8.

Turbulence Modelling
 Objective

Model the Reynolds stresses

−𝜌𝑢𝑣, −𝜌𝑢𝑢, etc.

in order to close the mean-flow equations.

 How Many Stresses

There are 6 distinct stresses: 𝑢2 , 𝑣 2 , 𝑤 2 , 𝑣𝑤, 𝑤𝑢, 𝑢𝑣

U(y)
uv

u2 U u2

uv

In simple shear, only 𝑢𝑣 is dynamically important

In complex flow, all 6 stresses are important

 Eddy-Viscosity Models
y

Total stress 
𝜏 = 𝜇
𝜕𝑈
− 𝜌𝑢𝑣

(in simple shear):
𝜕𝑦
viscous turbulent
U

𝜕𝑈
Eddy-viscosity model: −𝜌𝑢𝑣 = 𝜇𝑡
𝜕𝑦

𝜕𝑈
Total stress: 𝜏 = 𝜇eff
𝜕𝑦
Effective viscosity: 𝜇eff = 𝜇 + 𝜇𝑡
 Eddy-Viscosity Models

𝜕𝑈
−𝜌𝑢𝑣 = 𝜇𝑡
𝜕𝑦

• This is a model!

• 𝜇 is a physical property of the fluid

𝜇𝑡 is a hypothetical property of the flow

• 𝜇𝑡 varies with position

• At high Reynolds numbers, 𝜇𝑡 ≫ 𝜇

 Eddy-Viscosity Models

For
• Easy to implement
• Extra viscosity aids stability
• Theoretical basis in simple flows

Against
• Little foundation in complex flows
• Turbulence modelling reduced to a single scalar, 𝜇𝑡
 Consistency With the Log Law

Total stress is (i) constant, and (ii) dominated by turbulent stress:

𝜏 (turb) = 𝜏𝑤 ≡ 𝜌𝑢𝜏2

𝜕𝑈 𝑢𝜏
Logarithmic velocity profile: =
𝜕𝑦 𝜅𝑦

𝜏 (turb) 𝜌𝑢𝜏2
Definition of eddy viscosity: 𝜇𝑡 ≡ = = 𝜌(𝜅𝑢𝜏 𝑦)
𝜕𝑈/𝜕𝑦 𝑢𝜏 /𝜅𝑦

(Dynamic) eddy viscosity: 𝜇𝑡 = 𝜌(𝜅𝑢𝜏 𝑦)

Kinematic eddy viscosity: 𝜈𝑡 = 𝜅𝑢𝜏 𝑦

 General Stress-Strain Relationship
y

U(y)
In simple shear (𝑈(𝑦), 0,0):
𝜕𝑈
−𝜌𝑢𝑣 = 𝜇𝑡
𝜕𝑦

In arbitrary flow:
𝜕𝑈 𝜕𝑉
−𝜌𝑢𝑣 = 𝜇𝑡 +
𝜕𝑦 𝜕𝑥 𝜕𝑈𝑖 𝜕𝑈𝑗 2
𝜕𝑈 2 −𝜌𝑢𝑖 𝑢𝑗 = 𝜇𝑡 + − 𝜌𝑘𝛿𝑖𝑗
−𝜌𝑢𝑢 = 2𝜇𝑡 − 𝜌𝑘 𝜕𝑥𝑗 𝜕𝑥𝑖 3
𝜕𝑥 3
Reynolds’ Analogy For Other Turbulent Fluxes

Similar gradient-diffusion model for other turbulent fluxes

𝜕Φ
−𝜌𝑢𝑗 𝜙 = Γ𝑡
𝜕𝑥𝑗

𝜇𝑡
Γ𝑡 =
𝜎𝑡
eddy diffusivity
turbulent Prandtl number
 The Eddy Viscosity

𝜇𝑡
Kinematic eddy viscosity 𝜈𝑡 =
𝜌

𝜈𝑡 = 𝑢0 𝑙0 𝑢0 is a typical turbulent velocity

𝑙0 is a typical turbulent eddy size

e.g. In the log layer: 𝜈𝑡 = 𝜅𝑢𝜏 𝑦 = 𝑢𝜏 × (𝜅𝑦)

 Types of Eddy-Viscosity Model
𝜈𝑡 = 𝑢0 𝑙0

Classify, based on the number of additional scalar-transport equations.

