You are on page 1of 88

Turbulence Modeling

ANSYS Fluent Advanced Course


by Tomer Avraham
April 24, 2018 1
What is Turbulence?
Everyone who , at one time or another, has observed the efflux from a
smokestack has some idea about the nature of turbulence flow. How ever
it is very difficult to give a precise definition of turbulence. All one can do
is list some of the characteristics of turbulent flow:
Irregularity… Enhanced Diffusivity… Enhanced dissipation… Large
Reynolds numbers, Three-Dimensionality, Vorticity fluctuation and their
obvious stretching, tilting and fractalization Mechanisms,
Chaos…………… Turbulent flows are turbulent flows – H. Tennekes
and J. L. Lumley

April 24, 2018 2


The Future of CFD

April 24, 2018 3


The energy cascade concept and mechanisms

April 24, 2018 4


The energy cascade – Vortex Stretching Effect
 Turbulence may be considered as a state of fluid motion (that arises from an instability in
laminar flow) that features a wide range of length and time scales that organize themselves into
coherent vortical structures and via vortex stretching effect.

 Vortex stretching is a very known mechanism for the transfer of energy between small wave
(large energetic scales) numbers to high wave numbers (small scales).
A vortex tube subjected to strain from local velocity gradients of the flowfield will tend to
stretch, thereby shrinking its diameter. The consequence is that the energy associated with that
vortex is acting at a larger wave number (smaller scales).

 This so-called energy cascade process is effectively inviscid with an energy transfer rate set by
the large scales. At the so-called Kolmogorov small scales, this energy cascade stops due to the
ability of viscous forces at small scales balancing inertial forces and dissipating any turbulence
energy converting it into thermal energy or heat

April 24, 2018 5


Modeling of Turbulence:
Direct Numerical Simulations (DNS)
 Turbulence phenomena is very precisely described by a seemingly simple set of equations, the Navier-
Stokes equations, their nature is such that analytic solutions to even the most simple turbulent flows
can not be obtained and resorting to numerical solutions seems like the only hope.

 But the resourcefulness of the plea to a direct numerical description of the equations is a mixed blessing
as it seems the availability of such a description is directly matched to the power of a dimensionless
number reflecting on how well momentum is diffused relative to the flow velocity (in the cross-stream
direction) and on the thickness of a boundary layer relative to the body – The Reynolds Number.

 It is found that the computational effort in Direct Numerical Simulation (DNS) of the Navier-Stokes
equations rises as Reynolds number in the power of 9/4 which renders such calculations as prohibitive
for most engineering applications of practical interest and it shall remain so for the foreseeable future.

April 24, 2018 6


DNS: Kolmogorov's Route - Universality Assumptions
 Kolmogorov's First Universality Assumption

At very high, but not infinite, Reynolds number, all of the small-scale statistical properties are uniquely and
universally determined by the length scale ℓ, the mean dissipation rate (per unit mass) ε and the viscosity ν.

 Kolmogorov's Second Universality Assumption

In the limit of infinite Reynolds number, all small-scale statistical properties are uniquely and universally
determined by the length scale ℓ and the mean dissipation rate ε.

April 24, 2018 7


DNS: Kolmogorov's Route – length scales in turbulence
 four main sets of scales in a turbulent flow:

April 24, 2018 8


DNS and the Lambda2 criterion
 A three dimensional direct numerical simulation using high-order methods has been performed to study
the flow around the asymmetric NACA-4412 wing at a moderate chord Reynolds number
(Rec = 400,000), with an angle of attack of 5 degrees.
This flow regime corresponds approximately to the flow around a small glider. In addition to providing
highly accurate data, high-order methods produce massive amount of data enabling proper flow
visualization. For instance, in this study vortical structures emerging from tripping the flow to turbulence
are visualized using the lambda2 criterion. It is interesting to see how interaction of such vortical
structures from the turbulent boundary layer and the turbulent wake creates a natural art of its own.

https://youtu.be/aR-hehP1pTk

April 24, 2018 9


DNS – computational resources
 A direct numerical simulation consists in solving all essential scales of motion that are at least of order of
magnitude of the Kolmogorov scale computed as:

 As a rough guide, in order to describe a “minimal” sine curve on a full period, the number of grid-points
of the mesh is then given by:

where L is the size of the computational domain.

 Considering a turbulent flow at the Reynolds number Re = UbL/ν based on the bulk velocity Ub and the
characteristic length scale L, the above estimation can be easily simplified if one considers that the size
of the energetic big eddies Le = k3/2/e is roughly of order of magnitude of the characteristic geometrical
size of the flow itself leading to Nη = 64 Rt9/4

April 24, 2018 10


DNS – computational resources
 The computational time can be estimated if we assume that the turbulent Reynolds number is
proportional to the mean flow Reynolds number Rt = ζ Re, where ζ is an empirical coefficient usually
close to 1/10 in confined flows (and in usual Reynolds numbers range). It is then proportional to the
Reynolds number according to the law t ∝ 64ζ Rt 11/4 . These numerical order of magnitudes clearly show
that DNS (or even highly resolved LES) implies a huge numerical task and still remains difficult to reach
in practice at the present time even if considering the Moore’s law suggesting that the number of
transistors of a processor doubles every second years (10 years = factor 32) and of top nowadays
supercomputers :

April 24, 2018 11


Modeling of Turbulence:
Reynolds Averaged Simulation (RANS)
 in order to make predictions of engineering turbulent flows, it makes sense to attempt to predict the
statistics of the turbulence.

 This includes, first and foremost, the mean, representing a time-average for steady flows and an
ensemble- or phase-average for unsteady flows.

 The so-called Reynolds decomposition involved replacing the instantaneous flow variables with a mean
plus a fluctuation in the Navier-Stokes equations and applying the average to the entire equation.

 This yielded the RANS equations and introduced the so-called turbulence closure problem and Reynolds
stress tensor.

 different turbulence models goal is to relate the unknown Reynolds stress tensor to the mean velocity
field (actually derivatives of the velocity field) and other flow related quantities. These models can be
divided into two main categories: (a) eddy-viscosity models and (b) non-eddy viscosity models. Eddy
viscosity models invoke the Boussinesq approximation that enforces a linear relationship between the
Reynolds stress tensor and the mean strain-rate tensor with a so-called scalar eddy viscosity serving as
the isotropic proportionality factor.

April 24, 2018 12


RANS: the “closure problem”
 Applying a long-time averaging that is sufficiently large in comparison with the turbulence time scale
(and sufficiently small in comparison with the evolution time of the mean flow) on the instantaneous
equations leads to the averaged equations of conservation of mass and momentum of the flow.

 As a result, the motion equation contains the unknown turbulent stress that must be modeled to close
and solve the set of equations. This problem is known as the turbulence closure problem.

April 24, 2018 13


RANS – derivation and manifestation of the closure problem
 The starting point for this derivation is, of course, the NSE:

 Applying the Reynolds decomposition on NSE:

 Applying the averaging process on the above:

 Using a notion that the mean of pertubative pressure and velocity are zero:

April 24, 2018 14


RANS – derivation and manifestation of the closure problem

 In the left hand side:

 The second equality follows because u is independent of time and similarly, the third occurs because
∇, as an operator, is also independent of time. The last is again because the mean of u’ is zero.

