We are IntechOpen,

the world’s leading publisher of

Open Access books
Built by scientists, for scientists

5,400
Open access books available
132,000
International authors and editors
160M Downloads

Our authors are among the

154
Countries delivered to
TOP 1%
most cited scientists
12.2%
Contributors from top 500 universities

Selection of our books indexed in the Book Citation Index

in Web of Science™ Core Collection (BKCI)

Interested in publishing with us?

Contact book.department@intechopen.com
Numbers displayed above are based on latest data collected.
For more information visit www.intechopen.com
Chapter

Turbulence Models Commonly

Used in CFD
John Gorman, Suvanjan Bhattacharyya, Lijing Cheng
and John Abraham

Abstract

Here we provide an overview of some of the most commonly used turbulence

models used in current CFD modeling. We compare the governing equations,
applications of use, and results between the models. Finally, we provide our own
recommendations, based on more than two decades of collaborative research.

Keywords: computational fluid dynamics, numerical simulation, turbulent flow,

laminar flow, transitional flow, RANS, LES

1. Introduction

Calculation of turbulent flows is one of the most challenging problems in all of

science and mathematics. Exact solutions of turbulence have bedeviled researchers
for many decades and it is generally appreciated that there is no closed form
solution of any fluid flow problem except the most simple laminar situations.
Despite this fact, there are ways to complete calculations with sufficient accuracy so
that engineering and design decisions can be made. The accuracy of turbulent
calculations has gradually improved with more powerful computational resources
and with improvements to numerical modeling. Here we discuss the most com-
monly used methods to simulate turbulent flow and discuss the strengths and
weaknesses of each approach. The authors believe that particular methods are more
or less appropriate for a particular situation, depending on the characteristics of the
system, the computational resources available and the accuracy requirements. In
this chapter, we pay particular attention to turbulence models that are most com-
monly used by scientists and researchers; we also provide guidance to researchers
who are pondering different turbulent-modeling approaches.

2. Turbulence and CFD

The first problems handled by CFD were relatively simple, two-dimensional,

incompressible, steady state situations that often were limited to laminar flows. To
our best knowledge, the first three-dimensional CFD simulation was not completed
until 1967 [1]. Around the same time, the very first climate models were being
constructed, for modeling the circulation of fluids around the globe. Shortly there-
after, progress became much more rapid as both computational power and model-
ing approaches advanced. A key development was the incorporation of turbulence

1
Computational Fluid Dynamics

modeling into the CFD solutions. The first turbulence models accounted for turbu-
lence effects through a concept termed the “eddy viscosity”. Essentially, the eddy
viscosity (or turbulent viscosity) reflects an apparent increase in viscosity caused by
small-scale chaotic motions in a fluid. The simulations do not attempt to actually
capture small scale turbulent motions, rather they approximate their effect with an
increase in the fluid viscosity. As we will discuss, the concept of turbulence viscos-
ity plays a central role in Reynolds Averaged Navier Stokes (RANS) models. As we
will also show, other approaches do not rely extensively on the turbulent viscosity
concept.

2.1 RANS models

The first turbulent viscosity “eddy viscosity” models were developed in the
1960s and are classified as algebraic [2, 3], one-Equation [4], or two-Equation [5–7].
The basis for two equation models was the relationship between the turbulent
viscosity and local values of the turbulent kinetic energy k and turbulent
dissipation, ε. Since this approach soon became the dominant method (even for
today), it is worthwhile to discuss it in some detail. In essence, this group of
turbulence models neglect small scale and rapid turbulent motions and use an
average flow field (timewise average values in the velocities and pressure values) to
estimate the effects of turbulence.

2.1.1 k-ε models

The first major effort to simulate turbulence in the context of CFD was the so-
called k-ε model [5, 6]. This approach utilizes the fluctuating components of the
turbulent velocity in the three coordinate directions to obtain a turbulent kinetic
energy, from:

1
02 02 02

k¼ u þv þw (1)
2

That is, k is the additional turbulent energy that results from the time-
fluctuating turbulent motions. Accompanying the turbulent kinetic energy is a
turbulent dissipation ε which can be calculated as

κ 3=2
ε¼ (2)
0:3D

for flows in pipes with diameter D [7, 8]. The connection of turbulence kinetic
energy and turbulent dissipation will be provided, following the equations of
motion. In essence, the governing equations of motion are conservation of mass,
which under steady conditions is:

∂ui
¼0 (3)
∂xi

conservation of momentum, written as:

   
∂u j ∂p ∂ ∂u j
ρ ui ¼ þ ðμ þ μt Þ j ¼ 1, 2, 3 (4)
∂xi ∂xi ∂xi ∂xi

and the closure equations for turbulence:

2
Turbulence Models Commonly Used in CFD
DOI: http://dx.doi.org/10.5772/intechopen.99784
  
∂ðui kÞ ∂ μt ∂k
ρ ¼ Pk þ Pb ρε þ μþ (5)
∂xi ∂xi σ k ∂xi

ε2
  
∂ðui εÞ ∂ μt ∂ε ε
ρ ¼ μþ þ C1 ðPk þ C3 Pb Þ C2 ρ (6)
∂xi ∂xi σ ε ∂xi k k

The turbulent viscosity is calculated from

k2
μt ¼ ρCμ (7)
ε

The Pk is the production of turbulent kinetic energy from the shear strain rate
and Pb is the production of turbulent kinetic energy from buoyancy effects. The
production of turbulent kinetic energy is obtained from the time-averaged velocity
field from:
   
