weintroduce new length,ηηη
length,
newlength, ,, and
,and velocity
andvelocity
velocity
KE
KE
KE dissipation
dissipation
dissipation
KE dissipation
Smallest
Smallest
Smallest
Smallest scales
scales
scales
scales of
of
of motion
motion
motion
of motion Length
Length scale:
scale: ⌘⌘
3
3 33
u'u'
u'u' 2
2 22
vvv ηη
η vv
v Length scale: ⌘
≈≈≈≈εεε≈≈≈≈ννν 22 and
and
and ==11.1. .
= Velocity
Velocity scale:
scale: uu⌘⌘
IfLLLwe
xL scale ηη
u η
u⌘ dissipation
222
2 ννν with
rate u ⌘these Velocity scale: u⌘

xxx

⇠ ⌫ ⌘ and u⌘⌘ ⌘ Time


Time scale:
scale: tt⌘⌘
lds:
lds:
ds:
elds:

scales, ⇠ ⌫ 2 and ReRe ⌘⌘ == ⇡
⇡ 1
Time scale: t⌘
1
⌘⌘ 2 u 2 ⌫⌫
4 4 ⌘
ηηη≈≈≈≈((ν(ν(ν///ε/ε)ε)) ,, , vvv≈ νε
≈≈((νε
3 1 / 4 1/ 4
)))⌫.. . (4.2)
3 1 / 4 1/ 4

We
We
3 3

can
1 / 14 / 4

rearrange ✏ ⇠(νε
1/ 1/

these two equations


(4.2)
(4.2)
and solve,
Wecan canrearrange rearrangethese these
⌘ 2 two
two equations
equations and
and solve,
solve,
✓✓ 33 ◆◆14141 ⇣ ⌘
But, at this ⌫⌫scale the energy is 414immediately
3 4 1 ⇣ ⌘1
dissipated
11
⌫⌫ 222 by
eelarge
he large
large scales
scales
largescales ⌘⌘⇡⇡
scales (u’
(u’
(u’(u’and and
andLLxLxx))xare
)are fixed,
uu⌘⌘⌘ ⇡
arefixed,
fixed, multiple
(⌫✏)
multiple
⇡ multiple
(⌫✏)
1
instabilities
4 instabilities
instabilities occur

tt⌘⌘⌘occur
occur

viscosity ✏✏ and converted to internal energy ✏✏
ions with
ionswith
ons
tions with
withscales scales
scalesvvvand
scales andηηηsuch
and such
suchthatthat P~
thatP~P~εε.ε. .Using
Using
Usingthethe above
theabove
above
The local Reynolds Kolmogrov
Kolmogrov
Kolmogrov number scales
scales
scales has to
of
of
of be
motion
motion
motion close to 1 (whatever
ate
ate
mate useful
useful
te useful ratios
ratios
usefulratios of scales:
ofofscales:
comes scales:
ratios from large scales is immediately p
lost)
1/ 44
1/1/
1/ 4 p
ηηη ≈≈ ννν 3 ννL3LxLxx u44 ⌘ ⌘
4
333
ν 333
−−−
3/3/
t t

⌘ ⌫/✏
⌫/✏ 1/2
=== 4444 33x3Re ===Re
3/ 1/2
≈ ≈ 4444 Re
⌘ReL=
LL
− 3/4 4
⇡ 1 ⌘ == (4.3)
(4.3)
(4.3)
00
⇡⇡Re
ReL
LL
1/2
LxLxxx LLLxLxx εεε
L
L LxLxx u'u'u' L
⌫ tt L
L /u
xx /u 0
x

L 3 x
x
x L
LL
x
x x
x x x
that:
that:
appropriate Turbulence:lecture
Turbulence: lecture
to the large motions – ε ∝ u' / L . Kolmogorov (1941) – K41 – proposed
that:
that:
appropriate
that:
hat:
to the large motions Turbulence: lecture
Turbulence:
– ε ∝
lecture
u' 3
/ L . Kolmogorov (1941) – K41 – proposed
1.1.KE
KE Hypothesis
Hypothesis
dissipation
dissipation oflocal
of localisotropy:
isotropy: At Atsufficiently
sufficientlyhigh highRe, Re,the thesmall-scal
small-sca
that:
1.1.1. KE Hypothesis
1.Hypothesis
KE Hypothesis
dissipation
dissipation
Hypothesis ofoflocalof
local local
ofisotropy:
local
isotropy: isotropy:
isotropy:
At Atsufficiently
sufficientlyAt
At sufficiently
sufficiently
highRe,
high Re,the high
thesmall-scale
small-scaleRe, the small-sca
motions
motions are
are
statistically
Kolmogrov
statistically
Kolmogrov
1. Hypothesis of localisotropic
hypothesis
isotropic
hypothesis
isotropy: (say
(say Atfor forlength
sufficientlylengthhigh scales
scales Re,l<l l<lEIEI
the ).).
small-scale motions are
statistically
statistically
statistically
statistically isotropic
isotropic (sayfor
(say
isotropic
isotropic for length
length
(say
(say for
for scales
scales
length
length l<ll<lEIEI). ).
scales
scales l<lEI EI).
Kolmogrov
Kolmogrov hypothesis
hypothesis
statistically isotropic (say for length scales l<lEI).
2.2. 1 1
2.2. 11 Similarity
st
st Similarity
ststSimilarity
Similarity hypothesis:
hypothesis:
hypothesis:
hypothesis: InInevery In
everyturbulentIneveryeveryflow
turbulent turbulent
turbulent
flow atatsufficientlyflowat
flow
sufficiently at
high
high sufficiently
sufficiently
Re,the
Re, highRe
high
thestatistics
statistics R
2.
2. st1
1 Similarity
st Similarity
st
2. 1 Similarity hypothesis:
hypothesis:
hypothesis: In every In In every
every flow
turbulent turbulent
turbulent at sufficientlyflow at high sufficiently
Re, the statistics high R
ofthe
ofofthe
ofthe the small-scale
small-scale
small-scale
small-scale motions
motions motions
haveaauniversal
motions
have haveaauniversal
universal
have universal
form,uniquely
form, uniquely form,
form, uniquely
determined
uniquely
determined by νdetermined
andεε. .
byνdetermined
and byν
by
of the
of of
the the small-scale
small-scale
small-scale motions motions
motions
η η ≡ ν have
have a3 3universal
have
≡=(νf (⌫,/ ε✏) ε 1/ 4/ 4 aa universal
form, uniquely
universal form,
form, determined
uniquely by νdetermined
and ε. by ν
( /
η) ≡ (ν⌘3⌘: /:Lε
)
1
η ≡1 / 4(ν / ε ) u⌘⌘: LT
3 1 / 4 1 2 1 2 3
⌘ = f (⌫, ✏) u = f (⌫, ✏) t
⌘ = f (⌫, ✏) u⌘ ⌘= f (⌫, ✏) t⌘ ⌘= f (⌫,3 ✏) L )1 / 4u : LT 1
t⌘t⌘: :TT ⌫⌫: :LL2 TT 1 ✏✏: :LL2 TT 3
η ≡ (ν 1// 1/ε4 )4
(⌫, ✏) u⌘ = f (⌫, ✏) tv⌘v≡=≡(fεν
Thus:
⌘ = fThus: (εν
(⌫,) )ηη ≡ ( ν
✏) ≡ (ν✓✓/1/1/ ⌘ / ε
ε
33 : L 11 //u
)
)3◆4◆1 1
44 : LT 1 t : T
⌘ asasgiven
given ⌘ in inSection 2
Section4.1
⌫ : L T 1
4.1 ✏ : L 2
T 3

From
From Thus: Thus:
dimensionalanalysis,
Thus:
dimensional analysis,
vτ ≡=(εν )v1/v1≡1≡
4 (εν)⌫)1/⌫ 4◆ 1
/ 2/(2εν✓
3 4 4 4
as given in asgiven
1 as
1Section given
4.1 in ⇣⇣
inSectio
⌫ ⌘⌘
Secti11
⌫ 22
τ = ν ν ε ε ≡ ⇡εν u ⇡ (⌫✏) t ⇡ ⇣ ⌘
Thus: (
v ≡1 / 2(εν )⌫✏✏ 11/ 2/ 2
/ ) ⌘ u ⇡ (⌫✏) as given t ⇡ in Sectio
1/34 4 4

FromThus:
From dimensional
dimensional analysis, ( / v) ⌘ ( ) ⌘ 4 ⌘ 1
analysis,
η η
τ η = (ν /τετ)η⌘==⇡(ν (ν / /ε ε) )

u⌘ ⇡ (⌫✏)
1
4

t⌘ ⇡ ✏⌫✏ 2
η ✏ 11/equilibrium ✏
The region l<l is referred
The region l<lEI is referred to as τthe
EI to as τηη == ((νν //εε))
the universal
universal /22equilibrium range.
range.
The region l<lEI is referred to as the universal equilibrium range.
RecallThethat
The
Recall region
region
that l<lEIEIisisreferred
l<l
theReynolds
the Reynolds referred
numberbased
number to
based toas as on
on the
the these
these universal
universal
scales equilibrium
scalesisisequilibrium
unityand
unity andnotenote range.
range.
that
that
Thethat
The
Recall region
region l<lEI
l<l
the Reynolds is referred
EI is referred
number based to as
to asonthe the these universal
scales is equilibrium
universal unity and noterange. that
velocity
Recall
velocity gradients
that
gradients theofofthe
thedissipative
Reynoldsdissipative number eddiesare
eddies
Recall that the Reynolds number based on these scales is unity and n areO(
based O(1/1/on
ττη ).
η ).these
Recallscales
Recall alsothe
also theis ratios
unity
ratios ofand
of
velocity
Recallgradients
Recall that the
that theofReynolds
the dissipative
Reynolds number
number eddiesbased are O( 1/
based on
on τ η ).these
Recallscales also the is ratios
unityofand
smallest
smallest
velocity totolargest
largest
gradients scales(eqs.
scales (eqs.
of the (4.3)
(4.3) andand(4.4)).
dissipative (4.4)). eddies are O( 1/ τ ). Recall also
velocity
smallest to gradients
largest scales of the(4.3)
(eqs. dissipative
and (4.4)). eddies are O( 1/ τ ηη). Recall also
velocity gradients of the dissipative eddies are O( 1/ τ ). Recall also
velocity gradients of the dissipative eddies are O( 1/ τ ηη ). Recall also
3.3. Since,
Since, atsufficiently
at
smallestsufficiently
smallest to largescales
largest
large
to largest Re,there
Re,
scalesthere must
(eqs.
must
(eqs. beaarange
(4.3)
be
(4.3) range(4.4)).
and
and ofscales
of
(4.4)).scaleswhich
whichare
aremuch
much
3. Since, at sufficiently
smallest to large
largest Re,
scalesthere must
(eqs. be
(4.3)a range
and of scales which are much
(4.4)).
smallest
larger
larger thanηηto
than but
largest
but much
Wednesday, 23 October 13
scalesthan
muchsmaller
smaller (eqs.
than (4.3)
L,L,K41
and
K41also
also (4.4)). his:
proposed
proposed his:
Wednesday, 23 October 13
The region l<lEI is referred to as the universal equilibrium range.
Turbulence: lecture
Recall that the Reynolds number based on these scales is unity and note that

KEvelocity
dissipation
gradients of the dissipative eddies are O( 1/ τ η ). Recall also the ratios of

Kolmogrov hypothesis
smallest to largest scales (eqs. (4.3) and (4.4)).

3. Since, at sufficiently large Re, there must be a range of scales which are much
larger than η but much smaller than L, K41 also proposed his:
2nd similarity hypothesis, that in every turbulent flow at sufficiently large Re, the
statistics of the motions of scale l in the range η<<l<<L have a universal form which is
uniquely determined by ε. This is normally called the ‘inertial sub-range’.

Thus we can define various ranges as in figure below.

What is this range and how do we get to this?


Turbulence:
Turbulence:
Turbulence: lecture lecture
lecture
Decreasing
Decreasinglength
lengthscale
scale
KE
KE dissipation
dissipation
KE dissipation
Given
Given aalength
length scale
scale in
Kolmogrovin
Kolmogrov thth sihypothesis
inertial
ihypothesis
s inertial sub-range,
sub-range, l,l,corres
corres
Kolmogrov hypothesis
scales
scales can be
can be Let’s
Let’s define
expressed
define
expressed velocity
as:
velocity
as: and
and time-scales
time-scales that
that depe
dep
Let’s define velocity and time-scales that depend on local
length
length scale
scale ‘l’‘l’and
and dissipation
dissipation
length scale ‘l’ and dissipation
uu((ll))==((εεll)) ==vv((ll//ηη)) ∝∝uu’’((ll//LLxx))
11/ 3/ 3 11/ 3/ 3 1

ττ((ll))==((ll //εε)) ==ττηη((ll//ηη)) ∝∝ττoo((ll//


22 11/ 3/ 3 22/ 3/ 3

This
Thisisis
Rate of energy transfer T ~ O(u /t) 2 time-s
time-
(where
(where ττοοis
istime
time scale
Independent of localscaleof
of
Rate
large
large
length
Rate of
ofenergy
eddies
energy
eddies ––
scale transfer
and LL xx/u’).
/u’).
viscosity
transfer
This
TT~~O(u
This su
su
O(u /t/t
22

whichenergy
which Independent
energyisisIndependent
transferredfrom
transferred ofoflocal
from local length
length
eddies
eddies scalethan
scale
larger
larger andvis
and
than lvilt
Kinetic energyKinetic
tricklesenergy
down trickles
the scaledown at thethesamescalerate
at no
the
Kinetic energy trickles down
2/τ) if scale
the scale at the s
mightbe
might matterto
beexpected
expected what
tobe the
beO(u
O(u / τ
2length
matterwhat
matter ) if it’s
it’s
whatthe is
accomplished
accomplished
thelength
lengthscale prp
scaleisis
Wednesday, 23 October 13
Turbulence:
Turbulence: lecture
Turbulence: lecture
lecture
KE
KE dissipation
dissipation
Kolmogrov hypothesis
Kolmogrov hypothesis
E(k) = E(k, L x , ⌘, ✏), kk = 2⇡/l, where l is the length
E(k) = E(k, Lx , ⌘, ✏), = 2⇡/l, where l is the length scalescale
In non-dimensional form, E(k) = E(✏, k,
In non-dimensional form, E(k) = E(✏, k, kL , k⌘) kL x , k⌘)
x
Using dimensional analysis, k:L 1
1,
Using dimensional analysis, k : L , ✏ : L T
,
2 3
✏ : L 2T 3,
✏:L T ,
3 2
E : L 3T 2
E:L T
, E:L T
And saying that rate of energy transfer should be
And saying that rate of energy transfer should be
independent of 2/3
lengthscale and viscosity
independent of
E(k) = ✏ k lengthscale
5/3 and viscosity
g(kL, k⌘)
2/3 5/3
E(k) = ✏ 2/3k 5/3g(kL, k⌘)
E(k) = ✏ k g(kL, k⌘)

Wednesday, 23 October 13
energy
energytransfer
transfer which
whichdetermines
determineshow
howquickly
quicklyenergy
energyisisdissipated
dissipatedand
andviscosity
viscosity : :
1 Turbulence: lecture
3 41
3 4
,, isisthe
theKolmogorov
Kolmogorovmicro-scale.
KE dissipation
K
K
K
K micro-scale.