 A similar analysis can be conducted for the third term:

 The equation becomes:

April 24, 2018 15


RANS – derivation and manifestation of the closure problem

 The equation becomes:

 And could be also written by vector manipulation as:

 We define the Reynolds stress tensor:

 The final (steady-state) RANS is written as:

April 24, 2018 16


The law of the wall
 The law of the wall is associated with wall-bounded shear flows, and depending upon just how one
counts, these might be viewed as having two, three or even four different length scales represented in
their physical behaviors.

 Such flows are found in boundary layers, and thus also in pipes, ducts and channels.

April 24, 2018 17


The law of the wall
 let uτ denote a velocity scale for the inner region. This velocity shall be referred to as “friction velocity”.
This is defined as:

 If we observe that the two length scales are a large advective scale associated with the free-flow mean
velocity and a viscous scale corresponding to uτ, then for an intermediate scale to make sense, it must
be the case that:

 Within this range one argues that ν/uτ is too small to control flow dynamics, and h is too large to result
in effective interactions. Hence, y itself is then the only length scale for this region.

April 24, 2018 18


The law of the wall
 But there are two velocity scales, namely, the mean free-flow velocity and uτ. On purely dimensional
grounds they are related as:

 We define the dimensionless quantities:

 And the above differential equation becomes:

 Integrated to become the “log-law” that matches the inner to the outer layer:

April 24, 2018 19


Workshop I: Introduction
This short tutorial demonstrates in practice the technique for Creating the
desired wall spacing in ANSYS Mesh for matching a flat plate calculated y+
in ANSYS Fluent .

April 24, 2018 20


Part A: Predicting and calculating the desired wall
spacing
Taken from lecture 7 in ANSYS
Fluent introduction course

April 24, 2018 21


Part B: Creating the desired wall spacing in ANSYS Mesh
1) Unarchive the supplied
yPlus_example.wbpz.

The geometry is already supplied, it consists of a


simple rectangular domain, bottom wall of
which will be meshed with the theoretic values
from previous slide.

2) Open Mesh Component, define named


selections as shown on the next slide.

April 24, 2018 22


inlet outlet

wall_top

wall_bot

April 24, 2018 23


Part B: Creating the desired wall spacing in ANSYS Mesh
3) Select wall_top and wall_bot related edges,
right click -> insert -> sizing.

4) Define sizing parameters as shown, these


parameters will ensure that the stream wise
aspect ratio of the near wall region cells will
not exceed 5

5) Select inlet and outlet edges, right click ->


insert -> sizing.

6) Define sizing parameters as shown.

7) When applying bias please note that one of


the edges (inlet or outlet) will have to be
reversed.

April 24, 2018 24


Part B: Creating the desired wall spacing in ANSYS Mesh
Bias factor and number of divisions on inlet
and outlet edges were calculated in a way
that the first cell centroid will be situated at
the theoretical value of ~0.9 [mm]
calculated earlier, please note that in terms
of mesh it dictates first “layer” height at
~1.8 [mm].
First layer height

Remember that you also can obtain desired


first cell height wall spacing with the help of
local or global inflation controls. First cell height

April 24, 2018 25


Part B: Creating the desired wall spacing in ANSYS Mesh
8) Select the rectangular face, right click ->
insert -> face meshing.

This is done so that the resulting mesh will be


smooth and mapped.

9) Click “Generate Mesh” and observe the


resulting mesh. You can zoom in, rotate and
use the ruler on the screen in order to verify
the height of the first cell.

Resulting first layer height


is approx. 1.8 [mm], which
correspond to first cell
height of 0.9 [mm], which
both correspond to the
desired y+ of 50.

April 24, 2018 26


Part C: Fluent Setup
10) Close ANSYS Mesh, connect it to a new
Fluent component.
11) Update Mesh component if necessary.
12) Open Fluent setup in with double precision
option.

13) Models -> Viscous -> K-Epsilon.


14) Boundary conditions -> Inlet
15) Solution initialization -> hybrid -> Initialize.
16) Run calculation -> 100 Iterations ->
Calculate.

The run should converge after ~40 iterations.

April 24, 2018 27


Part C: Fluent Setup
The final step of this tutorial will be checking
what y+ values we actually have for our first cell
adjacent to wall_bot bc.

17) Results -> Plots -> XY Plot.


18) Set the parameters as shown and click plot.
19) Observe the y+ distribution next to wall_bot.

Please note that it takes about


0.4 [m] for the flow to develop
and then we can see that we get
the values that we expected from
theoretical calculation.

April 24, 2018 28


Conclusions
This tutorial demonstrated how to implement
theoretical y+ calculation in to the
computational grid and how to verify the y+
value of the grid after solution convergence.

As it was mentioned, it takes the flow some


distance to develop, that’s why we can see
values of up to 64 in terms of y+ in the first 0.4
[m] of the plate/wall.

You can try and see what values of y+ you will


receive for developed flow at the inlet.
You can achieve this by sampling the velocity
magnitude at the outlet to a profile and then
assigning this profile to the inlet.

April 24, 2018 29


Reynolds Averaged Simulation (RANS)
 different turbulence models goal is to relate the unknown Reynolds stress tensor to the mean velocity field
(actually derivatives of the velocity field) and other flow related quantities.

 These models can be divided into two main categories:


(a) eddy-viscosity models
(b) non-eddy viscosity models.

 Eddy viscosity models invoke the Boussinesq approximation that enforces a linear relationship between the
Reynolds stress tensor and the mean strain-rate tensor with a so-called scalar eddy viscosity serving as the
isotropic proportionality factor.

 These included zero-, one-, and two-equation models where 0, 1, or 2 additional modeled PDE’s are solved to
provide estimates for turbulent length and velocity scales to evaluate the eddy viscosity in a dimensionally
consistent fashion.

April 24, 2018 30


Mixing length theory and the Boussinesq hypothesis
 The Bousinesq Hypothesis stands in the basics of eddy-viscosity related turbulence modeling. The linear
Bousinesq hypothesis major claim is that the principal axis of the Reynolds stresses coincide with those of the
average strain:

 A hypothesis in the kinetic theory of (rare) gasses is that molecules passing through y=0 are holding their
characteristic momentum from the velocity layer they where coming from:

April 24, 2018 31


Mixing length theory and the Boussinesq hypothesis
 In the molecular level a decomposition is proposed: u=U+u’’ (Where U is defined by U(y) and u'' is molecular random
movement).

 The sudden flux of every property through y=0 is proportional to the normal to plane velocity normal to plane.
Concerning the description above it is v''. Hence the sudden change in momentum through a differential element dS may
be described as:

 After conducting an ensemble average this becomes:

 As by definition the stress acting on y=0 may be described as:

 This brings us to:

 For ideal gas the molecular velocity is following the maxwellian distribution, such that all directions are equally possible.
The average molecular velocity shall be the thermal velocity:

April 24, 2018 32


Mixing length theory and the Boussinesq hypothesis
 average half of the molecules follow to the positive side and the others to the negative. if we take the vertical
velocity these becomes:

 Now we integrate on a hemisphere:

 This means that the total molecules on the route for the positive direction:

 In their way from P to Q each molecule is " typical of where they come from", hence each molecule from P carries
about a negative momentum:

April 24, 2018 33


Mixing length theory and the Boussinesq hypothesis
 This means that the total momentum flux from to the negative side (to first Taylor expansion approximation):

 n the same grounds, the total momentum flux from to the positive side (to first Taylor expansion approximation):

 Summing both sides it becomes:

 Now we may write: where:

April 24, 2018 34


Mixing length theory and the Boussinesq hypothesis
 The assumptions that guarantees that a first Taylor expansion shall be valid require:

(a)

(b)

The analogy of the bousinesq hypothesis to the derived consequences from the theory of kinetic gasses:

molecule------------------------->Fluid parcel

Mean free path --------------->Mixing length

 It is very straightforward to write the following, derived directly from the above:

 The Boussinesq hypothesis then may be written as:

April 24, 2018 35


Shortcoming of the Boussinseq hypothesis
 It is possible to define lmix, but it is a property of the flow rather than the fluid (such as the case in the kinetic theory of
gasses) thus universality may not be expected.