∂ui ∂uj ∂ui ∂uκ ∂uκ
Pκ ¼ μt þ 3μt þ ρκ (8)
∂xj ∂xi ∂xj ∂xκ ∂xκ

The σ terms are corresponding Prandtl numbers for the transported variables.
The values of the constants and turbulent Prandtl numbers are specific to a partic-
ular k-ε model. The k-ε approach is likely the most widely used turbulent model,
even today. It is generally sufficient for flows that are wall bounded, with limited
adverse pressure gradients or separation.
Traditionally, the elements are not used to capture steep velocity and
temperature gradients near the wall. Rather, wall functions are employed to inter-
polate to the wall. Of course, the accuracy of this approach depends on the suitabil-
ity of a particular wall function to a problem. For example, wall functions often fail
when the flow experiences adverse pressure gradients and/or separation. On the
other hand, when small elements are deployed near the wall and/or when damping
equations are used to limit fluid motion in the boundary layer, integration can be
performed up to the wall. In our experience, if integration is to be performed up to
the wall (and wall function interpolation is avoided), the near-wall element should
have a size of y+1 for models that resolve the boundary layer. This guidance is not
used for models that use the law-of-the-wall to interpolate to the wall.
A popular modification of the traditional k-ε model is the RNG
(Renormalization Group) model. It was developed by [9] in an effort to handle
small flow phenomenon. The mechanism of multiple scale motions is achieved by
modifying the turbulent dissipation equation production term. In our experience, it
has somewhat better performance than the standard k-ε particularly for rotating
flows. The differences between the RNG and standard models is in the relationship
between the turbulent kinetic energy, turbulent dissipation, and turbulent viscosity.
With the RNG approach the turbulent viscosity is found from:

κ2
μt ¼ CμRNG ρ (9)
ε

and the new turbulent dissipation transport equation becomes:

  
∂ ∂ μt ∂ε ε
ðρui εÞ ¼ μþ þ ðCε1RNG Pκ Cε2RNG ρεÞ (10)
∂xi ∂xi σ εRNG ∂xi κ

With the following inputs

3
Computational Fluid Dynamics

η

η 1 4:38
Cε1RNG ¼ 1:42 (11)
ð1 þ βRNG η3 Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pκ
η¼ (12)
ρCμRNG ε

2.2 k-ω models

While the k-ε model has experienced success in computational modeling, it has
deficiencies in some situations. In particular, the k-ε model performs suitably away
from walls, in the main flow. However, it has issues in the boundary layer zone,
particularly with low Reynolds numbers. Here, Reynolds numbers refer to local
Reynolds numbers that decrease as one moves closer to the wall and the no-slip
condition exerts its influence (rather than to the Reynolds number based on mac-
roscopic dimensions such a pipe diameter or plate length).
A significant development in CFD was brought forward by the development of
k-ω model that replaced the transport equation for ε with a specific rate of turbu-
lence dissipation, ω [10]. The new equations are:
  
∂ðui kÞ ∂ μt ∂k
ρ ¼ ρPk þ ρPb ρβωk þ μþ (13)
∂xi ∂xi σ k ∂xi
  
∂ðui ωÞ ∂ μ ∂ω αω
ρ ¼ μþ t þ Pk βρω2 (14)
∂xi ∂xi σ ω ∂xi k

With a turbulent viscosity calculated as:

κ
μt ¼ ρ (15)
ω

2.3 Shear stress transport family of models

Recognizing that the k-ε and k-ω model each have strengths and weaknesses, a
new model was proposed that uses both of these approaches in a way that harnesses
their strengths [11]. This new approach, termed the Shear Stress Transport model
(SST), smoothly transitions from the k-ω model near the wall to the k-ε model in
the main flow. With the SST model, the governing equation for turbulent dissipa-
tion is recast into an ω form. The governing equations are:
  
∂ðρui kÞ ∂ μt ∂k
¼ Pk β1 ρkω þ μþ (16)
∂xi ∂xi σ k ∂xi
  
∂ðρui ωÞ ω 2 ∂ μt ∂ω 1 ∂k ∂ω
¼ α3 Pκ β2 ρω þ μþ þ 2ð1 F 1 Þρ (17)
∂xi κ ∂xi σ ω ∂xi σ ω2 ω ∂xi ∂xi

and the turbulent viscosity is found from

aρk
μt ¼ (18)
max ðaω, SF 2 Þ

As before, Pk is the production of turbulent kinetic energy and ω reflects the

specific rate of turbulent destruction. As noted earlier, the σ terms are turbulent
Prandtl numbers associated with their subscript. The function F1 is the aforemen-
tioned blending function that transfers the k-ω model near the wall to the k-ε model

4
Turbulence Models Commonly Used in CFD
DOI: http://dx.doi.org/10.5772/intechopen.99784

away from the wall from the wall. The S term is the magnitude of the shear strain
rate.
While ostensibly, the SST model is used for fully turbulent flows, it has shown
ability to capture both laminar and turbulent flow regimes [12]. However, in the
next section we discuss a set of modifications to the SST models that are specifically
designed to handled laminar/transitional/turbulent flow regimes that are
recommended.