Kolmogrov hypothesis 2 5

Hence: EE kk
Hence: EE k,k,L,
L, , , kk gg kL,
kL,kk KK by
bydimensional
dimensionalanalysis
analysisand
andcarry
carryon
32 35
K
K
3 3
on
independent
ofindependent
L and
2/3
of L and 5/3
with
withargument
argumentvery E(k) = ✏
verysimilar
similartotothat
thatwhich
whichleads k
leadstotothe
thelog
loglaw. g(kL, k⌘)
law. This
Thisargument
argumentworks
workswith
with
EE k,k,L,
L, , , K and
andnot
notwith
withe.g.
e.g.EE k,k,E,
E,L,L, . .
K
22
(a)
(a)Inner scaling: atatwavenumbers
Innerscaling: wavenumberskktoo toolarge
largecompared
comparedtoto L , ,we
wecan
canassume
assumethat thatthe
the
L
energy
energy spectrum
spectrumdoes
doesnot
notknow
knowhowhowlarge
largethe
theexperimental
experimentalsetup
setupisisi.e.
i.e.ininthe
thelimit
limit
where
wherekL>>2
kL>>2 we thatgg kL,
assumethat
weassume kL,kk K ggi kk K . .
K i K

(b)
(b)Outer scaling: atat wavenumbers
Outer scaling: wavenumberskktoo toosmall
smalltotoresolve
resolve KK we wecan
canassume
assumethat
thatthe
the
energy
energy spectrum
spectrum does
does notnot know
know what
what the
the value
value ofof isissince
sincethe
theenergy
energyisissimply
simply
transferred
transferredand
andnot
notdissipated,
dissipated,equally
equallyfor
forall
allvalues
valuesofofsmall
smallenough
enough . . Hence,
Hence,ininthe
the
limit
limitwhere
where kk K 22 , ,we
weassume
assumethat
that
K
gg kL,
kL,kk KK gg0 kL
kL ..
0

4
4
L 3
(b)(b)Outer scaling:at atwavenumbers
Outerscaling: wavenumbersk ktoo toosmall
smalltotoresolve
resolve K Kwewecan canassa
does Turbulence:
Turbulence:
energy spectrum does not know what thethevalue
energy spectrum not know what lecture
lecture valueofof is issince
sincethetheenerg
ene
KE transferred
transferred and
and
dissipation notnot dissipated,
dissipated, equally
equally for
for all
all values
values
as a ofof small
small Renough
ofenough . .H
KE dissipation The energy spectrum function λ
The
he limit
limit
energy
energy where
where
spectrum
spectrum ak
ask
as of 2
2
a function
function
K K
R R,λwe
of,λwe assume
assume that
that
10001000
Innerscaling
Inner scaling g gkL,
E (κ )
kL,k kK
k l0
Outer
g g kLkL. .
scaling
)E (κ ) K 0 0
2
η uη2
300 300

4 4
100 100
LL 3 3
(c)(c)Overlap
Overlap existsif ifReRe
regionexists
region
R =30
R =30
1 as Re
L L 1 as Re L L
. Thisregion
. This regionis iscalled
cal
λ λ Rλ=30
K K 100
300
range
range and
and is is where
where 1000

κηκη 31 31
κ l0 32

22 22
kk
L
L5/3
2/3 5/3
2/3
E(k)= ✏
E(k) =✏ k g(kL,K Kk⌘)
k g(kL, k⌘)
5 5 5 5
2 2 3 3 2 2 3 3
InIn
thethe inertial
inertial range,E Ek k
range, 3
k kgi gik kK
3
K
3
k kg 0g 0kLkL
3

g i g ik kK K g 0g 0kLkL g i g i g 0g 0 constant
constant (independentofofL Land
(independent and K ).K ).T
(b) Outer scaling: at wavenumbers k too small to resolve K we can ass
Turbulence:
Turbulence: lecture
lecture
energy spectrum does not know what the value of is since the energ
KE transferred
KE and
dissipation
dissipation not dissipated, equally for
The energy all values
spectrum as a of small
function ofenough
Rλ . H
The
he limit
energy
energy where ask
spectrum
spectrum as 2
a function
a function
K
of,λwe
of R Rλ assume that

10001000
Innerscaling
Inner scaling
E (κ )
g kL, k
k l0 g0
Outer
kL .
scaling
)E (κ ) K
2
η uη2
300 300

4
100 100
L 3
(c) OverlapR =30Rregion
=30
exists if Re L 1 as Re L . This region is called
λ λ Rλ=30
K 100
300
range and is where 1000

κηκη 31 31
κ l0 32

2 2
k
L 2/3 5/3
E(k)
E(k) = ✏= kC✏
2/3 5/3 k k⌘)
g(kL, K

5 5

In the inertial range, EThe famous


k gi k-5/3 spectrum
2 3 2 3
k 3
K k g 0 kL 3

The
g i kvalue
K g 0 of
kL C from
g i gexperiments
0 is about
constant (independent 1.5K ).
of L and T
Turbulence: lecture
KE dissipation
-5/3 spectrum for various
flows across different
Reynolds numbers

The plot is in “inner” scaling


and hence collapses at small
scales

Very fast roll-off after the


inertial range due to
viscous dissipation
Turbulence: lecture
KE dissipation
Kolmogrov hypothesis

Energy transfer T(l)


Production = P Dissipation = ε

Energy-containing Inertial Dissipation


range sub-range range

Lx η
lEI lDI

Decreasing length scale

ngth scale in this inertial sub-range, l, corresponding velo


Turbulence: lecture
KE dissipation
Kolmogrov hypothesis

This one-way cascade is in fact incorrect


There is some evidence for energy being transferred in
both directions
However, the “net” transfer is from large to small

There is a whole research industry trying to disprove


Kolmogrov hypothesis
ion (found by taking theTurbulence:
Turbulence: lecture
lecture
curl of the N-S eq.) is:
Use these tensor identities and the incompressibility condition, r · u
Vorticity
Vorticityand
andenstrophy
enstrophyequations
equations
as:
We
Wecan
cantake
takethe
thecurl
curlofofthe
theNSNSequations
equations and
andarrive
arriveatatthe
the
∂ωi ∂ωi ∂ui ∂ ωi
2
vorticity = ωj
+ u j transport
vorticity transport νequation
+equation
∂t ∂x j ∂x j ∂x j ∂x j
@!
+ u · r! = ! · ru + ⌫ r !.
2
@t
[1] [2] [3]
[1]--convection
[1] convectionofofvorticity
vorticity
m of the right-hand side of equation A.12 is responsible for the vortex-
[3]--diffusion
[3] diffusionofofvorticity
vorticity
e [1] is convection of vorticity, [2] is stretching or tilting of vorticity (ofte
uct equation A.12 with !, decompose the velocity gradient tensor as
[2]--Tilting
[2]
vortex Tilting orstretching
or
stretching’)stretching
and [3]of of vorticity
isvorticity
viscous diffusion. (We have ignored the
Also
Also known
known as vortex
as vortex stretching.
stretching. This
This happens when
happens when the rate
the rate of
of
mation
of the
termright-hand
that side, and use
exists in stratified the antisymmetry of ⌦ 1
to obtain:
flow.) Velocity derivatives are do
strain has a component along
strain has a component along vorticity.vorticity.
This is
This is crucial
crucial as this
as this is responsible for production of enstrophy
is responsible for production of enstrophy
mallest scales (more later).
and hence sustaining ✓turbulence
and hence sustaining turbulence

D 1 @ !i 2
!i !i = !i Sik !k + ⌫ !i .
Turbulence: lecture
. Use these tensor identities and the incompressibility condition, r ·
Vorticity and enstrophy equations
9 as:
It is more useful to look at kinetic energy equivalent for
vorticity: Enstrophy
@!
+ u · r! = ! · ru + ⌫ r !.
2
@t
Dot product of this equation with ! and decompose ru = S + ⌦
Also, use the fact that ⌦ is anti-symmetric,
m of the right-hand side of equation A.12 is responsible for the vortex
2
duct equation
2 Dt = !i Sij !j + ⌫!i r !i
1 D!A.12 with !, decompose the velocity gradient
2 tensor as

m of the right-hand side, and use the antisymmetry of ⌦ 1


to obtain:

Rate of change of enstrophy = Interaction + diffusion


✓ ◆
D 1 @ 2 !i
!! =! S ! +⌫! .
Turbulence: lecture
Enstrophy and dissipation equations

We can do the same sort of thing for dissipation

@2p

D
1 2⌫
2 Dt = Sij Sjk Ski 1
4 !i Sij !j Sij @xi @xj + ⌫Sij r Sij
2

vorticity pressure
Self Viscous
strain strain
interaction diffusion
interaction interaction

2
1 D!
2 Dt = !i Sij !j + ⌫!i r !i 2

vorticity
Viscous
strain
diffusion
interaction
Turbulence:
Turbulence:lecture
lecture
Strain-rotation interaction
Strain-rotation interaction
2
!i Sij !j = ! ! i (êi ·!
ˆ)
ii are the principal strain rates (i.e. eigen values of the strain rate tensor)
1 : : Extensive strain rate
1 >> 2 >> 3 1 Extensive strain rate
1 2 3 2 : : Intermediate
Intermediatestrain
strainrate
rate
2
1+ 2 + 3 = 0 3 : : Compressive strain rate
1+ 2+ 3 =0 3 Compressive strain rate

Irrespective of the eigen value, a principal strain rate will


Irrespective of the eigen value, a principal strain rate will
only contributeto
onlycontribute tovortex
vortexstretching
stretchingififthere
thereisiscomponent
component
of
ofstrain
strainalong
alongthe
thevorticity
vorticity
So,
So,we
wecan
canexamine
examinealignments
alignmentsbetween
betweenprincipal
principalstrain
strain
rates
ratesand
andvorticity
vorticity
Wednesday, 23 October 13
Turbulence: lecture
Strain-rotation interaction
2
eˆ1 · ω̂
eˆ2 · ω̂
eˆ3 · ω̂
1.5

This is a universal feature


pdf

1
Observed in most
turbulent flows
0.5
0 0.2 0.4 0.6 0.8 1
| eˆi · ω̂|

No preference for extensive


Parallel alignment with intermediate
Perpendicular with compressive
Turbulence: lecture
Turbulence: lecture
Invariants of
Invariants of velocity
velocity gradient
gradient tensor
tensor
@u11
@u @u11
@u @u11
@u
@x11
@x @x22
@x @x33
@x
P=
P = r r ·· uu
ru =
@u22
@u @u22
@u @u22
@u 11 22 22
@x11
@x @x22
@x @x33
@x Q=
Q = k⌦k
k⌦k kSk
kSk
@u33
@u @u33
@u @u33
@u
22
@x11
@x @x22
@x @x33
@x R=
R = det(ru)
det(ru)

ru

P,P,Q
Q and
and RR are
are the
the invariants
invariants of
of the
the velocity
velocity gradient
gradient tensor

PP == 00 for
for incompressible
incompressible flows
flows
Chakraborty et
Chakraborty et al.
al.(2005),
(2005),J.J.Fluid
Fluid Mech.
Mech.
Turbulence: lecture
Invariants of velocity gradient tensor
0.2

Discriminant
0.1

> 0, complex eigenvalues


Q/(⇥/ 2 )2

0
> 0 implies swirling regions (nodes)
< 0 implies Pstraining
= r·u regions (saddles)
-0.1

1
Q= k⌦k2 kSk2
-0.2 2
R = det(ru)
Q criterion more
-0.3
-0.02 -0.01 0 0.01 0.02 0.03
restrictive in vortex
R/(⇥/ )
2 3
identification
Tear-drop shape in Q-R space is considered universal feature
gridsmounted
grids mounted
entically
entically acrossit
zero,
across
zero, awind
aitwind tunnelimmediately
follows
tunnel
follows workingsection,
section,produce
immediately
working produce turbulence
that:turbulence
that:
pproximatelyisotropic
pproximately isotropicand
andhomogeneous
homogeneousin Turbulence:
Turbulence:
inthe
theframe lecture
lecture
Turbulence: lecture
framemoving
movingwith withthe
themean
mean
U. No mean velocity gradients exist ∂u∂u
(some ii way
. No mean velocity gradients exist
Decayingturbulence
turbulence ⟨s (some
⟨sijij ⟩⟩way= =downstreamof
⟨sijijssijij⟩of⟩the
downstream
⟨s +thegrid),
+⟨s grid),so
⟨s so
ijij ijij⟩⟩
ωω
ction
Decaying
Decaying
ctionterm
term(1) turbulence
(1)in
ineq.
eq.(3.3)
(3.3)isiszero
zeroand
andthe∂xturbulence
∂x
the jturbulence
j mustthus
must thusdecaydecaywith
withx.x.
= ⟨sijijssijConsider
= ⟨s ij⟩⟩ theturbulent
turbulentflow
flow (4.4)
(4.4)
Consider the
Consider the turbulent flow
Now it is customary to define a new downstream
downstream
downstream
variable MM
k, the of
of
of aa
averagea bi-planar
bi-planar
bi-planar grid.
grid.
grid.
fluctuating kinetic
Now it is customary to define a new variable k, the average fluctuating kinetic
gy per
gy per unit
unit mass,
mass, by:
by:
ItItisisisone
It oneof
one ofthe
of thesimplest
the simplestflows
simplest flows
flows
11 11 2x2x 11 we22 can consider.
UU k ≡ ⟨u u ⟩ = ⟨q ⟩ = we
we[⟨u can
can
⟩ + consider.
consider.
⟨u
k ≡ ⟨uiiuii⟩ = ⟨q ⟩ = [⟨u11⟩ + ⟨u22⟩ + ⟨u33⟩] 22 ⟩ + ⟨u22 ⟩] (4.5)
(4.5)
22 22 22
By dividing
By dividing equation
equation 4.1
4.1by andLet’s
by22and look
inserting
inserting at
this
this the kinetic
definition,
definition, energy
theequation
the equation forthe
for the
age kinetic
age kinetic energy
energy per
per unit
Let’s
Let’s
unit mass
mass of
look
look
of the
at
at the
the
the fluctuating
kinetic
kinetic
fluctuating motion
energy
energy
motion can
can be
be re-written
re-written
equation
equation for
equation for
for this this
this flow,flow,
flow,
as therms
s the !! turbulent
rms "" The
turbulentvelocity.
velocity. #
Thecharacteristic#time-scale
characteristic time-scalefor
forthe
theturbulence
turbulence $ $
∂ ∂ ∂
∂ ∂∂ 11 11 22
u'u' / /
du'
2 2 du'
22
+UUjj time
+
. . AAcharacteristic
characteristic kkfor
time =
for=theenergy
the −− ⟨pu
energycontaining⟨pu
containing ⟩δijijisis−
ii⟩δ
eddies
eddies