 The problem relies in the fact that lmix might not be smaller than variation in the average flow properties. This is due
to the spectral gap problem which is not evident in the molecular counterpart.

 Failure to predict flows with sudden and abrupt changes in the strain of the averaged flow. In the Taker-Reynolds
experiment on an almost isotropic flow a rate of strain flux is applied on a unified averaged flow (U,-ay,az), where a is
the a constant rate of strain. following some distance the strain is abruptly stopped.
While the experiment shows a gradual return to isotropy, the boussinesq hypothesis predicts a sudden return with the
exact moment of the abrupt strain stopped.

 Failure to give a reliable prediction to swirling flows, slows over curved surfaces separations etc...
The Boussinesq hypothesis ties between the average velocity tensor of the flow and the Reynolds stresses in a linear
relation. therefore even in the equations for the kinetic energy enters the influence of the strain tensor which is the
symmetric part of the velocity tensor after a decomposition to a symmetric an antisymmetric part.
The antisymmetric part – the rotation tensor doesn't appear in the equation for the kinetic energy nor in the
Boussinesq hypothesis.

April 24, 2018 36


Shortcoming of the Boussinseq hypothesis – final conclusions
 The set of flow situations in which the Boussinesq hypothesis has proven to be inadequate:

i) flows with sudden changes in mean strain rate

ii) flows over curved surfaces.

iii) flow in ducts or, in general, those containing secondary fluid motions, including boundary-layer
separation.

iv) flow of rotating and/or stratified fluids.

v) three-dimensional flows.

April 24, 2018 37


RANS – zero-equation models

 simply on dimensional grounds: and then set:

 In zero-equation models invoking the Boussinesq hypothesis: leads to:

 In zero-equation models it is typical to express ℓmix as: where δ(x) is the single characteristic length
(for example the diameter of a jet), and C1 is the closure constant to be determined such that computed mean
velocity profiles match corresponding measurments (or DNS).

 But there is no formula to ℓmix … In these sense zero-equation models are said to be “incomplete” as a major part of
the corresponding model must be determined entirely from experiment (or DNS).

 Values of mixing-length closure constant for various flows:

April 24, 2018 38


RANS – two-equation and eddy viscosity models

 1-equation and 2-equations models, incorporate a differential transport equation for the turbulent velocity
scale (or the related the turbulent kinetic energy) and in the case of 2-equation models another transport equation for
the length scale (or time scale).

 In this sense 2-equation models can be viewed as "closed" because unlike 0-equation and 1-equation models (with
exception maybe of 1-equations transport for the eddy viscosity itself) these models possess sufficient equations for
constructing the eddy viscosity with no direct use for experimental results.

 2-equations models do however contain many assumptions along the way for achieving the final form of the transport
equations and as such are calibrated to work well only according to well-known features of the applications they are
designed to solve.

 Nonetheless although their inherent limitations, today industry need for rapid answers dictates CFD simulations to be
mainly conducted by 2-equations models whose strength has proven itself for wall bounded attached flows at high
Reynolds number (thin boundary layers) due to calibration according to the law-of-the-wall.

April 24, 2018 39


The k-ε Turbulence Model

 The k-ε turbulence model still remains among the most popular, most known is the standard (Jones-Launder) k-ε
turbulence model.

 As in all eddy-viscosity turbulence model derivation initializes with the Boussinesq hypothesis:

 The second steps is to devise (on purely dimensional grounds) a relation between the eddy-viscosity and the two
chosen characteristics:

 Obtaining a transport equation for the total kinetic energy is a simple mathematical step of forming a dot product of
NSE with the velocity vector:

 A transport equation for the total kinetic energy could be written as:

 Decomposing the velocity vector according to Reynolds decomposition


and defining the turbulent kinetic energy as:

April 24, 2018 40


The k-ε Turbulence Model

 The construction of an energy transport equation for the mean flow by the same procedure as the total kinetic energy
transport equation was constructed (i.e. dot product of the mean velocity with RANS equations):

 The next steps consider time (or ensemble) averaging the total kinetic energy transport equation and
the subtraction of the mean flow energy transport equation:

1. Pressure work due to only turbulence.


2. Transport of turbulent kinetic energy due to fluctuations.
3. Diffusive transport of turbulence kinetic energy.
4. Turbulence production, or to be more precise the amplification of the Reynolds stress tensor by the mean strain.
5. Dissipation rate of turbulence kinetic energy.

April 24, 2018 41


The k-ε Turbulence Model
 Another important step evident to all such models is ad-hoc simplification relying on (some) physical justification for
each of the above terms to achieve a final transport equation:

 In the construction of an equation for the turbulence dissipation ε. To do so we first invoke local isotropy for the
dissipation.

 as far as the Reynolds decomposition is concerned it is somewhat harder to justify local isotropy as the fluctuating
term in RANS is not an actual representation of high wave-number (small spatial scale) behaviour in general (for LES it
does!).

 It is possible to derive the equation following the same route as for the turbulent kinetic energy to finaly achieve a
transport equation for the turbulence dissipation ε:

April 24, 2018 42


Final form of the k-ε Turbulence Model
 After constructing both equations and defining the relations between the transported variables to the eddy viscosity
the final form of the standard k-ε turbulence model may be presented:

 The last important step is to calibrate the model constants. In turbulence modeling calibration of the model is at least
as important as the derivation of the model itself. Calibration is achieved with the help of experimental and numerical
results of the type of flow that should be modeled. The calibration process is also the first step in which the range of
validity of the model would be revealed to close inspection and not just postulated from physical reasoning.

For the standard k-ε turbulence model the calibrated closure constants are:

April 24, 2018 43


Shortcomings of the k-ε Turbulence Model
 the model is essentially a high Reynolds model, meaning the law of the wall must be employed and provide velocity
"boundary conditions" away from solid boundaries (what is termed "wall-functions").
From a mathematical standpoint, even if one could impose Dirichlet conditions for ε on solid boundary, after meshing
it would still be difficult to numerically approach the problem due to what is termed in numerical analysis as
stiffness of the numerical problem, partially related to the high gradients.

 In order to integrate the equations through the viscous/laminar sublayer a "Low Reynolds" approach must be
employed. This is achieved as additional highly non-linear damping functions are needed to be added to low-
Reynolds formulations (low as in entering the viscous/laminar sublayer) to be able to integrate through the laminar
sublayer (y+<5). This again produces numerical stiffness and in case is problematic to handle in view of linear
numerical algorithms.

