“Essentially, all models are wrong,

but some are useful”

G. E. P. Box

George Edward Pelham Box

18 October 1919 – 28 March 2013. Statistician, who
worked in the areas of quality control, time-series
analysis, design of experiments, and Bayesian inference.
He has been called “one of the great statistical minds of
the 20th century”.

1
Leonardo da Vinci pioneered the flow visualization genre about 500
years ago. The illustrations to the left (Studies of water passing
obstacles and falling c. 1508-1509) represents perhaps the world's
first use of visualization as a scientific tool to study a turbulent flow.

The following da Vinci’s observation is close to the Reynold’s

decomposition.

“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 motions, one part of
which is due to the principal current, the other to the
random and reverse motion.”

The following da Vinci’s description is maybe the earliest reference

to the importance of vortices in fluid motion.

“So moving water strives to maintain the course pursuant to

the power which occasions it and, if it finds an obstacle in its
path, completes the span of the course it has commenced by
a circular and revolving movement.”

The following da Vinci’s observation is an analogy to the energy

cascade and coherent structures.

“...the smallest eddies are almost numberless, and large

things are rotated only by large eddies and not by small ones
and small things are turned by small eddies and large”

2
 The turbulent world around us

What is turbulence?
• Due to the extremely complex nature of turbulence and its incomplete understanding, there is
not a single accepted definition of turbulence.
• Leonardo da Vinci was so intrigued by turbulence that he depicted it in many of his
sketches (see previous slide). While observing the flow of water, he gave one of the very
first definitions of turbulence (if not the first one),

“...the smallest eddies are almost numberless, and large

things are rotated only by large eddies and not by small
ones and small things are turned by small eddies and large”

• Richardson [1] in 1922 stated that,

“Big whorls have little whorls,

which feed on their velocity;
And little whorls have lesser whorls,
And so on to viscosity”

3
[1] L. F. Richardson. “Weather Prediction by Numerical Process”. Cambridge University Press, 1922.
 The turbulent world around us

What is turbulence?
• Due to the extremely complex nature of turbulence and its incomplete understanding, there is
not a single accepted definition of turbulence.
• T. von Karman [1] who is known for his studies about Fluid Dynamics, quotes G. I. Taylor
with the following definition of turbulence in 1937,

“Turbulence is an irregular motion which in general makes

its appearance in fluids, gaseous or liquid, when they flow
past solid surfaces or even when neighboring streams
same fluid past or over one another.”

• J.O. Heinz [2] offers yet another definition for turbulence in 1959,

“Turbulent fluid motion is an irregular condition of the flow in which

quantities show a random variation with time and space coordinates,
so that statistically distinct average values can be discerned.”

[1] T. Von Karman. “Some remarks on the statistical theory of turbulence”. Proc. 5 th Int. Congr. Appl. Mech, Cambridge, MA, 347, 1938.
4
[2] J. O. Hinze. “Turbulence”. McGraw-Hill, New York, 1959.
 The turbulent world around us

What is turbulence?
• Due to the extremely complex nature of turbulence and its incomplete understanding, there is
not a single accepted definition of turbulence.
• A more modern and highly specific definition of turbulence is given by G. T. Chapman and
M. Tobak [1],

“Turbulence is any chaotic solution to the 3D Navier–Stokes equations that is sensitive to initial data and
which occurs as a result of successive instabilities of laminar flows as a bifurcation parameter is increased
through a succession of values.”

• S. Rodriguez [2], gives an even more modern definition linked to the use of approximations
to deliver solutions,

“Turbulent flows is the dynamic superposition of an extremely large number of eddies with random
(irregular) but continuous spectrum of sizes and velocities that are interspersed with small, discrete pockets
of laminar flow (as a result of the Kolmogorov eddies that decayed, as well as in the viscous laminar
sublayer and in the intermittent boundary). In this sense, turbulent flows are intractable in its fullest
manifestation; this is where good, engineering common sense and approximations can deliver reasonable
solutions, albeit approximate.”

[1] G. T. Chapman and M. Tobak. “Observations, Theoretical Ideas, and Modeling of Turbulent Flows — Past, Present and Future, in Theoretical Approaches to Turbulence”. Dwoyeret al.(eds),
Springer-Verlag, New York, pp. 19–49, 1985.
5
[2] S. Rodriguez. “Applied Computational Fluid Dynamics and Turbulence modeling”. Springer, 2019.
 The turbulent world around us

What is turbulence?
• Due to its complexity, a definition does not work properly for turbulence,
instead of it, it’s better to explain its characteristics.
• Tennekes and Lumley [1] in their book called “A First Course in Turbulence”,
list the characteristics of turbulence:
• Irregularity
• Diffusivity
• Dissipation
• Large Reynolds numbers
• Three-Dimensional Vorticity fluctuations
• Continuum
• Feature of a flow, not fluid

6
[1] H. Tennekes and J. L. Lumley. “A first course in turbulence”. MIT Press, 1972.
 The turbulent world around us

Turbulent flows have the following characteristics

• One characteristic of turbulent flows is their irregularity (or randomness). A fully deterministic
approach to characterize turbulent flows is very difficult. Turbulent flows are usually described
statically. Turbulent flows are always chaotic. But not all chaotic flows are turbulent. Magma
flowing can be chaotic but not necessarily turbulent.
• The diffusivity of turbulence cause rapid mixing and increased rates of momentum, heat, and
mass transfer. A flow that looks random but does not exhibit the spreading of velocity
fluctuations through the surrounding fluid is not turbulent.
• Turbulent flows are dissipative. Kinetic energy gets converted into heat due to viscous shear
stresses. Turbulent flows die out quickly when no energy is supplied.
• Turbulent flows always occur at high Reynolds numbers. They are caused by a complex
interaction between the viscous forces and convection.
• Turbulent flows are rotational, that is, they have non-zero vorticity. Mechanisms such as the
stretching of three-dimensional vortices play a key role in turbulence.
• Turbulence is a continuum phenomenon. Even the smallest eddies are significantly larger than
the molecular scales.
• Turbulence is a feature of fluid flow and is not a property of the flow. A liquid or a gas at high
Reynolds number will exhibit the same dynamics.

7
 The turbulent world around us

What is turbulence?
• For the purpose of this training, let us state the
following:
• Turbulence is an unsteady, aperiodic motion in
which all three velocity components fluctuate in
space and time.
• Every transported quantity shows similar
fluctuations (pressure, temperature, species,
concentration, and so on)
• Turbulent flows contains a wide range of eddy
sizes (scales):
• Large eddies derives their energy from the
mean flow. The size and velocity of large
eddies are on the order of the mean flow.
• Large eddies are unstable and they break-
up into smaller eddies.
• The smallest eddies convert kinetic energy
into thermal energy via viscous dissipation.
• The behavior of small eddies is more
universal in nature. 8
 The turbulent world around us

Buoyant plume of smoke rising from a stick of incense

Photo credit: https://www.flickr.com/photos/jlhopgood/
This work is licensed under a Creative Commons License (CC BY-NC-ND 2.0) 9
 The turbulent world around us

Von Karman vortices created when prevailing winds sweeping Von Karman Vortex Streets in the northern Pacific Photographed from
east across the northern Pacific Ocean encountered Alaska's the International Space Station
Aleutian Islands Photo credit: NASA
Photo credit: USGS EROS Data Center Satellite Systems Branch. Copyright on the images is held by the contributors. Apart from Fair Use,
Copyright on the images is held by the contributors. Apart from Fair permission must be sought for any other purpose.
Use, permission must be sought for any other purpose.