2.4 SST transitional models

The already discussed turbulent models were largely developed based on corre-
lations of canonical fully turbulent flow situations (such as flows over flat plates,
airfoils, Falkner-Skans flows, and flows in tubes and ducts). Of course, researchers
and engineers often experience situations where the flow is partially turbulent or
other situations where the flow changes so that for part of the time it is laminar and
other times turbulent. Consider for example pulsatile flow wherein the fluid veloc-
ity changes sufficiently so that for parts of the flow period, different flow regimes
occur. There are a number of approaches to handle these situations but with respect
to the RANS models, the approaches generally utilize the concept of turbulent
intermittency. Intermittency was originally defined as the percentage of time that a
flow was turbulent. However, more recently, turbulent intermittency has been used
as a multiplier on the rate of turbulent kinetic production [13–15].
Here we will set forth two current transitional models, both based on the SST
turbulence approach. The first method involves two extra transport equations. One
for the intermittency, γ, which is a multiplier to the turbulent production. The
transport equation for turbulent intermittency is:
  
∂ðργ Þ ∂ðρui γ Þ ∂ μt ∂γ
þ ¼ Pγ,1 Eγ,1 þ Pγ,2 Eγ,2 þ μþ (19)
∂t ∂xi ∂xi σ γ ∂xi

The P and E terms are, respectively, production and dissipation of intermittency.

An additional transport equation is required for the transitional momentum
thickness Reynolds number. This added equation is:
 
∂ðρReθt Þ ∂ðρui Reθt Þ ∂ ∂Reθt
þ ¼ Pθt þ σ θt ðμ þ μt Þ (20)
∂t ∂xi ∂xi ∂xi

Together, solution to Eqs. (19) and (20) determine the local state of turbulence.
They result in an intermittency that takes values between 0 and 1. For fully
laminar flow, γ = 0 and the model reverts to a laminar solver. When γ = 1, the flow
is fully turbulent. The turbulent production then is then multiplied by the local
value of the intermittency, γ. Interested readers are invited to review the develop-
ment of this model, including implementation for problems that involve heat
transfer [16–22].
Recently, the above two-equation model was modified to reduce the two transi-
tional transport equations to a single Equation [23] and that approach was later
adapted by [24] to accurately solve for situations in confined pipe/duct/tube flows.
Essentially, Eqs. (19) and (20) are replaced by a single intermittency equation
which is:
  
∂ðρui γ Þ ∂ μt ∂γ
¼ Pγ Eγ þ μþ (21)
∂xi ∂xi σ γ ∂xi

5
Computational Fluid Dynamics

As with the two-equation approach, the intermittency factor γ will take on

values between 0 and 1. Also, as before, The P and E terms represent, respectively,
the production and destruction in local value of intermittency.
For these intermittency models, the onset of turbulence is calculated by a series
of correlation functions. In particular, a local value of the critical Reynolds number
is determined from

Re θc ¼ CTU1 þ CTU2 exp ½ CTU3 TuL F PG ðλθL ފ (22)

Eq. (22) is used to identify the location of laminar-turbulent transition. It is based

on the local value of the momentum layer thickness. The C terms are correlation
constants and are based on comparison of numerically simulated results with exper-
imentation. An important term in Eq. (22) is the local value of the mid-boundary-
layer turbulence intensity (TuL). This value is attained at the midpoint of the bound-
ary layer as an output from an empirical formulation based on experimentation.
Local production of intermittency is calculated from:

Pγ ¼ F length
ρ
S
γ
Fonset
½1 ㊠(23)

As we have already noted, the term S is the shear strain rate. A new term that
appears in Eq. (23) is the so-called onset transition term (Fonset) which is calculated
using the following set of equations.
"    3 ! #
ReV RT
Fonset ¼ MAX min ,2 max 1 ,0 ,0 (24)
2:2
Re θc 3:5

Re V ¼ ρd2w S=μ (25)

Re T ¼ ρκ=μw (26)

Similarly, the local rate of destruction of intermittency is found by:

Eγ ¼ 0:06
ρ
Ω
γ
F turb
½50γ 1Š (27)

ðRT =2Þ4
Fturb ¼ e (28)
Pκ ¼ μt
S
Ω (29)

We have already noted that these transitional turbulence models were initially
developed for external boundary layer flows (flat plate boundary layers, airfoil
flows, Falkner-Skans flows, etc.). Insofar as we have adopted them for internal
flow, some modification was required. We recommend, at least for flows through
pipes, tubes, and ducts, that the initial constants determined in [23] be replaced by
alternative values from [24].
While we recommend the above approach for solving transitional flow problems,
this area of research is also heavily studied by other researchers who have provided
alternative approaches to handle such flows. We cite them here for readers who are
interested in those alternative but complementary viewpoints [25–33].