 The model suffers from lack of sensitivity to adverse pressure-gradient. It was observed that under such conditions it
overestimates the shear stress and by that delays separation.

April 24, 2018 44


The Realizable k-ε Turbulence Model
 A straightforward option to circumvent some of the standard drawbacks still in the framework of the same variables as
transport transport equations was made by invoking realizability constraints.

 There are a number of such constraints, the usual ones are that all normal stresses should remain positive and the
correlation coefficients for the shear stress should not exceed one:

 Remembering that the velocity gradient tensor may be decomposed to a symertic part (strain rate tensor) and an anti-
symmetric part (rotation tensor):

April 24, 2018 45


The Realizable k-ε Turbulence Model
 If it is assumed that the flow that on one axis the flow approaches a wall, the Boussinesq Hypothesis for the normal
stress becomes:

 It could be seen that if s11 is too large then: , hence nonphysical, i.e. non-realizable.

 It is customary at this stage to introduce the concept of "invariant", meaning something that is independent on a
coordinate system. For the above case this relates to rotation.

The Invariants are calculated via solving for the eigenvalues of the strain rate tensor. The eigenvalues of S correspond to
the strains in the principal axis, since we have applied the equation on the principle axis, S11 is replaced by the largest
eigenvalue such that:

 This simple modification to an eddy viscosity model ensures that the normal stresses stay positive.

April 24, 2018 46


The Realizable k-ε Turbulence Model
 Another realizability constraint appears when we require that if: , This shall be done smoothly.

 This is ensured by demanding:

 We also impose a requirement that when v1’2 approaches zero, the transport equation for it shall do so to.

 This is one of these cases where the normal stress goes to zero faster (O)(x4) than the parallel one (O(x2)) and creates
the state of turbulence called the two component limit.

April 24, 2018 47


Near Wall Treatment for ε-equation
Turbulence Models
 Standard wall functions:
Wall functions are applicable for a suitable range of y* and by that to the flow's Reynolds. As the purpose is to
allow not integrating through the viscous sublayer, the lower limit of y* is 11 hence standard wall functions
should not be used below that limit as the solution's accuracy might deteriorate in an uncontrolled manner.

 Scalable wall functions:


To avoid deterioration of the solution by standard wall functions in situations where it's unavoidable for the first
grid point to be located at y*<11, ANSYS Fluent proposes wall functions that produce consistent results for
arbitrary grid refinement by forcing the usage of the log law in conjunction with the standard wall functions
approach:

 Non-equilibrium wall functions:


sensitizing the log-law for mean velocity to pressure gradient effects and by the use of the two-layer-based
concept to compute the reciprocal relations between turbulence kinetic energy production, Gk, and its
dissipation ε, by non-equilibrium means.
Therefore to some extent the non-equilibrium formulation for the wall functions takes into account the effect of
pressure gradients on the distortion of the velocity profiles and by that account for some non-equilibrium effects.
April 24, 2018 48
Near Wall Treatment for ε-equation
Turbulence Models
 Enhanced Wall Treatment ε-Equation (EWT-ε):
by incorporating the two-layer model with enhanced wall functions a separation of the two regions is conducted via a
wall-distance-based, turbulent Reynolds number:

For regions where the above defined Reynolds number is above 200 the original standard k-ε is employed.
By the same token if the above Reynolds number is below 200, a one equation for the transport of turbulence kinetic
energy is employed (Wolfstein's k-equation).

 Menter-Lechner ε-Equation (ML-ε):


To avoid EWT-ε shortcomings such as regions with small turbulence kinetic energy that might also have a turbulent
Reynolds number below 200 and as such treated with a near-wall formulation even though they are actually away
from the wall.
The approach offers a y+ insensitive which is a not based on a two layer model assuming a sufficient resolution of the
boundary layer, predicts in a y+ independent manner the wall shear stress and wall heat flux. Such a
formulation switches gradually from wall functions to a low-Reynolds formulation when the mesh is refined. The
concept is achieved by adding a a source term to the transport equation of the turbulence kinetic energy that
accounts for near-wall effects.
April 24, 2018 49
Workflow for constructing eddy-viscosity 2-eq turbulence models

1. Invoke the Boussinesq hypothesis:

2. Devise a relation between the eddy-viscosity and the two chosen parameters:

3. Obtaine a transport equation for each parameter.

4. Surgically identify and simplify by physical reasoning the identified terms in the initial transport equation. This shall
leave the final equation in a natural transport form (Convection-Production-Destruction-Diffusion) with some added
constants.

5. Calibrate the model constants via measurements (or DNS) according to the law-of-the-wall.

April 24, 2018 50


The standard k-ω turbulence model
 The k-ω turbulence model (in Fluent: standard/BSL) alleviates one of the most important shortcomings of the k-ε
turbulence model as it could be shown by looking at the wall boundary conditions that the ω equation takes an
"elliptic" near wall behavior (partial differential wise), this means that it has an inherent nature of being able to
"communicate" with the wall and actually has Dirchlet (as in no-slip in this case) boundary conditions.

 The term "elliptic" is used with regards to a concept introduced to the ε-equation in the k-ε turbulence model in order
to avoid the need for damping functions for the viscous/laminar sublayer and serves as the basis for the v2-f 4-
equation turbulence model as the concept of elliptic relaxation which could be easily shown to be an inherent feature
of the k-ω turbulence model only by inspecting the ω-equation in the near wall region when combined with the
specified ω values at the wall :

 The implication of such behavior is the straightforward integration through the laminar sublayer without additional
numerically destabilizing damping functions or two more transport equation (which shall generally cause stabilization
issues due to reciprocity between the variables).

April 24, 2018 51


The k-ω shear-stress transport (SST) turbulence model
 one of the shortcomings of standard k-ω is that the model depends strongly on free-stream values of ω that are
specified outside the shear layer.

 To alleviate both shortcomings, that Jones-Launder k-ε near wall behavior and that of Wilcox k-ω ambiguity to
freestream values of ω in one formulation, F. Menter (ANSYS turbulence modeling lead) decided to blend
continuously the Wilcox k-ω to be applied in the near-wall region with the Jones-Launder k-ε applied towards the end
of the boundary layer.

 This is done in the BSL/SST k-ω turbulence model by invoking a blending-function relying on a limiter and designed to
switch the model between the modes:

April 24, 2018 52


The k-ω shear-stress transport (SST) turbulence model
 A second major drawback, is evident in almost all eddy-viscosity models, relating the Reynolds stress to the mean flow
strain and in fact is the major difference between such a modeling approach and a full Reynolds-stress model (RSM).

 The RSM approach accounts for the important effect of the transport of the principal turbulent shear-stress.

 The ingenious idea of Menter to include it in the revised k-ω model (termed the Baseline (BSL) model) is
related observed success in implementing what is termed as the Bradshaw's assumption, that the shear-stress in the
boundary layer is proportional to the turbulent kinetic energy:

 To make this desirable feature apply only and boundary layer, Menter uses yet again the blending function concept.

April 24, 2018 53


The k-ω shear-stress transport (SST) turbulence model
 The end result is Menter's k-ω SST, one of the most reliable RANS turbulence models:

April 24, 2018 54


The Spalart-Allmaras (SA) Turbulence Model
 Eddy viscosity transport equation turbulence models are a special kind of 1-equation models in as much that they
posses completeness (one differential equation for the eddy viscosity to be related directly to the Reynolds Stresses).