• Zero-equation models:
‒ constant eddy-viscosity models
‒ mixing-length models

• One-equation models:
‒ 𝑙0 specified algebraically; transport equation for some scale to determine 𝑢0

• Two-equation models:
‒ transport equations for 2 scales to determine 𝑢0 and 𝑙0 (𝒌 − 𝜺, 𝑘 − 𝜔, ...)
 Mixing-Length Models

y
Eddy viscosity: 𝜈𝑡 = 𝑢0 𝑙𝑚

lm
dU
lm dy

Mixing length 𝑙𝑚 specified geometrically U

𝜕𝑈
Velocity scale 𝑢0 = 𝑙𝑚 from the mean velocity gradient
𝜕𝑦

2
𝜕𝑈 𝜕𝑈 𝜕𝑈
Turbulent shear stress: 𝜏 (turb) = 𝜇𝑡 = 𝜌𝑢0 𝑙𝑚 2
= 𝜌𝑙𝑚
𝜕𝑦 𝜕𝑦 𝜕𝑦
 Consistency With the Log Law

2
(turb) 2
𝜕𝑈
𝜏 = 𝜌𝑙𝑚
𝜕𝑦

Log-law region: 𝜏 (turb) = 𝜏𝑤 = 𝜌𝑢𝜏2

𝜕𝑈 𝑢𝜏
=
𝜕𝑦 𝜅𝑦

2
𝑢𝜏
𝜌𝑢𝜏2 = 2
ρ𝑙𝑚
𝜅𝑦

𝑙𝑚 = 𝜅𝑦
 Mixing Length
Wall-bounded flows Free shear flows
𝑙𝑚 related to distance from the wall 𝑙𝑚 proportional to width of layer
y
y

mixing layer

𝑙𝑚 = 0.071𝛿

U
U
U y

jet

𝑙𝑚 = 0.098𝛿 (plane)

𝑙𝑚 = min( 𝜅𝑦, 0.09𝛿) 𝑙𝑚 = 0.080𝛿 (round)

U
y U

0.09  wake

𝑙𝑚 = 0.180𝛿
y
U
lm
U
 One-Equation Models

Eddy viscosity: 𝜈𝑡 = 𝑢0 𝑙

Length 𝑙 specified geometrically.

1/4 1/2
Velocity scale: 𝑢0 = 𝐶𝜇 𝑘
Transport equation for turbulent kinetic energy, 𝑘

𝑢03
Dissipation rate modelled as 𝜀 =
𝑙
 The 𝒌 − 𝜺 Model

(turb)
𝜕𝑈
1. Eddy-viscosity model: 𝜏 = 𝜇𝑡
𝜕𝑦

𝑘2 𝑘2
2. Formula for the eddy viscosity: 𝜇𝑡 = 𝐶𝜇 𝜌 𝜈𝑡 = 𝐶𝜇
𝜀 𝜀

3. Scalar-transport equations for:

𝑘 – turbulent kinetic energy
𝜀 – rate of dissipation of 𝑘
 The k and ε Transport Equations
𝜕 𝜕 𝜕𝑘
(𝜌𝑘) + ( 𝜌𝑈𝑖 𝑘 −Γ (𝑘) )= 𝜌(𝑃(𝑘) − 𝜀)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖
𝜕 𝜕 𝜕𝜀 𝜀
(𝜌ε) + ( 𝜌𝑈𝑖 𝜀 −Γ (𝜀) ) = 𝜌(𝐶𝜀1 𝑃(𝑘) − 𝐶𝜀2 𝜀)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖 𝑘
rate of
advection diffusion source
change

𝜇𝑡 𝜇𝑡
Diffusivity: Γ (𝑘) = 𝜇 + , Γ (𝜀) = 𝜇 +
𝜎𝑘 𝜎𝜀

𝜕𝑈𝑖
Production: 𝑃(𝑘) = −𝑢𝑖 𝑢𝑗
𝜕𝑥𝑗

Constants: 𝐶𝜇 = 0.09, 𝐶𝜀1 = 1.44, 𝐶𝜀2 = 1.92, 𝜎𝑘 = 1.0, 𝜎𝜀 = 1.3

 Transport Equations For Turbulence Variables

• Heavily modelled (especially the 𝜀 equation)

• Source-dominated, with a balance between:

