MMME3093

Computer Modelling
Techniques

Lecture 3
Viscous modelling and
turbulence models

Dr Cheng Xinwei
MMME3093 Computer Modelling Techniques 1
 MMME3093 Computer Modelling Techniques

Recap
• 1D diffusion-convection example
𝜕𝜑 𝜕𝜑
𝜌𝑢𝐴𝜑 𝑒 − 𝜌𝑢𝐴𝜑 𝑤 = Γ𝐴 − Γ𝐴
𝜕𝑥 𝑒
𝜕𝑥 𝑤

Central differencing scheme for convective terms

𝜑𝑃 + 𝜑𝐸 𝜑𝑊 + 𝜑𝑃
𝜌𝑢 𝑒 − 𝜌𝑢 𝑤
2 2

Upwind scheme for convective terms

𝜌𝑢 𝑒 𝜑𝑃 − 𝜌𝑢 𝑤 𝜑𝑊

MMME3093 Computer Modelling Techniques 2

 MMME3093 Computer Modelling Techniques

Recap
• Properties of discretisation schemes
• Conservativeness
• Boundedness
• Transportiveness
• Peclet number to determine strengths of convective effects
and diffusive effects
convection u u  dx
Pe = = =
diffusion  dx 

• Pressure-velocity coupling -> SIMPLE scheme

MMME3093 Computer Modelling Techniques 3

 MMME3093 Computer Modelling Techniques

Learning outcomes
• Introduction to viscous modelling
• Introduction to meshing and solution
• Reduce a geometry from 3D to 2D
• Turbulence models
• RANS k-ε equations

MMME3093 Computer Modelling Techniques 4

 MMME3093 Computer Modelling Techniques

Viscous modelling
• Viscous modelling is the modelling of flow based on viscosity.
• It depends on the Reynolds number to determine the nature of
flow and what model is suitable to model the flow.

Laminar Turbulent

• When a fluid flows smoothly in • When the flow motion is

parallel layers, with no chaotic and irregular.
disruption between layers. • Re > 4000
• Re < 2000
MMME3093 Computer Modelling Techniques 5
 MMME3093 Computer Modelling Techniques

Viscous modelling- Example

Simulate a flow in a cylindrical pipe without energy
variation

MMME3093 Computer Modelling Techniques 6

 MMME3093 Computer Modelling Techniques

Viscous modelling- Example

The geometry is cylindrical so I can check if there is any
symmetry in the geometry and flow
U

Rotation by 2p (360˚)
R

MMME3093 Computer Modelling Techniques 7

 MMME3093 Computer Modelling Techniques

Viscous modelling- Example

Before we start it is better to check the Reynolds number
𝜌𝐿𝑈 𝑖𝑛𝑒𝑟𝑡𝑖𝑎 𝑓𝑜𝑟𝑐𝑒𝑠
𝑅𝑒 = =
𝜇 𝑣𝑖𝑠𝑐𝑜𝑢𝑠 𝑓𝑜𝑟𝑐𝑒𝑠

Laminar flow In this case, we

changed the Reynolds
number by changing
the average velocity of
the flow
Turbulent flow

MMME3093 Computer Modelling Techniques 8

 MMME3093 Computer Modelling Techniques

Viscous modelling- Example

Hydraulic diameter for non-circular tubes
4𝐴
𝐷ℎ =
𝑃
Typical Reynolds number for turbulence to start

Conditions Re for turbulence Length scale

Flat plate – smooth wall 100,000 X position along length
Pipe – smooth wall 2000 Diameter
Rectangular pipe 1500 Hydraulic diameter
Cylinder 20 Diameter
Sphere 20 Diameter
Free jet 30 Diameter

MMME3093 Computer Modelling Techniques 9

 MMME3093 Computer Modelling Techniques

Viscous modelling- Example

• What about the grid/mesh?
• Can I imagine the main features of the flow inside the pipe?
• Do you think we’ll have a boundary layer?
• So is it a good idea to refine the mesh/grid in proximity of
the wall?
wall

inlet outlet

axis
Since the geometry is reduced from 3D to 2D, we will have
less cells and equations to solve.