LLx/u’
⟨q
anduujj⟩⟩+
⟨q
x/u’and
2ν⟨sijijuuii⟩⟩
+2ν⟨s
dtdt ∂t ∂t ∂xjj
∂x ∂xjj
∂x ρρ 22
overall decaycan
verall decay canonly
onlyoccur
occuras
asthese
theselarge
largeeddies
eddiesbreak
breakup ∂U
up∂U
tosmaller
to smaller
ii
ones,
ones,
time-scalesmust
ime-scales mustbebecomparable.
comparable. Thus:
Thus:
−⟨uiiuujj⟩⟩
−⟨u −2ν⟨s
− 2ν⟨sijijssijij⟩⟩ (4.6)
(4.6)
∂xjj
∂x
The role
The role of
of each
each of This
This
This
of these
these equation
equation
equation
terms
u'u' termsLL will be
will
22 simplifies
simplifies
besimplifies
examined
examined into,
in to, later.
to,
detail
detail later. First
Firstnotenotethat that
−−AA dd == (withA=O(1)
(with A=O(1)by
xx
byexpt.)
expt.)source: Turbulence for the 21st century, W. K. George
source:Turbulence
Turbulencefor
lternative form of
ternative form of this
this equation
(u'equation
) u'u' 22 can be
can be derived
derivedsource: leavingfor
by leaving
by the21st
the
the
the 21stcentury,
century,W.
viscous
viscous W.K.K.George
stress
stress George
in
in
grids
grids mounted
entically across
zero,
mounted aa wind
across it tunnel
follows
wind working section, produce
tunnel immediately that:turbulence
generatenew
pproximately isotropic
approximately
newturbulence,
isotropic and
and homogeneousTurbulence:
turbulence,
Turbulence:
e new turbulence, the kinetic energy equation
generate homogeneous in in the
thekinetic
the
the frame
kinetic
frame moving lecture
lecture
moving with
energyequation
energy
with the
the mean
mean
equationreduce
reduc
U. No mean velocity gradients exist ∂u
(some i way downstream
Decaying turbulence
U. No mean velocity gradients exist ⟨sij (some ⟩way= ⟨sij sijofof
downstream ⟩ the grid),
grid),ωso
the ⟨s
+ d so
ij ij ⟩
Decaying turbulence
ction term
uction term (1)
(1) in
in eq.
eq. (3.3)
(3.3) is
is zero
zero and ∂xturbulence
and the
the j
turbulence must
must thus
thus decay
d
decay with k==−ϵ
with x.
k x. −ϵ
= ⟨sij sConsider ⟩ d
dtdtthe turbulent flow (4.4)
ij
k = −ϵ
Note
Note that
that the
the time
timeMM derivative
derivative is isjust dt
just
Now it is customary to define a new variable k, the average fluctuatinggrid.
downstream anan ordinary
of a
ordinary derivative
bi-planar
derivative,
kinetic
pendence
pendence ofofby:anysingle
any singlepoint pointquantity
quantityon onposition.
position. This Thisisisofte
oft
gy per unit mass, Forisotropic
For isotropicturbulence, turbulence,this thiswritten
writtenasas
at for the time 1turbulence)
forisotropic
isotropic derivative
turbulence)
1
as:
as: is
1
It isjustone of
!
an theordinary
simplest flowsderi
"2
U
k ≡ ⟨ui ui ⟩ = ⟨q ⟩ = [⟨u
xx
2 we1 ⟩can
2 !
consider.
+dd⟨u 2 "
⟩3+ ⟨u ⟩] (4.5)
23
U

cee of any single 2 point 2 quantity


2
dt 22
on uu2 position.
2 3
= =−ϵ −ϵ Thi
dt
opic
By dividingturbulence)
where
where
equation 4.1 by 2 as:
and
Let’s
age kinetic energy per unit mass of the fluctuating
inserting
look at this
the definition,
kinetic
! motion "
the equation
energy for the
can be re-written
equationwhere, for d
this 3
flow, 3 3
where, kk≡≡ 2uu22
as
as the ! turbulent
the rms
rms " The
turbulent velocity.
velocity. #time-scale
The characteristic
characteristic time-scale for
for the turbulence u2
the turbulence
2 = −ϵ $

−u'
22 du'
du'
22

.. A
A
4
4 +
I UThis
This
can

is is
j personally
4 characteristic
characteristic timethe
k
time for
the
=
for the
simplest

simplest
testify
energy
thetestify
energy

1
equation
⟨pu
that
containing
containing
⟩δdt 2
equation
this
i ijis
eddies
eddies

1
isworks,
LLxx/u’
we
we
⟨q
/u’ and
2
u
since
j
andsince
can
can
⟩ + I
write,
write,
2ν⟨s u
arrived ⟩ yet,
yet,
ij i in Kinsale,
Kinsale
u' //
∂t I can personally
∂x ∂x that
ρ this works, 2 I arrived in Kinsale,
overall
dt
dt
voyage
voyage
decay can
across
across
only
j
occur the
as
the Atlantic
Atlantic
these large
we
we
j
eddies
cannot
cannot
from
from
overall decay can only occur as these large eddies break up∂Ubreak
America
America
up to
solve
solve
smaller
to smaller inin
ones,
ones,
this
this
mymy 4242 foot
foot sailboat
sailboat Wing
Wings
i
time-scales
time-scales must
must be u
be
3
comparable.
comparable. Thus:
Thus: u 2 −⟨ui uj ⟩
∂xj 3
− 2ν⟨sij sij ⟩
L L p
2 p (4.6)
✏=A = 15⌫ 2 Rearranging, Rearranging, k≡ u xx
⇠ ⇠ Re ReLLxx
The role of each Lxof−−AAtheseThis
22
u'
u' terms
LL
equation simplifies
will be examined 2 to, later. First note that
in detail
dd == xx
(with
(with A=O(1)
A=O(1) by
by expt.)
expt.)source: Turbulence for the 21st century, W. K. George
lternative form of this (u' equation
) u'
22 u' can be derived by leaving the viscous stress in
be defined by, as in §4: Re Lx = x
and Re λ = − n and we also see that x
=
Turbulence:
Turbulence:
kt ν lecture
x − xlecture
ν λ
imentally, it is found that 2
=A o
where xo is a virtual ori
Decaying
(Note turbulence
again the U
increasing separation of M
scales with increasing Re).
−n
w. Experimentally, it is found that kt = A x − xo where xo is a virtual origin –
2
U M
below. 10-3

2
u' /U 2 10-3
kt/U2
v' 2/U2
u'2/U2 kt/U2
v' 2/U2
10-4
10-4
log
scale log
scale

10 100
100 log
log
(x-x
(x-x o
)/M)/M
10 o
scale
scale
ν ν νLνLxx x
u'u' u'u'λλ λλ 1010 LLxx x
be
be defined
bedefined by,
definedby, as
by,as inin§4:
asin §4: ReLL L ===
§4: Re
Re and
and Reλλ λ===
andRe
Re and
and we
andwe also
wealso see
alsosee that ===
that
seethat
log Turbulence:
log Turbulence:
Turbulence:ννν
xx x
lecture
lecture
lecture
ννν λλλ
eincreasing
increasingseparation
separation ofofscales
scale
scale scaleswithwithincreasing
increasingRe). Re).
Decaying
Decaying
(Note
(Note
(Note again
again turbulence
turbulence
the
againthe
the increasing
increasing separation
increasingseparation
separation
− n− n
of
of scales
ofscales
scaleswith with increasing
withincreasing
increasingRe). Re).
Re).
ktkt xx−−xoxo
y,ititisisfound that 2 2==AA
foundthat k
k kwhere
where xxx
xx

−o−is
o
x
x xisa nvirtual
−−an nvirtualorigin
− origin––see seefigure
figure
Experimentally,
Experimentally,
Experimentally, UUitititis
isisfound
found
foundM10
10
M that t2t2t2===AAA
that
that oo100
o 100 log
where
wherelog xxx
where isis(x-x
ooois o)/M
aa(x-x o)/M origin
virtual
avirtual
virtual origin
origin–––
U
UU M
MM scale
scale

InInthe
below.
below.
below. themoving
movingframe
frame(x-x
(x-xo=U(t-t
o=U(t-to))
o))this
thiscan
canbe
bewritten
writtenasasktk=k
t=kt,ot,(t/t o)o)-nand
o(t/t
-n and

(5.1)
(5.1)
-3
1010 -3 ε=ε=ε ε
o (t/t
(t/t
o o )o ) . .The
-(n+1)
-(n+1) Theexponent
exponentnnisistypically
typicallyfound
foundtotobe
bearound
around1.3.
1.3.TT
10
10
10 -3-3
-3 Researchcontinues
Research continueseven
even
dependson
depends ongrid
gridgeometry.
geometry.
2 2/U
u'u'/U2 2 22
today
today to
to determine
determine what
what
2 2/U
u'u'2/U
22 ktk/U
t/U
2/U
v'v' 2/U
22
u' /U2
2/U
22/U 22 2
“n”and
“n” andkA
k k
/U
/UAare...
22 2
/U are...
tt t
v'
v'
v' /U
STUDENTexercise.
exercise.Show
Showthat
thatthe
thelength
lengthscale
scaledefined
definedby
byL=k t /ε/εincrea
L=kt3/2
STUDENT 3/2 incre
-4-4
1010 10 -4
10-4
-4
10number
Thereare
There aredifferent
different
andthe
and theReynolds
Reynolds numberbased
basedon
onktk1/2
1/2 andLLdecreases
t and decreaseswith
withtime.
time.
log
log log
log
resultsdepending
results dependingonon
log
scale
scale scale
scale
scale experiment
experiment
SinceRe
Since Redecreases
decreaseswith
withtime
time(or
(ordistance
distancedownstream),
downstream),eventually
eventuallythe
theefe
viscosity
viscosity will
will dominate
10
dominate and
and the
the 100)/M
turbulence
100
(x-x
turbulence )/Menters
log
enters a a (x-x
final
(x-x
final o)/M
)/M period of decay
o period of decay
1010 10 100
100 log
log (x-x
100o log
o log (x-x o )/M
10 scale
scale
scale scale
scale
mean
mean (although
ric
(although similar Turbulence: lecture
Turbulence:
similar considerations
considerations can
can be
be lecture
applied
applied to
to flows
flows that
that are
are axisymmet-
axisymmet-
ric in
in the
the mean,
mean, see see the
the exercises).
exercises). In In effect,
effect, this
this isis exactly
exactly the the same
same as
as assuming
assuming
Simple
the
Simple
the
tion
flow
flow is
is shear
homogeneous
shear flows
flows
homogeneous
flows
in
in the
the third
third direction.
direction. Also
Also we
we shall
shall restrict
restrict our
our atten-
atten-
tion to flows which are statistically
to flows which are statistically stationary,
stationary, so so that
that time
time derivatives
derivatives ofof averaged
averaged
quantities
Considercan
Consider
quantities 2Dbe
2D
can be neglected.
flows that
that have
neglected.
flows And,
And,
have of
of course,
aa planar
course,
planar we
we have
mean.
mean. Thealready
The
have flows
flows are
already agreed
agreed
are to
to confine
homogenousconfinein
homogenous in
our
the attention
third to
to Newtonian
direction and flows
flows at
at constant density.
our
the attention
third direction and are
Newtonian are stationary
stationary (i.e.
(i.e. time
constant derivatives
derivatives are
density.
time are zero).
zero).
It
It will
will bebe easier
easier to to abandon
abandon tensor
tensor notation
notation for for the
the moment,
moment, and and use
use the
the
symbols
symbols x, x,U,
U,uu for
for the
the streamwise
streamwise direction,
direction, mean
mean and and fluctuating
fluctuating velocities
velocities re-
re-
6.2.
6.2.
spectively,THE
THE AVERAGED
AVERAGED
and y, V, v for EQUATIONS
EQUATIONS
the cross-stream. Given all this, the mean
mean momentum
momentum 99
99
spectively, and y, V, v for the cross-stream. Given all this, the
equations
equations reducereduce to:to:
x-component:
x-component:
222 222 222
∂U
∂U ∂U
∂U 1
1 ∂P
∂P ∂⟨u
∂⟨u ⟩⟩ ∂⟨uv⟩
∂⟨uv⟩ ∂∂ UU ∂
∂ U
U
U
U ∂x +
+ VV ∂y ==−
−ρ ∂x −− ∂x − − ∂y + + νν ∂x222 ++ νν 22
∂y
∂y
(6.2)
(6.2)
∂x ∂y ρ ∂x ∂x ∂y ∂x ∂y 2
y-component:
y-component:
222 222 222
∂V
∂V ∂V
∂V 11 ∂P
∂P ∂⟨uv⟩
∂⟨uv⟩ ∂⟨v
∂⟨v ⟩⟩ ∂
∂ VV ∂
∂∂ VVV
U
U +
+ VV =−
= − −
− −
− +
+ νν 222 + +
+ ννν ∂y222 (6.3)
(6.3)
(6.3)
∂x
∂x ∂y
∂y ρρ ∂y
∂y ∂x
∂x ∂y
∂y ∂x
∂x ∂y
∂y
In
In addition,
addition, we
we have
have the
the two-dimensional
two-dimensional mean
mean continuity equation
continuity equation
equation which which reduces
which reduces
reduces
to:
to:
∂U
∂U + ∂V∂V = 0
+ =0 (6.4)
(6.4)
(6.4)
∂x
∂x ∂y∂y source:
source: Turbulence for the 21st century, W. K. George
source:Turbulence
Turbulence for
for the
the 21st
21st century,
century,W.
W.K.
K.George
George
Turbulence:
Turbulence:lecture
lecture
Simple shear
Simple flows:
shear Order
flows: of of
Order magnitude estimates
magnitude estimates
• Streamwise length
• Streamwise scale:
length L L
scale:
• Streamwise velocity
• Streamwise scale:
velocity Uo U
scale: oro orUs U
(depending on on
s (depending thethe
situation)
situation)
• Cross-stream length
• Cross-stream scale:
length scale:<<<<
L 6.2.LTHE AVERAGED EQUATIONS 101
6.2. THE AVERAGED EQUATIONS
considering specific problems and applying similarity techniques, the seemingly
• Cross-stream
• Cross-stream
velocity scale:
velocity Vs ⇠
scale: Vs ⇠Us (U/L)
s ( /L)
considering specific problems and applying similarity techniques, the see
arbitrary choices here become quite precise constraints (as we shall see).
arbitrary choices here become quite precise constraints (as we shall see).
We still haven’t talked about how to estimate the velocity scale for V , the
We still haven’t talked about how to estimate the velocity scale for
cross-stream mean velocity component. From the continuity equation we know
cross-stream mean velocity component. From the continuity equation we
that:

This
This completes
completes order
order of of
that: ∂V ∂U
= − ∂V ∂U (6.10)
∂y ∂x = −
∂y ∂x
From our considerations above, we know that:

magnitudes
magnitudes forfor mean
mean flows
flows
From our considerations above, we know that:
∂U ∆U∂Us
∆Us
∼ (6.11)
∂x L ∂x ∼ L
If there isIfnothere
meanis cross flow cross
no mean in theflow
external
in thestream,
externalthen the scale
stream, then for
the Vscale
is the
for V
same as thesamescale
as for
thechanges
scale forinchanges
V . Therefore,
in V . Therefore,
∂V Vs ∂V Vs
∼ (6.12)
δ ∂y ∼ δ

How about turbulence terms?


