turbulence model - stony brookzgao/ppt/ppt/turbulence.pdf · turbulence models a turbulence model...
TRANSCRIPT
![Page 1: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/1.jpg)
Turbulence Model
Theory and Implementation
![Page 2: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/2.jpg)
Turbulence
In fluid dynamics, turbulence is a flow regime characterized by chaotic
property changes.
Inertial forces versus viscous forces
Use Reynolds number to quantify
𝑅𝑒 =𝑢𝐿
𝜈
Laminar flow: low Reynolds number, viscous forces dominates, smooth,
constant motion
Turbulence: High Reynolds number, inertial forces dominates, eddies,
instabilities.
![Page 3: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/3.jpg)
![Page 4: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/4.jpg)
Turbulence models
A turbulence model is a computational procedure to complete
system of mean flow equations.
There is no single turbulence model, which can predict reliably
all kinds of turbulent flows
Computational effort versus the accuracy required by the
particular application.
![Page 5: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/5.jpg)
Reynolds Average
Time average: only for steady flow
Space average: only for homogeneous flow
Ensemble average: for any flow
Perform the same experiment many times, then take average.
They are equivalent when flow is steady and homogeneous.
Usually use 𝑢 , 𝑈, 𝑢 to indicate mean velocity
![Page 6: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/6.jpg)
Reynolds Decomposition
Any velocity field can be decomposed to mean flow and
turbulent fluctuations
𝑢 = 𝑢 + 𝑢′ 𝑢: velocity field
𝑢 : mean velocity
𝑢′: turbulent fluctuation
𝑢′ = 𝑢 − 𝑢 = 𝑢 − 𝑢 = 0
𝑢 𝑣 = 𝑢 𝑣
𝑢 + 𝑣 = 𝑢 + 𝑣
![Page 7: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/7.jpg)
Reynolds Averaged NS equaiton
Continuity equation: 𝜕𝑢𝑖𝜕𝑥𝑖
= 0
Take average: 𝜕𝑢𝑖
𝜕𝑥𝑖= 0
Mean flow satisfies divergence free condition
![Page 8: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/8.jpg)
Reynolds Averaged NS equaiton Momentum equation:
𝜕𝑢𝑖𝜕𝑡
+ 𝑢𝑗𝜕𝑢𝑖𝜕𝑥𝑗
= −1
𝜌
𝜕𝑝
𝜕𝑥𝑖+ 𝜈
𝜕2𝑢𝑖𝜕𝑥𝑗𝜕𝑥𝑗
+ 𝑓𝑖
Take average:
𝜕𝑢𝑖
𝜕𝑡+ 𝑢𝑗
𝜕𝑢𝑖𝜕𝑥𝑗
= −1
𝜌
𝜕𝑝
𝜕𝑥𝑖+ 𝜈
𝜕2𝑢𝑖
𝜕𝑥𝑗𝜕𝑥𝑗+ 𝑓𝑖
Substitute 𝑢 = 𝑢 + 𝑢′:
𝜕𝑢𝑖
𝜕𝑡+ 𝑢𝑗
𝜕𝑢𝑖
𝜕𝑥𝑗+𝜕(𝑢′𝑖𝑢
′𝑗)
𝜕𝑥𝑗= −
1
𝜌
𝜕𝑝
𝜕𝑥𝑖+ 𝜈
𝜕2𝑢𝑖
𝜕𝑥𝑗𝜕𝑥𝑗+ 𝑓𝑖
The equations are formally identical to the Navier-Stokes equations with the exception of the
additional term
![Page 9: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/9.jpg)
Reynolds Stress
Reynolds stress tensor consists in 3D of 9 components:
(𝑢′)2 𝑢′𝑣′ 𝑢′𝑤′
𝑣′𝑢′ (𝑣′)2 𝑣′𝑤′
𝑤′𝑢′ 𝑤′𝑣′ (𝑤′)2
To complete the RANS equations, we need turbulence model
One-Equation Model
Spalart-Allmaras
Two-Equation Models
Standard 𝑘 − 𝜖
RNG 𝑘 − 𝜖
Realizable 𝑘 − 𝜖
Reynolds Stress
Model
SST Transition Model
Increase in computational cost
![Page 10: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/10.jpg)
Boussinesq Hypothesis
−𝑢′𝑖𝑢′𝑗 = 2𝜈𝑇
𝜕𝑢𝑖
𝜕𝑥𝑗+
𝜕𝑢𝑗
𝜕𝑥𝑖−
2
3𝐾𝛿𝑖𝑗
𝜇𝑇 = 𝜌𝜈𝑇 stands for the eddy viscosity
Unlike the molecular viscosity, the eddy viscosity represents
no physical characteristic of the fluid, but it is a function of the
local flow condition.