10
 The turbulent world around us

Tugboat riding on the turbulent wake of a ship Trailing vortices

Photo credit: https://www.flickr.com/photos/oneeighteen/ Photo credit: Steve Morris. AirTeamImages.
This work is licensed under a Creative Commons License (CC BY-NC 2.0) Copyright on the images is held by the contributors. Apart from Fair Use,
permission must be sought for any other purpose.

Flow visualization over a spinning spheroid Flow around an airfoil with a leading-edge slat
Photo credit: Y. Kohama. Photo credit: S. Makiya et al.
Copyright on the images is held by the contributors. Apart from Fair Use, Copyright on the images is held by the contributors. Apart from Fair Use,
permission must be sought for any other purpose. permission must be sought for any other purpose. 11
 The turbulent world around us

Wind Tunnel Test of New Tennis Ball Wake turbulence behind individual wind turbines
Photo credit: NASA Photo credit: NREL's wind energy research group.
http://tennisclub.gsfc.nasa.gov/tennis.windtunnelballs.html Copyright on the images is held by the contributors. Apart from Fair Use,
Copyright on the images is held by the contributors. Apart from Fair permission must be sought for any other purpose.
Use, permission must be sought for any other purpose.

12
 The turbulent world around us

Vortices on a 1/48-scale model of an F/A-18 aircraft inside a Water Tunnel

Photo credit: NASA Dryden Flow Visualization Facility. http://www.nasa.gov/centers/armstrong/multimedia/imagegallery/FVF
Copyright on the images is held by the contributors. Apart from Fair Use, permission must be sought for any other purpose. 13
 The turbulent world around us

Shadowgraph Images of Re-entry Vehicles

Photo credit: NASA on the Commons. https://www.flickr.com/photos/nasacommons/
Copyright on the images is held by the contributors. Apart from Fair Use, permission must be sought for any other purpose. 14
 The turbulent world around us
Astrophysical, plasma, planetary and quantum turbulence

M8: The Lagoon Nebula A frame from the simulation of the two colliding Antennae galaxies.
Photo credit: Steve Mazlin, Jack Harvey, Rick Gilbert, and Daniel Verschatse. Photo credit: F. Renaud / CEA-Sap.
Star Shadows Remote Observatory, PROMPT, CTIO Copyright on the images is held by the contributors. Apart from Fair Use,
Copyright on the images is held by the contributors. Apart from Fair Use, permission must be sought for any other purpose
permission must be sought for any other purpose

Jupiter photo taken by Juno’s cam.

Photo credit: NASA / JPL / SwRI / MSSS / David Marriott
Copyright on the images is held by the contributors. Apart
from Fair Use, permission must be sought for any other
purpose 15
 Turbulence, does it matter?
Blower simulation using sliding grids

No turbulence model used (laminar, no K-epsilon turbulence model

turbulence modeling, DNS, unresolved http://www.wolfdynamics.com/training/turbulence/image2.gif

DNS, name it as you want)

http://www.wolfdynamics.com/training/turbulence/image1.gif

16
 Turbulence, does it matter?
Vortex shedding past square cylinder

URANS (K-Omega SST with no wall functions) – LES (Smagorinsky) – Vortices visualized by Q-criterion
Vortices visualized by Q-criterion www.wolfdynamics.com/wiki/squarecil/les.gif
www.wolfdynamics.com/wiki/squarecil/urans2.gif

Laminar (no turbulence model) – Vortices DES (SpalartAllmarasDDES) – Vortices visualized by

visualized by Q-criterion Q-criterion
www.wolfdynamics.com/wiki/squarecil/laminar.gif www.wolfdynamics.com/wiki/squarecil/des.gif 17
 Turbulence, does it matter?
Vortex shedding past square cylinder

Turbulence model Drag coefficient Strouhal number Computing time (s)

Laminar 2.81 0.179 93489

LES 2.32 0.124 77465

DES 2.08 0.124 70754

SAS 2.40 0.164 57690

URANS (WF) 2.31 0.130 67830

URANS (No WF) 2.28 0.135 64492

RANS 2.20 - 28246 (10000 iter)

Experimental values 2.05-2.25 0.132 -

Note: all simulations were run using 4 cores.

References:
D. A. Lyn and W. Rodi. “The flapping shear layer formed by flow separation from the forward corner of a square cylinder”. J. Fluid Mech., 267, 353, 1994.
D. A. Lyn, S. Einav, W. Rodi and J. H. Park. “A laser-Doppler velocimetry study of ensemble-averaged characteristics of the turbulent near wake of a square
cylinder”. Report. SFB 210 /E/100. 18
 Turbulence, does it matter?
Transitional flow past square cylinder with rounded corners – Re = 54000

Velocity magnitude
www.wolfdynamics.com/wiki/turb/media1.mp4

Turbulence model Drag coefficient Lift coefficient

DNS 0.06295 0.07524

LES 0.1146 0.03269

SAS 0.1058 0.0258

URANS (No WF) 0.1107 0.00725

Transition K-KL-Omega 0.059 -0.0104

Transition K-Omega SST 0.0987 -0.0143

Experimental values 0.045 to 0.075 -0.011 to -0.015

19
 Turbulence, does it matter?
Separated flow around a NACA-4412 airfoil

Inflow

Turbulence model 1
Turbulence model 2
Turbulence model 3
Turbulence model 4
Turbulence model 5
Experimental results

Turbulence model 1
Turbulence model 2
Turbulence model 3
Turbulence model 4
Turbulence model 5
Experimental results

• CFD has been around since the late 1970s, and after all these years is not that easy to compute
the flow around 2D airfoils.
• In particular, predicting the maximum lift and stall characteristics is not trivial.

References:
F. Menter. “A New Generalized k-omega model. Putting flexibility into Turbulence models (GEKO)”, Ansys Germany
A. J. Wadcock. “Investigation of Low-Speed Turbulent Separated Flow Around Airfoils”, NASA Contractor Report 177450 20
 Turbulence, does it matter?
Turbulence is not a trivial problem

“Turbulence is the most important unresolved problem of classical physics”

Richard Feynman

“Turbulence was probably invented by the devil on the seventh

day of creation when the good lord was not looking”
Peter Bradshaw (1994)

“Turbulence is the graveyard of theories”

Hans W. Liepmann (1997)

21
 Introduction to turbulence modeling

Turbulence modeling in engineering

• Most natural and engineering flows are turbulent, hence the necessity of modeling
turbulence.
• The goal of turbulence modeling is to develop equations that predict the time
averaged velocity, pressure, temperature fields without calculating the complete
turbulent flow pattern as a function of time.
• Turbulence can be wall bounded or free shear. Depending of what you want to
simulate, you will need to choose an appropriate turbulence model.
• There is no universal turbulence model, hence you need to know the capabilities and
limitations of the turbulence models.
• Due to the multi-scale and unsteady nature of turbulence, modeling it is not an easy
task.
• Simulating turbulent flows in any general CFD solver (e.g., OpenFOAM®, SU2,
Fluent, CFX, Star-CCM+) requires selecting a turbulence model, providing initial
conditions and boundary conditions for the closure equations of the turbulent model,
selecting a near-wall modeling, and choosing runtime parameters and numerics.

