CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Software User Forums > OpenFOAM > OpenFOAM Programming & Development

Output of velocity gradients

Register Blogs Community New Posts Updated Threads Search

Like Tree11Likes
  • 1 Post By wersoe
  • 3 Post By wersoe
  • 1 Post By romant
  • 6 Post By wersoe

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   July 11, 2011, 09:32
Default Output of velocity gradients
  #1
Member
 
Soeren Werner
Join Date: Mar 2009
Location: Wädenswil, Switzerland
Posts: 32
Rep Power: 17
wersoe is on a distinguished road
Dear Foamers,

I am quite new to OpenFOAM and try to modify one solver to write the velocity gradients as output. The solver used is interDynMFoam.

I started to add divergence of U as output as follows, and this is working good (green in code below)

The next step was to add the velocity gradients output, as you seen in red in code below.

First problem was, that I cant use muEff * (grad U+ gradU.T()) with an error, that it is not the same dimension. So I skipped it, as you can see in code below to think about later... any help is appreciated here.

The second problem is, that it is able to compile now, and even able to run after adding it fvSolution and to 0/gradU and 0/tau, however, no output is written at all.
The files 1/gradU and 1/tau are the same asa in 0/. Can anybody help with that?

Or is there any other, even more simple possibility to get the velocity gradients in a interFoam solver?


UEqn.H added:
Code:
surfaceScalarField muEff
    (
        "muEff",
        twoPhaseProperties.muf()
      + fvc::interpolate(rho*turbulence->nut())
    );

    fvVectorMatrix UEqn
    (
        fvm::ddt(rho, U)
      + fvm::div(rhoPhi, U)
      - fvm::laplacian(muEff, U)
      - (fvc::grad(U) & fvc::grad(muEff))
    //- fvc::div(muEff*(fvc::interpolate(dev(fvc::grad(U))) & mesh.Sf()))
    );

    UEqn.relax();

	volTensorField gradU 
		(fvc::grad(U));


if (pimple.momentumPredictor())
    {
        solve
        (
            UEqn
         ==
            fvc::reconstruct
            (
                (
                    fvc::interpolate(interface.sigmaK())*fvc::snGrad(alpha1)
                  - ghf*fvc::snGrad(rho)
                  - fvc::snGrad(p_rgh)
                ) * mesh.magSf()
            )
        );

	solve
	(
		fvm::ddt(d)
		==
		- fvm::div(phi,d)
	);
	
    }
own_interDynMFoam.C
Code:
-------------------snip/snap-------------------------
runTime.write();

	d=fvc::div(phi);
              tau=(gradU + gradU.T());

        Info<< "ExecutionTime = " << runTime.elapsedCpuTime() << " s"
-------------------snip/snap-------------------------
createFields.H
Code:
-------------------snip/snap-------------------------
//Creating field for gradU
Info<< "Reading field gradU\n" << endl;
    volTensorField gradU
    (
        IOobject
        (
            "gradU",
            runTime.timeName(),
            mesh,
            IOobject::MUST_READ,
            IOobject::AUTO_WRITE
        ),
        mesh
    );

//Creating field for tau
Info<< "Reading field tau\n" << endl;
    volTensorField tau
    (
        IOobject
        (
            "tau",
            runTime.timeName(),
            mesh,
            IOobject::MUST_READ,
            IOobject::AUTO_WRITE
        ),
        mesh
    );

//Creating field for divergence d
Info<< "Reading field d\n" << endl;
    volScalarField d
    (
        IOobject
        (
            "d",
            runTime.timeName(),
            mesh,
            IOobject::MUST_READ,
            IOobject::AUTO_WRITE
        ),
        mesh
    );
-------------------snip/snap-------------------------
Kummi likes this.
wersoe is offline   Reply With Quote

Old   October 21, 2011, 10:49
Default
  #2
Senior Member
 
Awais Ali
Join Date: Feb 2010
Location: Germany
Posts: 128
Rep Power: 17
owayz is on a distinguished road
Send a message via MSN to owayz
Hallo Soeren,
I want to know why do you want to calculate the gradients of velocity and write them?
I actually have an idea, I think you might also be trying to do the same. I want to do adaptive mesh refinement on the basis of velocity gradients. It would be really helpful for me if I could calculate the velocity gradients some how in each timestep I would then be able to do the adaptive mesh refinement. Tell me if you have the same idea.