 The Spalart-Allmaras Turbulence Model has been developed mainly for aerodynamic flows. The formulation blends
from a viscous sublayer formulation to a logarithmic formulation based on y+.

 In as such, no addition of highly non-linear damping functions for laminar/viscous sublayer modeling is in use.

 In the spalart-allmaras methodology, surgical physics based assumptions were made concerning each of the various
terms such as diffusion, production and destruction to the final aim of achieving a complete transport equation for the
eddy viscosity:

April 24, 2018 55


The Spalart-Allmaras (SA) Turbulence Model
 Diffusion:
Spalart-Allmaras representation of the diffusion terms kicks off with the classical diffusion operator as ∇·([νt/σ]∇νt)
where σ is a turbulent Prandtl number. Then, as to achieve an aerodynamic flow oriented diffusion behavior it is
pointed out that there is no reason for the integral of the eddy viscosity to be conserved due to cross terms between
∇k and ∇ε for example (as seen from the above derivation emanating from a 2-equation model), hence a non-
conservative term is added to the classical diffusion description:

 Production:
The representation of the production term is analogue to the production of turbulent energy, assuming that it shall
rise with an increase in total viscosity and with the increase of the mean vorticity.
The subtlety is in the consideration of the relation between the production and which form of mean vorticity is to be
chosen in the quest for most favorable effect. In the case of Spalart-Allmaras the favorable effect is to serve
aerodynamic flows in which turbulence is found only where vorticity is. Therefore the magnitude of vorticity is chosen
as the representation of the mean vorticity (it should be noted that future advancements of the original model to
account for various flow features may consider other forms for its representation) and the production takes the form:

where:
April 24, 2018 56
The Spalart-Allmaras (SA) Turbulence Model
 Destruction:
It is assumed for the eddy viscosity that the ability of a turbulent flow to transport momentum and the ability must be
directly related to the general ”level of activity“, therefore to the turbulent energy to construct the destruction term.
It's also claimed in the surgical process of deriving the destruction term that there is a "blocking effect" from a wall
that is felt at a distance by the pressure and acts as a destruction entity for the Reynolds shear stress, therefore the
use of a wall proximity parameter in the representation is mandatory under these assumptions.
The above assumption, and under proper calibration of the constants seem to reproduce an accurate log layer.

On the other hand, the skin friction it produces for a flat plate boundary layer is underestimated, a consequence of
the rate of decay of the destruction term in the outer portion of the boundary layer. Therefore, the Spalart-Allmaras
destruction term contains a function (which is equal to 1 in the log-layer) to control this rate of decay and takes the
final form of the destruction representation:

where d is the wall proximity parameter, cw1 the constant to be calibrated and fw is the control function.

 The final form (with constants to be calibrated):

April 24, 2018 57


Improvement to SA: rotation correction
 A drawback evident in almost all eddy-viscosity models is the inability to inherently account for rotation and
curvature.

 The route for altering the transport equation kicks off with the identification of the effect of curvature and rotation in
two types of extreme flows:

(a) thin shear flows with weak rotation (compared with the shear rate) or weak curvature (compared with the
inverse of the shear-layer thickness), highly impacting the level of the turbulent shear stress.

(b) homogeneous rotating shear flow and free vortex cores of which strong rotation reduces the turbulent
shear stress sharply.

 relationship between strain-rate and vorticity to be accounted for by a scalar quantity to handle curvature and
rotation for thin shear flows with weak rotation and also for homogeneous rotating shear flow and free vortex cores.

 Hence the production term of the eddy viscosity transport equation is multiplied by a "rotation function“:

April 24, 2018 58


Transition Modeling: Fluent/CFX Transition Model (LCTM)
 Modal analysis performed on the equations may achieve modal solutions characterized by initial exponential
growth, such as Klebanoff (K-type, classical), Novosibisrsk or Herbert (N-type or H-type, sub-harmonic) all typical to
low intensity of incoming turbulence.

 The initial breakthrough in the field of transition onset and process was the description both theoretically and
experimentally of what is termed Tollmien-Schlichting wave instabilities for a low incoming turbulence intensity
boundary layer:

April 24, 2018 59


Transition Modeling: Fluent/CFX Transition Model (LCTM)
 A different growth mechanism characterizing a type of transition onset typical to turbomachinery, where high
intensity incoming turbulence exist, is the bypass transition (as to bypassing the Tollmien-Schlichting waves instability):

April 24, 2018 60


Transition Modeling: Fluent/CFX Transition Model (LCTM)
 F. Menter, R. Langtry (Ansys) and S. Volker (GE) have devised a local correlation-based transition model, mainly based
on the following favorable features:

1. Calibrated prediction of transition onset and length.


2. Simulation of a diversity of transition mechanisms whether they are a consequence of modal or transient
growth.
3. Formulated locally, as quantities such as the integral thickness of the boundary layer. such non-local
calculations are problematic to carry out in commercial codes as they simply do not provide an
infrastructure for such calculations.
4. Do not affect the turbulence model while in a fully turbulent regime.

April 24, 2018 61


Transition Modeling: Fluent/CFX Transition Model (LCTM)
 Menter et al. LCTM is based on two additional transport equations for the intermittency-γ and the momentum
thickness (or the transition Reynolds number)-Reθt.

 The intermittency is used to trigger transition locally. The intermittency function is coupled with the k-ω SST Model.

 Another transport equation for the transition Reynolds number (momentum thickness) is added which captures the
non-local effects of changes in turbulence intensity and free-stream velocity outside the boundary layer.

 This equation also relates empirical correlations to the transition onset in the intermittency equation. This relation
allows for the model incorporation in commercial CFD package as "ready to use", without the need for interaction
from the user due to diverse geometry setups.

 As far as the mesh goes, the first grid must be located at y+=1 (wall units). A coarse mesh shall predict transition onset
further upstream with increasing y+ (highly undesirable).

April 24, 2018 62


Beyond Eddy-viscosity – Reynolds Stress Turbulence Models (RSM)
 In Reynolds stress transport equation models an additional set of 6 modeled PDE’s are solved to determine
the Reynolds stress tensor directly without resorting to an isotropic eddy viscosity assumption.

 The derivation of RSM is achieved by taking the second moment of the NS operator:

 Identifying and simplifying the outcome with ad-hoc physical reasoning brings about a “direct” tensorial transport
equation for the Reynolds stresses:

April 24, 2018 63


Beyond eddy viscosity – Reynolds Stress Turbulence Models (RSM)
 A drawback evident in almost all eddy-viscosity models is the inability to inherently account for rotation and
curvature.

 This drawback is resulted from relating the Reynolds stress to the mean flow strain and in fact is the major difference
between such a modeling approach and a full Reynolds-stress model (RSM).

 The RSM approach accounts for the important effect of the transport of the principal turbulent shear-stress.

 It was shown how curvature and rotation effects could be somewhat overcome in the framework of "first moment
models" such as the v2-f turbulence model by invoking the concept of elliptic relaxation, nevertheless the v2-f model
is still inferior to RSM for highly 3D swirling flows with strong secondary circulation as it holds only one attractive
feature of RSM (e.g. energy blocking).