MMME3093 Computer Modelling Techniques 10

 MMME3093 Computer Modelling Techniques

Viscous modelling- Example

Why is it a good idea to refine the mesh/grid in proximity of
the wall?

Real profile

Grid 4 points

Grid 5 points

Increasing the number of points near the walls will improve the solution

MMME3093 Computer Modelling Techniques 11

 MMME3093 Computer Modelling Techniques

Viscous modelling- Example

If the Reynolds number is low (below 3000), so we have laminar
flow, the equation to use are
𝜕𝜌 𝜕𝜌𝑢 𝜕𝜌𝑣
+ + =0
𝜕𝑡 𝜕𝑥 𝜕𝑦
𝜕𝑢 𝜕𝑢 𝜕𝑢 𝜕𝑝 𝜕2𝑢 𝜕2𝑢
𝜌 + 𝑢 + 𝑣 = 𝜌 𝑔𝑥 − +𝜇 +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2
𝜕𝑣 𝜕𝑣 𝜕𝑣 𝜕𝑝 𝜕2𝑣 𝜕2𝑣
𝜌 + 𝑢 + 𝑣 = 𝜌 𝑔𝑦 − +𝜇 +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 2 𝜕𝑦 2
𝜕𝑤 𝜕𝑤 𝜕𝑤 𝜕𝑤 𝜕𝑝 𝜕2𝑤 𝜕2𝑤 𝜕2𝑤
𝜌 + 𝑢 + 𝑣 +𝑤 = 𝜌 𝑔𝑧 − +𝜇 + +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧 2

𝜕(𝜌𝑇 ሻ 𝜕𝜌𝑢𝑇 𝜕𝜌𝑣𝑇 𝜕2𝑇 𝜕2𝑇

+ + = Γ + + 𝑆ℎ
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑥 2 𝜕𝑦 2

In FLUENT, you set the viscous model as laminar

MMME3093 Computer Modelling Techniques 12
 MMME3093 Computer Modelling Techniques

Viscous modelling- Example

If the Reynolds number is high (above 3000), so we have turbulent
flow. What to do?
𝜕𝜌 𝜕𝜌𝑢 𝜕𝜌𝑣
+ + =0
𝜕𝑡 𝜕𝑥 𝜕𝑦
𝜕𝑢 𝜕𝑢 𝜕𝑢 𝜕𝑝 𝜕2𝑢 𝜕2𝑢
𝜌 + 𝑢 + 𝑣 = 𝜌 𝑔𝑥 − +𝜇 +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2
𝜕𝑣 𝜕𝑣 𝜕𝑣 𝜕𝑝 𝜕2𝑣 𝜕2𝑣
𝜌 + 𝑢 + 𝑣 = 𝜌 𝑔𝑦 − +𝜇 +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 2 𝜕𝑦 2
𝜕𝑤 𝜕𝑤 𝜕𝑤 𝜕𝑤 𝜕𝑝 𝜕2𝑤 𝜕2𝑤 𝜕2𝑤
𝜌 + 𝑢 + 𝑣 +𝑤 = 𝜌 𝑔𝑧 − +𝜇 + +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧 2

𝜕(𝜌𝑇 ሻ 𝜕𝜌𝑢𝑇 𝜕𝜌𝑣𝑇 𝜕2𝑇 𝜕2𝑇

+ + = Γ + + 𝑆ℎ
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑥 2 𝜕𝑦 2

In FLUENT, you set the viscous model to turbulence

MMME3093 Computer Modelling Techniques 13
 MMME3093 Computer Modelling Techniques

Turbulence
Small
eddies

Large
eddies

• Turbulence causes the appearance in the flow of eddies with a

wide range of length and time scales interacts in a dynamically
complex way (irregularities).
• Turbulent eddies, which are the local swirling motions where
vorticity can be often very intense, is a commonly used term to
describe turbulence.
• Turbulent eddies of a wide range of sizes gives rise to vigorous
mixing and turbulent stresses.
MMME3093 Computer Modelling Techniques 14
 MMME3093 Computer Modelling Techniques