22
 Introduction to turbulence modeling

Why turbulent flows are challenging?

• Unsteady aperiodic motion.
• All fluid properties and transported quantities exhibit random spatial and
temporal variations.
• They are intrinsically three-dimensional due to vortex stretching.
• Strong dependence from initial conditions.
• Contains a wide range of scales (eddies).
• Therefore, in order to accurately model/resolve turbulent flows, the
simulations must be three-dimensional, time-accurate, and with fine enough
meshes such that all spatial scales are properly captured.
• Additional physics that makes turbulence modeling even harder:
• Buoyancy, compressibility effects, heat transfer, multiphase flows,
transition to turbulence, surface finish, combustion, and so on.

23
 Introduction to turbulence modeling

Reynolds number and Rayleigh number

• It is well known that the Reynolds number characterizes if the flow is laminar or turbulent.
• So before doing a simulation or experiment, check if the flow is turbulent.
• The Reynolds number is defined as follows,

Convective effects

Viscous effects

• Where U is a characteristic velocity, e.g., free-stream velocity.

• And L is representative length scale, e.g., length, height, diameter, etc.

24
 Introduction to turbulence modeling

Reynolds number and Rayleigh number

• If you are dealing with natural convection, you can use the Rayleigh number, Grashof number,
and Prandtl number to characterize the flow.

Specific heat Thermal expansion coefficient

Buoyancy effects

Viscous effects Thermal conductivity

Momentum diffusivity

Thermal diffusivity

25
 Introduction to turbulence modeling

Reynolds number and Rayleigh number

• Turbulent flow occurs at large Reynolds number.
• For external flows,

Around slender/streamlined bodies (surfaces)

Around an obstacle (bluff bodies)

• For internal flows,

• Notice that other factors such as free-stream turbulence, surface conditions, blowing, suction,
roughness and other disturbances, may cause transition to turbulence at lower Reynolds
number.
• If you are dealing with natural convection and buoyancy, turbulent flows occurs when

26
 Introduction to turbulence modeling

What happens when we increase the Reynolds number?

Creeping flow (no s eparation)

S teady flow Re < 5
• Easy to simulate
A pair of s table vortices
• Steady
in the wake
S teady flow
5 < Re < 40 - 46
• Relative easy to
simulate.
Laminar vortex s treet
(Von Karman s treet)
Uns teady flow
40 - 46 < Re < 150 • It becomes more
challenging when
the boundary layer
Laminar boundary layer up to
the s eparation point, turbulent
150 < Re < 300 transition to
Trans ition to turbulence
wake
Uns teady flow 300 < Re < 3 x 10 5 turbulent
• Unsteady
Boundary layer trans ition to 5 6
turbulent 3 x 10 < Re < 3 x 10
Uns teady flow • Challenging to
simulate
Turbulent vortex s treet, but the
6 • Unsteady
wake is narrower than in the
laminar cas e 3 x 10 > Re
Uns teady flow

Vortex shedding behind a cylinder and Reynolds number

27
 Introduction to turbulence modeling

What happens when we increase the Reynolds number?

Drag coefficient as a function of Reynolds number for Strouhal number for a smooth cylinder [2]
a smooth cylinder [1]

References:

1. Fox, Robert W., et al. Introduction to Fluid Mechanics. Hoboken, NJ, Wiley, 2010
2. Sumer, B. Mutlu, et al. Hydrodynamics Around Cylindrical Structures. Singapore, World Scientic, 2006
28
 Introduction to turbulence modeling

Vorticity does not always mean turbulence

Instantaneous vorticity magnitude field

http://www.wolfdynamics.com/training/turbulence/image6.gif

• The Reynolds number in this case is 100, for these conditions, the flow still is laminar.
• We are in the presence of the Von Karman vortex street, which is the periodic shedding of vortices caused by
the unsteady separation of the fluid around blunt bodies.
• Vorticity is not a direct indication of turbulence.
• However, turbulent flows are rotational, they exhibit vortical structures. 29
 Introduction to turbulence modeling

Turbulence modeling – Fluctuations of transported quantities

• We have defined turbulence as an unsteady, aperiodic motion in which velocity components and every
transported quantity fluctuate in space and time.
• For most engineering applications it is impractical to account for all these instantaneous fluctuations.
• Therefore, we need to somehow remove those small scales by using models.
• To remove the instantaneous fluctuations (or small scales), two methods can be used: Reynolds averaging
and Filtering.
• Both methods introduce additional terms that must be modeled for closure.
• We are going to talk about closure methods later. 30
 Introduction to turbulence modeling

Turbulence modeling – Fluctuations of transported quantities

Laminar flow in a pipe Turbulent flow in a pipe

• Turbulence has a direct effect on the velocity profiles and mixing of transported quantities.
• In the laminar case, the velocity gradient close to the walls is small (therefore the shear stresses
are lower).
• The turbulent case shows two regions. One thin region close to the walls with very large velocity
gradients (hence large shear stresses), and a region far from the wall where the velocity profile
is nearly uniform.
• In the illustration, the velocity profile of the turbulent case has been averaged (in reality, there
are fluctuations). 31
 Introduction to turbulence modeling

Turbulence modeling – Fluctuations of transported quantities

iscous sub ayer aminar o

urbu ent pro i e

iscous sub ayer aminar o
Averaged turbulent flow Instantaneous turbulent flow

• In the left figure, the velocity profile has been averaged.

• In reality, the velocity profile fluctuates in time (right figure).
• The thin region close to the walls has very large velocity gradients and is laminar.
• Far from the flows, the flow becomes turbulent.
32
 Introduction to turbulence modeling

Turbulence modeling – Fluctuations of transported quantities

Flow in a pipe. (a) Laminar, (b) Transitional, (c) Turbulent

• Turbulence has a direct effect on the velocity profiles and mixing of transported quantities.
• Case (a) correspond to a laminar flow, where the dye can mix with the main flow only via molecular diffusion,
this kind of mixing can take very long times.
• Case (b) shows a transitional state where the dye streak becomes wavy but the main flow still is laminar.
• Case (c) shows the turbulent state, where the dye streak changes direction erratically, and the dye has mixed
significantly with the main flow due to the velocity fluctuations. 33
 Introduction to turbulence modeling

Turbulence modeling – Fluctuations of transported quantities

• Illustration taken from Osborne Reynolds’ 1883 influential paper “An experimental investigation of the
circumstances which determine whether the motion of water in parallel channels shall be direct or sinuous and
of the law of resistance in parallel channels”.
• Water flows from the tank near the experimenter down to below the ground, through a transparent tube; and dye
is injected in the middle of the flow.
• The turbulent or laminar nature of the flow can therefore be observed precisely.