April 24, 2018 64


Shortcomings of Reynolds Stress Turbulence Models (RSM)
 The above tensor form equation brings into light the “closure problem” as it is seen that as higher and higher
moments of the set of RANS equations may be taken, more unknown terms arise and the number of equations never
suffices.

 despite the fact that second moment closures are claimed to “contain more physics” than their Boussinesq
hypothesis first moment closures counterpart, all the the advantageous added physics is somewhat hampered by the
amount of ad-hoc modeled terms needed to actually close the equations.

 Since the physics of the model does not really shine from these statistical terms it is not obvious that the models
actually contain as much physics as we would like when taking upon ourselves the penalty in computational resources
which are much more demanding in RSM.

April 24, 2018 65


The Scale Adaptive Simulation (SAS) Turbulence Model
 The methodology lies in the midst of RANS and LES and is especially attractive for flows of which strong instabilities of
the flow exist.

 F. Menter-Y. Egorov URANS – “Scale-Adaptive Simulation” (SAS) is based on J. Rotta (1970), exact transport equation
for kL uses a relation between the integral length scale:

and the diagonal two-point correlation tensor measured at a location x with two probes at distance:

April 24, 2018 66


The Scale Adaptive Simulation (SAS) Turbulence Model
 In Menter-Egorov SAS the original equations are variated to retain the “law-of-the-wall”.

 The formulation consists of two equations, the first for the kinetic energy and the second for the square-root of kl
(hence K-SKL)

 What distinguishes the KSKL model from other 2-equation closures is the fact that in the last, the turbulence length
scale will always approach the thickness of the shear layer, while for KSKL model, the behavior is such that it allows
the identification of the turbulent scales from the source terms of the KSKL model to a measure of both the thickness
of the shear layer but also for non-homogenous conditions, as the Von-Karman length scale is related to the strain-
rate, individual vortices have locally different time constants (inversely to turnover frequencies) and therefore from a
certain size dependable upon the local strain rate, they may not be merged to a larger vortex – unsteady features
appear! (at the expanse of computational resources of course…)

April 24, 2018 67


The law of the wall –
y+ according to Fluent turbulence models

April 24, 2018 68


RANS Turbulence Model Usage*
Model Behavior and Usage
Economical for large meshes. Good for mildly complex (quasi-2D) external/internal flows and boundary layer flows under pressure
Spalart-Allmaras gradient (e.g. airfoils, wings, airplane fuselages, missiles, ship hulls). Performs poorly for 3D flows, free shear flows, flows with strong
separation.
Robust. Widely used despite the known limitations of the model. Performs poorly for complex flows involving severe pressure gradient,
Standard k–ε separation, strong streamline curvature. Suitable for initial iterations, initial screening of alternative designs, and parametric studies.

Suitable for complex shear flows involving rapid strain, moderate swirl, vortices, and locally transitional flows (e.g. boundary layer
Realizable k–ε* separation, massive separation, and vortex shedding behind bluff bodies, stall in wide-angle diffusers, room ventilation).

Offers largely the same benefits and has similar applications as Realizable. Possibly harder to converge than Realizable.
RNG k–ε
Superior performance for wall-bounded boundary layer, free shear, and low Reynolds number flows compared to models from the k-e
Standard k–ω family. Suitable for complex boundary layer flows under adverse pressure gradient and separation (external aerodynamics and
turbomachinery). Separation can be predicted to be excessive and early.
Offers similar benefits as standard k–ω. Not overly sensitive to inlet boundary conditions like the standard k–ω. Provides more accurate
SST k–ω* prediction of flow separation than other RANS models.
Similar to SST k-w. Good for some complex flows if SST model is overpredicting flow separation
BSL k–ω
Physically the most sound RANS model. Avoids isotropic eddy viscosity assumption. More CPU time and memory required. Tougher to
RSM converge due to close coupling of equations. Suitable for complex 3D flows with strong streamline curvature, strong swirl/rotation (e.g.
curved duct, rotating flow passages, swirl combustors with very large inlet swirl, cyclones).

• Realizable k-e or SST k-w are the recommended choice for standard cases
* The above taken from ANSYS Fluent introductory course
April 24, 2018 69
Large Eddie Simulations (LES)

 In Large Eddy Simulation or LES where the large, energy-containing eddies (created by
the flow generating device) are simulated or resolved accurately on a fine mesh and
the effects of the smaller, dissipative eddies (which would require DNS-like resolution)
on the resolved scales are modeled using a so-called subgrid-scale stress (SGS)model.
This scale separation is theoretically implemented using a low-pass spatial filter. The
filtered LES equations introduce the SGS stress tensor which can be decomposed in a
number of different forms.

 LES has severe limitations in the near wall regions, as the computational effort
required to reliably model the innermost portion of the boundary layer (sometimes
constituting more than 90% of the mesh) where turbulence length scale becomes
very small is far from the resources available to the industry.

 Anecdotally, best estimates speculate that a full LES simulation for a complete
airborne vehicle at a reasonably high Reynolds number will not be possible until
approximately 2050…

April 24, 2018 70


LES SGS Modelling – Smagorinsky SGS model
 The most common SGS models consist of spatially filtering the NSE with some kind of shape filter (may it be cutoff, top
hat, etc...). The most common representation, is a linear stress-strain relation relying on the Boussinesq hypothesis and
the eddy viscosity concept.

 The first and possibly still the most popular is the Smagorinsky model.

 Among Smagorinsky model shortcomings is the fact that the Smagorinsky constant is always positive and uniform and as
a direct consequence so is the eddy viscosity, which is especially problematic for sheared laminar flows, rendering it
as inherently unfit for the prediction of laminar to turbulent flow transition.

 Another somewhat related deficiency is found in the effect of wall proximity and the value of the eddy viscosity in the
near wall region, as the model requires the aid of ad-hoc damping functions and by such the eddy viscosity behavior
which should be of the order of the third power of the wall normal proximity is far from attained.

April 24, 2018 71


LES SGS Modelling – Wall-Adapting Local Eddy-Viscosity (WALE)
SGS model
 An observation concerning the contribution of turbulence structures to the global dissipation led to a relation between
the eddy viscosity and both the strain rate and the rotation.

 By tuning this dependence on strain rate and rotation according to expected turbulence physics, the Smagorinsky
model specified in the above deficiencies are alleviated, the correct behavior of the eddy viscosity is attained and
almost no eddy-viscosity contribution is achieved for wall bounded laminar flows (such as Poiseuille).

 Most important, the formulation is local. Locality is a very important feature for modern industrial application oriented
CFD codes as they do not provide the infrastructure of computing parameters such as integral boundary layer
parameters, or allow integration of quantities along the direction of streamlines.

 Moreover, as industrial CFD simulations often require to be carried out on parallel computer invoking domain
decomposition methodologies, flow features such as boundary layers may be split and computed on separate
processors such that search or integration algorithms.

April 24, 2018 72


LES SGS Modelling – Dynamic SGS models
 In the original dynamic model as proposed by Germano, indirectly relying on scale-similarity concepts, another filter
(Test filter), dependable on the grid filter is added. This operation is equivalent to drawing information about the
unknown subgrid scales from the smallest resolved scales.

 The object of the procedure in the original dynamic model is to dynamically propose a local value for the Smagorinsky
constant (as a note all relations between the achieved stresses are algebraic).
 The aim is to select the model constant to depend as little as possible on the level of filtered velocity on which the
prediction is based.