∂y

How about turbulence terms? It follows It


immediately that the order
follows immediately that of
themagnitude of the cross-stream
order of magnitude velocity is:velo
of the cross-stream
δ
Vs ∼ ∆Us Vs ∼ ∆Us δ (6.13)
L L
We might Wehave expected
might have something like this iflike
expected something we had
this thought
if we hadabout
thoughtit. If the Vit.- If
about

It turns out they all scale as 2, sometimes, q 2


u
velocity were of thewere
velocity
be a “thinbe
shear
since bothsince
are some
same
flow”.
of order
On
measure
both are
as the
the same
the
some ofmeasure
U -velocity,
order
other hand,
as the Uhow
it
how theof flow
also makes
how spreads.
couldhow
-velocity, the could
sense
the flow Note
flow inthe
that V
thatNote
spreads.
anyflow
/U
a “thin shear flow”. On the other hand, it also makes sense that Vs /Us
o s s
equation

sense
in an
δ/L,
that 6.13
equatio
would notwould
be thenot
correct
be theestimate
correctfor Vs if there
estimate for Vwere an imposed
s if there were ancross-flow, since
imposed cross-flow
source:
then we wouldsource:
have
then we Turbulence
Turbulence
to consider
would for
and the
have toVconsider for
changes the
V 21st 21st
Vcentury,
andinchanges century,
separately W. W.
K. George
(exactly
in V separately for UGeo
K.
as(exactly). as
important
importantmost
mostofofthe
thetime,
time,and
andmust
mustlook
lookagain
againafter
afteryou
youhave
haveanalyzed
analyzedor
orsolved
solved
the
theequations
equationsto
tosee
seeififyour Turbulence:
yourestimates
estimateswere lecture
Turbulence: lecture
werecorrect
Turbulence: lecture
correct

Simple
Simple
6.2.3
6.2.3 shear
shear
The
The flows:
flows:
streamwise
streamwiseOrder
Order of
of
momentum
momentummagnitude
magnitude
equationestimates
estimates
Simple shear flows: Order of magnitude estimates
equation
Let’s
Let’s look
look now
now at
at the
the x-component
x-component ofof the
the mean
mean momentum
momentum equation
equation and
and write
write
below
beloweach
eachterm
termits
itsorder
orderofofmagnitude.
magnitude.
∂U
∂U ∂U
UU ++ VV ∂U
∂x
∂x ∂y
∂y
!! ""
∆U∆U ss δδ ∆U∆Uss
UUss ∆U
∆UssL
LL L δδ
22
11∂P
∂P ∂⟨u
∂⟨u ⟩⟩ ∂⟨uv⟩
∂⟨uv⟩ ∂∂22UU ∂∂22UU
== −− −− −− ++ νν 2 ++ νν 2
ρρ∂x
∂x ∂x
∂x ∂y
∂y ∂x∂x2 ∂y∂y 2
uu22 uu22 ∆U
∆U ss ∆U
∆U ss
?? νν 2 νν 2
LL δδ LL2 δδ 2
Now
Now thethe survival
survival ofof atat least
least one
one ofof the
the terms
terms on on the
the left-hand
left-hand side side ofof the
the
equation
equation isis the
the essence
essence
Divide of
of free
free shear
shear
through flow,
flow,
outsince
since
by the
the
U flow
flow
( is
Uis either
either
/L) being
being speeded
speeded
up
Divide through out by Uss ( Uss /L)
up or
or slowed
slowed down
down byby thethe external
external flow
flow oror surroundings.
surroundings. Since Since wewe have
have chosen
chosen
the
the primary
primary flow
flow direction
direction to to be
be x,x, then
then the
the largest
largest ofof these
these acceleration
acceleration (or (or
deceleration)
deceleration) terms
terms isis the
the first.
first. Therefore
Therefore to to see
see the
the relative
relative importance
importance ofof the the
remaining
remaining terms,
terms, wewe need
need to to re-scale
re-scale the
the others by
bydividing
otherssource:dividing all
Turbulence all
the estimates
the21st
for the estimates byGeorge
century, W. K.by
eration) terms is the first. Therefore to see the relative importance
Turbulence:
ining terms, we need to re-scale lecture
the others by dividing all the estima
Us /L. Doing this we have:
Simple shear flows: Order of magnitude estimates
∂U ∂U
U + V
∂x ∂y
∆Us
1
Us
1 ∂P ∂⟨u2 ⟩ ∂⟨uv⟩ ∂ 2U ∂ 2U
= − − − + ν 2 + ν 2
ρ ∂x ∂x ∂y ∂x ∂y
2 2 # $ # $
u u L ν ν L
?
Us ∆Us Us ∆Us δ Us L Us δ δ

o what do these mean? And how do we decide whether


How do we decide what to keep and what to throw out? they are of th
as our leading term, or much less? (Note that if any are bigger than o
itWe need
either some
means we information
have scaled itabout
wrong,Reynolds number
or that we guessedand
wrong
h term was theturbulence
largest.) intensities and oflength
The beginning scaleslies in remem
the answer
source: Turbulence for the 21st century, W. K. George
way
out all the
turbulence. turbulence Thereforeterms on
almost the right-hand
all the side
interesting of the
flows equa
ar
bout turbulence.
gicororhocus-pocus
ic
agic hocus-pocusabout Turbulence:
Therefore
Turbulence:
Turbulence:
about almost
this;you
this; lecture
yousimply all
lecture
lecture
U U the
L
s sL
simply interesting
can’thave
can’t flows
haveturbulen
have turbule
turbule
a
mber.
umber. But But what
what does
does this
this mean?
mean? NowNow we >>
we can
>> 1 1say.
can say. ItItcouldcou
ne
one turbulence
turbulence
Simple
Simple shear
shear term.
term.
flows:
flows: And
And
Order
Order if there
if
of there
of is
magnitude
magnitudeno
is no
ν turbulence,
ν turbulence,
estimates
estimates we we
we really
reall
really
ms
rms
.d.
inin the the mean
mean momentum
momentum equation
equation areare negligible,
negligible, which
which in
and
Almost
what and
hatand doesall
Almost
does all
thisflows
this tellof
flows
tell us? interest
of
us?interest
ItIttells are
tells us usat
are at
abouthigh
high
about δ!Reynolds
Reynolds
δ! Surprised? numbers
Surprised? numbers
I IIbetbetyoyy
Surprised? bet
ell
tellususaboutaboutu, Uu,
UsLright?
s
right?
L NotNot so.so. Look
Look at
Uat
U
s
the
δ the
δ right-hand
right-hand
L
right-hand
L sideside
side ofo
of
>> 1 s
non on the
the previous
previous page>>
page 1
andand thethe orders
orders of >>
of >>
magnitude
magnitude
magnitude below
below
below it
νν νν δδ
u<<UU s ,s ,sometimes
sometimes
2 2
uu
much
much less
less andand almost
almost never
never larger.
larger.
larger. ThenThen
Then t
e term If If
Obviously is the <<
the
one
<< 1second
then,
then, criterion
involving
1 the Reynolds is much shear more
stress,stringent,
∂⟨−uv⟩/
nce Obviously
ceObviously
termUsUiss Uthe sUs the
one second
involving criterion
the Reynoldsis much shearmore
stress,
stress,stringent
∂⟨−uv
stringent
∂⟨−uv
22
stimated
order
estimated
order
orderThe The as
1010
10 or
as
only [u
or
or
only [u/(U
Uless /(U
less for
turbulence
U δ
s sδ s
turbulence∆U
for
s our
∆U
L )](L/δ).
our
term
sterm
L s “thin”
)](L/δ).
“thin”
that
thatwill Hence
shear
willHence
shear
matter
matter there
layers.
isthere
layers.
the
is the can
can be
When be
be
When
When
production no
production U
no
no turbul
termUU
term
s δ/ν
turbu
turbu
ssδ/ν
δ/ν >
ess:
ofof
ofthe viscous >> >>
stresses
nless:
less: the viscous
the viscous
There
There cancanν ν turbulence
be beno no δ δ are
stresses
turbulenceareatcertainly
atall unless, negligible,
certainly
allunless, negligible,atat atleast
leasta
least
ofof
ofthe
themean
the meanmomentum
mean momentum δ δ equation u2u2 isisconcerned.
equation concerned.But ButifififUU
But
criterion
dpretty
criterion
prettymarginal is ismuch
much
marginalassumption,more
more stringent,

assumption, ∼stringent,
andand since
since
you
you L/δ
L/δ
might
might is istypically
wanttypically
want to to ret
reo
pretty marginal L L U U
s ∆U ∆U s want to re
rur“thin”
“thin”shear
analysis. Such shearlayers.
is layers.When
unfortunately WhenUU
the s
sδ/νδ/ν s>>1000,
1000,the thecontribution
contributio
analysis.
analysis. Such is s
unfortunately thecase caseininmany manyexper
many expe
exp
sok
esare
ook are
whatWe
what
quite We
we
low
quite lowstill
certainly
certainly
we don’t
still
have
have
Reynoldsdon’t know
negligible,
negligible,
know
learned
learned what
at least
what
at
about
about
numbers.
low Reynolds numbers. to
leastto do
as do
as
turbulence
turbulence with
far far as
with pressure...
as thethe
without
without x-componen
pressure...x-compone
ever
ever
ever solvin
solv
solv
um
quite
equation is concerned. But if U δ/ν ∼ 100 only, this is
tum
n. WeWe
. So,equation
know
know how
howis concerned.
the
the growth
growth But
of our
of ourif U
free
sfree δ/ν
shear ∼
shear 100
flow
flow only,
relates
relates
relatesthis to is
to
to
So,what
So, whatthen,
what then,youyouask,
ask,dodosource:
we do
s
source: doabout
weTurbulence about
forfor
Turbulence thethe the
21st the turbulenc
century,
21st turbulen
turbule
W.
century, K. George
W. K. George
George
had to
was
keep,
order
and
was of order one. Thus we
ititwas ofoforder one.
one.
we
Turbulence:
divided
Thus
Thus we
we
by
have
its
have
order
have already decided
Turbulence:already
already
lecture
of magnitude,
lecture
decided
decided how
how
U
we
s s ∆U