𝜕𝑈𝑖
production by mean flow: 𝑃(𝑘) = −𝑢𝑖 𝑢𝑗
𝜕𝑥𝑗

𝜕𝑢𝑖 2
dissipation by viscosity: 𝜀 = 𝜈 ෍( )
𝜕𝑥𝑗
𝑖,𝑗

• Turbulence said to be in local equilibrium if 𝑃 𝑘

=𝜀

• Many variants, including low-Reynolds-number models

• Other choices of dimensional scales (e.g. 𝑘 − 𝜔)

 Consistency With the Log Law

Shear stress: 𝜏 ≡ −𝜌𝑢𝑣 ≈ 𝜏𝑤 = 𝜌𝑢𝜏2 −𝑢𝑣 = 𝑢𝜏2

𝜕𝑈 𝑢𝜏 𝑢𝜏3
Production of 𝑘: 𝑃 (𝑘)
= −𝑢𝑣 = 𝑢𝜏2 × =
𝜕𝑦 𝜅𝑦 𝜅𝑦

Eddy viscosity: 𝜈𝑡 = 𝜅𝑢𝜏 𝑦

𝑢𝜏3
𝜀= 𝑃(𝑘) =
Local equilibrium: 𝑃(𝑘) = 𝜀 𝜅𝑦

𝑢𝜏4
𝜈𝑡 =
𝜀
2
𝑢𝜏4 −𝑢𝑣
𝑘2 𝐶𝜇 = 𝐶𝜇 =
𝒌 − 𝜺 model: 𝜈𝑡 = 𝐶𝜇 𝑘2 𝑘
𝜀

1/4
𝑢𝜏 = 𝐶𝜇 𝑘1/2
 Example

(a) The 𝑘 − 𝜀 turbulence model forms an eddy viscosity 𝜇𝑡 from fluid density 𝜌,
the turbulent kinetic energy (per unit mass) 𝑘 and its dissipation rate 𝜀. Write
down the basic physical dimensions of 𝜇𝑡 , 𝜌, 𝑘 and 𝜀 in terms of the
fundamental dimensions of mass M, length L and time T, and hence show, on
purely dimensional grounds, that any expression for 𝜇𝑡 in terms of the other
variables must be of the form
𝑘2
𝜇𝑡 = constant × 𝜌
𝜀

(b) The 𝑘 − 𝜔 turbulence model forms an eddy viscosity from 𝜌, 𝑘 and a quantity
𝜔 which has dimensions of frequency (i.e. T–1). Show, on dimensional grounds,
that any expression for 𝜇𝑡 in terms of the other variables must be of the form
𝑘
𝜇𝑡 = constant × 𝜌
𝜔
(a) The 𝑘 − 𝜀 turbulence model forms an eddy viscosity 𝜇𝑡 from fluid density 𝜌, the turbulent kinetic
energy (per unit mass) 𝑘 and its dissipation rate 𝜀. Write down the basic physical dimensions of 𝜇𝑡 , 𝜌, 𝑘
and 𝜀 in terms of the fundamental dimensions of mass M, length L and time T, and hence show, on
purely dimensional grounds, that any expression for 𝜇𝑡 in terms of the other variables must be of the
form
𝑘2
𝜇𝑡 = constant × 𝜌
𝜀
𝜇𝑡 ~ 𝜌, 𝑘, 𝜀

Dimensions:
force/area ML T −2 ΤL2
[𝜇𝑡 ] = = = ML−1 T −1
velocity/length L T −1 ΤL
[𝜌] = ML−3 Π = 𝜇𝑡 𝜌 𝑎 𝑘 𝑏 𝜀 𝑐

[𝑘] = [velocity]2 = L2 T −2 M0 L0 T 0 = ML−1 T −1 ML−3 𝑎

L2 T −2 𝑏
L2 T −3 𝑐

[𝑘] M: 0=1+𝑎 𝑎 = −1
[𝜀] = = L2 T −3
[time]
L: 0 = −1 + 3 + 2𝑏 + 2𝑐
T: 0 = −1 − 2𝑏 − 3𝑐
# variables: 4
0=1−𝑐 𝑐=1
# independent dimensions: 3
𝑏 = −2
# Π groups: 4 − 3 = 1
Π = 𝜇𝑡 𝜌−1 𝑘 −2 𝜀 = constant (𝐶𝜇 , say)
𝑘2
𝜇𝑡 = 𝐶𝜇 𝜌
𝜀
(b) The 𝑘 − 𝜔 turbulence model forms an eddy viscosity from 𝜌, 𝑘 and a quantity 𝜔 which has dimensions
of frequency (i.e. T–1). Show, on dimensional grounds, that any expression for 𝜇𝑡 in terms of the other
variables must be of the form
𝑘
𝜇𝑡 = constant × 𝜌
𝜔
𝜇𝑡 ~ 𝜌, 𝑘, 𝜔

