Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Turbulent Flow

Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

COMSOL Multiphysics ®

Turbulent Flow Modeling


Turbulent Flow
Determining the Flow Regime, Re
 At low 𝑅𝑅𝑅𝑅 (≪ 1): creeping (Stokes) flow
Viscous forces tend to damp out all flow disturbances
Reversible smooth flow pattern

 At intermediate 𝑅𝑅𝑅𝑅 ~1 − 2000∗ : laminar flow


Inertial forces become increasingly important
Viscous forces are confined to boundary layers, shear layers, and wakes
Regular, smooth flow pattern

 At high 𝑅𝑅𝑅𝑅 (> 4000∗ ): turbulent flow


Flow disturbances grow by nonlinear interactions producing a cascade of eddies (vortices)
Viscous damping is active everywhere but only has a significant effect on the
smallest eddies
Disordered (chaotic) flow pattern
*Critical Re of 2000 has been mentioned above as an example considering an internal channel flow
Reynolds Number Examples

𝑈𝑈0 𝑥𝑥
𝑅𝑅𝑅𝑅𝑥𝑥 =
𝜈𝜈
𝑈𝑈0 𝐿𝐿0
𝑅𝑅𝑅𝑅𝐿𝐿 =
𝜈𝜈
𝑈𝑈0 𝐷𝐷𝐻𝐻 4𝐴𝐴
𝑥𝑥 𝑅𝑅𝑅𝑅𝐷𝐷ℎ = , 𝐷𝐷𝐻𝐻 ≡
𝐿𝐿0 𝜈𝜈 𝑃𝑃
𝑈𝑈0 𝐷𝐷
𝑅𝑅𝑅𝑅𝐷𝐷 =
𝜈𝜈

𝐷𝐷
Determining the Flow Regime, Ma
 For 𝑀𝑀𝑀𝑀 = 0: formally incompressible flow
Speed of sound is infinite (hypothetical case), instantaneous spread of pressure disturbances

 For 0 < 𝑀𝑀𝑀𝑀 < 0.3: weakly compressible flow


Density changes due to pressure by max. 5%
Density changes can also occur due to the dependency on temperature

 For 𝑀𝑀𝑀𝑀 > 0.3: compressible flow


Thermodynamic effects not negligible
Transonic, supersonic flow with shock waves
Choosing the Right Fluid Flow Interface
𝑴𝑴𝑴𝑴 < 𝟎𝟎. 𝟑𝟑
𝑅𝑅𝑅𝑅 ≪ 1
𝑅𝑅𝑅𝑅~1 − 𝑅𝑅𝑒𝑒𝑐𝑐𝑐𝑐
𝑅𝑅𝑅𝑅 > 𝑅𝑅𝑒𝑒𝑐𝑐𝑐𝑐
𝑅𝑅𝑅𝑅 > 𝑅𝑅𝑒𝑒𝑐𝑐𝑐𝑐
𝑅𝑅𝑅𝑅 > 𝑅𝑅𝑒𝑒𝑐𝑐𝑐𝑐

𝑴𝑴𝑴𝑴 > 𝟎𝟎. 𝟑𝟑


𝑅𝑅𝑅𝑅 < 𝑅𝑅𝑒𝑒𝑐𝑐𝑐𝑐
𝑅𝑅𝑅𝑅 > 𝑅𝑅𝑒𝑒𝑐𝑐𝑐𝑐
𝑅𝑅𝑒𝑒 → ∞
Laminar, Transitional, and Turbulent Flow

u(y)
Log layer
u∞ u(y)
Buffer layer
Viscous sublayer

Laminar Transition Turbulent


Introduce turbulent viscosity term (𝜇𝜇 𝑇𝑇 )

 Reynolds Averaged Navier-Stokes (RANS) equations:


ρ
∂u
∂t
( )
+ ρ u ⋅ ∇u + ∇ ⋅ (ρ u′ ⊗ u′) = −∇P + µ ∇u + (∇u ) + F
T

ρ (∇ ⋅ u ) = 0
 Model the turbulent fluctuations using turbulent viscosity µT :
(
− µT ∇u + (∇u )
T
)

ρ
∂u
∂t
( )
+ ρ u ⋅ ∇u = −∇P + (µ + µT ) ∇u + (∇u ) + F
T

ρ (∇ ⋅ u ) = 0

In COMSOL: 9 different turbulence models to solve for 𝜇𝜇 𝑇𝑇 = spf.muT


Computational Cost for
Turbulence Models
 Algebraic yPlus and L-VEL
 Spalart–Allmaras
 k-epsilon
 k-omega, SST and
low Re k-epsilon
 Realizable k-epsilon
 v2-f
Increasing cost
Algebraic Turbulence Models
 Algebraic turbulence models are faster and more robust but, generally less accurate than
transport-equation turbulence models
 Reynolds number based on local velocity magnitude and wall distance

𝜌𝜌𝑈𝑈𝑈𝑈
̅ 𝑈𝑈 𝜌𝜌𝑢𝑢
̅ 𝜏𝜏 𝑦𝑦 𝜇𝜇̅
𝑅𝑅𝑅𝑅 = = = 𝑢𝑢+ 𝑦𝑦 + , 𝜇𝜇̅ + 𝜇𝜇 𝑇𝑇 = +
𝜇𝜇̅ 𝑢𝑢𝜏𝜏 𝜇𝜇̅ 𝜕𝜕𝑢𝑢 ⁄𝜕𝜕𝑦𝑦 +

