B.Tech/B.

E (Mechanical) 5th Semester Course 2023-24

Introduction to Turbulent Flows

(MEC3310/EMEC3310)
Module 4: Turbulent Flow

Syed Fahad Anwer

Professor
Department of Mechanical Engineering

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
 Introduction

• Contents
• Statistical description and concept of length and time scales, Equations for mean ow
and closure problem

• Reynolds’ stresses, turbulence modeling: Prandtls Mixing Length Model

• Velocity pro le forComputational
•
turbulentAerodynamics
ow in pipes, turbulent boundary layer.
Lab, Department of Mechanical Engineering
ZHCET, AMU, ALIGARH
fi
fl
fl
 Introduction
Reynolds (1883) pioneering dye experiment: The rst major studies
of uid ow in a pipe where transition from laminar to turbulent was
studied in detail.

• Laminar ow:
• Regular, orderly motions of uid elements.
• No perturbation is enhanced.
• Lower critical Reynolds number below which no turbulence is
observed.

• Turbulence ow:
• Random, disorderly motions of uid elements.
• In nitesimal perturbation is enhanced into a highly disorderly
motion.
• an upper critical one above which the laminar ow can no longer
be sustained.

• Transition ow: a minimal amplitude which perturbations have to

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fi
fl
fl
fl
fl
fl
fl
fl
fl
fi
 Turbulence problem:
“last unsolved problem in classical physics”
(Feynman, 1979)
 Introduction: Basic Features
Reynolds (1883) pioneering dye experiment: The rst major studies of uid
ow in a pipe where transition from laminar to turbulent was studied in
detail.

It shows that:

• All transport properties are random functions of space and time.

• Strong Mixing (Enhanced viscosity)

• Present wide range of length and time scales.

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fl
fi
fl
 Introduction:Physical Description
• Randomness

• Nonlinearity

• Vorticity

• Di usivity

• Dissipation

• Self Sustaining

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
ff
 Characteristics Of Turbulent Flow
• The most important characteristic of turbulent motion is the fact that velocity and pressure at a
point uctuate with time in a random manner.

• The mixing in turbulent ow is more due to these uctuations. As a result we can see more uniform
velocity distributions in turbulent pipe ows as compared to the laminar ows.

• Turbulent ow is di usive and dissipative . In general, turbulence brings about better mixing of a
uid and produces an additional di usive e ect. Such a di usion is termed as "Eddy-di usion ".
( Note that this is di erent from molecular di usion). At a large Reynolds number there exists a
continuous transport of energy from the free stream to the large eddies. Then, from the large eddies
smaller eddies are continuously formed. Near the wall smallest Eddie destroy themselves in
dissipating energy, i.e., converting kinetic energy of the eddies into intermolecular energy.

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fl
fl
fl
ff
ff
fl
ff
fl
ff
ff
fl
ff
fl
ff
 Turbulence: Issues
The Navier-Stokes equations are deterministic.

Turbulent ows are unpredictable.

• Velocity eld varies signi cantly and irregularly in both position and time

• Deterministic chaos – hypersensitivity to initial conditions

• Nonlinear dynamical system with very large number of degrees of freedom

• Importance of nonlinearity depends on Reynolds number

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fi
fl
fi
 Introduction: Kolmogorov Scales
• Kolmogorov’s 1st Similarity Hypothesis, the only factors in uencing the behaviour of the small scale motions are the
overall kinetic energy production rate (which equals to the dissipation rate) and the viscosity.

