CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > OpenFOAM Running, Solving & CFD

Matrix coefficients from boundary conditions how

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

Like Tree9Likes
  • 9 Post By hjasak

Reply
 
LinkBack Thread Tools Display Modes
Old   January 12, 2007, 06:58
Default Hello, I am trying to under
  #1
Senior Member
 
Thomas Jung
Join Date: Mar 2009
Posts: 100
Rep Power: 8
tehache is on a distinguished road
Hello,

I am trying to understand implementation of boundary conditions...

Simple question: Why is e.g. when using a fixedGradientFvPatchField gradient(internal/boundary)Coeffs called, but not value(Internal/Boundary)Coeffs ?

If I want to implement an implicit b.c. (flux depending on current value), how can I make Foam to call valueInternalCoefficients? Deriving from a different FvPatchField?

Thank you very much for any hint !
tehache is offline   Reply With Quote

Old   January 12, 2007, 09:55
Default Hello Thomas, First, a word
  #2
Senior Member
 
Hrvoje Jasak
Join Date: Mar 2009
Location: London, England
Posts: 1,758
Rep Power: 21
hjasak will become famous soon enough
Hello Thomas,

First, a word of warning: this is not easy. I hope you know your onions as far as finite volume method is concerned :-) I will limit my discussion to non-coupled boundary conditions: in other words, this does not apply for a cyclic, processor and other boundary types which formally create off-diagonal matrix coefficients.

Every boundary condition of what is left is implemented by adding a contribution to the diagonal and right-hand-side (source) of the cell next to it. The address of the cells is obtained by doing faceCells on the patch - just for your info.

Each boundary condition (of second order) can be expressed as a combination of a fixed value (Dirichlet) and fixed gradient b.c. Additionally, you have to worry about what operator the boundary condition works on: convection or diffusion.

When you look at the class fvPatchField, you will see for relevant functions:

- valueInternalCoeffs
- valueBoundaryCoeffs

- gradientInternalCoeffs
- gradientBoundaryCoeffs

The idea is that the first two give you the internal (read: diagonal coefficient) and boundary (read source) contribution for the convection term. The second two do the same for the diffusion term. The basic stuff will be found in the fixedValue, fixedGradient and mixed fvPatchFields.

Now that you got all that, we should tallk about your particular problem. I bet (might lose, but it is unlikely) that your boudnary condition is a combination of a fixed value and fixed gradient, where the value, gradient and value fraction (= mix) depend on some other conditions in the field. If this is the case, you should derive your boundary condition from mixed, set the refValue, refGradient and valueFraction based on what you wish to do and let the mixed b.c. do the rest of the job for you.

For vectors and tensors, there are some pretty comlicated ways of mixing things, e.g. separate mixed b.c. in the normal and tangential component like the stuff I did for the contact stress analysis with friction, but that's another story.

Hope you understand most of the above (otherwise there may be trouble) :-)

Enjoy,

Hrv
chegdan, ubald, marupio and 6 others like this.
__________________
Hrvoje Jasak
Providing commercial FOAM/OpenFOAM and CFD Consulting: http://wikki.co.uk
hjasak is offline   Reply With Quote

Old   August 5, 2007, 12:15
Default Dear Foamers, despites Hrvo
  #3
Member
 
sradl's Avatar
 
Stefan Radl
Join Date: Mar 2009
Location: Graz, Austria
Posts: 82
Rep Power: 9
sradl is on a distinguished road
Dear Foamers,

despites Hrvojes warning I've tried to implement my own boundary condition which shall represent convective heat transfer in the gas phase, i.e., a variable surface gradient of the temperature field in e.g. a solid. The gradient should be computed from

dT/dx=-alpha/lambda*(T_surface-T_gas,bulk)

with alpha, lambda, T_gas,bulk being constants and T_surface the temperature at the cell face of the boundary batch.