 Algebraic yPlus model prescribes 𝑢𝑢+ 𝑦𝑦 + from an extension of the logarithmic wall law and
solves for 𝑦𝑦 +
 L-VEL model prescribes 𝑦𝑦 + 𝑢𝑢+ from a different extension of the logarithmic wall law and
solves for 𝑢𝑢+
NOTE: No turbulent transport no wake interference and no in/outflow of turbulence
Transport-Equation Turbulence Models (2-Eqn.)
Transport-equation turbulence models solve for the transport of turbulence quantities.
 k- ε : Transport equations for 𝑘𝑘 and ε. General purpose model with wall functions for
smooth and rough walls.
 Realizable k- ε : Realizability constraints are built into the model parameters.
 k- ω : Transport equations for 𝑘𝑘 and ω. More accurate than the k-epsilon model close to
walls and in recirculation regions. Wall functions for smooth and rough walls but can also
resolve the flow close to walls.
 SST: Combines the k- ω model close to walls with the k- ε model in the outer flow region.
Can resolve the wall layer.
 Low Re k- ε : Wall resolved k- ε model.
Transport-Equation Turbulence Models (1 and 3 Eqn.)
 Spalart–Allmaras: Transport equation for 𝜐𝜐�𝑇𝑇 . Wall resolved turbulence model developed
for aerodynamic applications.
 v2-f: An extended low-Re k-ε model. In addition to 𝑘𝑘 and ε, it also solves for the wall normal
fluctuations (v2) and an elliptic relaxation function (f). Good predictive capabilities for flow
over curved surfaces. Can resolve the wall layer.
Wall Treatment
1
𝑢𝑢+ = log 𝑦𝑦 + + B
𝑢𝑢+ 𝑦𝑦 + 𝜅𝜅
𝑢𝑢+ = 𝑦𝑦 +

Wall offset

 Wall functions: A wall offset is applied. The computational domain always start in the log-
layer.
 Low-Reynolds-number treatment: The viscous sublayer must be resolved.
 Automatic wall treatment: A wall offset is applied. The computational domain can start
anywhere in the boundary layer.
Resolution for Low-Re Wall Treatment

Algebraic yPlus
L-VEL
Accurate wall stress May still converge

k-omega
Low-Re k-epsilon
SST
Algebraic Turbulence Models
 Turbulent viscosity evaluated from the
Reynolds number based on local speed and
wall distance
Algebraic yPlus
L-VEL

 Advantages:
Robust
Computationally inexpensive

 Disadvantage: Surface temperature and streamlines in a benchmark for electronic cooling.

Less accurate
Transport Equations for Turbulence: Two Equation Models
k-ε models
The standard k-ε model with
realizability constraints
The realizable k-ε model

k-ω model
The revised Wilcox k-ω model (1998) with
realizability constraints

SST model
Combines the k-ε model in the free stream
with the k-ω model close to walls

Low Re k-ε model


The AKN k-ε model Benchmark model of a NACA0012 airfoil
using the SST turbulence model.
Transport Equation Turbulence Models
Spalart-Allmaras
One-equation model with rotational
correction, developed for
aerodynamic applications

v2-f model
An extension of the k-ε model which accounts
for turbulence anisotropy by solving for the
wall-normal turbulence velocity fluctuations

Flow in a hydrocylone. A typical application where v2-f gives superior


results over two-equation models such as k-ε or SST.
Wall Treatment
Wall functions for smooth and rough walls
Robust, applicable for coarse meshes, but limited
accuracy
Supported by k-ε, Realizable k-ε and k-ω

Low-Reynolds-number treatment
Resolves the flow all the way down to
walls, accurate
Supported by all turbulence models except Low Re Wall functions
k-ε and realizable k-ε

Automatic wall treatment


Switches between low-Re treatment and wall functions,
accuracy according to local
Flow in a pipe elbow simulated with the k-ω model.
mesh resolution
Inherits the robustness provided by
wall functions
Default for all turbulence models except k-ε and realizable
k-ε
Large Eddy Simulation
 The larger three-dimensional, unsteady
eddies are resolved, whereas the effect of
the smaller eddies is modeled.
 Simulations must be three-dimensional and
time dependent.
 Computationally demanding
 The three current interfaces are based on
variational multiscale methods LES of turbulent channel flow at 𝑅𝑅𝑅𝑅𝑇𝑇 = 395

RBVM
RBVMWV
Smagorinsky
Detached Eddy Simulation
 Hybrid method between RANS and large
eddy simulation (LES)Simulations must be
three-dimensional and time dependent.
 RANS is used in the boundary layer and LES
is used elsewhere.
 Requires a less dense boundary layer mesh
compared with a pure LES which
substantially reduces the memory
requirement and computation
 Spalart–Allmaras turbulence model with the
following LES models
RBVM Flow over an Obstacle with DES

RBVMWV
Smagorinsky
Demo
Outlet

Model Definition
 Laminar flow in water
 Fully developed flow at the inlet
 Pressure condition at the outlet
 No-slip conditions at walls Inlet Walls

 Symmetry conditions at the two


lateral surfaces
Symmetry
 Why?
Typical expansion found in many
systems, e.g. in medtech
Symmetry
Benchmark with flow separation
Due to symmetry, we only have to model one eight of the model domain, provided that the flow is steady and
that the inlet flow is perpendicular to the inlet boundary.
Walls Symmetry Outlet

Model Definition
 Extension of previous laminar flow
example:
Elongate outlet section to avoid
recirculation zone close to outlet

 Turbulent flow in water


Inlet Symmetry
k-ε turbulence model

 Fully developed flow at the inlet


 Pressure condition at the outlet
 Wall functions at walls
 Symmetry conditions at the two
lateral surfaces It is possible to change the model settings from laminar flow to turbulent flow and also chose turbulence model.
Results
 Flow and pressure fields
 Length of recirculation zone u0 = 0.6[m/s]

 Note:
Recirculation reaches the outlet
-> elongate the outlet section

You might also like