• The dissipation rate is independent of viscosity but scales at which this energy is dissipated will depend on both the
dissipation rate and viscosity

2 3 2
• The dissipation rate per unit mass (ϵ) has dimensions m /sec and viscosity m /sec, so

ν 3 14
• η = ( ϵ ) , This is called the Kolmogorov Length scale and is the smallest hydrodynamic length scale in turbulent
ow

2 L
• KE per unit mass scales with U and Large eddy turn over time is U , So, dissipation rate per unit mass (ϵ) is
U2 U3
≈ ≈
L/U L
L UL
3

η ( ν )
3 4
Ratio of the Largest (L) to smallest scales (η) in the ow is given by ≈ = Re 4
•
Computational Aerodynamics Lab, Department of Mechanical Engineering
ZHCET, AMU, ALIGARH
fl
fl
fl
Introduction: Kolmogorov Scales
L
• Large eddy turn over time is tt =
U
ν 12 νL
• Time scale for small eddies using viscosity and dissipation tη = ( ) = ( 3)
ϵ U
• Ratio of the two time scales will be
tt UL 12
=( )
1
= Re 2
• tη ν

• So large scale Structures appear for more time than small scale structures
Computational Aerodynamics Lab, Department of Mechanical Engineering
ZHCET, AMU, ALIGARH
 Introduction: Basic Features
Multi-scale
•Signi cant energy over many scales – scale separation an idealisation.
•Large scale ow depends on geometry and external forcing
•Small scale ow – “universal character”
•Period – Scale – Energy relationship:

small scales, high frequency, low energy

large scales, low frequency, high energy
The small scale eddies are of the order of Kolmogrov length scale For air(υ = 1.5e-5 m /s) owing at
1

U=2.4m/s on a plate of length L=2.4 and width 1.2m, large scale is about 40mm while small scale is
about 0.05mm.

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fi
fl
fl
fl
 Turbulence physics: the energy cascade
π
Rate of extraction of
TKE from mean ow (Richardson 1922, Kolmogorov 1941)
per unit mass
Rate of dissipation of
TKE ϵ Extracts TKE from the mean ow UL
Inertia force ReL = >>1
ν

1 2
2
mu 1
π≡ ×
m tscale
u3
π≡
l
v2
ϵ ≈ νeijeij ≈ ν
ηk

vηk
Viscous force Reηk = ≈1
ν
Energy mopped up due to viscous dissipation

Energy dissipation at small scales = Energy injection at integral Scale

fl
fl
 Introduction: Basic Features
Vortical
• Turbulence is characterised by high degree of uctuating vorticity
• Kinetic energy density greatest at large scales
• Enstrophy at small scales

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fl
Navier-Stokes equations, incompressible, Newtonian
Why Turbulent ows are challenging?

• Unsteady aperiodic motion

• Fluid properties exhibit random spatial variations (3D)

• Strong dependence from initial conditions

• Contain a wide range of scales (eddies)

The implication is that the turbulent simulation MUST be always three-

dimensional, time accurate with extremely ne grids
Computational Aerodynamics Lab, Department of Mechanical Engineering
ZHCET, AMU, ALIGARH
fl
fi
 Velocity Pro le
• The turbulent boundary layer continues to grow in thickness, with a small
region below it called a viscous sub-layer. In this sub layer, the ow is well
behaved, just as the laminar boundary layer

Critical Re Recr = 300000

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fi
fl
 How to solve Turbulent Flow?

• Direct Numerical Simulation (DNS), Course mesh DNS

• Large Eddy Simulation (LES), Detached Eddy Simulation (DES)

• Reynolds Averaged Numerical Simulation (RANS), Unsteady

Reynolds Averaged Numerical Simulation (URANS)

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
 Direct Numerical Simulations (DNS)

• The objective is to solve the time-dependent NS equations

resolving.
• ALL the scale (eddies) for a su cient time interval so that the uid
properties reach a statistical equilibrium.

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
ffi
fl
Direct Numerical Simulation
DNS - pseudo-spectral calculation method
DNS - pseudo-spectral calculation method

40963 : world record (Earth Simulator, Japan, 2012, 1.08 – Peta Flops), 40 TB RAM
40963 : world record (Earth Simulator, Japan, 2021, 19.05 – Peta Flops), GPU NVDIA & AMD EPYC
Direct Numerical Simulation of
Turbulent Flow
❑ Computational effort= no. of computations that need to be
performed ≈ Re × Re × Re

❑ Typical engineering flow Re 6 8

≈ 10 − 10
❑ Too expensive!