Figures taken from the original reference (Osborne Reynolds, 1883). This work is in the public domain in its country of origin and other countries and areas
where the copyright term is the author's life plus 100 years or less. 34
 Introduction to turbulence modeling

Turbulence modeling – Fluctuations of transported quantities

iscous sub ayer aminar o

Laminar

urbu ent pro i e

Transitional

Turbulent

iscous sub ayer aminar o

Wall bounded turbulence Shear free turbulence

• Turbulent flows can originate at the walls. When this is the case, we talk about wall bounded
turbulence.
• Turbulent flows can also originate in the absence of walls (or far from walls). When this is the
case, we talk about shear free turbulence (usually jets, heated walls, atmospheric flows).
35
 Introduction to turbulence modeling

Turbulence modeling – Boundary layer

y ayer Note: The scales are
Laminar boundary Transition Turbulent boundary exaggerated for clarity
layer region layer

oundary layer thickness

u 0.99U

Logarithmic layer

uffer layer
Viscous sublayer

Critical length

Boundary layer (Laminar-Transitional-Turbulent flow)

• In this case, a laminar boundary layer starts to form at the leading edge.
• As the flow proceeds further downstream, large shear stresses and velocity gradient develop within the
boundary layer. At one point the flow becomes turbulent.
• The turbulent motion increases the mixing and the boundary layer mixing.
• What is happening in the transition region is not well understood. The flow can become laminar again or can
become turbulent.
• As for the pipe flow, the velocity profiles in the laminar and turbulent regions are different. 36
 Introduction to turbulence modeling

Turbulence modeling – Thermal boundary layer

herma boun ary ayer

oun ary ayer thickness

iscous boun ary ayer
Pr 1
Pr 1
Pr 1
T(y)

T wall

Thermal boundary layer vs. Viscous boundary layer Thermal boundary layer in function of Prandtl number (Pr)
Forced convection

Momentum and thermal boundary layer

• Just as there is a viscous boundary layer in the velocity distribution (or momentum), there is also a thermal
boundary layer.
• Thermal boundary layer thickness is different from the thickness of the viscous sublayer (momentum), and is
fluid dependent.
• The thickness of the thermal sublayer for a high Prandtl number fluid (e.g. water) is much less than the
momentum sublayer thickness.
• For fluids of low Prandtl numbers (e.g., air), it is much larger than the momentum sublayer thickness.
• For Prandtl number equal 1, the thermal boundary layer is equal to the momentum boundary layer. 37
 Introduction to turbulence modeling

Turbulence modeling – Thermal boundary layer

ra ity

ra ity umes iscous boun ary ayer

U

T(y)
T wall

T wall
herma boun ary ayer

Horizontal heated plate immersed in a quiescent fluid.

Natural convection Vertical heated plate immersed in a quiescent fluid.
Natural convection.

Natural convection in a heated plate

• As the fluid is warmed by the plate, its density decreases and a buoyant force arises which induces flow
motion in the vertical or horizontal direction.
• The force is proportional to , therefore gravity must be considered. 38
 Introduction to turbulence modeling

Turbulence modeling – Grid scales

Cell size
This cell is resolving the eddies

To resolve the boundary layer you

need very fine meshes close to the
Cell size walls
This cell is not resolving the eddies Cell size
This cell may be resolving the eddies

• Turbulence modelling aims at predicting velocity and transported quantities fluctuations without
calculating the complete turbulent flow pattern as a function of time.
• Everything below grid scales or sub-grid scales (SGS) is modelled or filtered.
• Therefore, if we want to capture all scales we need very fine meshes in the whole domain.

Bullet at Mach 1.5

Photo credit: Andrew Davidhazy. Rochester Institute of Technology.
39
Copyright on the images is held by the contributors. Apart from Fair Use, permission must be sought for any other purpose.
 Introduction to turbulence modeling

Energy spectrum and energy cascade

• Notice that this kind of graph is

local. It will be different for each
and every point in the domain.
• In the x axis the wave number is
plotted,

• The turbulent power spectrum represents the distribution of the turbulent kinetic energy across
the various length scales. It is a direct indication of how energy is dissipated with eddies size.
• The mesh resolution determines the fraction of the energy spectrum directly resolved.
• Eddies cannot be resolved down to the molecular dissipation limit. 40
 Introduction to turbulence modeling

Energy spectrum and energy cascade

• The energy-containing eddies are denoted by L0; L1 and L2 denotes the size of the eddies in
the inertial subrange such that L2 < L1 < L0; LN is the size of the dissipative eddies.
• The large, energy containing eddies transfer energy to smaller eddies via vortex stretching.
• Smallest eddies convert kinetic energy into thermal energy via viscous dissipation.
• Large eddies derive their energy from the mean flow.
• The size and velocity of large eddies are on the order of the mean flow

References:
41
S. Pope. Turbulent Flows. Cambridge University Press. 2014.
 Introduction to turbulence modeling

Energy spectrum and mesh resolution

• DNS simulations are quite expensive, they require a lot of grid points/cells in order to
resolve all the turbulent scales.
• Thus, in a DNS simulation the gridding requirements scales proportional to
or approximately proportional to for a single time step.
• And every time step should be sufficiently resolved in time (CFL condition less than 1,
and the ideal value should be less than 0.5).
• An alternative to DNS, is the use of large eddy simulations (LES).
• In a good resolved LES simulation, we aim at resolving 80% of the turbulent kinetic
energy.
• If the mesh requirements of a LES are too high, we can do a VLES (very large eddy
simulation).
• In a VLES we aim at resolving 50% of the turbulent kinetic energy.
• If LES requirements are still high (which is the case for most of the industrial
applications), we use RANS/URANS models.
• In RANS/URANS simulations the whole energy spectrum is modeled.
• The mesh spacing should be sufficiently to capture well integral scales and
model/resolve the boundary layer.
42
 Introduction to turbulence modeling

Energy spectrum and mesh resolution

• From the previous discussion, we can see that the mesh resolution determines the fraction of
the turbulent kinetic energy directly resolved.
• So, let us suppose that we want to resolve 80% of the turbulent kinetic energy k(l) in a LES
simulation. Then, approximately 15 cells will be needed across the integral length scale .
• In the same way, if you would like to solve 50% of the turbulent kinetic energy k(l) (VLES), you
will need approximately 4 cells across the integral length scale .

