Presented

by
Dr Ceri Morris
Computational Methods
Week 9

Turbulence

2020
What is turbulence?
• Flow exhibits some irregular, random, fluctuations
• Flow properties (i.e. pressure, velocity) vary in time and
space in a random manner
• Fully turbulent flow can be considered to consist of
eddies of different sizes
• An ‘eddy’ preludes precise definition, but it is conceived
to be a turbulent motion, localized over a region of size
l, that is at least moderately coherent over this region
What is turbulence?
• Some examples of simple turbulent flows are a jet entering a
domain with stagnant fluid and the wake behind objects such
as cylinders.
• Such flows are often used as test cases to validate the ability
of CFD software to accurately predict fluid flows.
What is turbulence?
Why is turbulence important?
• Most flows in nature and engineering are turbulent
o Most internal flows
₋ Pipes, heat exchangers, pumps, compressors,
turbines, engines etc.
o Most external flows
₋ Around and in wakes of aircraft, ships, trains, cars,
golf balls, footballs etc.
o In natural water flows
₋ Rivers, lakes and oceans
 Why is turbulence important?
One example:
• A DC-9 taking off in a snowstorm at
Denver encounters trouble, stalls and
flips over, killing twenty-eight people

• 2 explanations, both involve

turbulence

• Unruly vortex of air, spun up in the

wake of a jet landing on a near-by • A few grains of ice reported on the
runway failed to dissipate; other air plane’s wings after final de-icing.
currents nudged it into path of DC-9 Built up a turbulence powerful
and formed a fatal clot in the plane’s enough to bring the giant jet down
compressors.

http://cdn-www.airliners.net/aviation-photos/photos/4/1/1/0526114.jpg
 Kolmogorov Energy Spectrum
• Large eddies produced by
objects around or inside
which fluid flows
• The large eddies are
unstable and break up,
transferring their energy to
somewhat smaller eddies
• Smaller eddies undergo a
similar break-up process
and transfer their energy to
yet smaller eddies
• Energy cascade continues until the molecular viscosity is
effective in dissipating the kinetic energy as heat
 Kolmogorov Energy Spectrum
• Energy cascade, from large scale to small scale eddies
• Length scales:
o Largest eddies. Integral length scale
o Length scales at which turbulence is isotropic. Taylor microscale
o Smallest eddies. Kolmogorov length scale
Kolmogorov Energy Spectrum
• Kolmogorov’s theory describes how energy is
transferred from larger to smaller eddies; how
much energy is contained by eddies of a given size;
and how much energy is dissipated by eddies of
each size
• Hypothesised that small eddies are isotropic and
independent of geometry
• Flow dynamics are dominated by large eddies
• Shown empirically to be correct
Predicting Turbulence in CFD
• Direct Numerical Simulation (DNS)

• Large Eddy Simulation (LES)

• Reynolds Averaged Navier-Stokes (RANS)

 Direct Numerical Simulation (DNS)
• Navier-Stokes equations numerically solved, nothing modelled
• Whole range of turbulence scales must be resolved from smallest
Kolmogorov scale up to largest (integral)
dissipative (Kolmogorov) integral scale
• Requires a very fine mesh and very small time-step
• Result is exact, 3D, transient behaviour
• Great for very simple flows, but computationally intensive, very time-
consuming, huge amounts of data generated
• The overall cost of the computational effort is proportional to Re3
• Not suited to industrial applications with CPU resources available today
• Used for studying and understanding turbulence and for the verification
of RANS and LES turbulence models
Large Eddy Simulation (LES)
• Solves large eddies and models smaller eddies using a sub-grid
scale model
• Mathematically, velocity field is filtered into a resolved and sub-grid
part. The resolved part of the field represents the "large" eddies,
while the sub-grid part of the velocity represents the "small scales"
• Requires 3-D, transient modelling
• Requires spatial and temporal resolution of scales in “inertial
subrange”
• LES is a transient turbulence model that falls midway between
RANS and DNS models
• Main advantage is improved accuracy compared to RANS models
• Main disadvantage is the large computer time requirement
• Large data sets generated
Presented
by
Dr Ceri Morris
Computational Methods
Week 9

Reynolds Averaged Navier-Stokes

(RANS)