❑ For many engineering flows, we resort to computing merely the

mean flow or the filtered flow.
Flow over Backward Facing Step

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
Statistical and Filtered Methods
of Turbulence
❑ Statistical/Averaged Methods
❖ Reynolds-averaged Navier-Stokes- RANS (Incompressible Flows)
❖ Favre-averaged Navier-Stokes (Compressible Flows)
❖ Lagrangian Stochastic Methods

❑ Filtered Methods
❖ Bridging Methods
Partially-averaged Navier-Stokes, Partially Integrated NS
❖ Large Eddy Simulation

❑ All statistical/filtered methods require turbulence models

 Averaging: Time and Space
• Stationary Turbulence: Time Average
tl
1
tl→∞ 2tl ∫−t
ū = lim u(x0, t)dt
l

• Space Average for homogenous turbulence

l
1
l→∞ 2l ∫−l
⟨u⟩ = lim u( x , t0)dx

In our analysis, average of any quantity will be evaluated as a time average. Take a nite time
interval tl. This interval must be larger than the time scale of turbulence. Needless to say that it
must be small compared with the period ts of any slow variation (such as periodicity of the mean
ow) in the ow eld that we do not consider to be chaotic or turbulent.
Computational Aerodynamics Lab, Department of Mechanical Engineering
ZHCET, AMU, ALIGARH
fl
fl
fi
fi
Computational Aerodynamics Lab, Department of Mechanical Engineering
ZHCET, AMU, ALIGARH
 Reynolds Decomposition
• For a parallel ow, as in case of pipes, u(y, t) = ū(y) + u′( x , t)
• Generalising this concept, we can say

u ( x , t) = ⟨ u ( x , t)⟩ + u ′( x , t)

• Any Instantaneous quantity = Mean Component + Fluctuating Component

• Thus for parallel ows

u(y, t) = ū(y) + u′( x , t),

v(y, t) = 0 + v′( x , t)

w(y, t) = 0 + w′( x , t)
tl tl

∫−t ∫−t
u′dt = 0; u′v′dt ≠ 0
l l

Computational Aerodynamics Lab, Department of Mechanical Engineering




ZHCET, AMU, ALIGARH





fl
fl
Traditional Approach : Reynolds
Decomposition
Reynolds-Averaged Navier-Stokes Equation
Instantenous NS equations
2
∂U i ∂U i ∂p ∂ Ui
ρ + ρU k =− +µ
∂t ∂xk ∂xi ∂xk xk
∂ p 2
∂U i ∂U j
= −ρ
∂xk ∂xk ∂x j ∂xi
# of equations: 4. # of unknowns: 4 (U i , p )

Reynolds-averaged NS (RANS) equations

' '
φ = φ +φ where φ ∈ {U i , p} φ = Mean. φ = fluctuation
' '
∂ Ui ∂ Ui ∂ p ∂ 2
Ui ∂ ρu ui k
ρ + ρ Uk =− +µ +
∂t ∂xk ∂xi ∂xk xk ∂xk
2 ' '
∂ 2
∂U i ∂U j
p ∂ ρu u i k
= −ρ −
∂xk ∂xk ∂x j ∂xi ∂xi ∂x j
# of equations: 4. # of unknowns: 4+6=10.
{U i , p , ρu u '
i
'
k }
Reynolds Averaged Navier-Stokes (RANS):Rules
• Rules for averaging f and g are two dependent variables and if s denotes anyone of the independent variables x, y.
∂f ∂ ¯
f
∫ ∫
¯ ; =
f¯ = f¯; f + g = f¯ + ḡ;fg = f¯ḡ; fds = fds
∂s ∂s
For, incompressible ows, the Navier-Stokes equations can be rearranged in the form:

