Turbulence Modelling

Dr. Ahmed Ismail

 Who is the lecturer?
Dr. Ahmed Ismail (Lecturer of Fluid Dynamics)
• Queen Mary University of London, 2017-PRESENT
• University of Seville, 2012-2017.

Principle Investigator of electrospray Laboratories

• Research activities.

➢ Electrohydrodynamics. ➢ Multiphase flows in micro-scale.

Why do we need turbulence modelling?
The Road to model a turbulence Problem

Week 10
• Phenomenology of Turbulence
• Statistical Description of Turbulent flow
Week 11
• Reynolds Averaged Navier-Stokes Equations (RANS)
• Direct Numerical Simulation (DNS)
• Large Eddy Simulation (LES)
Problem to Solve!
What are you going to learn in this session?

Phenomenology of Turbulence

• Identify key terminologies in turbulence

• Describe how energy cascades from large eddies to small eddies

Basic Concepts

Viscosity

Descriptive (qualitative) definition: shear viscosity is a measure of the resistance to

sliding of the fluid elements

𝜏
Mathematical definition 𝜇=
of (dynamic) shear viscosity: 𝑑𝑢 strain
𝑑𝑦
Basic Concepts

Dynamic vs. kinematic viscosity

What is the difference?

Dynamic shear viscosity: 𝜇 (SI units: Pa)

𝜇
Kinematic shear viscosity: 𝜈= (SI units: m2/s) because of the unit since here is
𝜌 velocity which is needed for
momentum

Note: dynamic viscosity is usually considered if interested in forces on other

structures, while kinematic viscosity is used if the transfer of momentum within
the fluid itself is what we are looking for.
Basic Concepts

Reynolds number

Descriptive definition: It is a dimensionless quantity in fluid mechanics used to help

predict flow patterns in different fluid flow situations.

Mathematical definition 𝑖𝑛𝑒𝑟𝑡𝑖𝑎𝑙 𝑓𝑜𝑟𝑐𝑒 𝜌𝑢𝑑 𝑢𝑑

𝑅𝑒 = = =
of Reynolds: 𝑣𝑖𝑠𝑐𝑜𝑢𝑠 𝑓𝑜𝑟𝑐𝑒 𝜇 𝜈
Understanding Turbulence activity

How would you explain,

1- the difference between laminar and turbulent flows.

2- the transfer of energy in a turbulent flow.

to someone new to the topic?

➢ You will be split into two groups 1 & 2.

➢ Each group will discuss (take notes) one of the two points (the corresponding
number).
➢ You have 10 minutes to complete the activity in your groups, before coming back
to explain your findings to the others.
Phenomenology of Turbulence

Laminar :

➢ Smooth, regular, and orderly motion of fluid elements

➢ Perturbation is damped down by the viscous force (low Reynolds
number <Recr=2300)
Streamlines Velocity
profile
Turbulent:

➢ Perturbation is amplified when inertia is dominant (high Reynolds

number > Recr=2300)
➢ chaotic, strongly unsteady streamlines, and randomness of velocity
w. r. t. time and space.
➢ Enhanced mixing and diffusion.
In both cases:
The total shear stress is zero at the axis (by symmetry), and varies linearly with
radius (because the cross-channel shear stress gradient has to balance the constant
streamwise pressure gradient).
Phenomenology of Turbulence

“Turbulent” and “unsteady Laminar” are different concepts

Unsteady laminar Turbulent

Unsteady laminar flow has predictable time Turbulence is wildly chaotic, and is characterized by large
dependence (e.g. time-periodic flow) velocity fluctuations
Phenomenology of Turbulence

TURBULENCE ENERGY CASCADE

Largest eddies have large mechanical energy,

and are affected by inertial (inviscid)
Energy flux mechanisms

Smallest eddies have little mechanical

energy, and are strongly affected by viscosity

Energy cascade: eddies at a given level “break down” into

Example of an eddy
smaller eddies, providing them with mechanical energy
Phenomenology of Turbulence
What are you going to learn in the first part of the session?

Phenomenology of Turbulence

• Estimate and compare the different length scales in turbulence

• Analyze turbulence energy spectrum and dynamic of eddies

Length scales activity

What is the difference?

Integral length scale vs Kolmogorov length scale

Phenomenology of Turbulence
TURBULENCE: LENGTH SCALES
“Small” scales
Large scales
• Approximately
• Have a scale comparable homogeneous and
to system size isotropic

• Their dynamics
depends only
indirectly on the large
scales

• The largest scales of the turbulence are also called INTEGRAL

scales.
• The smallest scale of the turbulence is also called KOLMOGOROV
SCALE.