2020
RANS
• The Navier-Stokes equations provide a complete description of
turbulence – the problem is it is too expensive to solve and so
we approximate the turbulent terms
• For most engineering applications it is unnecessary to resolve
the details of the turbulent fluctuations. We only need to know
how turbulence affects the mean flow
• The objective of turbulence modelling is to develop equations
that will predict the time averaged velocity, pressure, and
temperature fields without calculating the complete turbulent
flow pattern. This saves us a lot of work
• In particular we need expressions for the Reynolds stresses
 RANS
• We use time-averaged equations that give us conservation
equations for the mean (average) values. These will successfully
model turbulent flow which overall, on average, does not vary
with time
• Important to understand: the time averaged flow pattern is a
statistical property of the flow
o It is not an existing flow pattern
o It does not usually satisfy the steady Navier-Stokes equations
o The flow never actually looks that way
• For a turbulence model to be useful it:
o must have wide applicability
o be accurate
o simple
o economical to run
RANS
• The time averaged equations contain six additional
unknowns in the momentum equations
• 𝜌𝑢′ 𝑢′ , 𝜌𝑢′ 𝑣 ′ , 𝜌𝑢′ 𝑤 ′ , 𝜌𝑣 ′ 𝑣 ′ , 𝜌𝑣 ′ 𝑤 ′ , 𝜌𝑤 ′ 𝑤 ′
• More unknowns than equations
• A turbulence model is required to close the system of mean
flow equations
• The main task of turbulence modeling is to develop
computational procedures of sufficient accuracy and
generality for engineers to be able to accurately predict the
Reynolds stresses and the scalar transport terms
• Turbulent flows are usually quite complex, and there are no
simple formulae for these additional terms
What is Turbulence?
𝑢

• Flow exhibits some

irregular,
irregular, random,
random, 𝑢′
fluctuations
fluctuations
𝑢 = 𝑢ത + 𝑢′
• Flow properties (i.e.
pressure, velocity) t
vary in time and 𝑢 = instantaneous velocity component
space in a random 𝑢ത = time average velocity component
space
𝑢 ′ = fluctuating velocity component
manner 1 𝑡+∆𝑡
𝑢ഥ′ = 0 i.e. ‫׬‬ 𝑢′ 𝑑𝑡 = 0
∆𝑡 𝑡

Likewise 𝑣 = 𝑣ҧ + 𝑣 ′ 𝑤=𝑤 ഥ + 𝑤′
𝑝 = 𝑝ҧ + 𝑝′ 𝑓𝑥 = 𝑓ഥ𝑥 + 𝑓𝑥 ′
 Incompressible Turbulent Flow
𝜕𝑢 𝜕𝑢2 𝜕𝑢𝑣 𝜕𝑢𝑤 1 𝜕𝑝 𝜕2𝑢 𝜕2𝑢 𝜕2𝑢
+ + + = 𝑓𝑥 − +ν + +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜌 𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧 2

Local Advective Body Pressure Stresses

acceleration acceleration Force Gradient

𝑢 + 𝑢′ ) 𝜕(ത
𝜕(ത 𝑢 + 𝑢′ )(ത𝑢 + 𝑢′ ) 𝜕(ത 𝑢 + 𝑢′ )(𝑣ҧ + 𝑣 ′ ) 𝜕(ത
𝑢 + 𝑢′ )(𝑤ഥ + 𝑤′)
+ + +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
′ 1 𝜕(𝑝ҧ + 𝑝′ ) 𝜕 2 (ത
𝑢 + 𝑢′ ) 𝜕 2 (ത𝑢 + 𝑢′ ) 𝜕 2 (ത
𝑢 + 𝑢′ )
= 𝑓ഥ𝑥 + 𝑓𝑥 − +ν + +
𝜌 𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧 2
 Incompressible Turbulent Flow
𝜕(𝑢ത + 𝑢′ ) 𝜕(ത 𝑢𝑢ത + 2𝑢′ 𝑢ത + 𝑢′ 2 ) 𝜕(ത 𝑢𝑣ҧ + 𝑢′ 𝑣ҧ + 𝑣 ′ 𝑢ത + 𝑢′ 𝑣 ′ ) 𝜕(ത 𝑢𝑤 ഥ + 𝑢′ 𝑤
ഥ + 𝑤 ′ 𝑢ത + 𝑢′ 𝑤 ′ )
+ + +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
′ 1 𝜕(𝑝ҧ + 𝑝′ ) 𝜕 2 (ത
𝑢 + 𝑢′ ) 𝜕 2 (ത 𝑢 + 𝑢′ ) 𝜕 2 (ത 𝑢 + 𝑢′ )
= 𝑓ഥ𝑥 + 𝑓𝑥 − +ν + +
𝜌 𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧 2