Dimensions:
force/area ML T −2 ΤL2
[𝜇𝑡 ] = = = ML−1 T −1
velocity/length −1
L T ΤL
[𝜌] = ML−3
2 2 −2 Π = 𝜇𝑡 𝜌 𝑎 𝑘 𝑏 𝜔 𝑐
[𝑘] = [velocity] =L T
1 M0 L0 T 0 = ML−1 T −1 ML−3 𝑎
L2 T −2 𝑏
T −1 𝑐

[𝜔] =
T M: 0=1+𝑎 𝑎 = −1
L: 0 = −1 + 3 + 2𝑏 𝑏 = −1
# Variables: 4
T: 0 = −1 + 2 − 𝑐 𝑐=1
# Independent dimensions: 3
# Π groups: 4 − 3 = 1
Π = 𝜇𝑡 𝜌−1 𝑘 −1 𝜔 = constant

𝑘
𝜇𝑡 = constant × 𝜌
𝜔
 Example
A modeled scalar-transport equation for 𝜀 is
D𝜀 𝜕 𝜈𝑡 𝜕𝜀 (𝑘)
𝜀
= + (𝐶𝜀1 𝑃 − 𝐶𝜀2 𝜀)
D𝑡 𝜕𝑥𝑖 𝜎𝜀 𝜕𝑥𝑖 𝑘
where D/ D𝑡 is the material derivative, 𝑃 𝑘 is the rate of production of 𝑘 and the
summation convention is implied by the repeated index 𝑖. 𝜎𝜀, 𝐶𝜀1 and 𝐶𝜀2 are constants.

In a fully-developed turbulent boundary layer,

3
𝑢 𝜏 −1/2
𝑃(𝑘) = 𝜀 = and 𝑘 = 𝐶𝜇 𝑢𝜏2
𝜅𝑦
where 𝜅 is von Kármán’s constant, 𝑢𝜏 is the friction velocity and 𝑦 is the distance from
the boundary. Using the 𝜀 equation and the eddy-viscosity formula from Classroom
Example 1, show that this implies the following relationship between coefficients:

(𝐶𝜀2 − 𝐶𝜀1 )𝜎𝜀 𝐶𝜇 = 𝜅 2

D𝜀 𝜕 𝜈𝑡 𝜕𝜀 𝜀
= + (𝐶𝜀1 𝑃(𝑘) − 𝐶𝜀2 𝜀)
D𝑡 𝜕𝑥𝑖 𝜎𝜀 𝜕𝑥𝑖 𝑘

(𝑘)
𝑢𝜏3 −1/2 2
𝑃 =𝜀= 𝑘 = 𝐶𝜇 𝑢𝜏
𝜅𝑦
D𝜀 𝜕𝜀 𝜕𝜀
Fully-developed: =0 = =0
D𝑡 𝜕𝑥 𝜕𝑧

Local equilibrium: 𝑃(𝑘) = 𝜀

d 𝜈𝑡 d𝜀 𝜀2
0= ( ) + (𝐶𝜀1 − 𝐶𝜀2 )
d𝑦 𝜎𝜀 d𝑦 𝑘

𝑢𝜏3 d𝜀 𝑢𝜏3 𝑢𝜏2 𝑘2

𝜀= =− 2 𝑘= 𝜈𝑡 = 𝐶𝜇 = 𝜅𝑢𝜏 𝑦
𝜅𝑦 d𝑦 𝜅𝑦 𝐶𝜇 𝜀

d 𝜅𝑢𝜏 𝑦 𝑢𝜏3 𝑢𝜏6 𝐶𝜇

0= (− ) + (𝐶𝜀1 − 𝐶𝜀2 ) 2 2 2
d𝑦 𝜎𝜀 𝜅𝑦 2 𝜅 𝑦 𝑢𝜏

𝑢𝜏4 d 1 𝑢𝜏4 𝐶𝜇
0=− ( ) + (𝐶𝜀1 − 𝐶𝜀2 ) 2 2
𝜎𝜀 d𝑦 𝑦 𝜅 𝑦
1 𝐶𝜇
0= + (𝐶𝜀1 − 𝐶𝜀2 )
𝜎𝜀 𝑦 2 𝜅2𝑦2
𝐶𝜇 1
(𝐶𝜀2 − 𝐶𝜀1 ) = (𝐶𝜀2 − 𝐶𝜀1 )𝜎𝜀 𝐶𝜇 = 𝜅 2
𝜅2 𝜎𝜀
 Example