2.5 Reynolds-stress models

Reynolds stress models (RSM) are quite different from the RANS approach that
was just discussed. For RSMs, transport equations are used for all components of the
Reynolds stress tensor and an eddy viscosity is not utilized. These models are
expected to be superior for situations with non-isotropic turbulence and flows with

6
Turbulence Models Commonly Used in CFD
DOI: http://dx.doi.org/10.5772/intechopen.99784

significant components of transport in three directions. There are a number of RSM

versions, some of which will be discussed here. The so-called SSG-RSM model
employed here utilizes the following momentum transport equation:

∂p0
 
∂u j ∂  ∂ ∂u j ∂ui ∂ 
ρ þρ ui u j ¼ þμ þ ρ ui u j þ B j (30)
∂t ∂xi ∂x j ∂xi ∂xi ∂xi ∂xi

The second-to-last term on the right-hand side represents the Reynolds stresses.
There is a pseudo-pressure term p’ that is calculated from the local static pressure p
and local velocity gradient from the following expression.

2 ∂uk
p0 ¼ p þ μ (31)
3 ∂xk

The Reynolds stresses are calculated by a collection of six equations for all
directional possibilities. The transport equations for Reynolds stresses are:
" ! #
2
∂u j ui ∂  ∂ 2 k ∂ui u j 2
ρ þρ uk ui u j μ þ ρCs ¼ Pij δij ρε þ Φij þ Pij,b
∂t ∂xk ∂xk 3 ε ∂xk 3
(32)

We note that a turbulence dissipation term, ε, appears in Eq. (32) and it has to be
solved from its own transport equation. We refer readers to [34, 35] for more
details.
A modification to the above is realized from the Baseline RSM (BSL RSM)
model. It differs from the SSG RSM in that the transport equation for ε is replaced
by a transport equation for ω. The new equation is:
  
∂ω ∂ðuk ωÞ ∂ μt ∂ω ω 2ρ 1 ∂k ∂ω
ρ þρ ¼ μþ þ α 3 Pk β3 ρω2 þ ð1 F1 Þ
∂t ∂xk ∂xk σ ω3 ∂xk k σ ω2 ω ∂xk ∂xk
þ Pωb
(33)

This approach blends between two different models that are used near the wall
and alternatively away from the wall. The modeling is accomplished using a
weighting function, similar to the SST:

ϕ3 ¼ F 1
ϕ1 þ ð 1 F1 Þ
ϕ2 (34)

Where the symbols ϕ correspond to any particular transport variable in the near
wall and far wall regions. Various constants change their values in the two regions,
so that:
The constants near the wall:

σk ¼ σω ¼ 2, β ¼ 0:075, α ¼ 0:553 (35)

The constants away from the wall:

σk ¼ 1, σω ¼ 1:168, β ¼ 0:0828, α ¼ 0:44 (36)

The last RSM version to be discussed is the Explicit Algebraic RSM (EARSM).
This approach includes a non-linear relationship between the local values of the
Reynolds stresses and the vorticity tensors. It is focused on flows with secondary

7
Computational Fluid Dynamics

motions and curvature [36]. The local values of the Reynolds stresses are calculated
using an anisotropy tensor which is based on algebraic equations [36]. This is
contrasted with RSM approaches that solve for the Reynolds stress components
using differential transport equations. The approach is to use higher order terms for
many of the flow phenomena. It was designed to handle secondary flow situations
and flows with extensive curvature and rotation. The governing equations are
complex and lengthy and for brevity sake, we refer interested readers to [36].

2.6 Scale adaptive models

So far, we have presented RANS-based models that perform conservation cal-

culations at each grid element. If turbulence is present, the impact of turbulence
appears via the eddy viscosity. Traditionally users either a priori specify that the
flow is laminar (so no eddy viscosity included) or the flow is turbulent (in which
case an eddy viscosity is determined and applied throughout the flow field). The
recent development of transitional modeling frees the researcher from having to a
priori predict the level of turbulence. With transitional modeling, the numerical
code automatically reverts to laminar flow in areas with low Reynolds numbers and
also automatically becomes a turbulent model in areas where the Reynolds number
is larger.
Regardless of the method that is selected, the coupled equations are solved for
each computational element and the turbulent viscosity is applied to the fluid in the
element under consideration.
In contrast to this approach, there is another major group of computational
techniques that are termed “scale adaptive models”. These are models that resolve
part of the turbulent motions but model flow features that are smaller than the
element size. Since there is less modeling and more actual resolution of fluid
motion, one might expect the scale-adaptive models to be more accurate than
RANS; and there are cases where that is so (particularly for free shear flows,
swirling flows, boundary layer separation, and jets). However, the RANS approach
can be more accurate than scale-adaptive methods in some situations, including
wall bounded flows. Also, RANS is less computationally expensive because the eddy
viscosity provides the link to the time-averaged flow field and the local turbulence
with a very simple calculation. In fact, for even problems of modest complexity,
scale adaptive models are more time consuming.
There are a number of established and new Scale-Adaptive Models that are used
in CFD simulations. We will not be exhaustive in this section by covering all the
existing models, rather we will focus on some of the models we think are most
useful and representative. Interested readers are directed to an excellent compre-
hensive discussion provided by [34, 37].