∂u ∂u 2 ∂uv ∂uw ∂p
ρ[ +
∂z ]
2
+ + = − + μ ∇ u
∂t ∂x ∂y ∂x

∂v ∂uv ∂v 2 ∂vw ∂p
ρ[ + + + ] =− 2
+ μ∇ v
∂t ∂x ∂y ∂z ∂x
2
∂w ∂uw ∂wv ∂w ∂p
ρ[ + + + ] =− 2
+ μ∇ w
∂t ∂x ∂y ∂z ∂z
∂u ∂v ∂w
And + + =0
∂x ∂y ∂z

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fl
 Averaging: Continuity Eqn
• Express the velocity components and pressure in terms of time-mean values and corresponding uctuations. In continuity equation, this
substitution and subsequent time averaging will lead to

∂(ū + u′) ∂(v̄ + v′) ∂(w̄ + w′)

+ + =0
∂x ∂y ∂z

∂(ū + u′) ∂(v̄ + v′) ∂(w̄ + w′)

+ + =0
∂x ∂y ∂z

∂(ū + u′) ∂(v̄ + v′) ∂(w̄ + w′)

+ + =0
∂x ∂y ∂z

∂(ū + ū′) ∂(v̄ + v̄′) ∂(w̄ + w̄′) ∂ū ∂v̄ ∂w̄ ∂ū′ ∂v̄′ ∂w̄′
+ + = 0 or + + + + + =0
∂x ∂y ∂z ∂x ∂y ∂z ∂x ∂y ∂z
∂ū ∂v̄ ∂w̄ ∂u′ ∂v′ ∂w′
∴ + + = 0 and by subtracting mean continuity from instantaneous equation ∴ + + =0
∂x ∂y ∂z ∂x ∂y ∂z

It is evident that the time-averaged velocity components and the uctuating velocity components, each satisfy the continuity equation
for incompressible ow.

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH


















fl
fl
fl
 Averaging: Momentum Eqn
• Express the velocity components and pressure in terms of time-mean values and corresponding uctuations for momentum equations

∂u ∂u 2 ∂uv ∂uw ∂p
ρ[ + + + ] = − + μ ∇ 2
u
∂t ∂x ∂y ∂z ∂x

∂ū ∂ū′ ∂(ū + u′)2 ∂(ū + u′)(v̄ + v′) ∂(ū + u′)(w̄ + w′) ∂p
ρ[ + + + + ] = − + μ ∇ 2
u
∂t ∂t ∂x ∂y ∂z ∂x

∂ū ∂ū′ ∂(ū2 + u ′2 + 2ūu′) ∂(ūv̄ + u′v̄ + v′ū + u′v′) ∂(ū w̄ + u′w̄ + w′ū + u′w′) ∂( p̄ + p′)
ρ[ + + + + ] = − + μ ∇2
u
∂t ∂t ∂x ∂y ∂z ∂x

∂ū ∂ū′ ∂(ū2 + u ′2 + 2ūu′) ∂(ūv̄ + u′v̄ + v′ū + u′v′) ∂(ū w̄ + u′w̄ + w′ū + u′w′) ∂( p̄ + p′)
ρ[ + + + + ] = − + μ ∇2
u
∂t ∂t ∂x ∂y ∂z ∂x

∂ū ∂ū′ ∂(ū2 + u ′2 + 2ūu′) ∂(ūv̄ + u′v̄ + v′ū + u′v′) ∂(ū w̄ + u′w̄ + w′ū + u′w′) ∂( p̄ + p′)
ρ[ + + + + ] = − + μ ∇2
u
∂t ∂t ∂x ∂y ∂z ∂x

∂ū ∂(ū2 + u ′2 + 2ūū′) ∂(ūv̄ + u′v̄ + v′ū + u′v′) ∂(ū w̄ + u′w̄ + w′ū + u′w′) ∂( p̄ + p′)
ρ[ + + + ] = − + μ ∇ 2
u
∂t ∂x ∂y ∂z ∂x