we are
are
s s /L, to make
how we are going to scale a
going
going toto scal
sca
he components of the vector equation, and so we must do exactly the same thin
the components ofofthe vector equation, and so we must do exactly the same t
ere.Simple
the
But first shear
componentswe mustflows:
the Order
vector
estimate of
equation,
the order magnitude
of magnitudeestimates
and so we must do exactly the
of each term, exactly a same
Simple
here.
efore.here. But
But shear
first
firstwe
we flows:
must
must Order
estimate
estimate of
the
the magnitude
order
order ofofmagnitude
magnitude estimates
ofofeach
each term,
term, exactl
exact
before.
Lets look
before. at the results,
transverse momentum equation,
Using
Lets our
look previous
atprevious
the transverse here’s what
momentumwe get: equation,
Using our results, here’s what
Using our previous results, here’s what we get: we get:
∂V ∂V ∂V ∂V
U U ∂V + + V V ∂V
∂xU∂x + ∂yV ∂y
∂x! " ∂y
∆Us∆U δ/L δ/L ! δ ∆U" δ/L
! δ "s∆U δ/L
Us U ∆Uss δ/L∆Us∆U δ ∆Us s δ/L
UsLs L L∆UssL δ δ
L L δ
2 2 2
1 ∂P1 ∂P ∂⟨uv⟩ ∂⟨uv⟩ ∂⟨v ⟩
∂⟨v 2
⟩ ∂ V∂ 2
V ∂ V∂ 2 V
= − − 1 ∂P
=ρ ∂y − − ∂⟨uv⟩ − − ∂⟨v+2 ⟩ ν+ ν2 ∂ 2+V ν+ ν2 ∂ 2 V
= −ρ ∂y − ∂x ∂x −∂y ∂y +∂xν∂x22 +∂yν∂y 22
ρ ∂y u2 2 ∂x u 2 ∂y
2 ∆U δ/L∂x ∆U δ/L ∂y
? ? u 2 u 2 ν
s∆Us δ/L
ν ∆U
s s δ/L
L u δ u νL2∆Us2δ/L νδ 2∆U2s δ/L
? L δ ν L2 ν δ 2
L δ L δ
NowNow dividing eacheach
dividing termterm
by Ubys ∆U /L,s /L,
Uss∆U exactly
exactlyas before,
as before, we obtain:
we obtain:
Now dividing each term by Us ∆Us /L, exactly as before, we obtain:
∂V ∂V ∂V ∂V
U U ∂V + + V V ∂V
∂x U∂x + ∂y V ∂y
∂x ! "
!
∂y "
δ δ ∆Us∆Uδs !δ " source: Turbulence for the 21st century, W. K. George
source: Turbulence for the 21st century, W. K. George
δ
Wednesday, 23 October 13
∆U δ
?1 ∂P ∂⟨uv⟩ ρ ⟩∂y ∂ V∂y ∂ V
∂⟨v
− UTurbulence:
s ∆U− U ∆U δ U L L
∂y lecture
= − s + ν
s 2 s + ν 2 s
ρ ∂y ∂x ∂x ∂y
mple equation has an u2equally simple
u2 interpretation.
∆U s δ/L ∆Us δ/L It says t
mean
Simple
less you haveshear
pressure
?flows:
seen
is
this Order
only due
L to
of
all before,
the
magnitude
the νleft-hand
δradial gradient
L 2 estimates
ν side is probably
δof the transve
2
of the mean 106 convection terms2CHAPTERare of order one! InFREE
6. TURBULENT factSHEAR
the only
FLO
NowLets
Reynolds look atterm
normal thebytransverse
s ⟨v ⟩. momentum
as before,equation,
dividing each
hat is of order onestress,
in
s U ∆U /L, exactly
the we obtain: 2
whole equation is ∂⟨v ⟩/∂y, and on
vecan
U
∂Vintegrate
already + V
equation
∂V that (u6.15
agreed 2
/U across
∆U
1 ∂P s ∂⟨uv⟩
s
the
)(L/δ) shear
had
2
∂⟨v ⟩ tolayer
be2 offrom
∂ V 2 a o
order
∂ V
g
= − − − + ν 2 + ν 2
nity (orterm
∂x
ulence anywhere
in the! else
∂y forρ ∂ythat matter)
x-momentum
" equation.to
∂x obtain:
∂y Of course,
∂x there
# $
∂y ca
δ ∆Us δ ! "
u2 u2 L ν δ ν
on withonly a single term ?equal to zero, unless we have badly over
L Us L Us ∆Us Us ∆Us2 δ Us L L Us δ
P (x, y) =there
er of magnitude. Fortunately P (x,is∞) the−pressure
ρ⟨v ⟩, term left to ba
Unless2you have seen this all before, the left-hand side is probably a surp
Everything
t order in δ/L ∼drops
u /U out
s ∆Uand
s , they-momentum
the only terms remaining equation are,
reduces
none of the mean convection terms are of order 2 one! In fact the only estim
ng of course the free stream value of ⟨v ⟩ to be2 zero. If the
term that is of order one in the whole equation is ∂⟨v ⟩/∂y, and only bec
2
stant (or zero)
we havemean
already velocity,
agreed that 1 ∂P
then
(u 2 ∂⟨v
/Us ∆UP (x,⟩∞)
s )(L/δ) had = P∞
to be = const
of order one to
a turbulence term 0in ≈the−x-momentum − equation. Of course, there cannot b
imply: ρ ∂y
equation with only a single term ∂y unless we have badly over-estim
equal to zero,
its order of magnitude. Fortunately there is the pressure term left to balance
imple equation has
to first aninequally
order δ/L u 2 simple interpretation.
P (x, y) = P∞ − ρ⟨v ⟩
∼ /Us ∆Us , the 2
y-momentum It
equation says
reduces that
to sim
mean pressure is only due to the radial1 ∂P gradient
∂⟨v 2 ⟩ of the transverse
2 0≈− − (
Reynolds normal
of these can be stress, ⟨v ⟩. into the
substituted ρ ∂y x-momentum
∂y equatio
e can integrate
This equation 6.15
simple equation across
has an equally the
source: shear
simpleTurbulence layer
for the 21stfrom
interpretation. Itcentury,
says W. aK. given
that George
the ch
momentum is considered. But without this small mean pressure gradient across
the
theflow
flow, would remain
be no parallel. Clearly
Turbulence: lecture whether an effect is negligib
the
pends
flow
on
would
which
remain
there would
question
parallel.
is being Turbulence: lecture
entrainment Clearly
and no whether
growth of the an effect
shear layer. is
Innegligib
other
words,
pends onthe flow would
which remain
question being asked.
is parallel. Clearly whether an effect is negligible or
asked.
Simple shear flows: Order
is beingof magnitude estimates
Simple shear flows: Order of magnitude estimates
not depends on which question asked.

The
The free
6.2.5 free shear
The free layer
shear layerequations
equations
6.3.
If we TWO-DIMENSIONAL
assume the Reynolds TURBULENT
number is always JETS
large enough that the viscous 107
terms
ssume
ssume the
the Reynolds
Reynolds number
6.3. TWO-DIMENSIONAL number is always JETS
TURBULENT large enough that the viscous 107 t
can be neglected, then as noted above, the pressure term in the x-momentum
neglected,
neglected,
equation canthen
then as
as noted
be evaluatednoted above,
above,
in terms
the pressure term in the x-momen
of P∞ and ⟨v 2 ⟩. Thus, to second-order in
22
equation
2 can be evaluated in
u /Us ∆Us or δ/L, the momentum equations terms of P ∞
∞ andfor a ⟩.
⟨v freeThus,
shear toflow second-order
reduce to ain
22
usingle
/Uss∆U ss or δ/L, the momentum equations for a free shear flow reduce to a
equation:
single equation: ! " ! "
∂U ! ∂U " 1dP∞ ∂ # $
!∂
U∂U + V ∂U = − dP − ∂ ⟨uv⟩ − ∂ #⟨u2 ⟩ − ⟨v 2 ⟩ $"
! " (6.18)
U ∂x + V ∂y = −⇢ dx∞ − ∂y ⟨uv⟩ − ∂x ⟨u22⟩ − ⟨v 22⟩

(6.18)
∂x ∂y dx ∂y ∂x
As we have seen above, the second term on the left-hand side (in curly brackets)
may Asorwemayhavenot seen
beabove, the second
important, term on the
depending left-hand
whether Us =side
∆U(in s orcurly
∆Usbrackets)
/Us <<
may or may this
1. Clearly not be important,
depends on thedepending on whether
velocity deficit Uss = ∆U
(or excess) ss or ∆U
relative ss/Ussfree
to the <<
1.stream.
Clearly The this depends
second term on the velocity
in brackets on thedeficit (or excess)
right-hand side isrelative to the free
also second-order
(in u2 /UThe
stream. s ∼ δ/L)
s ∆Usecond compared
term to theonothers,
in brackets and so could
the right-hand siderightly
is alsobesecond-order
neglected.
(in u22/U
It has been
ss∆Uretained
ss ∼ δ/L)for now sincetointhe
compared some flows and
others, it does not vanish
so could rightly withbeincreasing
neglected.
Itdistance
has been from the source.
retained for nowMoreover, it canflows
since in some be quite
it doesimportant
not vanish when withconsidering
increasing
2 2
integralsfrom
distance of thethemomentum equation,itsince
source. Moreover, can bethesource:
profiles
quite of ⟨u
important
Turbulence for ⟩
the and
when
21st ⟨v ⟩ do
consideringnot
source: Turbulence for the 21st century, W. K.George
century, W. K. George
Turbulent
efree
free stream
Turbulent
an
an
stream
ambient
ambient
jetsisis
jets are
are generated
assumed
generated
assumed
environment.
environment.
to
The
The
Turbulence:
to Turbulence:
byby a
havea concentrated
have
state
state
concentrated
zero
zero
ofthe
of theenvironment
lecture
lecture
source of
streamwise
source
streamwise
environment
ofvelocity,
momentum
velocity,
momentum
andthe
and theexternal
external
thenissuing
the
issuing
then the into
scale
into
boundary
boundary
scale
It
elocity
elocity is easy
and
and to see
gradients
gradients that the
of
of it
it entrainment
are
are the
the same,
same, velocity
so
so the
the calculated
equations
equations reduce
reduce to
to simpl
simpl
easy
onsistent
nnt7.16
7.16 Simple
with
to see that
Simple
conditions
conditions
and
and
our
with
the
the shear
shear
are
are
order
the
very
very
our
mean
mean
of
flows:
entrainment
flows:
important
important
order of Order
Order velocity
in
in
magnitude
continuity
continuity
magnitude
ofof
determining
determining
equation.
estimate
magnitude
calculated
magnitude
estimate
equation.
above
how
how
The
The above
(i.e., latter
latter estimates
estimates
the turbulent
the turbulent
(i.e.,
can
can be
be
flow evolves.
flow evolves.
multiplied
multiplied b
b
The
The simplest
simplest case
case to
to consider
consider is
is that
that
hat the integral makes no sense if the mean velocity U does in
in which
which the
the environment
environment is
is at
at rest
rest and
and
ld:
ed: integral
unbounded,
unbounded, makes
6.3.
so no
so all
all sense if the
TWO-DIMENSIONAL
the boundary
the ! mean
boundary
! velocity
conditions
conditions " U
aredoes
TURBULENT
" are JETS The
homogeneous.
homogeneous. The jetjet itself
itself can
can
creasing |y|. ∂Usource ∂Vor
gbe|y|.
be assumed
assumed to to issue
issue from
from aa line∂U
line source∂V or aa slot
slot asas shown
shown in in Figure
Figure 6.5. 6.5. The
The jetjet
is assumed
med to have to
zero have zero U
streamwise U velocity,
streamwise ++ then
velocity, =scales
=
the 0 0 the scales 2
then (6.21
(6.21
then evolves
then evolves and spreads
and
equation spreads
can be downstream
∂x
downstream
∂x
evaluated by
∂y
by
∂y entraining
in entraining
terms of Pmass mass
∞ and
from
from the Thus,
⟨v the
⟩. surroundings
surroundings
to secon
sadients
of it
which are
which are at
ofthe
are itatsame,
are
2 most
u most
theso
/Us ∆U insame,
the so the motion
equations
irrotational
ins irrotational
or δ/L, the
equations
reduce
motion
momentum to reduce
simply
induced by
inducedequations
tothe
simply
vortical
by the vortical fluidshear
fluid
for a free withinflow
within there
the
his
is isisBut
mean
continuity
jet.
jet. added
continuity
added
But no
no to equation
to
Add
equation.
new
new equation
equation.
this
The latter
momentum
momentum 7.16
to7.16The
the
isiscanthe
latter
the terms
momentum
added
added tocan
beterms
multiplied
to the can
be
thecan
flow
flow be combined
multiplied
be combined
equation,
by
downstream
downstream to
by ofof the
to obtain:
theobtain:
source, and
source, and
single equation:
itit isis this fact
! this fact
! thatthat distinguishes
" distinguishes
" ! the
the " jet jet from
from
# all other
all other $flows.
flows.
! As we
As we shall
shall see see
"
below, ∂U∂∂ ∂V
2
however,∂U ∂ ∂ ∂V
that ∂U
the ∂∂
rate ∂U
at which ∂∂ 1 #
dP
momentum2 ∂ 2
$
crosses ∂
any #
x-plane is $
not
below,
U however,
U
U+U 2
++ that
+
= UU
0 the
V + =
UV + + ⟨uv⟩rate0 V at
⟨uv⟩ which
+ + momentum
⟨u
∞ 2
= − ⟨u ⟩−− ⟨v⟨uv⟩ ⟩
(6.21)− ⟨v crosses
2
⟩⟩ =
(6.21)
=
− any
0
0 x-plane
2 is2 not
⟨u ⟩ − ⟨v ⟩(6.22 (6.22
quite∂x
quite ∂x∂x
constant
constant ∂y∂xat ∂y
at
∂y the
the ∂ysource
source
∂x ∂y ∂y value
value ∂yduedue∂x ∂x
to⇢the
to the small
dxsmall streamwise∂x
∂ystreamwise pressure gradient
pressure gradient
oionarising
arising7.16
equation from
from
the the
the
terms
7.16 theturbulence
turbulence
can be
terms normal
normal
combined
can be stresses.
stresses.
to
combinedobtain:to obtain:
n be
be integrated
integrated
The
The averaged
averaged across
Asacross
we have
continuity
continuity the
the seen flow
flow above,
equation
equation forcan
for theany
any
can be
be given
second
given term
integrated
integrated value
value onfrom
from ofthe
the
of x to
the to
xleft-handobtain:
side(where
obtain:
centerline
centerline (in curly
(where
itit∂isis∂zero
zero bybymay
∂ ∂ or
# may not
symmetry)
symmetry) #
to
to
$be important,
obtain
∂ ⟨v 2 ⟩2 = 0 2
obtain V V , , $
i.e.,
i.e., depending on whether U s = ∆U s or ∆
V + ⟨uv⟩ + ⟨u%%2 ⟩∞− (6.22) deficit
+ ∂y U V + 1. ∂x dd
⟨uv⟩
Clearly+ ∞ this ⟨u2 ⟩ −
depends ⟨v 2 ⟩ on = 0
the2 velocity (6.22) (or excess) relative t
∂y ∂y ∂x [U 2 + (⟨u2 ⟩ −%%⟨v 2 ⟩)]dy = 0 (6.23
stream.
dx The [U +
second term (⟨u ⟩ − ⟨vyy ⟩)]dy
∂U
∂U
in brackets = 0
on the right-hand side is (6.19) (6.23
also se
dx −∞
−∞ value of x V
V to = − −
= obtain: dỹ
dỹ (6.19)
s the flow
d across the(in for any
flow
2 given
u /Ufors ∆Uanys ∼ givenδ/L)value compared of00x ∂x ∂x
to obtain:
to the others, and so could rightly be
2for now since 2 in some flows it does not vanish with
e
e ItIthave
have assumed
It
assumed
immediately
immediately
% ∞ 2
has that
been
that
follows
follows UU
that
that , ,
retained ⟨u
⟨uthe
the 2
V⟩,
⟩,
V and
and
velocity
velocity ⟨v
⟨v 2
⟩⟩
at
at vanish
±∞
vanish
±∞ is
is as
as
given
given |y|
|y|by:
by: →→ ∞.
∞. (Remembe
(Remembe
d
[U These
These
2
sumptions + (⟨u two
two
⟩ − equations
equations
⟨v 2
distance
2 the next ⟩)]dy
2 from = 20the
time come
come inin handy
source.
someone handy
Moreover,
tries when
when
(6.23)
to it
tell we
we
canyou look
look
be quite
that for
fora self
self
important
small similarity
similaritywhen
co-flowin c
∞umptions [U the+ (⟨u next
integrals
⟩ −time
of
⟨v ⟩)]dy
the
someone= 0 tries
momentum d d %to
%∞
equation, ∞tell you(6.23)
since
that
the
a small
profiles of
co-flowin
⟨u 2
⟩ and ⟨
dx
and modest
−∞ background V turbulence
=
V∞∞turbulence
= −V
−V−∞ −∞ = − level in
= − U they)dy
(x,
U (x, external flow
y)dy flow areare(6.20)not im
(6.20) im
nd modest
t U , ⟨u ⟩, and2 background
vanish
⟨v 2
⟩ as rapidly
vanish as with
|y| → increasing
∞. level
dx
(Remember in
dx 0y0 source:
as the
source:
external
doTurbulence
the other
Turbulence forthe
for terms.
the 21stcentury,
21st century,W.
not
W.K.K.George
George
2 2
This idea is not the same as ‘self-similarity’ – or ‘self-preservation’ in Townsend’s
Turbulence: lecture
terminology – which we now discuss.
Self-similarity
Consider a quantity Q(x,y) – dependent on two independent variables (x & y). Define
characteristic scales for the dependent and independent variables, Qo(x) and δ(x),
respectively. So scaled variables can be defined by:

η ≡ y/δ(x) and f˜ (η, x) ≡ Q(x, y)/ Qo (x) .