April 24, 2018 73


LES: Inflow Conditions
 If the flow in the numerical domain is purely turbulent, a complete LES must account for turbulent structures in the
flow at the inflow region. Generally the effect of best applying inflow conditions could mean the difference between
seemingly colorful albeit worthless unsteady cartoon of CFD.

 Among the most popular methods for the generation of inflow conditions is to introduce an artificial perturbation
(based on Fourier series for example) to the mean flow.
There are two subtleties in the method. First, the perturbation must be tuned such that it is high enough to produce
turbulence, but not too high as to produce unphysical results. The second is that the numerical domain has to be built
such that the perturbation shall have enough downstream evolving space.

 Another method is to extract dependent variables downstream to the inflow, rescale them and dynamically apply
them to the inflow boundary, this way the introduction of inflow conditions is generated implicitly in the simulation
itself. The main shortcoming of the method is that special care must be taken to the actual rescaling process.

 An interesting method considers tripping the inflow region itself using blowing and suction at the inflow. Such
operations may generate bypass transition through generation of high levels of turbulence in the free stream, a
mechanism dominated by diffusion effects as turbulence is diffused into the boundary layer from high free stream
levels

April 24, 2018 74


LES: Outflow and other Computational Boundaries

 The consideration of other boundaries except of the obvious inflow conditions is extremely important for aeroacoustic
applications, subsonic flows and reacting flows, where unphysical numerical reflections from the boundary may totally
contaminate the results.

 Some other methods use characteristic wave relations to eliminate the effect.

 Another method is to deliberately include source terms in the equations themselves to increase damping of specific
features that are major contributors for such a reflective behavior from the computational boundaries.

 If the simulation allows it, generally when there is a dominant homogenous direction, periodic conditions could be
applied. In such a case the domain must be large enough (upstream-downstream) as to relaxing unwanted
downstream features.

April 24, 2018 75


LES: Spatial and Temporal Resolution
 In LES, especially for complex flows as the large-scale anisotropic features are to be accurately (as much as numerically
possible) captured. SGS modeling should apply to structures in the inertial range.

 This posses again a problem in as much as resolution of near wall region is concerned, the small eddies carry a great
amount of turbulent kinetic energy and the "spectral gap problem" renders no scale separation. The end result is that
the near wall resolution of LES compares with that of DNS..

 Mostly LES should be practiced using high-order spatial numerical methods. For flows dominated by vortices which
must be resolved for a long distance obtaining an engineering sufficient accuracy could prove very problematic with
low order methods. Such an example may be found in the calculation of helicopter loads which are strongly dependent
on the vortices generated by the tip of the rotor. As low order methods strongly dissipate the unsteady vortical
structures, a remarkably high and impossible to solve (on computational resources level) mesh should be generated to
capture the phenomenon.

 if higher-order spatial numerical schemes and finer grid resolutions are imposed to capture smaller turbulent
structures, the time scale should also be adjusted in order to accurately describe the advection and convection of
these structures.

 Time step choosing is a compromise. Trying to cut the time step to achieve a better temporal view would of course
increase the computational resources demand. Trying to do so on a coarser mesh would mean fine scale structures
would not be captured.

April 24, 2018 76


LES: Validation
 Is it possible to have a perfect LES such that the solution to LES equations shall be equal to spatially filtered value of
the true velocity field?... Actually no.

 As the spatially filtered value of the true velocity field is a random field, its future evolution is not determined by its
current state. So even if at some initial moment the LES solution equals the spatially filtered value of the true velocity
field, the last has just a statistical distribution and hence no specific value for the solution of the LES equation equals
the spatially filtered value of the true velocity field. Meaning the relation between the two entities is statistical.

 The immediate consequence is that a-priori testing comparing LES quantities obtained from a specific realization of the
true velocity field do not serve as a validation procedure.

 The procedure should have a set of equations for the resolved velocity field and a measure for some statistics as
temporal and wave number spectra (entirely not an easy task...). This could be done by periodically recording
complete data sets (memory issues shall arise) and by that obtain statistical information to achieve ensemble averages.

 The ultimate validation is achieved by incorporating the LES solution with statistical measurements extracted by the
previous described procedure (or some other) in order to get estimates on the statistics of the true turbulent velocity
field.

April 24, 2018 77


Hybrid Methods: Detached and Delayed Detached Eddy
Simulation (DES/DDES)
 Today's industry need for rapid answers dictates CFD simulations to be mainly conducted by Reynolds-Averaged
Navier-Stokes (RANS) simulations whose strength has proven itself for wall bounded attached flows due to calibration
according to the law-of-the-wall. However, for free shear flows, especially those featuring a high level of unsteadiness
and massive separation RANS has shown poor performance following its inherent limitations.

 In LES the large energetic scales are resolved while the effect of the small unresolved scales is modeled using a
subgrid-scale (SGS) model and tuned for the generally universal character of these scales. LES has severe limitations in
the near wall regions, as the computational effort required to reliably model the innermost portion of the boundary
layer (sometimes constituting more than 90% of the mesh) where turbulence length scale becomes very small is far
from the resources available to the industry.

 On the other hand, for free shear flows of which the large eddies are at the order of magnitude as the shear layer, LES
may provide extremely reliable information as it's much easier to resolve the large turbulence eddies in a fair
computational effort.

 As such, researchers have shifted much of the attention and effort to hybrid formulations incorporating RANS and LES
in certain ways. In most hybrid RANS-LES methods RANS is applied for a portion of the boundary layer and large eddies
are resolved away from these regions by an LES.

April 24, 2018 78


Hybrid Methods: Detached and Delayed Detached Eddy
Simulation (DES/DDES)
 Today's industry need for rapid answers dictates CFD simulations to be mainly conducted by Reynolds-Averaged
Navier-Stokes (RANS) simulations whose strength has proven itself for wall bounded attached flows due to calibration
according to the law-of-the-wall. However, for free shear flows, especially those featuring a high level of unsteadiness
and massive separation RANS has shown poor performance following its inherent limitations.

 In LES the large energetic scales are resolved while the effect of the small unresolved scales is modeled using a
subgrid-scale (SGS) model and tuned for the generally universal character of these scales. LES has severe limitations in
the near wall regions, as the computational effort required to reliably model the innermost portion of the boundary
layer (sometimes constituting more than 90% of the mesh) where turbulence length scale becomes very small is far
from the resources available to the industry.

 On the other hand, for free shear flows of which the large eddies are at the order of magnitude as the shear layer, LES
may provide extremely reliable information as it's much easier to resolve the large turbulence eddies in a fair
computational effort.

 As such, researchers have shifted much of the attention and effort to hybrid formulations incorporating RANS and LES
in certain ways. In most hybrid RANS-LES methods RANS is applied for a portion of the boundary layer and large eddies
are resolved away from these regions by an LES.

April 24, 2018 79


Hybrid Methods: Detached and Delayed Detached Eddy
Simulation (DES/DDES)
 One of the most popular hybrid RANS-LES models is Detached Eddy Simulation (DES) devised originally by
Philippe Spalart. The term DES is based on the Idea of covering the boundary layer by RANS model and
switching the model to LES mode in detached regions thereby cutting the computational cost significantly yet
still offering some of the advantages of an LES method in separated regions.