𝑡+∆𝑡 𝑡+∆𝑡
1 1
𝑢 + 𝑢′ ) =
(ത න 𝑢ത + 𝑢′ 𝑑𝑡 But 𝑢ത is constant and: න 𝑢′ 𝑑𝑡 = 0
∆𝑡 ∆𝑡
𝑡 𝑡
𝑢 + 𝑢′ ) = 𝑢ത
∴ (ത

𝑡+∆𝑡
1 2
𝑢𝑢ത + 2𝑢′ 𝑢ത + 𝑢′ 2 ) =
(ത 𝑢𝑢ത + 2𝑢′ 𝑢ത + 𝑢′ )𝑑𝑡
න (ത
∆𝑡
𝑡
But 𝑢ത 𝑢ത is constant and:
𝑡+∆𝑡 𝑡+∆𝑡
1 2ത
𝑢 𝑡+∆𝑡
න 2𝑢′ 𝑢ത 𝑑𝑡 = න 𝑢′ 𝑑𝑡 = 0 1
∆𝑡 ∆𝑡 න 𝑢′ 𝑢′ 𝑑𝑡 = 𝑢′ 𝑢′ ≠ 0
𝑡 𝑡 ∆𝑡
𝑡

∴ 𝑢ത 𝑢ത + 2𝑢′ 𝑢ത + 𝑢′ 2 = 𝑢ത 𝑢ത + 𝑢′ 𝑢′
Reynolds-Averaged Navier-Stokes (RANS)

𝜕𝑢ത 𝜕𝑢ത 𝑢ത 𝜕𝑢ത 𝑣ҧ 𝜕 𝑢ത 𝑤

ഥ 1 𝜕𝑝ҧ 𝜕 2 𝑢ത 𝜕 2 𝑢ത 𝜕 2 𝑢ത 𝜕𝑢′ 𝑢′ 𝜕𝑢′ 𝑣 ′ 𝜕𝑢′ 𝑤 ′
+ + + = 𝑓ഥ𝑥 − +ν + + − + +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜌 𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧 2 𝜕𝑥 𝜕𝑦 𝜕𝑧

𝜕𝑢ത 𝜕𝑢ത 𝜕𝑣ҧ 𝜕 𝑢ത 𝜕𝑤

ഥ 𝜕𝑢ത 𝜕𝑢ത 𝜕𝑣ҧ 𝜕𝑤
ഥ 𝜕𝑢ത 𝜕𝑢ത 𝜕𝑢ത
𝑢ത + 𝑢ത + 𝑢ത + 𝑣ҧ + 𝑢ത +𝑤
ഥ = 𝑢ത + + + 𝑢ത + 𝑣ҧ +𝑤
ഥ
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

From Continuity = 0

𝜕 𝑢ത 𝜕𝑢ത 𝜕 𝑢ത 𝜕𝑢ത Reynolds Stresses

+ 𝑢ത + 𝑣ҧ +𝑤
ഥ
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
1 𝜕𝑝ҧ 1 𝜕 𝜕𝑢ത 1 𝜕 𝜕𝑢ത 1 𝜕 𝜕𝑢ത
= 𝑓ഥ𝑥 − + 𝜇 − 𝜌𝑢′ 𝑢′ + 𝜇 − 𝜌𝑢′ 𝑣 ′ + 𝜇 − 𝜌𝑢′ 𝑤 ′
𝜌 𝜕𝑥 𝜌 𝜕𝑥 𝜕𝑥 𝜌 𝜕𝑦 𝜕𝑦 𝜌 𝜕𝑧 𝜕𝑧

Reynolds-Averaged Navier-Stokes Equation for incompressible Turbulent Flow in the x direction

Derived by replacing variables with mean and fluctuating components

Reynolds-Averaged Navier-Stokes (RANS)
• Most common method for modelling turbulence in CFD