∂ū ∂(ū2 + u ′2) ∂(ūv̄ + u′v′) ∂(ū w̄ + u′w′) ∂( p̄)

X-momentum: ρ[ + + + ] = − + μ ∇2
ū
∂t ∂x ∂y ∂z ∂x

As ū′ = 0; v̄′ = 0; ; w̄′ = 0; p̄′ = 0; ūv̄ = ū¯ v̄¯ = ūv̄; ū w̄ = ū¯ w̄

¯ = ū w̄;

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH






























































fl
 Averaging: Momentum Eqn
• Express the velocity components and pressure in terms of time-mean values and corresponding
uctuations for momentum equations

2
∂v ∂uv ∂v ∂vw ∂p
• [ ∂t ∂z ]
2
ρ + + + = − + μ ∇ v
∂x ∂y ∂x
2
∂v ∂uv ∂v ∂vw ∂p
ρ[ + + + ] = − + μ ∇2
v
∂t ∂x ∂y ∂z ∂x
2 ′2
∂v̄ ∂(ūv̄ + u′v′) ∂(v̄ + v ) ∂(v̄w̄ + v′w′) ∂( p̄)
Y-momentum: ρ[ + + + ] = − + μ ∇2
v̄
∂t ∂x ∂y ∂z ∂y
2 ′2
∂w̄ ∂(ūw̄ + u′w′) ∂(v̄w̄ + v′w′) ∂(w̄ + w ) ∂( p̄)
Z-momentum: ρ[ + + + ] = − + μ ∇2
w̄
∂t ∂x ∂y ∂z ∂z

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH










fl
 RANS Equation
2 ′2
∂ū ∂ū ∂ūv̄ ∂ūw̄ ∂p̄ ∂u ∂u′v′ ∂u′w′)
X-momentum: ρ[ + + + ] =− + μ ∇ ū − ρ[
2
+ + ]
∂t ∂x ∂y ∂z ∂x ∂x ∂y ∂z

∂v̄ ∂ūv̄ ∂v̄2 ∂v̄w̄ ∂p̄ ∂u′v′ ∂v ′2

∂u′w′
Y-momentum: ρ[ + + + ] = − + μ ∇ 2
v̄ − ρ [ + + ]
Reynold Stresses
∂t ∂x ∂y ∂z ∂y ∂x ∂y ∂z
2 ′2
∂w̄ ∂ūw̄ ∂v̄w̄ ∂w̄ ∂p̄ ∂u′v′ ∂u′w′ ∂w
Z-momentum: ρ[
∂z ] [ ∂x ∂z ]
2 Reynold Stresses
+ + + = − + μ ∇ w̄ − ρ + +
∂t ∂x ∂y ∂z ∂y

D V̄ ⃗ Reynold Stresses
ρ = − ∇ p̄ + ∇ . τij
Dt

τij = τijLaminar + τ′ijTurbulent

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
















 Turbulence Modeling
• De ne the Reynolds stresses in terms on known (averaged) quantities
1. Boussinesq hypothesis

• simple relationship between Reynolds stresses and velocity gradients through the eddy viscosity (similar to
molecular viscosity)

• isotropic (eddy viscosity is a scalar!)