If the scaled dependent variable is independent of x (i.e. there is a function f(η) such that
f˜ (η, x) ≡ f (η) ) then Q(x,y) is self-similar. Q(x,y) can be expressed as functions of single
independent variables – Qo(x), δ(x) and f(η).

With this,thewe
Of course, canmust
scales predict what
be chosen will happen
appropriately. And,atusually,
different downstream
self-similarity only
distances
occurs over by obtaining
limited ranges of x. Also,information at only
if Q(x,y) is governed by a one location
pde then Qo(x), δ(x) and
f(η) are all governed by odes.
f˜ (η, x) ≡ f (η) ) then Q(x,y) is self-similar. Q(x,y) can be expressed as functions of single
Turbulence: lecture
independent variables – Qo(x), δ(x) and f(η).
Self-similarity
Of course, the scales must be chosen appropriately. And, usually, self-similarity only
occurs over limited ranges of x. Also, if Q(x,y) is governed by a pde then Qo(x), δ(x) and
f(η) are all governed by odes.

So, in general, we look for solutions of our governing equations in which:

U = U1 + uo f(η), (η=y/δ)

uv = qo 2 g12 (η) , u 2 = qo 2 g1 (η ) and v 2 = qo 2 g2 (η) .

U1 is a translational velocity (almost always the free stream) and uo and δ are
characteristic velocity and length scales.

∂U uo
We may therefore write: = f ' (η ) (dashes denote differentiation w.r.t. η.
∂y δ
et showing jet source, coordinate system and typical
Turbulence: lecture
Turbulence: lecture
CHAPTER 6. TURBULENT FREE SHEAR FLOWS
ItItPlane
Plane
is
is easy
easy
jets
jets
toto see
see that
that the
the entrainment
entrainment velocity
velocity calculated
calculated
asy
easy to see
to see that
thatour the entrainment
theorder
entrainment velocity calculated
velocityestimate
calculated
nsistent
onsistent with
with our order of magnitude
of magnitude estimate above (i.e.,
above (i.e.,
t with
nt with our
our order of
ordermakes magnitude
of magnitude estimate
estimate above (i.e.,
abovevelocity
(i.e., U does
hat
that the integral
the integral makes no sense
nothesense if the
if the mean
mean==UUUvelocity UJETS
does velocity
integral
e integral|y|. 6.3.
makes
makes noTWO-DIMENSIONAL
sense if mean TURBULENT
UUo (x)
velocity
(x) (x,
does
(x,0)0) Centerline
Centerline velocity ofofthe
the jet
jet
creasing 6.3. no sense if the mean velocity
TWO-DIMENSIONAL U does
o TURBULENT JETS
gcreasing
|y|.
|y|.
g assumed |y|. (x)
(x) --Jet
Jet half
half width
width --UUo (x)/2
o (x)/2 ==UU(x,
(x, ))
s
is
med assumed
to have toto
zero have
have zero
zero
streamwise streamwise
streamwise
velocity, velocity,
velocity,
then the then
then
scales the
the scales
scales 2
med to have equation
zero can
streamwise be evaluated
velocity, in
then terms
the of
scales P ∞ and ⟨v 2⟩. Thus, to second
adients of it equation
are the can
same, be
so evaluated
the equationsin terms
reduce of P
to∞simply ⟩. Thus, to secon
and ⟨v
of it
sadients are the
of it areofthe it same,
2
uusame,
2/Us ∆U
/U
so
∆U
the
are thesossame,or
the
or
equations
so the
δ/L,
δ/L, the
equations
the
Look
reduce
Look
equations
momentum
reduce
momentum
toat
to the
simply
at the
reduce momentum
momentum
to
equations
simply
equations
simply
for
for aa free
free
equation,
equation,
shear
shear flow
flow re
re
mean
continuity continuity
mean continuityequation. s equation.
s
The
equation. The
latter can
The latter
be
latter can be
multiplied multiplied
by
can be multiplied by
by
continuity single
equation. equation:
The
single equation: latter can be multiplied by
!! !! " "" ! " ! "
∂U ∂U
∂V " ∂V ! ∂U " !∂ # $$"
∂U ∂V∂U ∂V∂U
∂U ∂U dP
1dP ∞ ∂∂ 2 2 ∂ #
U
plane
U ∂x U+
U+ ∂y
jet showing +
=+
jet
= 0
U0U =
+
=
source, +00V
coordinate
V = −
system
= − (6.21)

and−typical
(6.21)
− (6.21)
⟨uv⟩ −
(6.21)
⟨uv⟩ − ⟨u
⟨u 2⟩⟩−−⟨v
⟨v 2⟩

∂x ∂x
∂x
∂y ∂y ∂x
∂y∂x ∂y
∂y dx
dx ∂y
∂y ∂x
∂x
. ⇢
oon
ion 7.16
7.16 the
equation
equation the terms
7.16 the
terms
7.16As
can
the
wecan be
be combined
terms
terms
have can
seen
@U
As we have seen 2
be
combined
can be
2
to obtain:
combined
tothe
combined
above,@(U V )
above, the
to
obtain:to
second
second
obtain:
obtain:
term on
@huvi
term onthe
theleft-hand
left-handside
side(in
(incurly
curly
∂∂ may #
or
∂∂ or may not @U
$
be + @(U
important,
##2 $be important, $$
V ) =
depending @huvi
on whether UUss == ∆U ss or ∆U
∂∂
It UisV⟨uv⟩
. ++ +∂may

# 2may
⟨u ∂
⟩⟩−∂ not
⟨v @x
⟩depends + @y =
depending @y on whether ∆U or ∆
V easy
+
⟨uv⟩ to
+
1. see
⟨uv⟩ that
Clearly+
⟨u 2 the
− ⟨v
this ⟨u2 2
⟩ 2⟩=
entrainment
@x =− 00⟨v 2⟩on
2 velocity
@y
= 0
the (6.22)
calculated
(6.22) @y
velocity (6.22)
deficit (or excess) relative tt
+onsistent
∂y U
∂y V + ∂y1.∂x⟨uv⟩
Clearly+ ∂x this ⟨u ⟩ −
depends ⟨v ⟩ on= 0
the velocity (6.22)
deficit (or excess) relative
∂y∂y with ∂y ∂x
our order
Integrate
∂x with respect
of magnitude estimate above (i.e.,
to y,
that
the the
flow
stream.
integral
forZ
stream.
anymakes
TheIntegrate
The
no
second
second
sense if
22given value of x to obtain:
term
term
the with
in
meanin respect
brackets
brackets
velocity U
on
on to
the
the
does
Zothers,y,
right-hand
right-hand side
side is
is also
also sec
se
s the
across flowthe
d acrossdd|y|.
ncreasing
(in
forZ
the flow
u
any
flow
(in1 u /U
for
/U ss∆U
givenany
∆U ss ∼
value δ/L)
given
∼ of
δ/L) x compared
valueto obtain:
of
compared
1 for any given value of x to obtain: 1 x to
toto the
obtain:
the Z
others,
1 and
and so
so could
could rightly
rightly be
be
It has 22 retained for now since in some flows it 2does
been not vanish with
is 2assumed
%% ∞ 2 toIt has
have UUbeen
zerody
dy = 0
retained
streamwise
= 0 for now
velocity, since
thenMMin =
thesome
= scalesflows⇢U
⇢Uit 2dy
does
dy = not
= constant
vanish
constant with
d[U
[Ud 2++∞
radients dx
(⟨u
dx
(⟨u
of 22it − ⟨v
2
⟩⟩distance
− ⟨v 2⟩)]dy
distance
are 1the2 from
⟩)]dy
same, =
=
from 200the
so the
the source.
source.
equations Moreover,
Moreover,
reduce (6.23)
itsimply
(6.23)
to it can
can1 be
be quite
quite important
important when
when cc

[U 2+ (⟨u
1 2⟩ − ⟨v
[U + (⟨u ⟩ − ⟨v ⟩)]dy = 0 2⟩)]dy = 0 1(6.23)
(6.23) 2
Turbulence: lecture
Turbulence: lecture
CHAPTER 6. TURBULENT FREE SHEAR FLOWS
. It Plane jets
is easy Plane jets
to see that the entrainment velocity calculated
easy to see that the entrainment velocity U (x) calculated
= U (x, 0)above
Centerline
consistent with our order of magnitude estimate o (i.e.,velocity of the jet
nt with our order of magnitude estimate (x) -above
Jet (i.e.,
half widthU- U o (x)/2 = U (x, )
that the integral makes no sense if the mean
Uo (x) velocity
=UU does does
(x, 0) Centerline
e integral makes6.3. no sense if the mean velocity
TWO-DIMENSIONAL TURBULENT JETS velocity of the jet
ncreasing
ng |y|. |y|. In(x)the
- Jetself-similar
half width - Uoregion,
(x)/2 = U (x, )
is assumed
med to have to have
zero zero streamwise
streamwise velocity, velocity,
then the then the scales 2
scales
equation can be evaluated U in
= terms
U (x)fof (⌘),
P∞ andwhere⟨v ⟩. ⌘ Thus,
= y/ to(x)
secon
radients
s of it areofthe
it same,
are thesosame,
2 the equations Look at the
so the equations
reduce
u /Us ∆Us or δ/L, the momentum to reduce
o
simply momentum
to
equations for a free equation,
simply
shear flow r
econtinuity
mean continuity
equation. equation.
The latterThe
canlatter can be multiplied
be multiplied 2
by by
single equation:
Z huvi = Uo g(⌘)
! ! " "1 ! " ! "
2 2
M
f plane=∂U⇢U ∂V
U+ o
∂U (x)
(x)
U jet showing +
jet
∂V∂U f (⌘)
= source,
0U = 0V
∂U
coordinate
+
d⌘ 1
systemdP
= − and
∞ ∂
−typical
(6.21) Since M is a conserved,
(6.21)
⟨uv⟩ −
∂ #
⟨u2 ⟩ − ⟨v 2 ⟩
$