Ratio of integral length scale to grid length scale

Ratio of characteristic eddy size to integral length scale

6.1 1.0

1.6 3.9

0.42 15

0.16 38

Characteristics turbulent kinetic energy and length scales of

the energy spectrum. For a rigorous explanation of these
results, please refer to Turbulent Flows by S. Pope 43
Cumulative turbulent kinetic energy against lengths scale of eddies
 Introduction to turbulence modeling

Energy spectrum and mesh resolution

• In LES simulations, it is a good practice to have at least 10-15 cells across the integral length
scale .
• If you are interested in VLES, 4 to 6 cells per integral length scale are recommended (this
also applies to RANS/URANS).
• To resolve an eddy with a length scale l (where l is the smallest scales that can be resolved with
the grid or ), at least a couple of cells need to be used in each direction.
• Remember, eddies cannot be resolved down to the molecular dissipation limit (it is too
expensive).
k
o

Slope 5 3

o k

eso e spectrum o e e spectrum

nner region
Anisotropic Isotropic
ous sublayer Flow dependent Homogeneous
uffer Universal Log law layer 44
Layer
 Introduction to turbulence modeling

Integral length scale and grid length scale

• The integral length scale can be roughly estimate as follows,
• Based on a characteristic length, such as the size of a bluff body or pipe
diameter.
• From correlations.
• From experimental results.
• From a precursor RANS simulation.

• Remember, turbulent kinetic energy peaks at integral length scale .

• Therefore, these scales must be sufficiently resolved in LES/DES simulations, or
capture (be able to track) in RANS/URANS simulations.
• After identifying the integral scales, try to cluster enough cells in the domain regions
where you expect to find the integral scales (or large eddies).
• In other words, put sufficient cells in the wake or core of the flow.
• In RANS/URANS/VLES simulations, it is acceptable to use a minimum of 5 cells
across integral length scales.
• LES simulations have higher requirements.
45
 Introduction to turbulence modeling

Integral length scale and grid length scale

• The integral length scales can be computed from a precursor RANS simulation as follows.
• You will need to use a two-equation model ( family or family).
• Depending on the model selected, you can compute as follows,

or where

• The ratio of integral length scale to grid length scale can be computed as follows,

where can be approximated as follows

This approximation is accurate if the aspect
ratios are modest (less than 1.2)

• The recommended value of should be

• Where 5 should be considered the lowest limit of resolution (for RANS/URANS and VLES) and
10 is the desirable lower limit (for LES/DES).
• Higher values can be used if computer power and time constrains permits.
• This is a very rough estimate, which is likely problem dependent.
• Remember, in well resolved LES simulations equal mesh resolution should be provided in all
directions. 46
 Introduction to turbulence modeling

Turbulence near the wall - Law of the wall

y ayer Note: The scales are
Laminar boundary Transition Turbulent boundary exaggerated for clarity
layer region layer

oundary layer thickness

u 0.99U

Logarithmic layer

uffer layer
Viscous sublayer

Critical length

Actual profile - Physical velocity profile

• Near walls, in the boundary layer, the velocity changes rapidly.

• In turbulence modeling in CFD, the most important zones are the viscous sublayer and the log-law layer.
• The buffer layer is the transition layer which we try to avoid as much as possible.
• Turbulence modeling in CFD requires different considerations depending on whether you solve the viscous
sublayer, model the log-law layer, or solve the whole boundary layer (including the buffer zone).
47
 Introduction to turbulence modeling

Turbulence near the wall - Law of the wall

integral scales

inertial scales

dissipation scales

Non-dimensional profile Actual profile - Physical velocity profile

• The use of the non-dimensional velocity and non-dimensional distance from the wall , results in a
predictable boundary layer profile for a wide range of flows.
• Under standard working conditions this profile is the same, however, under non-equilibrium conditions
(production and dissipation of turbulent kinetic energy not balanced), rough walls, porous media, buoyancy,
viscous heating, strong pressure gradients, and so on, the profile might be different.
48
 Introduction to turbulence modeling

Turbulence near the wall - Law of the wall

Where y is the distance normal to the wall, U

is the shear velocity, and relates the mean
velocity to the shear velocity

or wall distance units is a very

important concept when dealing
with turbulence modeling,
remember this definition as we are
going to use it a lot.

49
 Introduction to turbulence modeling

Near-wall treatment and wall functions

• When dealing with wall turbulence, we need to choose a near-wall treatment.
• If you want to resolve the boundary layer up to the viscous sub-layer you need very
fine meshes close to the wall.
• In terms of , you need to cluster at least 6 to 10 layers at .
• But for good accuracy, usually you will use 20 to 30 layers.
• This is the most accurate approach, but it is computationally expensive.

50
 Introduction to turbulence modeling

Near-wall treatment and wall functions

• When dealing with wall turbulence, we need to choose a near-wall treatment.
• If you are not interested in resolving the boundary layer up to the viscous sub-layer,
you can use wall functions.
• In terms of , wall functions will model everything below or the target
value.
• This approach use coarser meshes, but you should be aware of the limitations of the
wall functions.
• You will need to cluster at least 5 to 8 layers to resolve the profiles (U and k).

51
 Introduction to turbulence modeling

Near-wall treatment and wall functions

• When dealing with wall turbulence, we need to choose a near-wall treatment.
• You can also use the y+ insensitive wall treatment (sometimes known as continuous
wall functions or scalable wall functions). This kind of wall functions are valid in the
whole boundary layer.
• In terms of y+, you can use this approach for values between .
• This approach is very flexible as it is independent of the y+ value, but is not available
in all turbulence models
• You also should be aware of the limitations this wall treatment method.

52
 Introduction to turbulence modeling

Near-wall treatment and wall functions

• Generally speaking, wall functions is the approach to use if you are more interested in
the mixing in the outer region, rather than the forces on the wall.
• If accurate prediction of forces or heat transfer on the walls are key to your simulation
(aerodynamic drag, turbomachinery blade performance, heat transfer) you might not
want to use wall functions.
• The wall function approach is also known as high-RE (HRN).
• The approach where you do not use wall functions is known as low-RE (LRN).
• Wall functions should be avoided if . This is the transition region, and
nobody knows what is going on there.
• The low-RE approach is computational expensive as it requires clustering a lot cells
near the walls.
• To get good results with LRF, you will need to cluster at least 10 layers for
• If you do not have any restrictions in the near-wall treatment, use wall functions.
• Wall functions can be used in RANS, DES and LES.
• If you are doing LES, it is highly recommended to use wall functions. Otherwise, your
meshing requirements will be very similar to DNS. 53
 Introduction to turbulence modeling

Removing small scales

Steady mean value Unsteady mean value

• We have seen that turbulent flows are characterize by instantaneous fluctuations of velocity,
pressure, and all transported quantities.
• In most engineering applications is not of interest resolving the instantaneous fluctuations.
• To remove the instantaneous fluctuations or small scales, two methods can be used:
• Reynolds averaging
• Filtering
• Both methods introduce additional terms that must be modeled for closure. 54
 Introduction to turbulence modeling

Removing small scales

• Two methods can be used to eliminate the need to resolve the small scales:
• Reynolds averaging (RANS/URANS):
• All turbulence scales are modeled.
• Can be 2D and 3D.
• Can be steady or unsteady.
• Filtering (LES/DES):
• Resolves large eddies.
• Models small eddies.
• Intrinsically 3D and unsteady.