2. Reynolds stress transport models

• equations derived directly manipulating the NS equations

• still contain unknown (undetermined) quantities
• no assumption of isotropyvery complicated and expensive to solve
3. Non-linear Eddy viscosity models (Algebraic Reynolds stress)

4. Model directly the divergence of the Reynolds Stresses

•

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fi
 Boussinesq Hypothesis
The Boussinesq hypothesis, often called the “Boussinesq approximation”—
relate turbulent shear stress to the mean ow strain rate. (This is also termed
“gradient transport,”)Thus, for example, we might hypothesize that:
μt 1 ∂⟨Ui⟩ ∂⟨Uj⟩
Rij = − ⟨u′i u′j⟩ = 2 ⟨Sij⟩ With ⟨Sij⟩ = ( + )
ρ 2 ∂xj ∂xi
∂⟨Ui⟩ ∂⟨Ui⟩ 1 ∂⟨p⟩ ∂ (μ + μt) ∂⟨Ui⟩
∂t
+ ⟨Uj⟩
∂xj
=−
ρ ∂xi
+
∂xj ( ρ ∂xj )

How to de ne eddy viscosity? Not a property of uid

2
• Dimensional Arguments: Units are m /s, de nes two out of three scales: velocity, length and time

• Physical arguments: Asymptotic analysis and consistency with experiments

• Numerical arguments: simple and easy to compute

Computational Aerodynamics Lab, Department of Mechanical Engineering


ZHCET, AMU, ALIGARH
fi
fl
fi
fl
 Classi cation
Based o dimensiona analysis, μt ca b determined from a
turbulence time scal (or velocity scale) and a length scale.
Each model tries to calculate μt di erently
The various models (about 200) are classi ed in terms of number of transport
equations solved in addition to the RANS equations:

• zero-equation/algebraic models: Mixing Length, Cebeci-Smith, Baldwin-

Lomax, etc

• one-equation models: Wolfstein, Baldwin-Barth, Spalart-Allmaras, k-model, etc

• two-equation models: k-ε, k-ω, k-τ, k-L, etc.
• three-equation models: k-ε-A
• four-equation models: v -f model 2

Computational Aerodynamics Lab, Department of Mechanical Engineering

•
ZHCET, AMU, ALIGARH
n
fi
e
l
ff
fi
n
e
 Algebric / Zero Equation Model: Prandtl’s Mixing
Model
2 2
μt = ρLmixS = ρLmix 2⟨Sij⟩⟨Sij⟩
• From dimensional arguments and analogy with molecular transport

• De nition of L is di erent for each problem (boundary layes, mixing layers,

etc.)

• Eddy viscosity is zero if the velocity gradients are zero

• No “history” e ect; purely local

• L can be made “universal” using ad hoc functions of distance from the

•
walls, pressure gradients, etc.
Computational Aerodynamics Lab, Department of Mechanical Engineering
ZHCET, AMU, ALIGARH
fi
ff
ff
 Algebric / Zero Equation Model: Prandtl’s Mixing
Model
• Consider a fully developed turbulent
boundary layer. The stream wise mean
velocity varies only from streamline to
streamline. The main ow direction is
assumed parallel to the x-axis (Fig 11)

• The time average components of velocity

are given by ⟨u⟩ = ⟨u(y)⟩, ⟨v⟩ = 0, ⟨w⟩ =
0 . The uctuating component of
transverse velocity transports mass and
momentum across a plane at y from the 1

wall. The shear stress due to the

uctuation is given by
∂⟨u⟩
•
τt = − ρ⟨u′v′⟩ = μt
∂y
Computational Aerodynamics Lab, Department of Mechanical Engineering


ZHCET, AMU, ALIGARH
fl
fl
fl
 Algebric / Zero Equation Model: Prandtl’s Mixing
Model
• Fluid, which comes to the layer yl from a layer
(yl − l) has a positive value of v′. If the lump of uid
retains its original momentum then its velocity at its
current location yl is smaller than the velocity
prevailing there. The di erence in velocities is then

( ∂y )
∂⟨u⟩
Δu1 = ⟨u(yl)⟩ − ⟨u(yl − l)⟩ or Δu1 ≈ l
yl

• The above expression is obtained by expanding the

function ⟨u(yl − l)⟩ in a Taylor series and neglecting
all higher order terms and higher order derivatives. l
is a small length scale known as Prandtl's mixing
length . Prandtl proposed that the transverse
displacement of any uid particle is, on an average, l.