e. ∂x ∂x
∂y ∂y∂x 1 ∂y ⇢ dx ∂y ∂x
tion 7.16 the7.16
o equation
2
terms the can
termsbe combined
As we have seen can be
above,
to obtain:to obtain:
combined
the second dUon
term 1 d side (in curl
o the left-hand
Uo (x) (x) =# constant @U 2
@(U V ) @huvi=
w. ∂ ∂
It is easy to
may
∂ ∂ or
see that
may ∂
the
not
#entrainment
+
$be important,$ = U
depending
o dx 2
on whether dx Us = ∆Us or ∆
V + ⟨uv⟩ + ⟨u 2
⟩ − ⟨v 2
⟩ 2 = 0 2 velocity
@x @y calculated
(6.22) @y
+ U
∂y V + 1. ⟨uv⟩
∂x +
Clearly this ⟨u ⟩ −
depends ⟨v ⟩on = 0
the velocity (6.22)
deficit (or excess) relative
∂ySubstitute
consistent with ∂y all self-similar
our order ∂xof magnitude profiles
estimatein to theonmomentum
above (i.e., and cros-
stream.
that the integral makes no TheIntegrate
sense if the with
second term in respect
brackets
mean velocity U doesto
the y,
right-hand side is also se
ss the flow forZ any anysstream velocity equations Z 1 and so could rightly be
2 given value of xvalue
to obtain:
ed across d|y|.
ncreasing (in
the flow 1 /U
u fors ∆U ∼ δ/L)
given compared
of x to obtain:
to the others,
It has been
2 retained ✓ for now Zsince ⌘ in ◆ flows it 2does not vanish with
some
m
∞ is assumed
% to have U zerody 1streamwise
d= 0 velocity, thenM the=scales ⇢U dy = constant
d 2 ∞ 2 2 2 0 0
[U + (⟨u
gradients dx
of it2⟩ distance
− ⟨v
are the⟩)]dy
1
[U + (⟨u ⟩ − 2⟨vdx
from
2 same, =so20the f
the source.
⟩)]dy = 0
+
equationsf reducef(6.23)
Moreover, d⌘
to =
it can
simply 1g
be quite important when
(6.23)
∞ 2
hedxmean −∞ continuity integrals
Wednesday, 23 October 13
of
equation. theThe momentum
latter can equation,
be 0multiplied since
by the profiles of ⟨u ⟩ and ⟨
Turbulence: lecture
Turbulence: lecture
CHAPTER 6. TURBULENT FREE SHEAR FLOWS
. It Plane jets
is easy Plane jets
to see that the entrainment velocity calculated
easy to see that the entrainment velocity calculated ✓ aboveZ (i.e., ◆
consistent with our order of magnitude estimate ⌘
nt with our order of magnitude estimate
that the integral makes no sense if the
1 d
above (i.e.,
mean 2
velocity 0U does 0
e integral makes6.3. no Uo (x)
sense if the mean velocity
TWO-DIMENSIONAL =UUfdoes
(x,+
TURBULENT 0) f JETSf d⌘
Centerline = gof the jet
velocity
ncreasing |y|. 2(x)dx- Jet half width -0 Uo (x)/2 = U (x, )
ng |y|.
is assumed
to have to have zero streamwise velocity, then the scales 2
med zero streamwise velocity,
equation can be evaluated For
then self-similarity,
the
in termsscales
of P∞ and ⟨v ⟩. Thus, to secon
radients
s of it areofthe
it same,
are thesosame,
2 the equations Look at the
so the equations
reduce
u /Us ∆Us or δ/L, the momentum to reduce
simply momentum
to
equations for a free equation,
simply
shear flow r
econtinuity
mean continuity
equation. equation.
The latterThe
canlatter d can be multiplied
be multiplied by by(x) ⇠ x
single equation: =S 1/2
! ! " " ! " dx ! o (x) ⇠ x
U "
∂U ∂U
∂V ∂V∂U ∂U 1dP ∞ ∂ ∂ # $
f plane U+
U jet showing +
jet
= source,
0U = 0V
coordinate
+ system
= − and −typical
(6.21) (6.21)
⟨uv⟩ − ⟨u2 ⟩ − ⟨v 2 ⟩
e. Self-similarity
∂x ∂x
∂y in∂xmean and
∂y ∂y shear ⇢ dx stress
∂y implies ∂x that eddy
tion 7.16 the7.16
o equation terms viscosity
the can
termsbe combined
As we have seen can
@U
should
2
be
above, also
to obtain:
combined
the
@(U V )
be self
to obtain:
second term similar
@huvithe left-hand side (in curl
on
w. ∂ ∂
It is easy to
may
∂ ∂ or
# may not
see that ∂
the #
entrainment
+
$be important, $ =
depending on whether
p Us = ∆Us or ∆
V + ⟨uv⟩ + ⟨u 2
⟩ − ⟨v 2
⟩ 2 = 0 2 velocity
@x @y calculated
(6.22) @y
+ U⌫
∂y
consistent
∂y
VT +
(x)
with
1.
∂y
= ⟨uv⟩
∂x U
Clearly
our o +
(x)
order (x)
this
of
∂x
⟨u ⌫ ˆ⟩ −(⌘)
depends
T
magnitude
⟨v ⟩on =
estimate
0
the ⌫
velocity
above T (x) ⇠
(6.22)
deficit
(i.e., x
(or excess) relative
stream.
that the integral makes no TheIntegrate
second
sense if the with
termmean respect
in velocity
brackets Upon to
the y,
does right-hand side is also se
ss
ed the
Localflow
across d|y|.
ncreasing
forZ(in
Reynolds
the flow any
u 2 given value of x to obtain:
/U ∆U
number, ∼
1 for any given value
s s δ/L)Re o (x) of x ⌫to ⇠
compared= obtain:Z increases
Uo to the others,
x, 1 and sowith coulddistance
rightly be
Ithave
has U been retained
2 streamwise for now since in some flows it 2does not vanish with
m
∞ is assumed
% to zero dy = 0 velocity, thenM the= scales
Uo ⇢U dy = constant
d Turbulent
[U 2
gradients ∞
+ (⟨u dx 2
are Reynolds
of it2⟩ distance
− ⟨v 2
the⟩)]dy
2 from
same, = 0
so
2
1 ⟩ − ⟨v ⟩)]dy = 0
number,
the
the source.
equations Re T (x)(6.23)
Moreover,
reduce =simply
to it ⌫can
T 1
is
be independent
quite important of x
when
∞ [U + (⟨u (6.23) 2
hedxmean −∞ continuity integrals
Wednesday, 23 October 13
equation.of the Themomentum
latter can equation,
be multiplied since
by the profiles of ⟨u ⟩ and ⟨
convected
5.2 in a uniformChannel
Two-dimensional free-stream.
Flow
5.2 Two-dimensional Channel Flow Turbulence: lecture
The laminar (fully developed) equivalent yields the well-known parabolic – Poiseuille –
5.2
TheTwo-dimensional Channel Flow
Wall-bounded flow : Channel
profile and in (fully
laminar that case the shear
developed) stress is linearly
equivalent distributed.
yields the well-knownDefine the bulk
parabolic velocity –
– Poiseuille
as
The andand
profile assume
U laminar in thatthe
(fully topthe
case and
developed) bottom
shear walls
stress
equivalent are no-slip
is yields
linearly boundaries.
distributed.
the well-known DefineWe discuss
the
parabolic bulk the –
velocity
– Poiseuille
as U andandassume
corresponding
profile fullythe
in that top
theand
developed
case bottom
shear stresswalls
turbulent are Reynolds
(high
is linearlyno-slip boundaries.
number)
distributed. We
thediscuss
flow.
Define the
bulk velocity
corresponding
as U and assume fully
thedeveloped turbulent
top and bottom walls(high Reynolds
are no-slip number) flow.
boundaries. We discuss the
corresponding fully developed turbulent (high Reynolds number) flow.

δ
δ y
δ y
y
x
x
Consider the various elements of u' i u' j . Using cartesian x,y,z coordinates for the (1,2,3)
x
Consider the various elements of u' i u' j . Using cartesian x,y,z coordinates for the (1,2,3)
directions and (u,v,w) for the corresponding velocities, we note first that u' w' and v' w'
Consider the various elements of u' i u' j . Using cartesian x,y,z coordinates for the (1,2,3)
directions and (u,v,w) for the corresponding velocities, we note first that u' w' and v' w'
are identically zero by symmetry about the z=0 plane (w is as likely to be negative as
directions and (u,v,w)
are identically zero byfor the corresponding
symmetry velocities,
about the z=0 plane (wweisnote first that
as likely w' and v'
to beu'negative asw'
positive). The mean flow is in the x-direction so V=W=0 and U depends only on y.
are identically
positive). Thezero
meanbyflow
symmetry
is in theabout the z=0soplane
x-direction V=W=0(w isand
as likely to be negative
U depends only on y.as
Thus the mean momentum equations simplify to:
positive). The mean
Thus the mean flow is in
momentum the x-direction
equations simplifysoto:
V=W=0 and U depends only on y.
Turbulence: lecture
Channel flow 1 ∂P ∂ uv ∂U 2
0=− − +ν 2 ,
ρ ∂x ∂y ∂y
1 ∂P ∂ v 2
0=− − ,
ρ ∂y ∂y
1 ∂P P
0=− P is independent of z
ρ ∂z
where Pw is the mean pressur
Identical to the laminar flow equation you have seen before, except
now we have the expression
turbulent into (5.3a) yields:
stresses
P = Pw (x ) − ρ v (y) ,
2

ass continuity is automatically satisfied). Notice that these differ from


Integrate
mean pressure at the
P = walls
P (x ) and
− ρ v 2
depends
(y) , only on x. Substituting th
e only by the turbulentwstress terms in (5.1a) and (5.1b). Clearly P is in
equation 2
5.3a)
s the
d, fromyields:
mean pressure at the walls and depends only on x. Substitutin
(5.3b)
Substitute
into in ∂Pw
(5.3a) yields: dP dτ dU
= = , where τ (y) = µ − ρ uv is t(
equation 1 ∂x dx dy dy
∂Pw dP dτ
= = ,
τ
he mean pressure at the walls
expression into (5.3a) yields: and depends only on x. Substituting
ields: Turbulence: lecture
to (5.3a) yields: the∂Pfriction
=

dP Reynolds
= number is de
Channel∂flow
w
Pw dP dτ ,
= =∂P , ∂x dx dy (5.4)
∂x dx dy dP dτ
w
= = , from the wall in viscous le
distance
dU
∂x dx dy
where τ (y) = µ − ρ uv is the total mean shear stress. In (5.4), since the
LHS depends
dy on x and RHS depends on y
v is the total mean shear stress. In (5.4), since the LHS is a
dU
function only of x and Must
the be
RHS a
only constant
of y both terms must be constant. It f
µ − ρ uv is the total mean shear stress. In (5.4), since the LHS y
hedy
RHS only of y both terms must be constant. It follows that
dU
τ = τ w y / δ , where τ w = µ , just as for the laminar flow equivalent. In
ofdUx and the RHS only ofdyy yboth =0 terms must be constant. It follow
µ , just as for the laminar flow equivalent. In this case,
But,
y =0 totalthe
dyhowever,
dU stress
totalincludes a Reynolds
stress includes shearstress
a Reynolds stresscontribution.
contributionThe ski
Since the latter is like a Reynolds n
here τ w = µ , just as for the laminar flow equivalent.
τ w The skin friction
In this
s includes
Skin a
coefficientReynolds
dy
friction
is defined
y =0
stress
coefficient: contribution.
C f = Reynolds
as usual by:viscous stress
2 . number:
is compared with the
1
ρU
τw 2

usualstress
total =
by: C fincludes 2 . a Reynolds as Re = U (2δ )The
stress contribution.
defined / ν . skin frict
1
2ρU
τ w _______________________________
τ
defined as usual by: C
We can define viscous fscales:= the
2 .friction velocity, u = w
and viscous le
2 ρU
τ
ρ
1
τw
coefficient is defined as usual by: C f = 2 .
Turbulence: lecture
ρU1
2

Wall-bounded flows
τw
We can define viscous scales: the friction velocity, uτ = and viscous length scale,
ρ
ν
δν = . Note that the Reynolds number based on these is identically unity, whereas

u τδ
the friction Reynolds number is defined by: Re τ = . And we can measure the
ν
distance from the wall in viscous lengths – wall units – denoted by:

+ y uτ y
y ≡ =
δν ν
When non-dimensionalised using the viscous scales, it is referred to
asSince
“inner” scaling since the scaling represents the view point with
the latter is like a Reynolds number it will provide a measure of how large the
respect to the wall
viscous stress is compared with the total stress. The channel Reynolds number is often
defined as Re = U (2δ ) / ν .
“Outer” length scale: “Outer” velocity scale: Ucl , Ue , U1
_____________________________________________________________________________
Turbulence: lecture
Turbulence: lecture
Wall-bounded
Wall-bounded flows
flows
Fully developed turbulent channel flow can be specified by ⇢, ⌫, and u⌧ .

There
Thereare
aretwo
twoindependent
independentnon-dimensional groups: yy++ and
non-dimensionalgroups: andy/y/ (y/ = yy++/Re
(y/ = /Re⌧⌧))
⇣⇣yy ⌘⌘
Mean
Mean velocity
velocity can
can be
be written
written as,
as, UU =
= uu⌧⌧FFoo ,,ReRe⌧⌧
FFoo isis aa universal
universal function
function that
that isis yet
yet to
to be
be determined
determined

Instead of
Instead of this
this form,
form, lets lets examine
examine the
the gradient
gradient of
of mean
mean
velocity as
velocity as itit isis more
more dynamically
dynamically relevant
relevant
dU uu⌧⌧ ⇣⇣ ⌘⌘
dU + yy
+
The gradient
The gradient of
of mean
mean velocity
velocity depends
depends on,on, == yy ,,
dy
dy yy
where, isis aa non-dimensional
where, non-dimensional universal
universal function
function to
to be
be determined
determined
Turbulence: lecture
Turbulence: lecture
Wall-bounded flows:
Wall-bounded flows:Mean
Mean velocity
velocity profiles
profiles
Prandtl postulated that high Reynolds numbers, close to the wall (y/ << 1)
that there is an “inner” layer in which the mean velocity profile is determined
by the viscous scales and is independent of and U1
⇣⇣ + yy⌘⌘ + yy
yy+ ,, ! (y
! (y+ ),), for
for <<<<11

+
dU uu⌧⌧
dU + dU
dU +
11 +
=
= 1 (y
(y ) In non-dimensional
+ ) In non-dimensional form,
form, +
= +
= 1 (y
(y +
))
dy y 1 dy+
dy yy+ 1
dy y
The integral
The integral of
of this
this equation
equation isis the
the Law
Law of of the
the wallwall
Z
Z y 1
y
+
+

U++
= f (y++
), where, f (y++
) = 1 (y0 0
)dy0 0
U = fww(y ), where, fww(y ) = y0 0 11(y )dy
00 y
The critical
The critical point
point isis not
not the
the integral
integral but
but thethe
+ +
hypothesis that
hypothesis that UU+ depends
depends only
only on
on yy+ ..
Turbulence:
Turbulence: lecture
lecture
Turbulence: lecture
Wall-boundedflows:
Wall-bounded flows:Mean
Meanvelocity
velocityprofiles
profiles
Wall-bounded flows: Mean velocity profiles
Viscous
Viscous
Viscous
sub-layer
sub-layer
sub-layer
No-slip condition
No-slip gives fw (0) = 0, while the viscous stress law gives f0 w0 (0) = 1
No-slip condition
condition gives
gives ffww(0)
(0) =
= 0,
0, while
while the
the viscous
viscous stress
stress law gives ffww0 (0)
law gives (0) =
= 11

+
Taylor’s
Taylor’s expansion
expansion for
for small
small yy+
+
Taylor’s expansion for small y
+ + + 2
f w
ffww(y(y
++ ) = y
+
+ + O(y +
+ 2 )
2
(y )) == yy + + O(y
O(y ))

+
• For
•• For y
+
+ < 5, the departure is minimal.
For yy < < 5,
5, the
the departure
departure is
is minimal.
minimal.
+
• For
•• For y
+
+ > 10 the deviation is 25%
For yy > > 10
10 the
the deviation
deviation is
is 25%
25%
• “Bu↵er”
•• “Bu↵er” layer beyond viscous sublayer
“Bu↵er” layer
layer beyond
beyond viscous
viscous sublayer
sublayer

Wednesday, 23 October 13
Turbulence: lecture
Turbulence: lecture
Wall-bounded
Wall-boundedflows:
flows:Mean
Meanvelocity
velocityprofiles
profiles
Di↵erent layers in wall flows,
The
Thelog
loglaw
law
• Inner layer y/ < 0.15
++
The
Theouter
outerpart
partofofthis
+thislayer
layercorresponds
correspondstotolarge largeyy
• Viscous sublayer y < 5
Analogous
AnalogoustotoKolmogorov’s
Kolmogorov’shypothesis,
hypothesis,ininthis
thisouter
outerpart
partofofthe
theinner
innerlayer,
layer,
p
Suppose
• the
Suppose theviscosity
Outer parthas
viscosity of little
has layer y + > 3 Re⌧ - beginning of log r
littlee↵ect
inner e↵ect
++ p 11
Therefore,
Therefore,
• Bu↵er layer 5 < y < 3 Re⌧ +1
1 (y(y ) )!! constant
constant = =

++ p
dU
dU
• Log layer== 1 1
3 Re⌧ for <for ++
yyy< 0.15Re
+>>
>> 11 ⌧ (y/
and
and y/
y/ =<< 0.15)
<< 1 1
dy + +
dy + y y +
• Outer layer y + > 0.15Re⌧ (also called wake region in BLs
++ 11 ++
integrate,
integrate, U U = =  ln
lnyy ++BB

- -von
vonKarman
Karmanconstant
constant==0.39 0.39(latest
(latest“universal”
“universal”number).
number)
BB==4.34.3 Marusic
Marusicetetal.
al.(2013),
(2013), J.J.Fluid
FluidMech.
Mech.
Turbulence: lecture
Turbulence: lecture
Wall-bounded
Wall-bounded flows:
flows:Mean
Mean velocity
velocity profiles
profiles
The
The log
log law
law
yy++

U+

yy
y ++=
y =100
100 ==0.15
0.15

yy
Turbulence: lecture
Wall-bounded flows: Mean velocity profiles
Di↵erent layers in wall flows,
• Inner layer y/ < 0.15
+
• Viscous sublayer y < 5
+
p
• Outer layer y > 3 Re⌧ - beginning of log region
+
p
• Bu↵er layer 5 < y < 3 Re⌧
p +
• Log layer 3 Re⌧ < y < 0.15Re⌧ (y/ = 0.15)

These numbers are constantly changing


Turbulence: lecture
Wall-bounded flows: Mean velocity profiles (defect form)
+
In the outer layer, ( , y/ ) is independent of ⌫.
This implies, tends to a function of y/
(y + , y/ ) ! o (y/ )

Substituting this in to the law of the wall,


Z 1
Uo U 1 0 0
= FD (y/ ) where, FD = 0 o (y )dy
u⌧ y/ y
At sufficiently high Reynolds number, p
+
There is an overlap between inner (y/ < 0.15) and outer layers (y > 3 Re⌧ ).

y dU +
= 1 (y )= o (y/ ) for ⌫ << y <<
u⌧ dy
Turbulence: lecture
Wall-bounded flows: Mean velocity profiles (defect form)

y dU +
= 1 (y )= o (y/ ) for ⌫ << y <<
u⌧ dy

This argument was put forward by Millikan (1938)


and is an alternative derivation of the log law
More importantly, it establishes the defect law as well

Uo U y
= ln + B1
u⌧
1
+ ln 1
= ln(Re δ ) + B + B1 (8.1)
uτ Turbulence:
κ uτ κ lecture
so that, since Reδ is a large quantity, uτ/U1 must be a small one and thus the surface
Wall-bounded flows: Mean velocity profiles
friction falls only slowly with distance downstream. uτ/U1 is typically 0.03-0.05.
Law of the wake
Note that the outer layer profile can be expressed in terms of the law of the wake – i.e.

