CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Main CFD Forum

decay 2d turbulence - the initial field

Register Blogs Members List Search Today's Posts Mark Forums Read

Like Tree1Likes

Reply
 
LinkBack Thread Tools Display Modes
Old   February 25, 2013, 19:21
Default decay 2d turbulence - the initial field
  #1
New Member
 
Trần Đěnh Thắng
Join Date: Mar 2012
Location: Moscow Russia, Hanoi Vietnam
Posts: 24
Rep Power: 5
mp121209 is on a distinguished road
Send a message via ICQ to mp121209 Send a message via Yahoo to mp121209 Send a message via Skype™ to mp121209
I've seen some videos that demonstrates the process of 2d turbulence decay. Here are some of them:
http://www.youtube.com/watch?v=boWeyfYs2EQ
http://www.youtube.com/watch?v=pNCst5-zPEQ
http://www.youtube.com/watch?v=-fDwnyi_EkM

I want to simulate this process by using the Euler system of equations. For this goal, I must have the initial field (rho, u, w, p).

Does anyone give me some ideas about this?

Thanks for helping!
mp121209 is offline   Reply With Quote

Old   February 26, 2013, 04:45
Default
  #2
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 1,645
Rep Power: 23
FMDenaro will become famous soon enough
Quote:
Originally Posted by mp121209 View Post
I've seen some videos that demonstrates the process of 2d turbulence decay. Here are some of them:
http://www.youtube.com/watch?v=boWeyfYs2EQ
http://www.youtube.com/watch?v=pNCst5-zPEQ
http://www.youtube.com/watch?v=-fDwnyi_EkM

I want to simulate this process by using the Euler system of equations. For this goal, I must have the initial field (rho, u, w, p).

Does anyone give me some ideas about this?

Thanks for helping!
try to see in the books of Lesieur ...
just to say that if you use the inviscid flow equation, you have no decay due to dissipation. Furthermore, in 2D you can have some inverse cascade as well as you can see small vortical structures forming bigger ones (e.g., the vortex merging)
FMDenaro is offline   Reply With Quote

Old   February 26, 2013, 05:12
Default
  #3
New Member
 
Trần Đěnh Thắng
Join Date: Mar 2012
Location: Moscow Russia, Hanoi Vietnam
Posts: 24
Rep Power: 5
mp121209 is on a distinguished road
Send a message via ICQ to mp121209 Send a message via Yahoo to mp121209 Send a message via Skype™ to mp121209
Can you give me the link to that book you sad?
I know that there is no decay due to physical dissipation in inviscid flow, but it's impossible to neglect the grid viscosity which is mesh-dependent factor. I don't know what kind of the final field will be, I'm interested in generating the initial field to initialize the test case.
Thanks for replying!
mp121209 is offline   Reply With Quote

Old   February 26, 2013, 05:22
Default
  #4
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 1,645
Rep Power: 23
FMDenaro will become famous soon enough
http://www.amazon.com/Turbulence-Flu...der_1402064349
FMDenaro is offline   Reply With Quote

Old   February 26, 2013, 06:29
Default
  #5
New Member
 
Trần Đěnh Thắng
Join Date: Mar 2012
Location: Moscow Russia, Hanoi Vietnam
Posts: 24
Rep Power: 5
mp121209 is on a distinguished road
Send a message via ICQ to mp121209 Send a message via Yahoo to mp121209 Send a message via Skype™ to mp121209
OK, that book explains the theory of turbulence very good and fully. But I still need particular specifically formulas for generating the initial field to initialize a simulation case. In the book I can't find out them.
mp121209 is offline   Reply With Quote

Old   February 26, 2013, 06:34
Default
  #6
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 1,645
Rep Power: 23
FMDenaro will become famous soon enough
see also:

http://onlinelibrary.wiley.com/doi/1...d.613/abstract

in the references you will find a paper of Lesiuer describing the details
FMDenaro is offline   Reply With Quote