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH

fl
ff
fl
 Algebric / Zero Equation Model: Prandtl’s Mixing
Model
• Consider another lump of uid with a negative value of v′.
This is arriving at yl from (yl + l) If this lump retains its
original momentum, its mean velocity at the current lamina
yl will be somewhat more than the original mean velocity
of yl. This di erence is given by
The di erence in velocities is then

( ∂y )
∂⟨u⟩
Δu2 = ⟨u(yl + l)⟩ − ⟨u(yl)⟩ or Δu2 ≈ l
yl

• The velocity di erences caused by the transverse motion

can be regarded as the turbulent velocity components at
yl.

• We calculate the time average of the absolute value of this

1 ∂⟨u⟩
uctuation as | u′| = ( | Δu1 + Δu2 | ) = l (
2 ∂y ) yl

Computational Aerodynamics Lab, Department of Mechanical Engineering


ZHCET, AMU, ALIGARH

fl
ff
ff
ff
fl
 Algebric / Zero Equation Model: Prandtl’s Mixing
Model
• Suppose these two lumps of uid meet at a layer yl. The
lumps will collide with a velocity 2u′ and diverge. This
proposes the possible existence of transverse velocity
component in both directions with respect to the layer at
yl. Now, suppose that the two lumps move away in a
reverse order from the layer yl with a velocity 2u′. The
empty space will be lled from the surrounding uid
creating transverse velocity components which will again
collide at yl . Keeping in mind this argument and the
continuity for uctuating velocities, we may state that
∂⟨u⟩
| v′| ≈ | u′| ⇒ | v′| = Const | u′| ⇒ | v′| = Const l (
∂y ) yl
along with the condition that the moment at which u′ is
positive, v′ is more likely to be negative and conversely
when u′ is negative. Possibly, we can write at this stage
2

( ∂y )
2 ∂⟨u⟩
⟨ | u′v′| ⟩ = − Const1 | u′| | v′| ⇒ ⟨ | u′v′| ⟩ = − Const2 l

Computational Aerodynamics Lab, Department of Mechanical Engineering












ZHCET, AMU, ALIGARH





fl
fi
fl
fl
 Algebric / Zero Equation Model: Prandtl’s Mixing
Model
• Constants Const1 & Const2are di erent proportionality
constants. However, these constants can now be included in still
2

( ∂y )
∂⟨u⟩2
unknown mixing length⟨u′v′⟩ ≈ − l

• The turbulent shear stress τt we may write after comparing this

expression with the eddy viscosity
∂⟨u⟩ ∂⟨u⟩ ∂⟨u⟩
∂y ( ∂y ) ( ∂y )
2
τt = ρl = μt where the apparent

viscosity and the apparent kinematic viscosity is given by may

2 ∂⟨u⟩ 2 ∂⟨u⟩
be expressed as μt = ρl or νt = l
∂y ∂y

• The decision of expressing one of the velocity gradients with

sign and the other without sign was to assign the τt the same
•
sign as the velocity gradient.

Computational Aerodynamics Lab, Department of Mechanical Engineering



ZHCET, AMU, ALIGARH
ff
 Algebric / Zero Equation Model: Prandtl’s Mixing
Model
• Note that the apparent viscosity and
consequently, the mixing length are not
properties of uid. They are dependent
on turbulent uctuation. But how to
determine the value of “l” the mixing
length? Several correlations, using
experimental results for τt have been
proposed to determine “l”. However, so
far the most widely used value of
mixing length in the regime of isotropic
turbulence is given as l = χy. Where y
is the distance from the wall and χ is
known as von Karman constant and for
parallel ow is 0.4

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fl
fl
fl
 Structure of Boundary Layer
Universal Law (Velocity Pro le)

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH
fi
 Thanks

Computational Aerodynamics Lab, Department of Mechanical Engineering

ZHCET, AMU, ALIGARH