Turbulence

Turbulent flow

Laminar flow

Results show that at any point in a turbulent flow, there is a

mean value of flow variables and a fluctuating component. This
is known as Reynolds decomposition.
𝑢 = 𝑢ത + 𝑢′ 𝑣 = 𝑣ҧ + 𝑣 ′
Where u, v are the velocities at a point, 𝑢ത , 𝑣ҧ are the mean velocities and 𝑢′ , 𝑣 ′
are the mean fluctuations from the mean velocity.
MMME3093 Computer Modelling Techniques 15
 MMME3093 Computer Modelling Techniques

Turbulence
• The largest turbulent eddies interact with the mean flow and
subsequently draw energy from it. This is known as vortex
stretching.
• Consequently, turbulent kinetic energy is transferred
progressively from larger to smaller eddies. This is known as
energy cascade.
• The total kinetic energy per unit mass, k at a given location is
1 2
𝑘 = 𝑢′ + 𝑣′2 + 𝑤′2
2
• Kinetic energy is then dissipated to thermal internal energy (heat)
at the smallest scales of turbulence.
• These smallest scales are dominated by viscous effects and are
named as the Kolmogorov scales, where lengths are of order of
0.1 mm to 0.01 mm.
MMME3093 Computer Modelling Techniques 16
 MMME3093 Computer Modelling Techniques

Turbulence
The total kinetic energy per unit mass, k gives an absolute value but it is
useful to know what proportion of the flow of kinetic energy this is.
Therefore, turbulence intensity, T is calculated
2 1/2
𝑘
3
𝑇=
𝑈
Nevertheless, it is difficult to measure local fluctuations i.e. from point to
point. As such, the turbulence intensity can be approximated using the
following.
−1/8
𝐼 = 0.16 𝑅𝑒ℎ
By knowing the Reynolds number of the flow, we can then determine the
range of turbulence intensity.
𝐼 < 1%, indicates laminar flow
𝐼 > 10%, indicates very intense turbulence
MMME3093 Computer Modelling Techniques
1% < 𝐼 < 10%, typical turbulence in practice 17
 MMME3093 Computer Modelling Techniques

Turbulence
• The motion of the smallest scales depend on
• Rate at which the larger eddies supply energy
• Kinematic viscosity
• Imagine what would happen if every cell in a domain is of the
size of the Kolmogorov scales?
• Computational time will be infinite.
• So, how can we represent turbulence or the viscous dissipation?
• Turbulence fluctuations will be included in the Navier-Stokes
equations and turbulence models are needed to account
for the extra turbulence effects generated from turbulent flow.

MMME3093 Computer Modelling Techniques 18

 MMME3093 Computer Modelling Techniques

Turbulence models

Turbulence models

Reynolds-Averaged
Direct Numerical Large Eddy
Navier Stokes
Simulation (DNS) Simulation (LES)
(RANS)

MMME3093 Computer Modelling Techniques 19

 MMME3093 Computer Modelling Techniques

Turbulence models
Direct numerical simulation (DNS)
• Navier-Stokes equations are solved for all of the motions in a
turbulent flow.
• Thus, very fine spatial grids are required such that the
Kolmogorov length scales at which energy dissipation takes
place can be resolved.
• Time-steps must be sufficiently small to resolve the period of
the fastest fluctuations.
• Computationally demanding.
• Only applicable to fundamental research.

MMME3093 Computer Modelling Techniques 20

 MMME3093 Computer Modelling Techniques

Turbulence models
Large eddy simulation (LES)
• Tracks the behavior of larger eddies by applying space
filtering prior to computation.
• Large eddies are resolved directly using the Navier-Stokes
equations.
• Meanwhile, the effects of small eddies on the flow (mean flow
plus large eddies) are modelled using sub-grid scale model.
• Computing demand is still high but less than DNS.
• Currently, can be used to solve complex problems.