• Various expressions for approximating Reynolds Stresses

• 𝜌𝑢′ 𝑢′ , 𝜌𝑢′ 𝑣 ′ , 𝜌𝑢′ 𝑤 ′ , 𝜌𝑣 ′ 𝑣 ′ , 𝜌𝑢′ 𝑤 ′ , 𝜌𝑤 ′ 𝑤 ′

Common turbulence models
• Classical models. Based on Reynolds Averaged Navier-Stokes
(RANS) equations (time averaged):
o Zero equation model: mixing length model
o One equation model: Spalart-Almaras
o Two equation models: k- ε models (standard, RNG,
realizable), k-ω model (standard, BSL, SST)
o Seven equation model: Reynolds stress model
• The number of equations denotes the number of additional
PDEs that are being solved
 Boussinesq hypothesis
• Many RANS turbulence models are based upon the Boussinesq hypothesis
• It was experimentally observed that turbulence decays unless there is
shear in isothermal incompressible flows
• Turbulence found to increase as the mean rate of deformation increases
• Boussinesq proposed in 1877 that the Reynolds stresses could be linked to
the mean rate of deformation 𝝏ഥ𝒖 𝝏ഥ 𝒖
𝜌𝑢′ 𝑢′ = −𝜼 +
𝝏𝒙 𝝏𝒙
• A new quantity appears: the turbulent (eddy) viscosity
• The eddy viscosity is not homogeneous, i.e. it varies in space
• directions This
It is, however, assumed to be isotropic. It is the same in all directions.
assumption is valid for many flows, but not for all (e.g. flows with strong
separation or swirl)
 Isotropic eddy viscosity models
• The following models are based on the Boussinesq hypothesis:
o Mixing length model
o Spalart-Allmaras model
o Standard k-ε model
̶ Realizable k-ε model
̶ k-ε RNG model
o Standard k-ω model
̶ k-ω BSL model
̶ k-ω SST model
• We will discuss the k-ε and k-ω models
The k-ε model
• The k-ε model focuses on the mechanisms that affect
the turbulent kinetic energy k
• ε is the dissipation rate of k
• Equations needed for k and ε
• k determines the energy in the turbulence
• ε determines the scale of the turbulence

𝒌𝟐
𝜂=ρ𝑪𝝁
𝝐
k-ε model discussion
• Advantages:
o Relatively simple to implement
o Leads to stable calculations that converge relatively easily
o Reasonable predictions for many flows
• Disadvantages:
o Poor predictions for:
̶ swirling and rotating flows
̶ flows with strong separation
̶ axisymmetric jets
̶ certain unconfined flows
̶ fully developed flows in non-circular ducts
o Valid only for fully turbulent flows
o Simplistic ε equation
k-ω model
• This is another two equation model. In this model ω is the
specific rate of dissipation (k/ε)
• This model solves two additional PDEs:
o A modified version of the k equation used in the k-ε model
o A transport equation for ω (specific rate of dissipation)
𝒌
𝜂=𝝆
𝝎
• Its numerical behavior is similar to that of the k-ε models
• Performs better in near wall regions
• It suffers from some of the same drawbacks, such as the
assumption that eddy viscosity is isotropic
• Performs less well in free stream regions
Reynolds stress model
• RSM closes the Reynolds-Averaged Navier-Stokes equations by
solving additional transport equations for the six independent
Reynolds stresses
• Closure also requires one equation for turbulent dissipation
• Isotropic eddy viscosity assumption is avoided
• Resulting equations contain terms that need to be modeled
• RSM is good for accurately predicting complex flows
• Accounts for streamline curvature, swirl, rotation and high strain
rates
o Cyclone flows, swirling combustor flows
o Rotating flow passages, secondary flows
o Flows involving separation
DNS v LES v RANS
Captions are available for this resource
• This video has been automatically captioned by a machine.
• We believe this captioning will be more than 75% accurate.
• To see the captions click on “CC” on the bottom right of the
video viewer. You can choose how these captions appear.
• When refreshing your knowledge you can skip to the
relevant section of video or captions using a key word
search.

• Provision of this captioning does not impact on your rights to

request additional support or higher accuracy captions via
Disability Services.
• Feedback on the captions should initially be given to the
session lead.