2.6.1 Scale-adaptive SST models

One of the primary decisions that models are faced with is whether to perform
calculations in steady or unsteady mode. Typically with numerical simulation,
unsteadiness is driven by either timewise changes in boundary conditions or it is
related to unsteady phenomena that occur in an otherwise steady scenario. A classic
example is the Karmen Vortex Street that occurs in a wake region of a blunt object.
Figure 1, shown below, illustrates this phenomenon.
Researchers have often conjectured that if a RANS model is performed with
sufficiently small elements and time steps, the unsteady features of the flow would
naturally be resolved. But in fact, this is not true. It is important to note that steady
state calculations using RANS models will often provide very accurate information

8
Turbulence Models Commonly Used in CFD
DOI: http://dx.doi.org/10.5772/intechopen.99784

Figure 1.
Unsteady wake region, even though oncoming flow is steady state.

about averaged quantities (like drag), these simulations will miss details in the
rapidly fluctuating downstream wake region. This issue was explored in depth
in [35] where time-averaged results of drag obtained from unsteady RANS
simulations were compared with calculations from steady RANS calculations
(using the SST transitional model that was previously described). It was found
that the steady state calculations were able to accurately capture drag forces
but were only partially adept at capturing vortex movement in the downstream
wake region.
With this discussion as background, it is now time to turn attention to the
governing equations of scale-adaptive RANS models. The model to be discussed
here uses the SST approach for the underlying governing equations (in the
literature it is often termed the SAS-SST model). The scale-adaptive approach
modifies the ω transport equation based on [37]. In particular, a new transport
equation is presented that incorporates the turbulent length scale L and is set
forth here:
pffiffiffi
Φ¼ kL (37)

and
 2 !  
∂Φ ∂U i Φ ΦPk Lt ∂ μt ∂Φ
þρ ¼ C1 C2 ρC3 k þ (38)
∂t ∂xi k Lk ∂xi σ Φ ∂xi

Values of the various constants can be found in [34, 37] and are not repeated
here for brevity. The term Lt is a novel modification; it refers to the von Karmen
length scale. Figures 2 and 3 are provided that show a comparison of downstream
wake regions for an unsteady RANS calculation using the SST model (Figure 2) and
a simulation using the scale-adaptive SST modification. Results are obtained from
[34]. It can be seen that the standard SST model does capture a periodic release of
eddies from the downstream side of a circular cylinder (shown in blue). In both
images, the flow is left-to-right. The color legend is keyed to the local values of the
turbulent length scale. Clearly the scale-adaptive approach provides a much wider
range of turbulent eddy sizes.

9
Computational Fluid Dynamics

Figure 2.
Calculations of turbulent length scale for flow over a circular cylinder, based on an unsteady SST model.

Figure 3.
Calculations of turbulent length scale for flow over a circular cylinder, based on a scale-adaptive unsteady SST
model.

2.6.2 LES WALE model

Another common approach to dealing with these types of problems is based on

the so-called “large eddy simulation”. To the best knowledge of the authors, the first
articulation of a LES model was [38] and the models have been updated in the
intervening decades. Here we focus on one popular and current LES method (the
Wall-Adaptive Local Eddy, or WALE LES model). The general processes of LES
modeling are the same, regardless of which variant is used. LES models involve the
filtering of eddies that are smaller than the size of the computational elements. The
algorithm incorporates an eddy viscosity for flow scales that are not resolved.
For this model, the tensor-form of the Navier Stokes equations is:

∂ρui u j
    
∂ ∂p ∂ ∂ui ∂u j ∂τij
ðρui Þ þ ¼ þ μ þ þ (39)
∂t ∂xi ∂x j ∂xi ∂x j ∂xi ∂xi

10
Turbulence Models Commonly Used in CFD
DOI: http://dx.doi.org/10.5772/intechopen.99784

where τij is the small-scale stress defined as

1
τij ¼ ρui u j þ ρui u j ¼ 2μsgs Sij þ δij τkk (40)
3

And the Sij term indicates the strain rate tensor for large scale motions. The
small-scale eddy viscosity μsgs is found from

3=2
Sdij Sdij
μsgs ¼ ρðCw ΔÞ2 
5=4 (41)
5=2
Sij Sij þ Sdij Sdij

The term Cw is a constant and the symbol ∆ = (element volume)1/3. The tensor
Sijd is calculated from the strain-rate and vorticity tensors, as shown here

1 
Sdij ¼ Sik Skj þ Ωik Ωkj δij Smn Smn Ωmn Ωmn (42)
3

And the vorticity tensor Ωij is defined as

 
1 ∂ui ∂u j
Ωij ¼ þ (43)
2 ∂x j ∂xi

3. Results from various CFD model calculations

Now that the main CFD models have been presented, we turn attention to
comparisons of the results from different models. There are comparisons available
in [7, 8, 34, 35, 37, 39–46] and a very small subset of those comparisons will be
provided here. We have selected the classic problem of flow over a square blockage.
This canonical problem has the features that elucidate the strengths and weaknesses
of the particular models. For instance, some important parameters relate to the
time-averaged interactions between the fluid and the solid structure (drag force).
Also, there are significant unsteady phenomena, particularly in the wake region that
provide a challenging test for the models. In addition, this is a problem with
extensive experimental work that will serve as the basis for evaluating the results.
To begin we refer to Figure 4 which shows the solution domain (similar to [35]).

Figure 4.
Geometry for flow over a square cylinder.

11
Computational Fluid Dynamics

Figure 5.
Computational mesh used for square cylinder simulation.