Regards,
Awais Ali
owayz is offline   Reply With Quote

Old   October 21, 2011, 11:44
Default
  #3
Member
 
Soeren Werner
Join Date: Mar 2009
Location: Wädenswil, Switzerland
Posts: 32
Rep Power: 17
wersoe is on a distinguished road
Dear Awais ,

I manage to write the gradients since some weeks...
We are working in the field of biotech, and using CFD to optimize and/or develop bioreactors in which cells growth, which are quite sensitive to any stress. So, we would like to use the gradients to caculate shear strain and normal strain...

However, we know, that these values (shear strain, vel gradients, ...) are quite much dependent on the size of the mesh... so we should be able to compare just same cases with different working parameters, e.g. stirrer speed...

Later on, we would like to compare between different reactor sizes, which means different meshes, too. Therefore, we would need to adapt the mesh until the change in the gradients is small enough...

If you could help me with adapting the mesh, I would much appreciate...


Here the changes in the solver, I use interDyMFoam:

file interDyMFoam.C
Code:
.
.
.
#include "UEqn.H"
#include "shear.H" // added
.
.
.
]

file shear.H
Code:
volScalarField mu
    (
        "mu",
        twoPhaseProperties.mu()
    );

volTensorField gradU = fvc::grad(U);   // is not written, and needs not to be declared in createFields.H if declaration (volTensorField) is done here

//--------------------------------------------------------------------------------------------------------------------
	D = alpha1*(gradU+gradU.T());				// needs to be declared in createFields.H
	Tau = alpha1 * mu * (gradU+gradU.T());			// 
	Strain = alpha1 * pow(2,0.5) * mag(symm(fvc::grad(U))); //
	TauStrain= mu * Strain;
file ../createFields.H
Code:
.
.                   //just add as much as you defined new
.
//Creating field for Strain
Info<< "Reading field Strain\n" << endl;
    volScalarField Strain
    (
        IOobject
        (
            "Strain",
            runTime.timeName(),
            mesh,
            IOobject::MUST_READ,    //is read, needs to be in folder 0
            IOobject::AUTO_WRITE   //is written
        ),
        mesh
    );
.
.
.

Hope this helps, dont hesitate to ask again...

If anybody finds this is really not good, please inform me, since this is my own hacking, and I did it the first time with OF... thanks

Best, Sören
arvindpj, meth and wbywbywby6 like this.
wersoe is offline   Reply With Quote

Old   October 22, 2011, 20:53
Default
  #4
Senior Member
 
Awais Ali
Join Date: Feb 2010
Location: Germany
Posts: 128
Rep Power: 17
owayz is on a distinguished road
Send a message via MSN to owayz
Thanks for your reply Soeren,
Well if you are using interDyMFoam you must have seen the dambreak tutorial. My idea is that if you want to refine on the basis of strain and as you are able to write the strain at every step. You just need to redefine the dynamicMeshdict. Instead of alpha1 you need to refine on the basis of mag or gradients or strain. If hope this helps you and best of luck to you.
But I have one problem. The solver that I have modiefied according to the information that you have provided asks me to provide a file for magnitude of gradients of U. And even if I provide a file it somehow is writing the same file every time step again.
If you can suggest me something which could help me.
Regards,
Awais Ali
owayz is offline   Reply With Quote

Old   October 24, 2011, 10:57
Default
  #5
Senior Member
 
Bernhard Linseisen
Join Date: May 2010
Location: Heilbronn
Posts: 183
Blog Entries: 1
Rep Power: 16
Linse is on a distinguished road
Quote:
Originally Posted by owayz View Post
Hallo Soeren,
I want to know why do you want to calculate the gradients of velocity and write them?
I actually have an idea, I think you might also be trying to do the same. I want to do adaptive mesh refinement on the basis of velocity gradients. It would be really helpful for me if I could calculate the velocity gradients some how in each timestep I would then be able to do the adaptive mesh refinement. Tell me if you have the same idea.

Regards,
Awais Ali
Seems we are rowing the same boat here. At least this is one of the issues I plan to do in close future. So this thread is on my "watchlist" by now. ;-)
I'll keep you posted on any developments from my side...
Linse is offline   Reply With Quote

Old   October 24, 2011, 11:09
Default
  #6
Member
 
Soeren Werner
Join Date: Mar 2009
Location: Wädenswil, Switzerland
Posts: 32
Rep Power: 17
wersoe is on a distinguished road
Quote:
Originally Posted by owayz View Post
Well if you are using interDyMFoam you must have seen the dambreak tutorial. My idea is that if you want to refine on the basis of strain and as you are able to write the strain at every step. You just need to redefine the dynamicMeshdict. Instead of alpha1 you need to refine on the basis of mag or gradients or strain. If hope this helps you and best of luck to you.
Thanks, will try it someday... Rite now i didnt care about refinement at all. So, I know the dambreak, but didnt have a closer look to the refinement at the surface...