MMME3093 Computer Modelling Techniques 21

 MMME3093 Computer Modelling Techniques

Turbulence models
Reynolds-averaged Navier-Stokes (RANS)
• Mean flow and the turbulence effects on mean flow properties
are of interest.
• Navier-Stokes equations are Reynolds-averaged before
applying this model.
• Extra terms (known as Reynold stresses) appear in the
Reynolds-averaged flow equations are due to turbulent
fluctuations.
• These extra terms can be modelled with classical turbulence
models, i.e. k-ε model, Reynolds stress model.
• Computing resources for reasonably accurate flow are the
modest.

MMME3093 Computer Modelling Techniques 22

 MMME3093 Computer Modelling Techniques

Reynolds-averaged Navier-Stokes (RANS)

RANS models are developed to predict the extra terms
(Reynolds stresses) and to close the system mean flow
equations, with additional equations.

RANS turbulence models

Linear eddy Non-linear eddy Reynolds stress

viscosity model viscosity model model (RSM)

• Zero-equation model • Explicit nonlinear

• One-equation model constitutive relation
• Two-equation model • v2-f models

MMME3093 Computer Modelling Techniques 23

 MMME3093 Computer Modelling Techniques

Turbulence models in ANSYS Fluent

The numbers in
brackets indicate Increasing
additional computational
conservation time
equations that need
to be solved

MMME3093 Computer Modelling Techniques 24

 MMME3093 Computer Modelling Techniques

Reynolds-averaged Navier-Stokes equations

• For turbulent flow, you need to include the fluctuations into the
Navier-Stokes equations.
• This is known as time-averaged or Reynolds-averaged.
• Recall the Navier-Stokes equations derived in the first lecture
𝜌𝒖 𝑑𝑖𝑣 𝒖 = −𝑔𝑟𝑎𝑑 𝑝 + 𝜇𝛻 2 𝑢
• Apply the time-averaged quantities
ഥ + 𝑢′ 𝑑𝑖𝑣 (ഥ
𝜌 𝒖 𝒖 + 𝑢′ ሻ = −𝑔𝑟𝑎𝑑 𝑝ҧ + 𝜇𝛻 2 (ഥ
𝒖 + 𝑢′ ሻ
• Rearranging (derivation is shown during lecture)
𝜌ഥ ഥ = −𝑔𝑟𝑎𝑑 𝑝ҧ + 𝜇𝛻 2 𝒖
𝒖 𝑑𝑖𝑣 𝒖 ഥ − 𝜌 𝑑𝑖𝑣 ∙ 𝑢′ 𝒖′

• Known as Reynold stresses, which is due to turbulent motion

(analogy with viscous stresses).
• Leads to closure problem and can be solved by additional models.
MMME3093 Computer Modelling Techniques 25
 MMME3093 Computer Modelling Techniques

Reynolds-averaged Navier-Stokes equations

• The equation from previous slide, which is known as Reynolds-
averaged Navier-Stokes (RANS) equations, can be re-written as
𝜕𝜌𝑈 𝜕𝑃
+ 𝑑𝑖𝑣 𝜌𝑈𝑼 = − + 𝑑𝑖𝑣 𝜇 𝑔𝑟𝑎𝑑 𝑈
𝜕𝑡 𝜕𝑥
𝜕 𝜌𝑢′2 𝜕 𝜌𝑢′𝑣 ′ 𝜕 𝜌𝑢′𝑤 ′
+ − − − + 𝑆𝑀𝑥
𝜕𝑥 𝜕𝑦 𝜕𝑧

• The time-average transport equation for an arbitrary scalar φ is

𝜕𝜌𝛷
+ 𝑑𝑖𝑣 𝜌𝛷𝑼 = 𝑑𝑖𝑣 𝛤𝑔𝑟𝑎𝑑 𝛷
𝜕𝑡
𝜕 𝜌𝑢′𝜑′ 𝜕 𝜌𝑣′𝜑′ 𝜕 𝜌𝑤′𝜑′
+ − − − + 𝑆𝛷
𝜕𝑥 𝜕𝑦 𝜕𝑧

MMME3093 Computer Modelling Techniques 26

 MMME3093 Computer Modelling Techniques

Reynolds-averaged Navier-Stokes (RANS)