Figure 6.
Drag coefficients for flow over a square cylinder and comparison with experiments. Reynolds numbers range
from 1 to 10,000,000.

12
Turbulence Models Commonly Used in CFD
DOI: http://dx.doi.org/10.5772/intechopen.99784

Figure 7.
Drag coefficients for flow over a square cylinder and comparison with experiments, Reynolds numbers ranging
between 10 and 10,000.

A number of computational meshes were used and an example mesh is shown in

Figure 5. The images are provided in a series of increasing magnification. Image (a)
is the most global view, part (b) is focused on the square obstruction, and image (c)
reveals details of the elements in the near-wall region, near a corner of the cylinder.
With this mesh, we present results for a large number of computational
methods. We note here that in reality appropriate meshes may differ depending on
the turbulence model that is used. For instance, a mesh that is suitable for a k-w
simulation may not be appropriate for SST, and vice versa. We recommend that
mesh independent studies be carried out for each turbulence mode that is
employed. The results, set forth in Figures 6 and 7, provide the drag coefficient on
the square cylinder (large aspect ratio). Each model has its own color. Literature-
based values from experiments are also included (shown as gray x symbols).
In the above calculations, which were first set forth in [35], the SST and
transitional-SST models were most accurate (when compared with existing experi-
ments) for calculating the drag coefficient. On the other hand, since these
approaches were RANS, they lose some local detail and flow structure. For example,
in Figure 8 which is provided below, we show velocity vectors, overlaid atop a
velocity contour image. It is evident from the upper part of the figure that there are
the expected stagnation locations at the leading edge, and in the wake region. There
is also a slow-moving recirculation zone above and below the cylinder that are a
result of flow separation at the leading corners. However, the lower images show a
focus on the flow patterns at the leading edge. It is seen that with the SST RANS
model, there are no small-scale eddies at this location. But for the LES model, there
are two LES results that are obtained at two different instances in time. These
sequential images show the time-varying flow field. While a RANS model like the
SST is excellent for full-body drag, it does not capture some small flow structures.
Researchers thus need to consider their computational needs before selecting a CFD
model.

13
Computational Fluid Dynamics

Figure 8.
(upper image) velocity contour and vectors for SST model and (lower image) side-by-side comparison of
leading-edge flow for SST and WALE LES models.

The last result to be presented is shown in Figure 9. There, instantaneous

results are displayed for the SST model. There, clearly, the unsteady nature of
flow in the downstream wake region are evident. If the simulation of Figure 9
was carried out with a steady state SST solver, there would still be timewise
changes in the flow field but they would have a different frequency than the
unsteady calculations.
In order to elucidate the iteration-by-iteration fluctuations in drag that result
from a steady state solver (compared to an unsteady simulation), Figure 10 is
prepared. This figure shows the timewise (iteration wise) fluctuations in drag force
on the square cylinder first with a steady state SST solution and then with a truly
unsteady solution. The steady state results are calculated using a “false transient”
approach wherein the algorithm steps forward to new iterations using a non-
physical time. The figure has two call outs that provide focus on different parts of
the graph. The important conclusion is that the average value of unsteady fluctua-
tions of drag obtained by the steady state algorithm are an excellent match that that

14
Turbulence Models Commonly Used in CFD
DOI: http://dx.doi.org/10.5772/intechopen.99784

Figure 9.
Streamline patterns and velocity contours for Re = 100,000 flow over a square cylinder. Images at a sequence of
time instances, using SST model.

attained from the unsteady calculations. On the other hand, the period is very
different between the two.

4. Concluding remarks

This chapter has presented a brief overview of a large number of turbulence

models. While there is no “correct” turbulence model, there are models that are
better suited for particular situations.
For flows that are truly laminar with no regions of intermittency or turbulence, a
laminar solver can be used. However, if there is a potential for any turbulent flow,
caution is warranted. For flows that are fully turbulent, particularly wall bounded
flows, the SST model is recommended. In our experience it is more able to capture
flow phenomena compared to other RANS models. It also has excellent perfor-
mance for a wide range of thermal-transport situations.
If regions of mixed flows (laminar/transitional/turbulent) are expected, of if the
flows might change in time (pulsatile flows for example), the SST transitional

15
Computational Fluid Dynamics

Figure 10.
SST solution that began as steady state and then was changed to unsteady.

model is recommended. This new approach is rapidly becoming more common in

the CFD community and could replace fully turbulent models in the future.
For situations where small scale and short-lived flow must be captured, we
recommend the scale-adaptive SST model or the LES model. They are more com-
putationally expensive but the scale adaption enables small features to be calcu-
lated. We also direct readers to two further excellent resources [47, 48] for more in
depth discussion.

Author details

John Gorman1, Suvanjan Bhattacharyya2, Lijing Cheng3 and John Abraham1*

1 University of St. Thomas, St. Paul, Minnesota, USA

2 Department of Mechanical Engineering, Birla Institute of Technology and Science

Pilani, Pilani, Rajasthan, India

3 Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

*Address all correspondence to: jpabraham@stthomas.edu

© 2021 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms
of the Creative Commons Attribution License (http://creativecommons.org/licenses/
by/3.0), which permits unrestricted use, distribution, and reproduction in any medium,
provided the original work is properly cited.