Old   February 26, 2013, 06:57
Default
  #7
New Member
 
Trần Đěnh Thắng
Join Date: Mar 2012
Location: Moscow Russia, Hanoi Vietnam
Posts: 24
Rep Power: 5
mp121209 is on a distinguished road
Send a message via ICQ to mp121209 Send a message via Yahoo to mp121209 Send a message via Skype™ to mp121209
Can you tell the name of Lesiuer's paper describing the details of turbulence initial field for numerical simulations?
I couldn't download the journal.
mp121209 is offline   Reply With Quote

Old   February 26, 2013, 07:05
Default
  #8
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 1,645
Rep Power: 23
FMDenaro will become famous soon enough
Lesieur, M. Staquet C, Le Roy P, Comte P. The mixing layer and its coherence examined from the point of
view of two-dimensional turbulence. Journal of Fluid Mechanics 1986; 192:511–534.


you cna find the case of the initialization for a 2D micing layer
FMDenaro is offline   Reply With Quote

Old   February 26, 2013, 07:59
Default
  #9
New Member
 
Trần Đěnh Thắng
Join Date: Mar 2012
Location: Moscow Russia, Hanoi Vietnam
Posts: 24
Rep Power: 5
mp121209 is on a distinguished road
Send a message via ICQ to mp121209 Send a message via Yahoo to mp121209 Send a message via Skype™ to mp121209
Do you know that I don't need an initial field for 2D shear layer. I need an initial field for the exactly given above examples: decay of 2D turbulence flow with vortices. I've found out some articles, which describe the initial field for some different kind of Euler equations. They used the vortex-velocity transport equations to solve the problem. With that kind of Euler equations it's possible to define the exact field for all the vortices. I use the Euler equations in FV method to solve the problem, I need the initial state for all components of the velocity. This is my problem!
Thanks for helping!
mp121209 is offline   Reply With Quote

Old   February 26, 2013, 09:14
Default
  #10
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 1,645
Rep Power: 23
FMDenaro will become famous soon enough
Quote:
Originally Posted by mp121209 View Post
Do you know that I don't need an initial field for 2D shear layer. I need an initial field for the exactly given above examples: decay of 2D turbulence flow with vortices. I've found out some articles, which describe the initial field for some different kind of Euler equations. They used the vortex-velocity transport equations to solve the problem. With that kind of Euler equations it's possible to define the exact field for all the vortices. I use the Euler equations in FV method to solve the problem, I need the initial state for all components of the velocity. This is my problem!
Thanks for helping!

if you want to simulate only the homogeneous/isotropic decay, you can simply work in the x,y plane by using the initial 1D distribution (extended in the homogeneous 2D meaning and adjusting the coefficients to satisfy the continuity contraint) described here:

On the application of congruent upwind discretizations for large eddy simulations

Authors
Andrea Aprovitola, Filippo Maria Denaro

Publication date
2004/2/10

Journal name
Journal of Computational Physics

Volume
194

Issue
1

Pages
329-343

Publisher
Academic Press
FMDenaro is offline   Reply With Quote

Old   February 26, 2013, 17:25
Default
  #11
New Member
 
Trần Đěnh Thắng
Join Date: Mar 2012
Location: Moscow Russia, Hanoi Vietnam
Posts: 24
Rep Power: 5
mp121209 is on a distinguished road
Send a message via ICQ to mp121209 Send a message via Yahoo to mp121209 Send a message via Skype™ to mp121209
I want to solve the problem in the x,y plane, with the limited space [LxL] and periodic boundary conditions. I need the initial with vorticies like in the above videos. But I can't find out some formulas for this purpose. I've read the article you sad, but it didn't help me.
I think that in the videos, the vortex-velocity form of the Euler equations was used. It's possible to give the exact values of vorticies at the cell centers. I use the FV approach to solve the the Euler system of equations. It's no need to have a vorticity field at the beginning time. But I need the beginning values for the velocity components.
I have read many works, but I can't find out some of them helpful for me!
mp121209 is offline   Reply With Quote