In grid-generated turbulence there is no mean velocity shear, and hence no turbulence

production and minimal diffusion. The 𝑘 and  transport equations reduce to
d𝑘 d𝜀 𝜀2
= −𝜀 , = −𝐶𝜀2
d𝑡 d𝑡 𝑘
where 𝑡 (=distance/mean velocity) is the travel time downstream of the grid. By
substituting into these equations, show that they admit a solution of the form
𝑘 = 𝑘0 𝑡 −𝑚 , 𝜀 = 𝜀0 𝑡 −𝑛
where 𝑘0 , 𝜀0 , 𝑚 and 𝑛 are constants, and find 𝐶𝜀2 in terms of 𝑚 alone.

(This rate of decay for 𝑘 provides a means of determining 𝐶𝜀2 experimentally.)

In grid-generated turbulence there is no mean velocity shear, and hence no turbulence production and
minimal diffusion. The 𝑘 and 𝜀 transport equations reduce to
d𝑘 d𝜀 𝜀2
= −𝜀 , = −𝐶𝜀2
d𝑡 d𝑡 𝑘
Show that they admit a solution of the form
𝑘 = 𝑘0 𝑡 −𝑚 , 𝜀 = 𝜀0 𝑡 −𝑛
Find 𝐶𝜀2 in terms of 𝑚 alone.
d𝑘 d𝜀 𝜀2
= −𝜀 = −𝐶𝜀2
d𝑡 d𝑡 𝑘
𝜀02 𝑚−2𝑛
−𝑚𝑘0 𝑡 −𝑚−1 = −𝜀0 𝑡 −𝑛 −𝑛𝜀0 𝑡 −𝑛−1
= −𝐶𝜀2 𝑡
𝑘0
𝑚𝑘0 𝑛𝑘0
= 𝑡 𝑚+1−𝑛 = 𝑡 𝑚+1−𝑛
𝜀0 𝐶𝜀2 𝜀0

𝑚𝑘0 𝑛𝑘0
= = 𝑡 𝑚+1−𝑛
𝜀0 𝐶𝜀2 𝜀0

𝑛
𝑚+1−𝑛 =0 𝑚=
𝐶𝜀2
𝑛=𝑚+1 𝑛
𝐶𝜀2 =
𝑚
𝑚+1
𝐶𝜀2 =
𝑚
 Limitations of Eddy-Viscosity Models

For
• Simple to code
• Extra viscosity aids stability
• Theoretical basis in some simple but common flows

Against
• Only capable of predicting one Reynolds-stress component accurately
• Lack justification in complex flows
• Fail to predict common properties such as anisotropy
 Turbulence Anisotropy
𝜕𝑈 𝜕𝑉
y −𝜌𝑢𝑣 = 𝜇𝑡 ( + )
𝜕𝑦 𝜕𝑥
v
u 𝜕𝑈 2
w −𝜌𝑢2 = 2𝜇𝑡 − 𝜌𝑘
𝜕𝑥 3
2
𝜕𝑉 2
−𝜌𝑣 = 2𝜇𝑡 − 𝜌𝑘
𝜕𝑦 3
𝜕𝑊 2
U −𝜌𝑤 2 = 2𝜇𝑡 − 𝜌𝑘
𝜕𝑧 3
𝜕𝑈
Eddy-viscosity model: −𝑢𝑣 = 𝜈𝑡
𝜕𝑦
2
𝑢2 = 𝑣2 = 𝑤2 = 𝑘
3

Experiment: 𝑢2 ∶ 𝑣 2 ∶ 𝑤 2 = 1.0 ∶ 0.4 ∶ 0.6

Advanced Turbulence Models
Direct
Numerical
Simulation

Large-Eddy
increasing complexity Simulation

Reynolds-Stress
Transport
Models

Non-Linear
Eddy-Viscosity
Models

two-equation
Eddy-Viscosity one-equation
Models mixing length
constant
 Reynolds-Stress Transport Models
• Also known as differential stress models or second-order closure

• Solve individual transport equations for each Reynolds stress

−𝜌𝑢𝑣, −𝜌𝑢𝑢, etc.