16
Turbulence Models Commonly Used in CFD
DOI: http://dx.doi.org/10.5772/intechopen.99784

References

[1] Hess J, Smith A. Calculation of [10] Wilcox D. Multiscale model for

potential flow about arbitrary bodies. turbulence flows. AIAA Journal. 1988;
Progress in Aerospace Sciences. 1967;8: 26:1311-1320.
1-138.
[11] Menter F. Two-equation eddy-
[2] Smith A, Cebeci T. Numerical viscosity turbulence models for
simulation of the turbulent boundary engineering applications. AIAA Journal.
layer equations, Douglass Aircraft D1994;32:1598-1605.
Division Report DA 33735, 1967.
[12] Sparrow E, Tong J, Abraham J. Fluid
[3] Baldwin B, Lomax H. Thin layer flow in a system with separate laminar
approximation and algebraic model for and turbulent zones. Numerical Heat
separated turbulent flows. AIAA paper Transfer A. 2008;53:341-353.
no. 78-257. 1978.
[13] Menter F, Esch T, Kubacki S.
[4] Spalart P, Allmaras S. A one-equation Transition Modelling Based on Local
turbulence model for aerodynamic Variables. In: 5th Int. Symposium on
flows. AIAA paper no. 92-0439, 1992. Engineering Turbulence Modeling and
Measurements; 2002: Mallorca, Spain.
[5] Jones W, Launder B. The prediction
of laminarization with a two equation [14] Menter F, Langtry R, Likki S,
model of turbulence. International Suzen Y, Huang P, Volker S. A
Journal of Heat and Mass Transfer. 1972; Correlation-Based Transition Model
15:301-314 Using Local Variables, Part i – Model
Formulation. In: Proceedings of ASME
[6] Launder B, Spalding D. The
Turbo Expo Power for Land, Sea, and
numerical computation of turbulent Air; 14–27 June 2004: Vienna, Austria.
flows. Computer Methods in Applied
Mechanics and Engineering. 1974;3:
[15] Menter F, Langtry R, Likki S,
269-289.
Suzen Y, Huang P, Volker S. A. In:
Proceedings of ASME Turbo Expo
[7] Gorman J, Sparrow E, Abraham J,
Power for Land, Sea, and Air. 14–17
Minkowycz W. Evaluation of the
June 2004: Vienna, Austria.
efficacy of turbulence models for
swirling flows and the effect of
turbulence intensity on heat transfer. [16] Abraham J, Tong J, Sparrow E.
Numerical Heat Transfer Part B: Breakdown of laminar pipe flow into
Fundamentals. 2016;70:485-602 transitional intermittency and
subsequent attainment of fully
[8] Sparrow E, Gorman J, Abraham J, developed intermittent or turbulent
Minkowycz W. Validation of turbulence flow. Numerical Heat Transfer B. 2008;
models for numerical simulation of fluid 54:103-115.
flow and convective heat transfer.
Advances in Heat Transfer. 2017;49: [17] Abraham J, Sparrow E, Tong J, Heat
397-421. transfer in all pipe flow regimes –
Laminar, transitional/intermittent, and
[9] Yakhot V, Orszag S, Thangam S, turbulent. International Journal of Heat
Gatski T, Speziale C. Development of and Mass Transfer. 2009;52:557-563.
turbulence models for shear flows by a
double expansion. Physics of Fluids. 4; [18] Minkowycz W, Abraham J,
1510;1992. Sparrow E. Numerical simulation of

17
Computational Fluid Dynamics

laminar breakdown and subsequent [26] Tam L, Ghajar A. Effect of inlet

intermittent and turbulent flow in geometry and heating on the fully
parallel plate channels: Effects of inlet developed friction factor in the
velocity profile and turbulence transition region of a horizontal tube.
intensity. International Journal of Heat Exp Therm. Fluid Sci. 1997;15:52-64.
and Mass Transfer. 2009;523:
4040-4046. [27] Tam H, Tam H, Ghajar A, Ng, W,
Wong I, Leong K, Wu C. The effect of
[19] Lovik R, Abraham J, Minkowycz W, inner surface roughness and heating on
Sparrow E. Laminarization and friction factor in horizontal micro-
turbulentization in a pulsatile pipe flow. tubes. Proceedings of ASME-JSME-
Numerical Heat Transfer A. 2009;56: KSME joint fluids Engineering. 2011;
861-879. 24-29.

[20] Abraham J, Sparrow E, Tong J, [28] Tam H, Tam L, Ghajar A. Effect of

Bettenhausen D. Internal flows which inlet geometries and heating on the
transist from turbulent through entrance and fully-developed friction
intermittent to laminar. International factors in the laminar and transition
Journal of Thermal Sciences. 2010;49: regions of a horizontal tube. Exp.
256-263. Therm. Fluid Sci. 2013;44:680-696.