𝜇𝑇 is not homogeneous in space but assumed to be isotropic
If the turbulent-viscosity hypothesis is accepted, all that remain
is to determine 𝜈𝑇
![Page 11: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/11.jpg)
Mixing-Length Hypothesis
Prandtl(1925) visualized a simplified model for turbulent motion in
which fluid particles coalesce into lumps and move as a unit.
Simply apply molecular theory to turbulence
The task of specifying 𝜇𝑇 is generally approached through
specifications of 𝑢∗ and 𝑙∗
Mixing-length theory Molecular theory
𝜇𝑇 = 𝜌𝑢∗𝑙∗ 𝑢∗: turbulent velocity scale
𝑙∗ : mixing length
𝜇 = 𝜌𝑢 𝜆
𝑢 : average molecular speed
𝜆: mean free path
![Page 12: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/12.jpg)
𝑘 − 𝜖 Model
𝑘: turbulent kinetic energy
𝜖: energy dissipation rate
𝑢∗ = 𝐶𝜇𝑘1/2, 𝑙∗ = 𝑘3/2/𝜖, 𝐶𝜇 = 0.09
𝜇𝑇 = 𝜌𝑢∗𝑙∗ = 0.09𝜌𝑘2/𝜖
𝜇𝑇 only depends on 𝑘 and 𝜖, no need to specify mixing length
![Page 13: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/13.jpg)
Standard 𝑘 − 𝜖 Model
𝜕𝑘
𝜕𝑡+ 𝛻 ∙ 𝑘𝑢 −
𝜈𝑇
𝜎𝐾+ 𝜈 𝛻𝑘 = 𝑃𝑘 − 𝜖
𝜕𝜖 𝜕𝑡
+ 𝛻 ∙ 𝜖 𝑢 −𝜈𝑇
𝜎𝜖 + 𝜈 𝛻𝜖 =
𝜖 𝑘(𝐶1𝑃𝑘−𝐶2𝜖)
𝑃𝑘 = 0.5𝜈𝑇𝜕𝑢𝑖
𝜕𝑥𝑗+
𝜕𝑢𝑗
𝜕𝑥𝑖
2
is the production of turbulent kinetic
energy
![Page 14: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/14.jpg)
Boundary Condition Inlet
𝑢 = 𝑢𝑖𝑛, 𝑘 = 𝐶𝑏𝑐|𝒖|2, 𝜖 = 𝐶𝜇
𝑘3/2
𝑙0
Outlet
𝑛 ∙ 𝛻𝑢 = 0 , 𝑛 ∙ 𝛻𝑘 = 0, n ∙ 𝛻𝜖 = 0
Inlet Outflow
Wall boundary
![Page 15: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/15.jpg)
Wall Boundary
Slip Boundary: the normal component of the velocity is set
equal to zero, while tangential slip is permitted
Non-slip Boundary: both normal component and tangential
component set to be zero
For viscous flow non-slip is realistic, however slip boundary is
often used if the resolution near the wall is not high enough
and the wall effects is modeled by wall function.