• The INERTIAL RANGE includes all the scale affected by fluid’s inertia
(i.e. their dynamics is governed by the inertia term of the NS
equations): from the integral scale down, excluding the smallest
(dissipative) scales.
Phenomenology of Turbulence

THE KOLMOGOROV SCALE

• The breakdown of eddies in a turbulent

Dissipative eddies
flow cannot proceed indefinitely (i.e. the
smallest eddies cannot have molecular
size!)

• The breakdown process proceeds until

the kinetic energy of the eddies is
dissipated into heat by viscous effects->
the properties of smallest eddies must
depend on the fluid viscosity
• The smallest eddies must dissipate energy at
Simulation of dissipative eddies
the same rate at which turbulence is produced
(at the large scales)-> the properties of the
smallest eddies must depend on the average
rate of turbulent dissipation (=rate of
Smallest eddies are strongly affected by viscosity production)
(they are small with small velocity, their Reynolds
number is therefore small)
Phenomenology of Turbulence
ESTIMATE OF THE KOLMOGOROV SCALE

Rate of dissipation  at the smallest eddy scale, Largest Smallest

eddy eddy
 2 Length  
 ~ 2 ------ (1)

 Velocity u0
Time t t
Kinematic viscosity

 
At the smallest eddy scale, Re ~ 1 ~1 ~ ------ (2)
 
1ൗ
3 4
𝜈
From 1 & 2, 𝜂~ Kolmogrov Length Scale
𝜀
Phenomenology of Turbulence
ESTIMATE OF THE KOLMOGOROV SCALE

Rate of Extraction  is equal to the rate of dissipation, Example of a pump creating

a turbulent flow,
 ~
Rate of Extraction of K.E from mean flow per unit mass at large eddy scale, 
u02 u03
~ ~ 1 Pump Liquid
t  3 power mass
Turnover time Re 
1ൗ
𝜈 3 4  4
 3
 −3 / 4
𝜂~ 4
~ 3 3 ~ Re
𝜀  u
0 

Energy spectrum activity

The kinetic energy transfers through a cascade of eddies

but

How is the kinetic energy distributed across the different length scales?
Phenomenology of Turbulence
THE TURBULENCE ENERGY SPECTRUM
• In turbulence theory, it is customary to consider the so-called power spectrum,
which is the distribution of kinetic energy per mass across the various length
scales. For this, we need to define a wavenumber.

Large eddies Small eddies

Wavelength L
velocity

velocity
Wavelength L

x x
The wavenumber k of a wave is k=2π/L, Large eddies have long wavelength and
22
where L is the wavelength small wavenumber

Small eddies have large wavenumber

 Phenomenology of Turbulence

Large scales (small

wavenumbers) have large
• The energy spectrum is a log-log
energy
plot giving the turbulent kinetic
energy content corresponding to
the wavenumber of each turbulent Small scales (large
wavenumbers) have small
scale.
energy

• The energy spectrum is based on

experimental data.

• The curve, scaled appropriately, is Zero energy beyond a cut-

Energy
universal: experiments in different off wavenumber
conditions give the same result (corresponding to
Kolmogorov scale)

Wavenumber
Phenomenology of Turbulence
Result: for sufficiently high Reynolds numbers, the energy spectrum of the
intermediate turbulent scales (inertial range) decays like k-5/3. This result, which can
be obtained with dimensional analysis, is confirmed by experiments.

For intermediate scales, viscosity is not important

slope – 5/3

dissipation rate

Apply dimensional analysis to get:

For a range of values of k corresponding to

the inertial scales, the plot is a straight line!
The “-5/3” scaling of turbulence
Phenomenology of Turbulence
Dynamic of Eddies
Phenomenology of Turbulence
Dynamic of Eddies

Momentum eqn.,

 du    2
  + (u .)u  = −p +  u
 dt 

Vorticity is the curl of velocity vector,

 
 = u
Phenomenology of Turbulence
Dynamic of Eddies

Take the curl of both sides in the momentum eqn.,

d    
  (  u ) +   (u .)u  = −  (p) +  (  u ) 2

 dt 
 

0 (curl of scalar)

      

  (u .u ) − u  (  u ) = − ( .)u − (u .) 
0 (curl of scalar)
 Phenomenology of Turbulence
Dynamic of Eddies


 du    2
  + (u .)u  = −p +  u (Momentum equation)
 dt 


   D  1   
+ (u . ) =  = ( .)u +  
2
(Governing equation for vorticity)
t Dt 
Vorticity change Viscous
in space and dissipation of
time Vorticity
Phenomenology of Turbulence
Dynamic of Eddies
The rate of change of the angular momentum of vortex element produces torque due to viscous forces,