Before Reynolds-averaged

After Reynolds-averaged

MMME3093 Computer Modelling Techniques 27

 MMME3093 Computer Modelling Techniques

Reynolds-averaged Navier-Stokes (RANS)

Two-equation model (k-ε model)
• Uses 2 equations to describe the turbulence
fluctuations:
• (1) The energy in the turbulence is determined by
turbulent kinetic energy (k)
• (2) The rate of dissipation of the turbulent kinetic energy
is calculated by turbulent dissipation (ε)
• Examples:
• Standard, Renormalisation Group (RNG), Realisable

MMME3093 Computer Modelling Techniques 28

 MMME3093 Computer Modelling Techniques

RANS k-ε turbulence model

Replace k and ε into the time-average equation
𝜕𝜌𝛷 𝜕 𝜌𝑢′𝜑 ′ 𝜕 𝜌𝑣′𝜑 ′ 𝜕 𝜌𝑤′𝜑 ′
+ 𝑑𝑖𝑣 𝜌𝛷𝑼 = 𝑑𝑖𝑣 𝛤𝑔𝑟𝑎𝑑 𝛷 + − − − + 𝑆𝛷
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
With some re-arrangement, the time-average equation for k becomes
𝜕(𝜌𝑘ሻ 𝜇𝑡
+ 𝑑𝑖𝑣 𝜌𝑘𝑼 = 𝑑𝑖𝑣 𝑔𝑟𝑎𝑑 𝑘 + 2𝜇𝑡 𝑆𝑖𝑗 . 𝑆𝑖𝑗 − 𝜌𝜀
𝜕𝑡 𝜎𝑘
The time-average equation for 𝜀 is
𝜕(𝜌𝜀ሻ 𝜇𝑡 𝜀 𝜀2
+ 𝑑𝑖𝑣 𝜌𝜀𝑼 = 𝑑𝑖𝑣 𝑔𝑟𝑎𝑑 𝜀 + 𝐶1𝜀 2𝜇𝑡 𝑆𝑖𝑗 . 𝑆𝑖𝑗 − 𝐶2𝜀 𝜌
𝜕𝑡 𝜎𝜀 𝑘 𝑘

In words, the equations are

Rate of Transport Transport Rate of Rate of
change of + of k or 𝜀 by = of k or 𝜀 by + production - destruction
k or 𝜀 convection diffusion of k or 𝜀 of k or 𝜀
MMME3093 Computer Modelling Techniques 29
 MMME3093 Computer Modelling Techniques

Choices of turbulence model

This flow has sharp This flow is not With the sharp interior
edges, with a flow from sharply separated bend, there will be
right. There will be due to the more separation and RSM
separation at the streamlined structure. will be more suitable.
trailing edge and
vortex structure with A k-ε model can be k-ε can be applied if
increasing intensity used. However, RSM there a large number
with velocity. Since the may be applied if the of cells (small cell
flow is anisotropic on velocity increases. sizes) but there may
the trailing edge, RSM be poor turbulence
is more suitable. presentation in the
corner.
MMME3093 Computer Modelling Techniques 30
 MMME3093 Computer Modelling Techniques

Conclusions
This lecture covers the following aspects of numerical
modelling technique:
• Geometry simplification from 3D to 2D
• Turbulence models
• RANS equations
• Choice of turbulence model

MMME3093 Computer Modelling Techniques 31

 MMME3093 Computer Modelling Techniques

References
• Lecture notes from UK Campus
• Versteeg, H.K. and Malalasekera W. An introduction to
computation fluid dynamics: the finite volume method.
Pearson, 2007

MMME3093 Computer Modelling Techniques 32