Documente Academic
Documente Profesional
Documente Cultură
100
Outline for Part-I
101
Introduction
102
Introduction (continued)
• What is turbulence?
– Fluid flow occurs primarily in two regimes: laminar and
turbulent flow regimes.
– Laminar flow:
• smooth, orderly flow restricted (usually) to low values of
key parameters- Reynolds number, Grashof number,
Taylor number, Richardson number.
– Turbulent flow:
• fluctuating, disorderly (random) motion of fluids
103
Introduction (continued)
105
Introduction ( continued)
106
Introduction (Continued)
• Energy Cascade Concept:
• Fluctuations are sustained by vortex stretching under the
action of shear.
107
Laminar Flow Examples
• Velocity (fluctuations): u
• Length (eddy size):
• Time, τ = /u
• Turbulence Reynolds
number
– Ret = u /ν
• Turbulent kinetic energy:
Large eddies in a turbulent k~ 3u2/2
boundary layer (Tennekes • Dissipation rate: ε ~ u3/
and Lumley, 1992): • Kolmogorov scales
– τK = (ν/ε)1/2
~ Lt = boundary layer thickness
– K = (ν3/ε)1/4
– uK = (νε)1/4
110
Introduction: Why turbulence modeling?
112
Transport Equations: Conserved scalar
Jn ∂
Je = Jw + (J w )∆ x
∂x
∆y Similarly for Jn, Js, Jtop, Jbot
Jw ( ρ φ∆Vol.) Je
æ ∂φ ö
J = çm φ e − Ar
∆x è ∂x
Js Convection + Diffusion
= ρUA; mass flow rate
m
• Control volume balance: {time rate of change} = {net flux through the surfaces}
∂ (ρ φ ∆ x ∆ y ∆ z )
= (J e − J w ) + (J n − J s ) + (J top − J bot ) + source − sin(k)
∂t
∂ (ρφ )
+ ∇
"⋅ (ρu"
φ) = ∇
! "⋅ (Γ∇
"φ ) + Sφ
!
∂t! convention diffusion net
Time Rate source
113 of Change
Governing Equations: Conservation of Mass
∂ρ æ ∂ρ ö Dρ
= ∇ ⋅ (ρ U ) = ç + U ⋅ ∇ρ + ρ ⋅ ∇U = + ρ∇ ⋅ U
∂t è ∂ t Dt
114
• Incompressible fluids: ∇.(U) =0 <=> D(ρ)/dt =0
Navier Stokes Equations
∂
• e.g. (m U )east = (m U )west + (m U )∆x
∂x
116
Energy Equation
∂ (ρh ) ∂ (ρU j h ) ∂ (q j ) ∂P ∂P
+ =− +Φ+ + Uj
∂t ∂x j ∂x j ∂t ∂x j
∂T
dh = C p dT; q j = −k ( Heat flux in j - direction);
∂x j
Φ = Viscous dissipation function, h = enthalpy, T = Temperature
∂u i
Φ = τij
∂x j
2
τij = 2µSij − ∇ ⋅ U δij (does not include pressure)
3 ~
117
Stress Tensor
• Newtonian Fluids
éτ xx τ xy τ xz ù
ê
[τ ]
ij = êτ yx τ yy τ yz : Symmetric 2nd Order Tensor
êτ τ τ (ij = ji, e.g. xy = yx)
ë zx zy zz
• Momentum equation
∂ (ρU i ) ∂ (ρU i U j ) ∂P ∂ (2µSij )
+ =− + + ρg i
∂t ∂x j ∂x i ∂x j
1 æç ∂Ui ∂Uj ö
s ij = ç +
2 è ∂x j ∂x i
• Thermal energy equation cp , cv = constant
∂ ∂ ∂ æç ∂T ö ∂U i
(ρc pT ) + (U j .ρc pT ) = ç
kh + τ ij + ΦT
∂t ∂x j ∂x j è ∂x j ∂x j
119
Averaging Techniques: Reynolds Averaging
120
Averaging Techniques: Reynolds Averaging
• Phase Averaging:
∆t
< U (x , t ) >=
1 2
U ( x , t + τ )d τ
∆t − ∆t
2
– t = window width
121
Averaging Techniques: Favre averaging
122
Averaging Rules: Averaging
• < U + V > = <U> + <V>; < <U> > = <U>; <U><V>> = <U><V>
Structure of turbulent flow in a pipe (a) Shear stress (b) Average velocity
126
Typical shear flows: Mean flow
Experimental rough-pipe velocity profiles, showing the Composite plot of the profile-shift parameter ∆B(k+) for
downward shift ∆B of the logarithmic overlap layer various roughness geometries, as complied by Clauser (1956)
Boundary layer velocity profiles for rough walls Notation v* = U*=U
1 +
U+ = ln y + + B − ∆B( k s ) ; U+ = U/U*; y* = yU*/
κ
1 1 y
∆B ≅ ln(1 + 0.3k s+ ) for k s+ > 60; U + = ln( ) + 8.5
129 κ κ ks
Fluctuating Velocities in a boundary Layer
130
Classical Models: Assumptions
• Dimensional analysis:
t = lchuch ; t = t/ ; not a fluid property
131
Classical Models: Assumptions
• In general:
2
τ = 2 µSij − µ∇ ⋅ uδ ij − Pδ ij ; Laminar Stresses
l
ij
3
τ ijt = − ρ (u i u j ) = 2 µ tSij − µ t ∇ ⋅ uδ ij − ρkδ ij ;
2 2
turbulent stresses
3 3
u i u i = (u 2 + v 2 + w 2 ); turbulent kinetic energy
1 1
k=
2 2
1 æ ∂ U ∂U j ö
Sij = ç i +
2 çè ∂x j ∂x i ; rate of deformation (strain) tensor
ì0, when i ≠ j
δ ij = í ; Kronecker’s delta
1, when i = j
132
Zero Equation models: Mixing length models