Old   May 7, 2013, 00:39
Default
  #12
New Member
 
Angus Hutchison
Join Date: May 2013
Posts: 6
Rep Power: 4
AHutchison is on a distinguished road
@mp121209 did you have any luck with your simulation?
AHutchison is offline   Reply With Quote

Old   May 7, 2013, 10:36
Default
  #13
Member
 
Reza
Join Date: Mar 2009
Location: Appleton, WI
Posts: 99
Rep Power: 8
triple_r is on a distinguished road
I don't know if you are still interested in this or not, but you can find an initial velocity field for homogenous turbulence in the following paper:

Numerical Experiments in Homogeneous Turbulence, by Robert S. Rogallo

It is a very old NASA report and a very good reference for DNS codes, so it is available on many sites. Just google the title and you should be able to get it. I can't remember which page :-p but it was near the end of the paper. It uses Fourier transform though, so you should be prepared to take inverse FFT of a spectral distribution, just a heads up :-)

Good luck.
triple_r is offline   Reply With Quote

Old   October 16, 2013, 07:45
Default
  #14
New Member
 
Dhruv Mehta
Join Date: Jun 2012
Posts: 22
Rep Power: 5
anzillo is on a distinguished road
Hello

Thank you for the information on Rogallo's code. I would be thankful if you could tell me how can I modify the code to generate a 2D field instead of the 3D field that the code is supposed to generate.

Thank you !
anzillo is offline   Reply With Quote

Old   October 16, 2013, 09:53
Default
  #15
Member
 
Reza
Join Date: Mar 2009
Location: Appleton, WI
Posts: 99
Rep Power: 8
triple_r is on a distinguished road
Why two-dimensional? Turbulence is intrinsically three-dimensional.

Anyhow, if you want a 2D initial field, you can either generate a 3D field and apply any averaging operator on the third dimension, or just generate the 2D spectral distribution first (Rogallo's distribution makes sure, for example, continuity is satisfied, so when modifying the equations, do not just remove the third component. You'll probably need to change the remaining two components as well) and then perform a two-dimensional IFFT.

It has been very long since last I saw that paper. Let us know if you have any specific problems converting it to 2D and I might be able to help.

At the end, again, I have to emphasize the inherent three-dimensionality of turbulence.
triple_r is offline   Reply With Quote

Old   October 16, 2013, 10:18
Default
  #16
New Member
 
Dhruv Mehta
Join Date: Jun 2012
Posts: 22
Rep Power: 5
anzillo is on a distinguished road
Thank you for your response.

I agree that turbulence is 3-D in nature. However, there are few numerical tests conducted on the decay of 2-D turbulence. I just wanted to verify a 2-D code and thus required the field.

I will try what you suggested and would let you know in case it works.

Thank you again for your prompt response.

Best !
anzillo is offline   Reply With Quote

Old   October 18, 2013, 09:47
Default
  #17
New Member
 
Dhruv Mehta
Join Date: Jun 2012
Posts: 22
Rep Power: 5
anzillo is on a distinguished road
Hello Reza

I modified the code to generate a 2D field that satisfies the continuity equation in wavenumber space, i.e. kx*ux + ky*uy = 0;

The field is indeed isotropic and divergence free (to machine precision, set to 10-8) in this case. However, on taking the inverse FFT of this field, the physical fields that results is not divergence free at all.

I am making a mistake in using the inverse FFT (I use MATLAB's iffn(symmetric) ). I would be thankful if you could help me with this.

To generate the 2D field I used: u = a*ky*(1/k) i - a*kx*(1/k) j

where, sqrt( kx^2 + ky^2 ) = k, i and j are the x and y unit vectors, u here is defined in wavenumber space.

This definition of u satisfies continuity as ux/uy = -ky/kx. Also, a is a function of the desired energy spectrum as used by Rogallo. I use

a = ( E(k) / 2*pi*k)^(0.5) * e^(iota*phi)

Thank you.