Using the core of the mixed or fixedGradient boundary condition, this can be done quite easily. The remaining problem is to get the surface temperature (T_surface) out of the matrix for evaluating the gradient field in the "evaluate" member function.
Ironically, this evaluate function should calculate the value at the patch boundary, so this might be trivial

Can anybody give me a hint how to conveniently get this surface scalarField for the temperature? Has anybody ever implemented something similar?

br
Stefan Radl
sradl is offline   Reply With Quote

Old   August 6, 2007, 04:55
Default Hi Stefan, I believe you sh
  #4
Senior Member
 
Thomas Jung
Join Date: Mar 2009
Posts: 100
Rep Power: 8
tehache is on a distinguished road
Hi Stefan,

I believe you should calculate your surface temperature, which is a result from your prescribed gradient, like this:

scalarField::operator=
(
(this->patchInternalField() + gradient()/this->patch().deltaCoeffs())
);


... but this is just another noobs opinion

regards,

Thomas
tehache is offline   Reply With Quote

Old   August 6, 2007, 16:05
Default Hi, thanks for your opinion
  #5
Member
 
sradl's Avatar
 
Stefan Radl
Join Date: Mar 2009
Location: Graz, Austria
Posts: 82
Rep Power: 9
sradl is on a distinguished road
Hi,

thanks for your opinion, but it does not work :-)

I can get the patchInternalField, which should represent the temperature in the cell centers of the patch cells. However, this seems to be a tensor field (the compiler complains about that when I try to subtract it from a scalarField)!! Consequently, I'm unable to calculate the difference between T_gas,bulk and this patchInternalField, which is clearly to me.

The question now is: how is this tensor interelated with the scalarField of the patch-temperature and how can I get it?

Any suggestions?

br
Stefan Radl
sradl is offline   Reply With Quote

Old   August 7, 2007, 05:50
Default Hmmm... I am calculating a gr
  #6
Senior Member
 
Thomas Jung
Join Date: Mar 2009
Posts: 100
Rep Power: 8
tehache is on a distinguished road
Hmmm...
I am calculating a gradient at the surface like this (in updateCoeffs(), for a radiation boundary condition. epslam is a factor containing thermal conductivity and emissivity. *this gives the field values at the surface. I think this is pretty much the same you want to do, for me it works.

gradient() = 5.67051e-8*epslam_*(-(*this)*(*this)*(*this)*(*this)+TRef_*TRef_*TRef_* TRef_);

in evaluate(), then, I do this, using underrelaxation:

scalarField::operator=
(
(1-relax_)*(*this)+relax_*(this->patchInternalField() + gradient()/this->patch().deltaCoeffs())
);

Works, and reults compare well with results from other programs.

regards,

Thomas
tehache is offline   Reply With Quote

Old   August 7, 2007, 06:15
Default Dear Thomas, thanks for tha
  #7
Member
 
sradl's Avatar
 
Stefan Radl
Join Date: Mar 2009
Location: Graz, Austria
Posts: 82
Rep Power: 9
sradl is on a distinguished road
Dear Thomas,

thanks for that hint. Let me ask you some more questions.

1) What have you used as a basis for your boundary condition c-file (maybe fixedGradientFvPatchFields.C ?)

2) What are the declarations for gradient, TRef? (maybe Field<type> or scalarField)

3) Are you compiling your boundary condition file together with the solver or are you creating a library and then linke this library to the solver?

I'm really wondering why you can have a scalar field in the evaluation function as in the fixedGradientFvPatchFields.C you have a Field<type>. I'm really confused about that my compiler calls for a tensor instead of a scalar. Would be great if you can send me your code to my email adress

stefan <dot> radl <at> tugraz <dot> at

to sort out this nasty issue.

br
Stefan Radl
sradl is offline   Reply With Quote

Old   August 7, 2007, 16:03
Default Dear Thomas, thanks for you
  #8
Member
 
sradl's Avatar
 
Stefan Radl
Join Date: Mar 2009
Location: Graz, Austria
Posts: 82
Rep Power: 9
sradl is on a distinguished road
Dear Thomas,

thanks for your explanation, with your guidance I did it!! A first test with a single cylindrical slab agrees with the analytical solution, so your/my implementation should be correct.