• “Source” is a balance between:

‒ production by mean-velocity gradients (and, sometimes, body forces)
‒ redistribution by pressure fluctuations
‒ dissipation by viscous action

For
• Production and advection are exact and don’t need modelling

Against
• many terms still do need modelling
• computational expense: 6 extra transport equations
• lack of a stabilising diffusion term
 Non-Linear Eddy-Viscosity Models

Simple eddy-viscosity model:

𝜕𝑈
𝐬𝐭𝐫𝐞𝐬𝐬 ∝ 𝐯𝐞𝐥𝐨𝐜𝐢𝐭𝐲 𝐠𝐫𝐚𝐝𝐢𝐞𝐧𝐭: 𝜏turb = 𝜇𝑡
𝜕𝑦

Non-linear eddy-viscosity model:

𝐬𝐭𝐫𝐞𝐬𝐬 = 𝑪𝟏(𝐯𝐞𝐥𝐨𝐜𝐢𝐭𝐲 𝐠𝐫𝐚𝐝𝐢𝐞𝐧𝐭) + 𝑪𝟐(𝐯𝐞𝐥𝐨𝐜𝐢𝐭𝐲 𝐠𝐫𝐚𝐝𝐢𝐞𝐧𝐭)𝟐 + ⋯

For
• Qualitatively correct response of turbulence to certain types of strain
• Little more computationally expensive than linear models

Against
• Doesn’t accurately reflect the real physical processes
• Little justification in complex geometries
 Wall Boundary Conditions

• Near a solid boundary:

‒ there are very large velocity gradients
‒ wall-normal fluctuations are selectively damped
‒ viscous and turbulent stresses are comparable

• Alternative approaches:
‒ low-Reynolds-number turbulence models
‒ wall functions
 Low-Re Turbulence Models

Try to ensure correct asymptotic behaviour as 𝑦 → 0:

2𝜈𝑘
𝑘 ∝ 𝑦2, 𝜀~ 2 ~constant, 𝜈𝑡 ∝ 𝑦 3 (𝑦 → 0)
𝑦

Viscous modifications to eddy viscosity and transport equations

𝑢𝜏 𝑦
Resolve flow right to the boundary: 𝑦+ ≡ ≤1
𝜈

Alternative approach: blend with one-equation model as a two-layer model

 Wall Functions
Equilibrium (log law):
control volume
𝑈𝑃 1 𝑦𝑃 𝑢𝜏
= ln( 𝐸𝑦𝑃+ ) 𝑦𝑝+ =
𝑢𝜏 𝜅 𝜈
near-wall Up
 node 1/4 1/2
assumed velocity 𝜈𝑡 = 𝜅𝑢𝜏 𝑦 𝑢𝜏 = 𝐶𝜇 𝑘𝑃
yp
profile

w Non-equilibrium:
1/4 1/2
𝜈𝑡 = 𝜅𝑢0 𝑦 𝑢0 = 𝐶𝜇 𝑘𝑃
𝜕𝑈
𝜏𝑤 = 𝜌(𝜅𝑢0 𝑦)
𝜕𝑦
𝜕𝑈 𝜏𝑤 /𝜌 1
=
𝜕𝑦 𝜅𝑢0 𝑦
𝜏𝑤 /𝜌 𝜅𝑢0 𝑈𝑝
𝑈= ln( 𝐶𝑦) 𝜏𝑤 = 𝜌 𝑦 𝑢
𝜅𝑢0 ln( 𝐸 𝑃𝜈 0 )

𝑈𝑃 𝜌(𝜅𝑢0 𝑦𝑝 )
Implemented via an effective wall viscosity: 𝜏𝑤 = 𝜇𝑤 𝜇𝑤 = 𝑦 𝑢
𝑦𝑃 ln( 𝐸 𝑃𝜈 0 )
 Summary (1)

• A turbulence model or turbulence closure is a means of

specifying Reynolds stresses (and other turbulent fluxes)

• The most popular type is an eddy-viscosity model

• The eddy viscosity may be specified:

‒ geometrically (e.g. mixing-length models)
‒ by solving transport equations (e.g. 𝑘 − 𝜀 model)
 Summary (2)

• Advanced turbulence models include:

‒ Reynolds-stress transport models (RSTM)
‒ non-linear eddy-viscosity models (NLEVM)
‒ large-eddy simulation (LES)

• Wall boundary conditions require either:

‒ low-Reynolds-number (i.e. viscous) modifications and a fine grid
‒ wall functions and a coarse grid