• Both methods introduce additional terms in the governing equations that must be
modeled.
• The final goal of turbulence modeling is to find the closure equations to model the
additional terms (usually a stress tensor).
55
 Introduction to turbulence modeling

Overview of turbulence modeling approaches

MODEL

Increasing modelling and mathematical

Increasing computational cost
RANS
Reynolds-Averaged Navier-Stokes equations

URANS

complexity
Unsteady Reynolds-Averaged Navier-Stokes equations

SAS
Scale Adaptive Simulations
Scale-Resolving Simulations

DES
SRS

Detached Eddy Simulations

LES
Large Eddy Simulations

DNS
Direct Numerical Simulations
56
 Introduction to turbulence modeling

Overview of turbulence modeling approaches

RANS LES – Instantaneous field

RANS/URANS DES/LES DNS

• Solve the time-average NSE. • Solve the filtered unsteady NSE. • Solves the unsteady laminar NSE.
• All turbulent spatial scales are • Sub-grid scales (SGS) are filtered, • Solves all spatial and temporal
modeled. grid scales (GS) are resolved. scales; hence, requires extremely
fine meshes and small time-steps.
• Many models are available. One • Aim at resolving the temporal scales,
equation models, two equation hence requires small time-steps. • No modeling is required.
models, Reynolds stress models,
• For most industrial applications, it is • It is extremely computational
transition models, and so on.
computational expensive. However, expensive.
• This is the most widely approach for thanks to the current advances in
• Not practical for industrial flows.
industrial flows. parallel and scientific computing it is
becoming affordable. • It is intrinsically 3D and asymmetric.
• Unsteady RANS (URANS), use the
same equations as the RANS but • Many models are available.
with the transient term retained.
• It is intrinsically 3D and asymmetric.
• It can be used in 2D and 3D cases. 57
 Introduction to turbulence modeling

Overview of turbulence modeling approaches

LES – Smagorinsky – nut URANS – SA – nut URANS – kw – nut
http://www.wolfdynamics.com/training/turbulence/image9.gif http://www.wolfdynamics.com/training/turbulence/image10.gif http://www.wolfdynamics.com/training/turbulence/image11.gif

LES – U mean URANS SA – U mean URANS kw – U mean

• The instantaneous field of each method might be slightly different, but when we average the solution, we
obtain similar fields.
• All cases use the same mesh and have approximately the same computational overload.
• Unsteadiness makes turbulence simulations long. 58
 Introduction to turbulence modeling

Overview of turbulence modeling approaches

• Most of the times we are not interested in the unsteadiness evolution.
• Therefore, we must compute the statistics of the time evolution of the quantities of interest.
• On the other hand, if we use RANS models (steady simulations) we will get the mean results, which is ok for
most engineering applications.
• Sometimes the iterative evolution of a RANS simulation might fluctuate; therefore, it is advice to compute the
descriptive statistics in a given window.

Mean value 2.495

Standard deviation 0.286
Variance 0.0822
RMS 2.512

Mean value
The initial transient was not
considered to compute the
descriptive statistics

59
 Introduction to turbulence modeling

Turbulence modeling – Starting equations

NSE

Additional closure equations for the turbulence models

• Equations cannot be derived from fundamental principles.

• All turbulence models contain some sort of empiricism.
• Some calibration to observed physical solutions is contained in the turbulence
models.
• Also, some intelligent guessing is used.
• A lot of uncertainty is involved! 60
 Introduction to turbulence modeling

Incompressible RANS equations

• Let us derive the Reynolds average equations for an incompressible flow.
• This is our starting point,

• And we want to arrive to the following RANS equations,

Note: if you drop the time derivative, we call it RANS (steady equations), and if you retain the time derivative, we call it
URANS (unsteady equations). 61
 Introduction to turbulence modeling

Incompressible RANS equations

• To derive the incompressible RANS equations we need to apply Reynolds average to
the primitive variables of the starting equations, as follows,

• Let us recall the following averaging rules,

62
 Introduction to turbulence modeling

Incompressible RANS equations

• Let us recall the Reynolds decomposition for the primitive variables of the starting
equations,

• By substituting the previous equations into our starting equations, using the previous
averaging rules, and doing some algebra, we arrive to the incompressible
URANS/RANS equations,
If we retain this term we talk about URANS equations and if we drop it we talk about RANS equations Reynolds stress tensor

Notice that all quantities are averaged (the overbar over the field quantities) 63
 Introduction to turbulence modeling

Incompressible RANS equations

• The RANS equations are very similar to the starting equations.
• The only difference is that all quantities are averaged (the overbar over the field
quantities).
• And the appearance of Reynolds stress tensor

• The RANS approach to turbulence modeling requires the Reynolds stresses to be

appropriately modeled.
• It is possible to derive its own governing equations (six new equations as the tensor is
symmetric), but it is much simpler to model this term.
• Probably this is the most physically sound RANS model (RSM or Reynolds stress
model) as it avoids the use of the Boussinesq hypothesis that we will study next.
64
 Introduction to turbulence modeling

Incompressible RANS equations

• A common approach used to model the Reynolds stress tensor, is to use the
Boussinesq hypothesis.
• By using the Boussinesq hypothesis, we can relate the Reynolds stress tensor to the
mean velocity gradient such that,

• In the previous equation, denotes the strain-rate tensor.

• is the identity matrix.
• is the turbulent eddy viscosity.
• or is the turbulent kinetic energy.
• At the end of the day we want to determine the turbulent eddy viscosity. Each
turbulence model will compute this quantity in a different way.
• The turbulent eddy viscosity is not a fluid property, it is a property needed by the
turbulence model. 65
 Introduction to turbulence modeling

Incompressible RANS equations

• In the Reynolds decomposition, the quantities of interest can be averaged as follows.
• For steady RANS we can use time averaging as follows,

where

• For unsteady RANS or URANS we can use ensemble averaging as follows,

where

• For ensemble average the number of experiments of the ensemble must be large
enough to eliminate effects of fluctuations. This type of averaging can be applied to
any flow (steady or unsteady).