• von Karman
uch = lmix (dU/dy) ; lmix = (dU/dy) / (d2U/dy2)]
– lmix = y for y < yo ; (inner region)
lmix = for y > yo ; (outer region) ;
= shear layer thickness, 0.20
133
Zero Equation models: Mixing length models
135
Eddy Viscosity Distribution
− κB é z Z2 ù + κu
µ t = µκe ê e − 1 − Z − Z = κu =
ë 2 ν*
µ t ,outer ≈ Cρ U e δ * C ≈ 0.016
0.016ρU e δ *
µ t ,outer ≈
é æ y ö
6
ù
ê1 + 5.5ç δ÷
ë è ø
136
Zero Eq. Models: Baldwin-Lowmax model
Comparison of computed and measured boundary layer velocity profiles and shape factor for nonzero pressure
gradient; Cebeci-Smith model (After Wilcox, 1993)
139
Zero Eq. Models: examples
140
Zero Eq. Models: examples
Velocity profiles and pressure distribution Measurements and calculations (Cebeci-Smith model),
------basic model; with curvature effects; __________ with low-Rθeffects, ……. With
extrapolated pressure distribution (After Celik and Patel, 1984):
141
Zero Eq. Models: assessment
• Mathematically very simple, numerically robust
• Acceptable results even with separation and non-equilibrium flows (e.g. one and
1/2 Eq. Model of Johnson-King)
• Not good for complex flow effects such as curvature, history effects.
• Comment: These models will work well only for the cases for which they have
been fine tuned (note a different constant for each flow in Table ??).
142
One Eq. models:
Badwin-Bart Model
æ ∂υ t ∂υ t ö
= t; Pυ t = (same as in Menter model); Dissυ t = c 2 ρ ç
ç ∂x ∂x
è j j
Comment: no distance from the surface is required
Prandtl’s Model
∂U ∂U
= k; t = k1/2 l Pk = −τ ∂x = − u i u j ∂x ; Dissk = = cD k3/2 / l ; l has to be supplied
t i i
ij
j j
• The length scale relations are empirical, and not so suitable for more general,
complex industrial flows.
• The troublesome -equation is not solved. This is a big advantage if length scale
relations are satisfactory.
145
The Standard k-εε Model
A phenomenological approach
k-equation ε-equation
D(ρk ) D(ρε)
= Diff ( k ) + Pk − ρε = Diff ( ε ) + ρPε − ρD ε
Dt Dt
∂ éæ µ t ö ∂k ù ∂ éæ µ t ö ∂ε ù
Diffusion = Diff ( k ) = êç µ + ÷÷ Diff ( ε ) = ê ç µ + ÷
∂x j ëêçè σ k ø ∂x j ∂x j êëçè σε ÷ø ∂x j
∂u i ∂u C P C ε
Pr oduction = Pk = −ρ u i , u j = τijt i Pε = ε1 k ; D ε = ε 2 ; k = Time scale
ε
∂x j ∂x j k/ε k/ε
k2
∂u i ∂u j Eddy Viscosity = ν t = Cµ
Dissipation = ε = ν , ε
∂x j ∂x i
æ ∂u ∂u ö 2 Empirical constants are determined from asymptotic
τijt = ρν t ç i + j − ρkδij
ç ∂x ∂x 3 analysis, experiment s, and optimization.
è j i
Cµ = 0.09 C ε1 = 1.44 Cε 2 = 1.92 σ k = 1.0 σ ε = 1.3
146
Examples: k-ε Model
Comparison of the k-ε model with the flat plate boundary layer data
of Klebanoff (1955). [After Jones and Launder (1972).]
147
Examples: k-ε Model
Flow over 35° swept wing; Re = 2.42 × 106. [After Johnston (1987).]
Note: experiments indicate separated flow. All models predict attached flow.
148
Examples: k-ε Model
149
Examples: k-ε Model
150
Assessment of the Standard k-ε Model
•Robust, easy to apply, the most commonly used method
•Easily extendible to complex geometries and multi-
dimensions
• Fairly good results for many engineering applications,
especially good for trend analysis
•Isotropic and inherently assumes “local” equilibrium,
hence good for high Reynolds number flows.
•Not good for flows involving significant curvature,
rotation, buoyancy, sudden acceleration, separation, and
low Re regions
151
Summary: Part-I
•An introduction to fluid flow regimes, governing
equations, and turbulence modeling is presented
•The zero- and one-equations models using the mixing-
length or eddy-viscosity concept are reviewed.
•These models are good for calculations of boundary or
thin shear-layer flows such as wakes, jets, flow over
airfoils etc. with simple geometry but without separation.
•The standard k-ε model is introduced using a
phenomenological approach. This model is a popular
model, easy to use and robust, but has many
shortcomings.
•All need further improvements for accurate predictions
152