Quote:
Originally Posted by owayz View Post
But I have one problem. The solver that I have modiefied according to the information that you have provided asks me to provide a file for magnitude of gradients of U. And even if I provide a file it somehow is writing the same file every time step again.
Well, its quite hard to say anything...
First, you have to provide the file because there is a
Code:
IOobject::MUST_READ,    //is read, needs to be in folder 0
in createFields.H
Try to change it to NO_READ or something similar, I do not know the right keyword, so I just provided a file...
And here the error may be caused: gradU is a volTensorField.
So, in createFields it must be defined as volTensorField - and so it must be defined in 0/gradU:
Code:
.
.
.
dimensions [ 0 0 -1 0 0 0 0 ]
internalField uniform ( 0 0 0 0 0 0 0 0 0 )
.
.
.
This is valid for 3D, change it accordingly to 2D...

Hope this helps, Sören
wersoe is offline   Reply With Quote

Old   November 28, 2011, 11:10
Default
  #7
Senior Member
 
Awais Ali
Join Date: Feb 2010
Location: Germany
Posts: 128
Rep Power: 17
owayz is on a distinguished road
Send a message via MSN to owayz
Dear Sören,
//--------------------------------------------------------------------------------------------------------------------
D = alpha1*(gradU+gradU.T()); // needs to be declared in createFields.H
Tau = alpha1 * mu * (gradU+gradU.T()); //
Strain = alpha1 * pow(2,0.5) * mag(symm(fvc::grad(U))); //
TauStrain= mu * Strain;

I have a question. In the above code what does the "symm" function do or why are you using it. Does it make any difference if I calculate the gradients without using the symm function.

Regards,
Awais
owayz is offline   Reply With Quote

Old   November 28, 2011, 11:40
Default
  #8
Senior Member
 
romant's Avatar
 
Roman Thiele
Join Date: Aug 2009
Location: Eindhoven, NL
Posts: 374
Rep Power: 21
romant is on a distinguished road
Quote:
Originally Posted by owayz View Post
Dear Sören,
//--------------------------------------------------------------------------------------------------------------------
D = alpha1*(gradU+gradU.T()); // needs to be declared in createFields.H
Tau = alpha1 * mu * (gradU+gradU.T()); //
Strain = alpha1 * pow(2,0.5) * mag(symm(fvc::grad(U))); //
TauStrain= mu * Strain;

I have a question. In the above code what does the "symm" function do or why are you using it. Does it make any difference if I calculate the gradients without using the symm function.

Regards,
Awais
"symm" gives you the symmetric tensor of fvc::grad(U) since it is symmetric. I think it might be reduce computational and memory usage.
owayz likes this.
__________________
~roman
romant is offline   Reply With Quote

Old   September 14, 2013, 01:34
Default Extract div(phi, U) tensor
  #9
JFM
Member
 
JFM's Avatar
 
John Anonymous
Join Date: Jan 2011
Location: Melbourne Australia
Posts: 37
Rep Power: 15
JFM is on a distinguished road
Send a message via Skype™ to JFM
Good day everyone

I am using OpenFOAM to assess a hydraulic structure which is designed to operate at critical condition (Fr=1). Previously I had assessed this structure using two commercial 2D FVM model, however the results were not acceptable.

As a result I have created a 3D interFoam model of the structure and are aiming to prove that OpenFOAM is a suitable tool for designers. I need to determine the influence, if any, of the cross momentum terms from the convective velocity field, div (phi, U). The terms I am particularly interested in are:
x-direction: u.du/dx + v.du/dy + w.du/dz
y-direction: u.dv/dx + v.dv/dy + w.dv/dz
z-direction: u.dw/dx + v.dw/dy + w.dw/dz
(at this stage I am not interested in the RAS or LES stresses)

I have not found anyway (in the forum or other references) to write this tensor out from the model. However, it appears that there is a way to extract these components from the model and your post provides some directions on how to do this. Could someone provide some direction on how this could be achieved in my case, please include file / dictionary names that may need to be amended.

I have been using OpenFOAM for around a year but still have limited confidence when it comes to manipulation of source files therefore any assistance will be greatly appreciated - hopefully I can return the favour one day.