⇣ ⇡y ⌘
Π
U = f (y )+ κ
+ + y w(y/ ) = 2sin2 (8.2)
w
δ 2
where w is the wake function and is often assumed to be universal; it is defined to
satisfy the conditions w(0)=0 and w(1)=2. Π is called the wake strength parameter. In
terms of (8.1), B1=2Π/κ and is approximately 2.75, but much lower in pipes and channels
In terms
(typically 0.2 of defect
– 0.7). Figureprofile,
8.1 showsBthe
1 = 2⇧/ layer profile plotted as u+ vs. ln(y+).
boundary
Strong for boundary layers, but, weak for pipes and channels
Turbulence: lecture
Skin friction and Reynolds number
Uo U y + 1 +
= ln integrate,
+ B1 U = ln y + B
u⌧ 
Adding these two together
✓ ◆ ✓ ◆
Uo 1 U⌧ 2⇧
= ln + B + B1 =
u⌧  ⌫ 
" ✓ ◆ 1# ✓ ◆
Uo 1 Uo 2⇧
= ln Reo + B + B1 =
u⌧  u⌧ 
For a given Reo , can be solved for Uo /u⌧
✓ ◆2
2⌧w u⌧
cf = = 2
⇢Uo2 Uo

Turbulence: lecture
= const., uτ ≡ U 1
Cf
= const.
dx 2
Effect of surface roughness
A rough surface changes the universal log-law region (see Section 5.2) to:

+ 1
u = ln(y + ) + C(h + , geometry)
κ
where h is the typical element height and the constant must now depend on the nature
factor
of for a fully
the surface. In rough
terms case
of thewith fixed
usual h/d would
smooth remain independent
wall log-law, of Reynolds
we can therefore write:
number (as for the fixed x/ks+lines
1 in fig.8.4)
+ and, as
+ in
+ boundary layers, the particular
u = ln(y ) + Co − ∆U (h , geometry) (8.4)
κ
value would depend on the size of the roughness.
where ∆U is the roughness function given by ∆U + = C(h+ , geom.) − Co and Co=B in the

previous
Note alsonotation. The
that, often, figure
the rough(from Raupach
surface et al) displaces
effectively contains experimental data showing
the flow vertically a little
(or a∆lot!),
how U varies with
so one h. normally used (y-d), where d is a ‘zero-plane displacement’,
would
rather than y for the vertical distance.
Assuming that, for large enough h+, viscosity becomes irrelevant – i.e. the flow has
‘Reynolds number
Note, finally, that ifsimilarity’ – ν/uare
the elements τ is large
an inappropriate length
enough, there may scale
be noand so we
log-law writeat
region the
all
Turbulence: lecture
previous notation. The figure (from Raupach et al) contains experimental data showing
how ∆U varies with h.
Effect of surface roughness
Assuming that, for large enough h+, viscosity becomes irrelevant – i.e. the flow has
‘Reynolds number similarity’ – ν/uτ is an inappropriate length scale and so we write the
law-of-the-wall as:
+ 1 y
u = ln + c(h + ,geom.) .
κ h
We can re-express this in terms of yo, a roughness length, so that:

+ 1 y
u = ln . (8.5)
κ yo

(This is the way that meteorologists often express the log law.) Note that yo contains all
the information about the surface morphology – changes in nature, shape, spacing, etc.,
of the roughness elements will change yo.
Turbulence: lecture
Effect of surface roughness
Using (8.4) and (8.5) it can be shown that
∆U+ ≈ ln(h+)

igure 8.3
Problem is same h + can have different drag
Variation of the roughness function, ∆U+, with h+=huτ/ν.
Turbulence: lecture
values for the von Karman constant, κ, and B (i.e. Co)). So yo remains defined for a
smooth surface but is flow-dependent, unlike in the fully rough limit.
Effect of surface roughness
Typically, the wall is aerodynamically smooth if 0 < h+<5
and fully rough flow for h+>70 (with yo/h=30)
according to classical sand-grain experiments in pipe flows. There is therefore an
‘admissable roughness’ height; roughness of a smaller size (than h+=5, say) will not
affect the flow.
U1 hadm 7
Schlichting expresses this as: hadm<100ν/U1 or, more precisely, < .
ν Cf '
Turbulence: lecture
Turbulence: lecture
Effect of
Effect of surface
surface roughness
roughness
CD varies with L/✏ for the same ReL

✏✏ (or
(or kks):
s
): Equivalent
Equivalent sandgrain
sandgrain roughness
roughness

We don’t
We don’t know
know its
its value
value from
from geometry!
geometry!

So, we
So, we develop
develop correlations
correlations
between surface
between surface parameters
parameters
and equivalant
and equivalant sandgrain
sandgrain
roughness that
roughness that will
will allow
allow us
us to
to
predict drag
predict drag

Wednesday, 23 October 13
Turbulence: lecture
where the energy production is a maximum must be in a region where the shear stress
−u' v' is increasing, so must be nearer the wall than the log-law region, because −u' v' is
Boundary layers: Effects of pressure gradient
constant in the latter but zero at the wall.

dθ 1 du e Cf Momentum
8.2
+ (2θ + δ*) =
Effects of pressure gradient. integral
dx ue dx 2 equation
The momentum integral equation can be rewritten:
dθ uτ 2 2θ dP1 dU1
= 2 1 + Ψ 1+ * = U1
dx U1 δ dx dx

δ * dP1 δ *U1 dU1


where the quantity Ψ = =− 2 is a parameter which measures the
τ w dx uτ dx
importance of the pressure gradient. If it’s positive (pressure increasing, velocity
decreasing) the layer grows more quickly so that boundary layer assumptions become
less and less valid and much of the above does not apply. Note that for constant
pressure gradient, increasing δ∗ and decreasing uτ (because of falling velocities) will
cases. Turbulence: lecture
profile has zero slope at the wall and the surface stress is zero.

Boundary layers: Effects of pressure gradient


Note what happens to the shape factor, H=θ/δ. For zero pressure-gradient this is around
1.3 – but depending on Reynolds number of course. Favourable pressure gradient
causes the velocity profile to steepen and thus decreases H and increases Cf. and vice
versa for adverse gradients.

Note also what happens to the shear stress profile – from the momentum equation, at
the wall, we must have:
∂τ dP1
=−
∂y dx
so the shear stress increases with y in an adverse pressure gradient, as shown in Figure
8.2 As the pressure gradient (Ψ) increases from zero the velocity profile becomes less
full and the shear stress profile develops a peak, with a lower value at the wall. This
value gradually decreases with increasing until separation is reached, when the velocity
profile has zero slope at the wall and the surface stress is zero.
PhD opportunities

• Back up slides
f˜ (η, x) ≡ f (η) ) then Q(x,y) is self-similar. Q(x,y) can be expressed as functions of single
Turbulence: lecture
independent variables – Qo(x), δ(x) and f(η).
Self-similarity
Of course, the scales must be chosen appropriately. And, usually, self-similarity only
occurs over limited ranges of x. Also, if Q(x,y) is governed by a pde then Qo(x), δ(x) and
f(η) are all governed by odes.

So, in general, we look for solutions of our governing equations in which:

U = U1 + uo f(η), (η=y/δ)

uv = qo 2 g12 (η) , u 2 = qo 2 g1 (η ) and v 2 = qo 2 g2 (η) .

U1 is a translational velocity (almost always the free stream) and uo and δ are
characteristic velocity and length scales.

∂U uo
We may therefore write: = f ' (η ) (dashes denote differentiation w.r.t. η.
∂y δ
f˜ (η, x) ≡ f (η) ) then Q(x,y) is self-similar. Q(x,y) can be expressed as functions of single
Turbulence: lecture
independent variables – Qo(x), δ(x) and f(η).
Self-similarity
Of course, the scales must be chosen appropriately. And, usually, self-similarity only
occurs over limited ranges of x. Also, if Q(x,y) is governed by a pde then Qo(x), δ(x) and
f(η) are all governed by odes.

So, in general, we look for solutions of our governing equations in which:

U = U1 + uo f(η), (η=y/δ)

uv = qo 2 g12 (η) , u 2 = qo 2 g1 (η ) and v 2 = qo 2 g2 (η) .

U1 is a translational velocity (almost always the free stream) and uo and δ are
characteristic velocity and length scales.

∂U uo
We may therefore write: = f ' (η ) (dashes denote differentiation w.r.t. η.
∂y δ
U isisaatranslational
translational
U11characteristic velocity
velocity
velocity (almost
(almost
and length Turbulence:
always
always the
scales. lecture
Turbulence: lecture
thefree
freestream)
stream)and andδδ are
anduuooand are
characteristicvelocity
characteristic velocityand
andlength
lengthscales.
scales.
Self-similarity
Self-similarity
∂U uo
We may therefore write:∂∂UU uu = f ' (η ) (dashes denote differentiation w
Wemay
We maytherefore
thereforewrite:
write: ==∂yoo ff' '((η δη)) (dashesdenote
(dashes denotedifferentiation w.r.t.ηη..
differentiationw.r.t.
∂∂yy δδ
∂U dU1 duo uo dδ ∂ 2
U uo
∂∂UU dU = 11 du
dU du+oo f (ηuuo)o d−dδδ ηf ' (η) ∂∂ UU uuoo 2 = 2 f '' (η ) .
and22
∂==x ++
dx ((ηη))−−
ffdx δηηffdx
' '((ηη)) and
and = ∂22yff''''((ηηδ))..
22 =
∂∂xx dx
dx dx
dx δδ dx
dx ∂∂yy δδ

The
The
The
continuity
continuity
continuity
equation
equation
equation
yields
yields
yields
yy ηη ηη
∂∂UU yy
dU
dU
∂δ∂UU 11ηη −− dU du
du ηη
ddδ
δ ηη
dfd'δ'δddη
VV == −− dy
dy == −
VV == −−− δ dy ==−−δδ
dy δδ
dU 11 oo fd
ηη −− δδ
fd η
du
η
duo+
o+ u η
η
fdooηη ++ uuoo
u
fd f ηη.η.ff''ddηη..
∂∂xx ∂∂xxdx
dx dx
dx dx
dx dx
dx dx
dx dx
dx
00 0 0 0
0 0 0 00 0 0

dd dd
((ηηff)η
ηηff' '== Since
Since ff' '=
)η−− f=f this((ηηreduces
this f))−− ff this
freduces to:
this reduces
to:reduces to:
to:
ddηη ddηη

yy ηη
yy ∂∂UU dU
dU11 d(δδuu ))
ηηd( ddδδ
∂∂UUVV == −− dU dU d(δδηuηuoo−)−)
dy11==−−δδ d(
dy oo
fdηηddδ
fd +δ+ uuoo ηηff..
VV == −− dy ==−−δδ∂∂xx ηη −− dx
dy dx fd ηη ++ uuoo ηηff ..dx
fddx
dx dx
∂∂xx 0 0 dx
dx dx
dx 0 0 dx
dx
00 00
0 0 0
0
Turbulence: lecture
Self-similarity
Substituting these into the boundary layer equation (6.6a) yields:
bstituting these into the boundary layer equation (6.6a) yields:
Substituting these in the momentum equation,
dU1 duo uo dδ duo 2 uo d(δuo )
uo
dU (f − ηf ' ) + U1du f −u dδ .ηf ' + uo du f − u d(δu.f)' fdy
uo 1 dx
(f − ηf ' ) + U1 dx
o
f− δ
o dx
.ηf ' + uo dx f − δ o dx o .f ' fdy (6.9)
o 2
dx dx 2 δ dx dx δ dx
dδ qo (6
+ qo (g1 ' − g2 ' ) + 2 g12 ' = 0
2

2 dδ dx qo δ
+ qo (g1 ' − g2 ' ) + g12 ' = 0
dx δ
To be consistent with self-preservation, the coefficients must be zero or proportional to

o beone another, since


consistent withthe equation must bethe
self-preservation, a function of η only.
coefficients mustSince we know
be zero that, for
or proportiona
example, since
e another, the shear
the stress termmust
equation is not be
zero, we can multiply
a function all coefficients
of η only. by δ/qothat,
Since we know so fo 2

that the following parameters must all be constants:


ample, the shear stress term is not zero, we can multiply all coefficients by δ/qo2 so
at the following parameters must all be constants:
δ dU1 δ duo uoU1 dδ δ duo uo d(δuo )
2 uo , 2 U1 , 2 , 2 uo , 2 . (6.10)
qo dx qo dx qo dx qo dx qo dx
δ dU1 δ duo uoU1 dδ δ duo uo d(δuo )
2 uo , 2 U1 , 2 , 2 uo , 2 . (6.10)

Vous aimerez peut-être aussi