D
I  = torque (due to viscous forces)
Dt
 represent the angular velocity

D  DI Tviscous I represent the moment of inertia

=− + I = f (r 2 )
Dt I Dt I
For large eddies,
Vortex stretching Radial length
scale

r1 r2 Vortex stretching
It is the lengthening of vortices in 3D fluid flow, associated
with a corresponding increase of the component of vorticity
in the stretching direction, due to the conservation of angular
momentum.
Phenomenology of Turbulence
Dynamic of Eddies
Comparing the vorticity equation with the angular momentum equation,

D  DI Tviscous
= +
Dt I Dt I 1
= 
2

D  1   
 = ( .)u +  2 1  
By analogy ( .)u is the term responsible of the vortex
Dt  stretching


Vorticity change Vortex Viscous

in space and Stretching dissipation of
time Vorticity
Phenomenology of Turbulence
Useful books:

Authors: Tennekes & Lumley

31
Statistical description of turbulence

Velocity

Time
What are you going to learn in the second part of the session?

Statistical Description of Turbulent flow

• Analyze and represent variables decomposition in a turbulent flow

• Calculate statistical quantities in a turbulent flow
• Identify Turbulent kinetic energy and relation to the variance
Variables decomposition in a turbulent flow

Fluctuation
amplitude
For a statistical treatment of turbulence, it is useful to
decompose the velocity vector field into its time-average
mean + its fluctuating component: mean
Stationary

Velocity
Turbulence

Time

Flow over a
cylinder in
3D

34
Variables representation in a turbulent flow

What is the difference?

Time Average vs Space average vs Ensemble average

 Variables decomposition in a turbulent flow

Time average:
System
time Scale
The average is usually interpreted as a time average at a given location:

Fluctuation

Velocity
Non-stationary time Scale
Turbulence
1 t 0 +T
u( x0 ) =
T t0
u( x0 , t )dt
Time

NOTE: The averaging interval 𝑇 is large in comparison to all the time-scales of the
fluctuating motions, but small in comparison to the time scale over which the mean
ഥ does make sense.
flow varies. With this interpretation, taking the time-derivative of 𝒖
36
 Variables representation in a turbulent flow

Space average:

The average could be also over a certain distance:

mean

Velocity
1 x0 + X
u(t0 ) =
X x0
u( x, t0 )dx

Ensemble average:
x
Averaging from a large number of experiments
conducted under identical conditions.

Velocity
x 37
Typical statistics

For N equally-spaced velocity data points u1, u2, u3, …, uN

• (Time-averaged) mean:

Over bar denotes time-averaging

• Fluctuation about the mean:

• Variance:

38
Typical statistics

The standard deviation is the square root of the variance:

standard deviation

Probability

Velocity
Mean 39
Typical statistics

The standard deviation is also called root mean square, or r.m.s.

To remember the definition: you need to take the root (R) or the mean (M) of the
square (S) of the fluctuations

40
Typical statistics

For any turbulent quantities, let us say for example, u and v :

u = u + u u =u

u+v =u +v uv = u v + u v

u u
=  udx =  u dx
xi xi

41
Statistics exercise

In the table below you will find values of streamwise velocity

sampled by a sensor (e.g. by an anemometer). Calculate mean,
variance, and r.m.s. of the signal.
Time Streamwise velocity u
t=1 7
t=2 6
t=3 5

Mean streamwise velocity:

Velocity variance:

R.m.s. of velocity:
42
Statistics exercise

??

Property: the mean value of the fluctuating velocity field is identically zero

Proof:

Statistically speaking, “positive” fluctuations cancel out with “negative” fluctuations

43
Typical statistics

Isotropic Turbulence: Directional independence of turbulent statistics

v
u
u 2 = v2 = w2 w

Homogenous Turbulence: Position independence of turbulent statistics

44
Turbulent kinetic energy

We are often interested in characterizing the “strength” of the turbulence, i.e. how big the
velocity fluctuations are. The quantity to use in this case is the turbulent kinetic energy.

Definition: the turbulent kinetic energy is given by the relation

(to remember it, think about the usual definition of the kinetic energy of a point mass 1/2mV2)

For isotropic turbulence, all the components of velocity are identical. Therefore:
The variance

For isotropic turbulence, I just need to measure one single component of velocity.
45