Regards
JFM
JFM is offline   Reply With Quote

Old   September 14, 2013, 09:15
Default
  #10
Member
 
Soeren Werner
Join Date: Mar 2009
Location: Wädenswil, Switzerland
Posts: 32
Rep Power: 17
wersoe is on a distinguished road
You should use something like this in a seperate file:

-----------grad.H--------------------------

volScalarField X=U.component(0);
volScalarField Y=U.component(1);
volScalarField Z=U.component(2);


volScalarField XX=gradU.component(0);
volScalarField YY=gradU.component(4);
volScalarField ZZ=gradU.component(8);

volScalarField XY=gradU.component(1);
volScalarField YZ=gradU.component(5);
volScalarField ZX=gradU.component(6);

volScalarField YX=gradU.component(3);
volScalarField ZY=gradU.component(7);
volScalarField XZ=gradU.component(2);

Variable = X*XX + ...



Include it in the solver, dont forget to create the fields in createFields.H...
Just see my previous post...

Then you will be able to write it during the calculation.
It would be much easier and more comfortable if you calculate this solution after the simulation is done...
There are a couple of tools which could be the basis, need to be changed and recompiled too...

Best, Sören
wersoe is offline   Reply With Quote

Old   March 31, 2014, 21:57
Default
  #11
Senior Member
 
Huang Xianbei
Join Date: Sep 2013
Location: Yangzhou,China
Posts: 302
Rep Power: 14
huangxianbei is on a distinguished road
Hi:
I just use this code
Code:
volTensorField graU = fvc::grad(UMean);
    volScalarField graUy(graU.component(1));
in calculateFields.H to calculate dU/dy in postChannel utility, and in the collapse.H
Code:
scalarField graUyValues(channelIndexing.collapse(graUy));
makeGraph(y, graUyValues, "graUy", path, gFormat);
while when I check the results, I found the results are wrong for the value is of e-23 order, I'm wondering if this can be used in a post-process?
huangxianbei is offline   Reply With Quote

Old   December 19, 2015, 05:15
Default
  #12
Member
 
Avdeev Evgeniy
Join Date: Jan 2011
Location: Togliatty, Russia
Posts: 69
Blog Entries: 1
Rep Power: 21
j-avdeev will become famous soon enough
Send a message via Skype™ to j-avdeev
Quote:
Originally Posted by owayz View Post
//--------------------------------------------------------------------------------------------------------------------
D = alpha1*(gradU+gradU.T());
Hello.
Can someone explain what gradU.T() means?
Why here gradU+gradU.T(), but not just gradU?
As I understand we want to get gradient field of U here.
j-avdeev is offline   Reply With Quote

Old   January 6, 2016, 09:17
Default
  #13
Senior Member
 
romant's Avatar
 
Roman Thiele
Join Date: Aug 2009
Location: Eindhoven, NL
Posts: 374
Rep Power: 21
romant is on a distinguished road
Quote:
Originally Posted by j-avdeev View Post
Hello.
Can someone explain what gradU.T() means?
Why here gradU+gradU.T(), but not just gradU?
As I understand we want to get gradient field of U here.
The .T() operator gives you the transpose of the matrix http://foam.sourceforge.net/docs/cpp...61ae7a972b4f96
__________________
~roman
romant is offline   Reply With Quote

Old   June 24, 2016, 10:54
Default
  #14
New Member
 
gnompher
Join Date: Jun 2016
Posts: 6
Rep Power: 10
Mowgli is on a distinguished road
I use the same code to calculate velocity gradients of the field in my case, but I get 0 gradients, which seems like the 0 file for gradU is being duplicated. Even with a NO_READ condition, I get 0 gradients. ANy idea as to why fvc::grad(U) outputs 0 gradient ?
Mowgli is offline   Reply With Quote

Reply


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 Off
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Problem with assigned inlet velocity profile as a boundary condition Ozgur_ FLUENT 5 August 25, 2015 05:58
Emergency:UDF for a time dependent parabolic velocity zumaqiong Fluent UDF and Scheme Programming 12 March 25, 2010 13:00
Velocity gradients from previous timesteps Ed Kay FLUENT 0 March 4, 2010 09:04
How to output unsteady velocity profiles? Jason FLUENT 3 August 8, 2007 10:40
Velocity Gradients at Inviscid Flow Lada FLUENT 0 January 15, 2001 08:20


All times are GMT -4. The time now is 04:00.