The problem was that I declared the variable as "Field<type>", i.e., valid also for matrices. Now everything is set to a scalarField and it works. I've not used the underrelaxation, the simulation is stable without it.

br
Stefan Radl
sradl is offline   Reply With Quote

Old   August 26, 2009, 13:25
Default diagonal and source of cell?
  #9
New Member
 
Sam Lee
Join Date: Mar 2009
Location: Shanghai, P.R.China
Posts: 4
Rep Power: 8
gameoverli is on a distinguished road
Hi, Hrvoje Jasak and others

I read your post concerning the boundary condition using following functions

- valueInternalCoeffs
- valueBoundaryCoeffs

- gradientInternalCoeffs
- gradientBoundaryCoeffs

I understood that convection term is using value of the boundary while the diffusion term is using the gradient of the boundary. I think that is why those function names are used? but, I am confused about the concept of diagonal and source of cell which differentiates two functions in one group. would you please explain that a little bit?

any hint will be appreciated.

thanks all
gameoverli is offline   Reply With Quote

Old   December 11, 2010, 08:44
Default
  #10
Member
 
George Pichurov
Join Date: Jul 2010
Posts: 39
Rep Power: 7
jorkolino is on a distinguished road
Quote:
Originally Posted by hjasak View Post
Hello Thomas,

Every boundary condition of what is left is implemented by adding a contribution to the diagonal and right-hand-side (source) of the cell next to it. The address of the cells is obtained by doing faceCells on the patch - just for your info.

Each boundary condition (of second order) can be expressed as a combination of a fixed value (Dirichlet) and fixed gradient b.c. Additionally, you have to worry about what operator the boundary condition works on: convection or diffusion.

When you look at the class fvPatchField, you will see for relevant functions:

- valueInternalCoeffs
- valueBoundaryCoeffs

- gradientInternalCoeffs
- gradientBoundaryCoeffs

The idea is that the first two give you the internal (read: diagonal coefficient) and boundary (read source) contribution for the convection term. The second two do the same for the diffusion term. The basic stuff will be found in the fixedValue, fixedGradient and mixed fvPatchFields.

Hrv
I have tried hard to instantiate an object of class fvPatchField without any success. It requires type, and I have tried anything from scalar and vector to volScalarField etc. Can you provide some real example of how to access the matrix coefficients of a scalar variable, named LMA?

Last edited by jorkolino; December 20, 2010 at 03:40.
jorkolino is offline   Reply With Quote

Old   October 16, 2011, 18:26
Default instantiate a virtual class
  #11
Member
 
ubald's Avatar
 
Nicolas Lussier Clément
Join Date: Apr 2009
Location: Montréal, Qc, Canada
Posts: 47
Rep Power: 8
ubald is on a distinguished road
Hi jorkolino,

fvPatchField is a virtual class. Meaning that you cant instantiate it. To do so you will need to create a class from this one like all the other patch class.


Wath you are tiring to do by instantiate fvPatchField is like tiring to create a object character in a RPG video game. The game as character object but you cant create a "character" you will need to choose which kind of character...
fvPatchField is there to define what is a patch so every patch will be child of fvPatchField.

Ther is a auto pointer class "autoPtr" that will aloud you to instantiate a virtual class. In some case it is useful but I'm guessing in your case it will be useless.

For the question about LMA, I'm sorry to tel you that I don't understand the question and I probably don't know the answer.

Hope you got the
answer you wanted and that this helped you some.

Regards
Nicolas L.


ubald 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
Burgerbs equation non constant Boundary Conditions Initial Conditions arkangel OpenFOAM Running, Solving & CFD 1 October 2, 2008 14:48
Addressing matrix element and reuse of system matrix marziolettich OpenFOAM Running, Solving & CFD 2 February 19, 2008 06:04
Integral boundary conditions turbulent intensitylength boundary conditions olesen OpenFOAM Running, Solving & CFD 0 July 27, 2006 07:18
matrix coefficients dominik Main CFD Forum 0 June 10, 2004 03:47
Elemtary matrix to CSR global matrix xueying Main CFD Forum 2 September 24, 2002 09:44


All times are GMT -4. The time now is 10:20.