![Page 16: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/16.jpg)
Law of the Wall
Boundary Layer is divided to: viscous sublayer and logarithmic sublayer
In viscous sublayer, viscous effects dominate and flow becomes laminar
High resolution is needed to resolve the steep velocity profile
Alternatively, one can use “wall function” to bridge the first node and the wall
𝑢+ = 𝑢/𝑢𝜏, 𝑦+ = 𝑦𝑢𝜏/𝜈
Viscous sublayer
![Page 17: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/17.jpg)
Wall Functions
To use wall functions, the first node should locate as
30 < 𝑦+ < 100, which is in logarithmic region and out of
viscous sublayer
Therefore, the log-law can be applied to the first node.
This could either be made through an added source term
simulating the wall friction or a modified viscosity ensuring the
correct friction.
This is more robust than specify the velocity
![Page 18: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/18.jpg)
Wall Functions
𝑛 ∙ 𝛻𝑘𝑝 = 0: 𝑘 at first interior node 𝑝
𝑢𝜏 = 𝐶𝜇1/4 𝑘𝑝: friction velocity at node 𝑝
𝜇𝑒 =𝜌𝑢𝜏𝑦𝑝𝜅
ln (𝐸𝑦𝑝+)
: effective viscosity at node 𝑝
𝜖 =𝑢𝜏
3
𝜅𝑦𝑝: dissipation rate at node 𝑝
𝑦𝑝: distance between node p and wall
𝑦+ =𝑦𝑢𝜏
𝜈: distance in wall coordinates
![Page 19: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/19.jpg)
𝑘 − 𝜖 Model Discussion Advantages:
Relatively simple to implement.
Leads to stable calculations that converge relatively easily.
Reasonable predictions for many flows.
Disadvantages:
Poor predictions for:
swirling and rotating flows,
flows with strong separation,
axisymmetric jets,
certain unconfined flows
fully developed flows in non-circular ducts.
Valid only for fully turbulent flows, no transition
![Page 20: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/20.jpg)
Summary
1. Calculate characteristic Reynolds number and determine whether flow is turbulent.
2. If the flow is transition range, consider use of the turbulence transition models
3. Estimate wall-adjacent cell y+ before generate mesh
4. Prepare wall functions except for low-Re model
5. Begin with RKE and change to the other models if needed
6. Use RSM for highly swirling, 3-D, rotating flows
![Page 21: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/21.jpg)
Numerical Results
𝑘 − 𝜖 model fail to generate vortex, because this is a transition process Re = 100
Use eddy
viscosity
Without
eddy
viscosity
![Page 22: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/22.jpg)
Velocity magnitude (DNS) Velocity magnitude (RANS)
𝑘 − 𝜖 fails to
predict
separation and
reattachment
Numerical Results
![Page 23: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/23.jpg)
Eddy viscosity ratio is
about 180 larger than
molecular viscosity Vorticity Eddy viscosity
Numerical Results
![Page 24: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/24.jpg)
Vorticity Equation
Vorticity is the curl of the velocity field: 𝜔 = 𝛻 × 𝑢 𝑑𝜔
𝑑𝑡+ 𝑢 ∙ 𝛻 𝜔 = 𝜔 ∙ 𝛻 𝑢 + 𝜈𝛻2𝜔
𝜔 ∙ 𝛻 𝑢 is the vortex stretching term, very important in turbulent flow
In 2-dimension, vortex stretching term vanishes
𝜔 ∙ 𝛻 𝑢=𝜕𝑣
𝜕𝑥−
𝜕𝑢
𝜕𝑦
𝜕𝑢
𝜕𝑧,
𝜕𝑣
𝜕𝑥−
𝜕𝑢
𝜕𝑦
𝜕𝑣
𝜕𝑧, 0 = (0,0,0)
In 2-dimension, vorticity evolves as a conserved scalar
(qualitatively different from 3d problem)
![Page 25: Turbulence Model - Stony Brookzgao/ppt/ppt/turbulence.pdf · Turbulence models A turbulence model is a computational procedure to complete system of mean flow equations. There is](https://reader036.vdocuments.site/reader036/viewer/2022081721/5b60b59e7f8b9a4a488b911c/html5/thumbnails/25.jpg)
Discussion on Turbulence Model