Regards
Dhruv
anzillo is offline   Reply With Quote

Old   October 18, 2013, 10:07
Default
  #18
Member
 
Reza
Join Date: Mar 2009
Location: Appleton, WI
Posts: 99
Rep Power: 8
triple_r is on a distinguished road
Hi,

Multi-dimensional IFFT can be a real headache sometimes, specially when you are generating values in frequency domain and you are hoping to get a real-valued function in physical space.

I'm not familiar with MATLAB's ifftn function, so I'll take a look at that as soon as a license becomes available for me to use :-)

However, by then, are you taking care of the wrap-around? Also, if you have to give MATLAB the whole spectral response, you need to make sure it satisfies the symmetry rules to guarantee a real function in the physical domain. These symmetries mainly come from the fact that:

\hat{f}_{(-k_x, -k_y)}=\hat{f}_{(k_x, k_y)}^*

where \hat{f} is the Fourier transform of f and a^* is the conjugate of a complex number a.

Because of this symmetry and wrap-around, some on the values have to be real, for example:

\hat{f}_{-(0,0)}=\hat{f}_{(0,0)}^*\Rightarrow\hat{f}_{(0,0)}\in\Re

and you can get some specific relations for (i,j) where i and j are equal to -\frac{N}{2}, 0, \frac{N}{2}-1, where N is the size of the grid that you are using.

Last edited by triple_r; October 18, 2013 at 10:08. Reason: LATEX Error
triple_r is offline   Reply With Quote

Old   October 18, 2013, 10:17
Default
  #19
New Member
 
Dhruv Mehta
Join Date: Jun 2012
Posts: 22
Rep Power: 5
anzillo is on a distinguished road
Hello

Thank you for the explanation. As advised, I will go through some notes on FFT.

Regards
anzillo is offline   Reply With Quote

Old   October 18, 2013, 10:44
Default
  #20
Member
 
Reza
Join Date: Mar 2009
Location: Appleton, WI
Posts: 99
Rep Power: 8
triple_r is on a distinguished road
Yeah, as I mentioned, you need to make sure the frequency-domain field that you are generating is as MATLAB describes it "conjugate symmetric". Here is an octave code that I was able to dig from my archives that makes sure the field is symmetric:

Code:
for i = 1:M
    for j = 2:N/2
        E(i, j) = temp(i, j);
        E(mod(M-i+1, M)+1, mod(N-j+1, N)+1) = conj(temp(i, j));
    end
end
for i = 2:M/2
    E(i, 1) = temp(i, 1);
    E(mod(M-i+1, M)+1, 1) = conj(temp(i, 1));
end
for i = 2:M/2
    E(i, N/2+1) = temp(i, N/2+1);
    E(mod(M-i+1, M)+1, N/2+1) = conj(temp(i, N/2+1));
end

E(1,1) = real(temp(1,1));
E(M/2+1,1) = real(temp(M/2+1,1));
E(1,N/2+1) = real(temp(1,N/2+1));
E(M/2+1,N/2+1) = real(temp(M/2+1,N/2+1));
In this code, temp is a randomly generated field of complex numbers that can come from your relations for uhat and vhat, then E is the symmetric version of temp that will result in a real function after an IFFT. I don't know if it will work as it is in MATLAB, but if there is a need for change, it should be minimal.

Hope this helps.
triple_r is offline   Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
pisoFoam - unstable pressure residual Industrial_CFD OpenFOAM Running, Solving & CFD 21 February 24, 2013 16:39
problem with Min/max rho tH3f0rC3 OpenFOAM 7 February 23, 2013 06:37
Error while running rhoPisoFoam.. nileshjrane OpenFOAM Running, Solving & CFD 8 August 26, 2010 12:50
lift and drag on ship superstructures vaina74 OpenFOAM Running, Solving & CFD 3 June 8, 2010 12:30
ForcesCoeffs ronaldo OpenFOAM 4 September 14, 2009 07:11


All times are GMT -4. The time now is 21:44.