66
 Introduction to turbulence modeling

Incompressible RANS equations

• We just outlined the incompressible RANS.
• The compressible RANS equations are similar, to derive them we use Favre average
(which can be seen as a mass-weighted averaging).
• If we drop the time derivative in the governing equations we are dealing with steady
turbulence.
• On the other hand, if we keep the time derivative we are dealing with unsteady
turbulence and we should use ensemble averaging.
• The derivation of the LES equations is very similar, but instead of using averaging we
filter the equations in space, and we solve the temporal scales.
• LES/DES models are intrinsically unsteady and three-dimensional.
• We will briefly address filtering and the LES equations.
• Remember, the main goal of the turbulence models is to find the turbulent eddy
viscosity to model the Reynolds stress tensor.
• Let us take a look at the governing equations of the RANS model.
67
 Introduction to turbulence modeling

Turbulence model overview

• It is called because it solves two additional equations for modeling the
turbulence, namely, the turbulent kinetic energy and the specific rate
of dissipation

• These are the closure equations of the turbulence problem using Reynolds average.
• These are not physical properties. They kind of represent the generation and
destruction of turbulence.
• Recall that

• In the model, the turbulent eddy viscosity is computed as follows,

68
 Introduction to turbulence modeling

Turbulence model overview

• In its simplest form, the following equations are applied in the first cell center of the
logarithmic layer,

• These equations are here referred to as the standard wall function for the
method. They relate the local wall shear stress (through the shear velocity ) to
the mean velocity, turbulence kinetic energy k, and rate of dissipation .
• A small implicit system must be solved in an iterative way in order to compute the
friction velocity .
• First, we compute the friction velocity, then we compute , and then we find .
• By adopting a mesh where the first cell center is located in the logarithmic layer, it is
possible to use the law-of-the-wall to specify the boundary condition for the
dependent variables so that the turbulence model equations are not
solved close to the wall.
• Each turbulence model formulation has different wall function equations. 69
 Introduction to turbulence modeling

Turbulence model overview

• The boundary and initial conditions recommended for this model are,

where L is the Length of the domain.

• Alternatively, you can use the following equations:

where is the viscosity ratio and is the turbulence intensity

(recommended values in the next slide).

70
 Introduction to turbulence modeling

Turbulence model overview

• If you are totally lost, you can use these reference values. They work most of the
times, but it is a good idea to have some experimental data or a better initial estimate.

Low Medium High

I 1.0 % 5.0 % 10.0 %

1 10 100

where is the viscosity ratio and is the turbulence intensity.

• By the way, use these guidelines for external aerodynamics only.

• Similar estimates are available for internal flows.

71
 Introduction to turbulence modeling

Turbulence model free-stream initial conditions

• At the walls, you can use the following values

where and is the distance to the first cell center normal to the wall.

• The values are the same stated in the original reference.

• If you use a different turbulence model, just refer to the original reference or visit the
following site:

https://turbmodels.larc.nasa.gov/

• You can also visit our website and use the calculator to estimate turbulence
quantities,
http://www.wolfdynamics.com/tools.html?id=110
72
 Turbulence modeling in OpenFOAM®

Turbulence model boundary conditions at the walls

• As for the free-stream boundary conditions, you need to give the boundary conditions
for the near-wall treatment and domain boundaries.
• When it comes to near-wall treatment, you have three options:
• Use wall functions (values up to 600 are acceptable if they do not cover more
that 20% of the surface and are not located in critical regions):

• Use y+ insensitive wall functions, this only applies with the SST model:

• Resolve the boundary layer (no wall functions):

73
 Turbulence modeling in OpenFOAM®

Turbulence model boundary conditions at the walls

• If you are planning to use wall functions ( ),
the following values are good choices.
• For you can use the wall function kqRWallFunction with the following value:

or

• For you can use the wall function omegaWallFunction with the following value:

or the free-stream value

where and is the distance to the first cell center normal to the
wall.
• To avoid potential erroneous arithmetic operations, it is recommended to use a small
value instead of 0.
74
 Turbulence modeling in OpenFOAM®

Turbulence model boundary conditions at the walls

• If you are planning to use y+ insensitive wall functions ( ),
the following values are good choices.
• For you should use the wall function kqRWallFunction or kLowReWallFunction
with the following value:

or

• For you should use the wall function omegaWallFunction with the following value:

or the free-stream value

where and is the distance to the first cell center normal to the
wall.
• To avoid potential erroneous arithmetic operations, it is recommended to use a small
value instead of 0. 75
 Turbulence modeling in OpenFOAM®

Turbulence model boundary conditions at the walls

• If you are planning not to use wall functions ( ),
the following values are good choices.
• For you should use fixedValue or kLowReWallFunction with the following value:

• For you should use fixedValue with the following value:

• You can also use fixedValue with a large value.

• To properly resolve the boundary layer, it is recommended to cluster at least 10 layers
close to the walls.
• To avoid potential erroneous arithmetic operations, it is recommended to use a small
value instead of 0. 76
 Turbulence modeling in OpenFOAM®
Typical wall functions boundary conditions
• Remember, you can only use wall functions if the primitive patch (the patch
type defined in the boundary dictionary), is of type wall.

Field Wall functions – High RE Resolved BL – Low RE

nut(–)WallFunction* or nutUSpaldingWallFunction** or
nut nutUSpaldingWallFunction** (with 0 or a nutLowReWallFunction or fixedValue (with
small number) 0 or a small number)

kqRWallFunction or kLowReWallFunction kLowReWallFunction or fixedValue (with

k, q, R
(with inlet value, 0, or a small number) inlet value, 0, or a small number)

epsilonWallFunction (with inlet value) or

epsilon epsilonWallFunction (with inlet value) zeroGradient or fixedValue (with 0 or a small
number)

omegaWallFunction** or fixedValue (both

omega omegaWallFunction (with a large number)
with a large number)

nuTilda – fixedValue (with 0 or a small number)

* $WM_PROJECT_DIR/src/TurbulenceModels/turbulenceModels/derivedFvPatchFields/wallFunctions/nutWallFunctions
** For y+ insensitive wall functions (continuous wall functions) 77
 Introduction to turbulence modeling

Short description of some RANS turbulence models

Model Short description

This is a one equation model. Suitable for external aerodynamics, tubomachinery and high
speed flows. Good for mildly complex external/internal flows and boundary layer flows under
Spalart-Allmaras pressure gradient (e.g. airfoils, wings, airplane fuselages, ship hulls). Performs poorly with flows
with strong separation.

This is a two equation model. Very robust and widely used despite the known limitations of the
model. Performs poorly for complex flows involving severe pressure gradient, separation, strong
Standard k–epsilon streamline curvature. Suitable for initial iterations, initial screening of alternative designs, and
parametric studies. Can be only used with wall functions.

This is a two equation model. Suitable for complex shear flows involving rapid strain, moderate
swirl, vortices, and locally transitional flows (e.g. boundary layer separation, massive separation,
Realizable k–epsilon and vortex shedding behind bluff bodies, stall in wide-angle diffusers, room ventilation). It
overcome the limitations of the standard k-epsilon model.

This is a two equation model. Superior performance for wall-bounded boundary layer, free
shear, and low Reynolds number flows compared to models from the k-epsilon family. Suitable
Standard k–omega for complex boundary layer flows under adverse pressure gradient and separation (external
aerodynamics and turbomachinery).

This is a two equation model. Offers similar benefits as the standard k–omega. Not overly
sensitive to inlet boundary conditions like the standard k–omega. Provides more accurate
SST k–omega prediction of flow separation than other RANS models. Can be used with and without wall
functions. Probably the most widely used RANS model.