[21] Abraham J, Sparrow E,

[29] Tam H, Tam L, Ghajar A, Sun C,
Minkowycz W. Internal-flow Nusselt
Leung H. Experimental investigation of
numbers for the low-Reynolds number
the single-phase friction factor and heat
end of the laminar-to-turbulent
transfer inside the horizontal internally
transition regime. International Journal
micro-fin tubes in the transition region.
of Heat and Mass Transfer. 2011;54:
Proceedings of ASME-JSME-KSME joint
584-588.
fluids Engineering. 2011;24-29.
[22] Gebreegziabher T, Sparrow E,
Abraham J, Ayorinde E, Singh T. High- [30] Tam H, Tam L, Ghajar A, Sun C,
frequency pulsatile flows encompassing Lai W. Experimental investigation of
all flow regimes. Numerical Heat single-phase heat transfer in a
Transfer A. 2011;60:811-826. horizontal internally micro-fin tube
with three different inlet configurations.
[23] Menter F, Smirnov P, Liu T, Proceedings of the ASME 2012 Summer
Avancha R. A one-equation local Heat Transfer, 2012.
correlation based transition model.
Flow, Turbulence and Combustion. [31] Tam H, Tam L, Ghajar A, Sun C,
2015;95:583-619. Leung H. 2011. Experimental
investigation of the single-phase friction
[24] Abraham J, Sparrow E, Gorman J, factor and heat transfer inside the
Zhao Y, Minkowycz W. Application of horizontal internally micro-fin tubes in
an intermittency model for laminar, the transition region. Proceedings of
transitional, and turbulent internal ASME-JSME-KSME Joint Fluids
flows. Journal of Fluids Engineering. Engineering 2011;24-29.
2019;141: paper no. 071204.
[32] Everts M, Meyer J. Relationship
[25] Ghajar A, Madon K. Pressure drop between pressure drop and heat transfer
measurements in the transition region of developing and fully developed flow
for a circular tube with three different in smooth horizontal circular tubes in
inlet configurations. Exp Therm. Fluid the laminar, transitional, quasi-
Sci. 1992;5:129-135. turbulent and turbulent flow regimes.

18
Turbulence Models Commonly Used in CFD
DOI: http://dx.doi.org/10.5772/intechopen.99784

International Journal of Heat and Mass enhancement in tube flow with twisted
Transfer. 2018;117:1231-1250. tape insert, Progress in Computational
Fluid Dynamics, an International
[33] Everts M, Meyer J. Heat transfer of Journal, 2017; 17(3): 193-197.
developing and fully developed flow in
smooth horizontal tubes in the [42] S Bhattacharyya, H Chattopadhyay,
transitional flow regime. International S Bandyopadhyay, S Roy, A Pal, S
Journal of Heat and Mass Transfer. Bhattacharjee, Experimental
2018;117:1331-1351. investigation on heat transfer
enhancement by swirl generators in a
[34] Menter F. Best Practices: Scale- solar air heater duct, International
Resolving Simulations in ANSYS CFD, Journal of Heat and Technology, 2016;
Published by ANSYS, 2015. 34(2): 191-196.

[35] Olsen L, Abraham J, Cheng L, [43] M W Alam, S Bhattacharyya, B

Gorman J, Sparrow E, Summary of Souayeh, K Dey, F Hammami, M
forced-convection fluid flow and heat Rahimi-Gorji, R Biswas, CPU heat sink
transfer for square cylinders of different cooling by triangular shape micro-pin-
aspect ratios ranging from the cube to a fin: Numerical study, International
two-dimensional cylinder. Advances in Communications in Heat and Mass
Heat Transfer. 2019;51:351-457. Transfer, 2020;112:104455

[36] Wallin S, Johansson A. An explicit [44] S Bhattacharyya, H Chattopadhyay,

algebraic Reynolds stress model for Computational of studies of heat
incompressible and compressible transfer enhancement in turbulent
turbulent flows. Journal of Fluid channel flow with twisted strip inserts,
Mechanics. 2000;403:89-132. proceedings of CHT-15. 6th
International symposium on advances in
[37] Menter F, Egorov Y. Scale-adaptive computational heat transfer, 2015:
simulation method for unsteady flow 209-220. DOI: 10.1615/ICHMT.2015.
predictions. Part 1 Theory and model IntSympAdvComputHeatTransf.160
description. Journal of Flow Turbulence
and Combustion. 2010:85:113-138. [45] S Bhattacharyya, H Chattopadhyay,
R Biswas, DRE Ewim, Z Huan,
[38] Smagorinsky J. General circulation Influence of inlet turbulence intensity
experiments with the primitive on transport phenomenon of modified
equations. Monthly Weather Reviews. diamond cylinder: A numerical study,
1963;91:99-165. Arabian Journal for Science and
Engineering, 2020; 45(2):1051-1058.
[39] Tong J, Abraham J, Tse J,
Minkowycz W, Sparrow E. New archive [46] S Bhattacharyya, S Das, A Sarkar, A
of heat transfer coefficients from square Guin, A Mullick, Numerical simulation
and chamfered cylinders in cross flow. of flow and heat transfer around
International Journal of Thermal hexagonal cylinder, International
Sciences. 2016;105:218-223. Journal of Heat and Technology, 2017;35
(2):360-363
[40] Tong J, Abraham J, Tse J, Sparrow E.
Impact of chamfer contours to reduce [47] Turbulence Modeling Resource
column drag. Engineering and (NASA). https://turbmodels.larc.nasa.
Computational Mechanics. 2015;168:79-88. gov

[41] S Bhattacharyya, H Chattopadhyay, [48] ANSYS CFX Theory Guide,

AC Benim, Simulation of heat transfer Published 2009, ANSYS, Inc.

19