April 24, 2018 80


Hybrid Methods: Detached Eddy Simulation (DES)
 One of the most popular hybrid RANS-LES models is Detached Eddy Simulation (DES)
devised originally by Philippe Spalart. The term DES is based on the Idea of
covering the boundary layer by RANS model and switching the model to LES mode in
detached regions thereby cutting the computational cost significantly yet still
offering some of the advantages of an LES method in separated regions.

 The formulation of the hybridization of the model is fairly straight forward:

 This means that as Δ is max(ΔX, ΔY, ΔZ) this modification of the S-A model, changes
the interpretation of the model as the modified distance function causes the model
to behave as a RANS model in regions close to walls, and as an eddy-viscosity based
LES (Smagorinsky, WALE, etc'...) manner away from the walls.

 In DES the hybrid formulation has a limiter switching from RANS to LES as the grid is
reduced. The problem with natural DES is that an incorrect behavior may be
encountered for flows with thick boundary layers or shallow separations. It was
found that when the stream-wise grid spacing becomes less than the boundary layer
thickness the grid may be fine enough for the DES length scale to switch the DES to
its LES mode without proper "LES content", i.e. resolved stresses are too weak
("Modeled Stress Depletion" or MSD"), which in turn shall reduce the skin friction
and by that may cause early separation. The phenomenon is termed Grid Induced
Separation (GIS).

April 24, 2018 81


Hybrid Methods: Detached Eddy Simulation (DES)
 As a consequence of the original DES deficiencies an advancement to the model was devised, termed Delayed-DES
(DDES). In the Fluent DES-SST formulation a DES limiter "shield" is added to maintain RANS behavior in the boundary
layer without grid dependency.

 A function is defined to ensure that the solution will be a RANS solution even if the grid spacing is smaller than the
boundary layer thickness (so it will be 1 in the LES region where the length scale defined above is much smaller than 1,
and 0 elsewhere while not sensitive in situations of high proximity to the wall when the length scale exceeds 1:

 Now an alteration to the DES length scale is proposed such that under specific coefficient values (which the above
function is not so sensitive to even in the case of a different formulation of DES other than spalart-Allmaras, in ANSYS
Fluent case the k-ω SST Model ):

 In this formulation, when the function is 0, the length scale dictates RANS mode to operate, and when the function is 1
natural DES (P. Spalart 1997) applies. The difference lies in the fact that on contrary to natural DES formulation where
the length scale depends solely on the grid, in the DDES formulation it depends also on the eddy-viscosity. This means
that the revised formulation will "insists" upon remaining on RANS mode if the grid is inside the boundary layer and if
massive separation is encountered, the functions value will switch to LES mode a much more abrupt manner than the
switch in the natural DES formulation, rendering the "grey area" narrower which is highly desirable.

April 24, 2018 82


Shielded Detached Eddy Simulation (SDES)
 The SDES formulation is yet another variation of DES. The improvement is in the shielding function and the interaction
with the grid scale. This is emphasized in the turbulence model by an additional sink term in the turbulence kinetic
energy equation:

 The shielding function in the SDES formulation (namely – fs) provides more shielding then the corresponding shielding
function in the DDES formulation (F-DDES), this means that the original shielding based on the mesh length scale can
be reduced and is therefore defined in SDES as:

 The first part in the above is the conventional LES mesh length scale, the second is again based on the maximum edge
length as in the DES formulation and the 0.2 in the above ensures that for highly stretched meshes the grid length scale
is a fifth of that of DDES and another implication is the reduction of the eddy-viscosity in LES mode by a factor of 25 as
it is dependent quadratically upon the grid size. This is an important artifact as it improves the RANS to LES transition of
DES models.

 In engineering flows, flow characteristics of shear flows is much more encountered than that of decaying isotropic
turbulence (DIT). As the last is the basis for the calibration of the DES/DDES constant, for shear flows the Smagorinsky
constant is reduced and this is achieved by setting the constant in SDES to 0.4.
Now if we combine the above explained effect of the grid scale on the eddy viscosity with the modified constant a
reduction by a factor of nearly 60 is achieved for separated flows on stretched grids which is favorably affects the RANS
to LES transition.
April 24, 2018 83
Stress-Blended Eddy Simulation (SBES)
 The intention of the SBES methodology is to resolve the following issues (F. R. Menter 2016):

 Exhibit an asymptotic shielding of the RANS boundary layers.

 perform an explicit switch to user-specified LES model in LES region.

 Allowance of rapid ‘transition’ from RANS to LES regions Allow practitioners to be able to clearly
distinguish regions where the models run in RANS and regions where the model runs in
LES mode.

 Allow Wall-modeled LES capability once in regions of sufficient numerical resolution and an
upstream trigger into LES-mode for WMLES simulations.

April 24, 2018 84


Stress-Blended Eddy Simulation (SBES)
 SBES is not a new hybrid RANS-LES model, but a modular approach to blend existing models to achieve optimal
performance. In this sense SBES is a modular approach which allows the CFD practitioner to use a pre-selected
RANS and another pre-selected LES model instead of the mix of both formulations within one set of equations.

 The above become handy in certain fields of which the modeling sophistication is to be extended from what was
originally practiced with a specific and validated LES to include parts of the domain which can only be covered by
RANS models without having to replace the trusted LES.

 SBES model concept is built on the SDES formulation. In addition, SBES is using the shielding function to explicitly
switch between different turbulence model formulations in RANS and LES mode.

 For the general case one of the (RANS or LES) models is not based on the eddy viscosity concept the general
formulation is presented either in modeled stress tensor:

 For the case where both RANS and LES models are based on the eddy viscosity concepts, the formulation simplifies
to:

April 24, 2018 85


Scale-Resolving Simulation (SRS) in ANSYS CFD codes

April 24, 2018 86


References
 ANSYS FLUENT: Introductory FLUENT Notes
 Turbulence Modeling for CFD (David C. Wilcox)
 TENZOR Blog - https://cfdisraelblog.wordpress.com/
 Large Eddy Simulation, Dynamic Model, and Applications - Charles Meneveau (Department of Mechanical Engineering
Center for Environmental and Applied Fluid Mechanics Johns Hopkins University)
 Turbulence: Subgrid-Scale Modeling (Scholarpedia) doi:10.4249/scholarpedia.9489
 Wall-modeled large eddy simulation resource (university of Maryland)
 Turbulence Modeling Resource (NASA Langley Research Center)
 Improved two-equation k-omega turbulence models for aerodynamic flows (F. Menter 1992)
 Transition Modelling for Turbomachinery Flows (F. Menter, R.B. Langtry – ANSYS 2012)
 Development of DDES and IDDES Formulations for the k-ω Shear Stress Transport Model (F. Menter, M. Gritskevich, A.
Gritskevich, J. Schütze)
 The Scale-Adaptive Simulation Method for Unsteady Turbulent Flow Predictions. Part 1/2: Theory and Model
Description/Application to Complex Flows (F. Menter et al. 2010)
 The DESIDER Project - http://cfd.mace.manchester.ac.uk/desider/index2.html
 The State of the Art of Hybrid RANS/LES Modeling for the Simulation of Turbulent Flows (Bruno Chaouat 2017)
 Introductory Lectures On Turbulence - Physics, Mathematics and Modeling (J. M. McDonough - University of Kentucky)

April 24, 2018 87


April 24, 2018 88