78
 Introduction to turbulence modeling

Final remarks on RANS turbulence models

• In terms of meshing requirements, RANS simulations are not computational
expensive.
• Therefore, it is recommended to use a RANS approach for steady
simulations.
• If you are conducting URANS simulations (unsteady runs), low-RE meshes
can impose strict time-step restrictions. Therefore, high-RE meshes are
recommended.
• Nevertheless, you can use large CFL numbers in URANS simulations.
• If you are dealing with thermal boundary layers, it is recommended to use a
low-RE approach.
• Remember, avoid as much as possible the buffer layer
• Use RANS simulations as starting point for LES/DES.

79
 y+ and wall distance units

Turbulence near the wall - Law of the wall

• Let us revisit the definition of y+.
• Remember, y+ or wall distance units normal to the wall is a very important concept
when dealing with wall bounded turbulence modeling, remember this definition as it is
used a lot.

wall shear stresses

Where y is the distance normal to the wall, U

is the shear velocity, and relates the mean
velocity to the shear velocity
80
 y+ and wall distance units

wall distance units normal to the wall

• We never know a priori the value (because we do not know the friction velocity).
• What we usually do is to run the simulation for a few time-steps or iterations, and then
we get an estimate of the value.
• After determining where we are in the boundary layer (viscous sub-layer, buffer layer
or log-law layer), we take the mesh as a good one or we modify it if is deemed
necessary.
• It is an iterative process and it can be very time consuming, as it might require
remeshing and rerunning the simulation.
• Have in mind that it is quite difficult to get a uniform value at the walls.
• Try to get a mean value as close as possible to your target.
• Also, check that you do not get very high maximum values of (more than a 1000)
• Values up to 300 are fine. Values larger that 300 and up to a 1000 are acceptable is
they do not covert a large surface (no more than 10% of the total wall area), or they
are not located in critical zones.
• Use common sense when accessing value. 81
 y+ and wall distance units

Estimating normal wall distance

• At meshing time, to estimate the normal wall distance to the first cell center, we use
the well know y+ definition,

• Where we set a target y+ value and then we isolate the quantity y (normal wall
distance to the first center). This will be distance that we will use when generating the
boundary layer mesh.

• So if you choose a low y+, you

will have a mesh that is
clustered towards the wall. And
if you choose a large y+ value,
you will have a coarse mesh
towards the walls.

82
 y+ and wall distance units

Estimating normal wall distance

• At meshing time, to estimate the normal wall distance to the first cell center, we use
the well know y+ definition,

• The problem is that at meshing time we do not know the value of the shear velocity,

• So, how do we get an initial estimate of this quantity?

83
 y+ and wall distance units

Estimating normal wall distance

• To get an initial estimate of the distance from the wall to the first cell center ,
without recurring to a precursor simulation, you can proceed as follows,

1.
(Skin friction coefficient of a flat plate, there are
2. similar correlations for pipes)

3.

4.
Your desired value
5.

• You will find a simple calculator for the wall distance estimation in the following link:
http://www.wolfdynamics.com/tools.html?id=2
84
y+ and wall distance units

Wall distance units

• Similar to , the wall distance units can be
computed in the stream-wise ( ) and
span-wise ( ) directions.
• The wall distance units in the stream-wise
and span-wise directions can be computed
as follows:

• And recall that is computed at the cell

center, therefore:

where
Viscous length 85
 y+ and wall distance units

Wall distance units and some rough estimates

• Similar to , the wall distance units can be computed in the span-wise ( ) and
stream-wise ( ) directions.
• Typical requirements for LES are (these are approximations based on different references):

Wall resolving

Wall modeling
86
 y+ and wall distance units

Wall distance units and some rough estimates

• DES and RANS simulations do not have stream-wise and span-wise wall
distance units requirements as in LES simulations. Therefore, they are more
affordable.
• If you are conducting DES simulations it is highly recommend to resolved the
boundary layer, but you can also use wall functions.
• LES wall functions are valid across the whole boundary layer, even in the
buffer layer).
• The upper limit of y+ for LES and DES simulations should be less than
300 ( ).
• Remember, it is strongly recommended to use wall functions with LES
simulations. Otherwise your meshing requirements will be close to those of
DNS.
• If you are doing DNS, y+ should be less than 1. The spanwise and
streamwise values should be less than 10, but ideally close to 1.

87
 y+ and wall distance units

Wall distance units and some rough estimates

• If you are using wall functions with RANS models, you should avoid the
buffer layer.
• In RANS, the typical range values for wall functions is between
• But values of y+ up to 600 are also acceptable (depending on the Reynolds
number).
• If you want to resolve the viscous layer of the boundary layer using
RANS, , you should cluster at least 10 cells in viscous region.
• If you want to use y+ insensitive wall functions with RANS
models,
• Remember the y+ insensitive wall functions are only valid with y+ insensitive
turbulence models ( SST ).
• There are no strict requirements when it comes to the span-wise and stream-
wise wall distance units in RANS, but as a general rule you can use
and values as high as 300 times the value of

88
 y+ and wall distance units

A few mesh resolution guidelines and rough estimates

• The mesh is everything in CFD, and when it comes to turbulence modeling it is
extremely important to have meshes with good quality and acceptable resolution.
• Some general guidelines for meshes to be used with RANS/DES/LES:
• Resolve well the curvature.
• Allow a smooth transition between cell of different sizes (at least 3 cells).
• Identify the integral scales and try to cluster at least 5 cells in the domain regions
where you expect to find the integral scales.

• Some guidelines specific to RANS meshes:

• When it comes to RANS, the most important metric for mesh resolution is the
value. Identify your wall treatment a-priory and mesh your domain according to
this requirement.
• If you are doing 3D simulations, there are no strict requirements when it comes to
the span-wise and stream-wise directions, but as a general rule you can
use and values as high as 300 the value of .

89
 y+ and wall distance units

A few mesh resolution guidelines and rough estimates

• Some guidelines specific to DES meshes:
• The mesh requirements are very similar to those of RANS meshes.
• It is extremely important to resolve well the integral length scales.
• DES simulations are intrinsically 3D.
• Try to avoid the use of symmetry (axial and planar).
• Some guidelines specific to LES meshes:
• When it comes to LES meshes, it is recommended to use wall functions.
Otherwise the meshing requirements are similar to those of DNS.
• It is recommended to use values in the range of . LES uses wall
functions that can deal with the buffer layer.
• In LES, it is extremely important to resolve well the stream-wise and span-wise
directions. Recommended values are: and
• LES simulations are intrinsically 3D.
• Try to avoid the use of symmetry (axial and planar).
• Use hexahedral meshes.
90
 Final remarks – General guidelines

References
• Turbulent Flows
S. B. Pope
• Turbulence Modeling for CFD
D. C. Wilcox
• Turbulence: An Introduction for Scientists and Engineers
P. A. Davidson
• Large Eddy Simulation for Incompressible Flows
P. Sagaut
• A First Course in Turbulence
H. Tennekes and J. L. Lumley
• Boundary-Layer Theory
H. Schlichting
• Turbulence Modelling - http://www.tfd.chalmers.se/~lada/comp_turb_model/
Lars Davidson
• https://turbmodels.larc.nasa.gov/

91