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

3 discretisation solving questions

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

Reply
 
LinkBack Thread Tools Display Modes
Old   April 26, 2008, 10:21
Default Hello, I would be happy to
  #1
kar
Senior Member
 
Kārlis Repsons
Join Date: Mar 2009
Location: Latvia
Posts: 111
Rep Power: 8
kar is on a distinguished road
Hello,

I would be happy to read some answer to these 3 things currently confusing me:

1. Whats the point of ddtSchemes subdictionary (userguide u-109), if actual discretisation scheme is what's written by constructing fvMatrix?

2. May I call OpenFOAM a implementation of *linear* FVM? Is there anything really quadratic (or more) about it? (what are cubicCorrection, fourth?)

3. Is it necessary to use fvm::{Su/Sp/SuSp}, if matrices are somehow preconditioned? For example, if momentum equation includes volume and body force (or some heat sources)?

Thank you.
kar is offline   Reply With Quote

Old   May 3, 2008, 13:21
Default Ok, one question: are phi (sur
  #2
kar
Senior Member
 
Kārlis Repsons
Join Date: Mar 2009
Location: Latvia
Posts: 111
Rep Power: 8
kar is on a distinguished road
Ok, one question: are phi (surface field representing mass / volume flux) and U & mesh.Sf() supposed to be equal after PISO loop (U is velocity, of course)? Please answer to this!
kar is offline   Reply With Quote

Old   May 3, 2008, 20:30
Default Phi does not necessarily to be
  #3
Member
 
roy fokker
Join Date: Mar 2009
Posts: 44
Rep Power: 8
dbxmcf is on a distinguished road
Phi does not necessarily to be equal, however, it should obey the continuity after the 2 PISO correction steps, which is the key issue of the Pressure-Velocity correction procedure.
dbxmcf is offline   Reply With Quote

Old   May 4, 2008, 06:34
Default Another "mystery" related to t
  #4
kar
Senior Member
 
Kārlis Repsons
Join Date: Mar 2009
Location: Latvia
Posts: 111
Rep Power: 8
kar is on a distinguished road
Another "mystery" related to this is a steady solution of cavity - when using linear interpolation for convection, I get this ugly solution:



but, the problem is, when use some limited scheme, say GammaV 1, they do not disappear! When solving in transient regime, this problem vanishes -- why? Could you explain, please?
kar is offline   Reply With Quote

Old   May 4, 2008, 11:26
Default Hi, I don't know the exactly p
  #5
Member
 
roy fokker
Join Date: Mar 2009
Posts: 44
Rep Power: 8
dbxmcf is on a distinguished road
Hi, I don't know the exactly problem you have, my suggestion might be wrong, but I just think something "might" affect your solution. Then continuity equation is:

ddt(rho)+div(rho, U)=0

If you have compressible flow, phi should be rho*U which is interpolated to the face? Is your fluid density a constant? If not, then ddt(rho) might have an effect?
dbxmcf is offline   Reply With Quote

Old   May 4, 2008, 12:01
Default A simple case, where rho is co
  #6
kar
Senior Member
 
Kārlis Repsons
Join Date: Mar 2009
Location: Latvia
Posts: 111
Rep Power: 8
kar is on a distinguished road
A simple case, where rho is constant. Just for testing, so I can see if things work well, but they don't in fact.

Currently I've managed to crash icoFoam (and my own solvers) (see "OpenFOAM crashes when input is a steady solution!!").

A more constructive question would be: how to calculate phi field, if *only* U field is given (say, for initial conditions)? U&Sf result is different from solved one...
kar is offline   Reply With Quote

Old   May 4, 2008, 17:28
Default Hi, I think the U field is int
  #7
Member
 
roy fokker
Join Date: Mar 2009
Posts: 44
Rep Power: 8
dbxmcf is on a distinguished road
Hi, I think the U field is interpolated to the control volume surface while calculating the phi. Since your rho is a constant, the phi is using the interpolated surface U times the surface area, that is why phi is coded as "surfaceScalarField" when used in the convection term (both in the pressure correction equation and the pseudo whole N-S equation for the PISO loop), I think a complete phi expression (for the compressible flow) should be:

linearinterpolate(rho*U) & mesh.Sf()

Correct me if I am wrong, thanks!
dbxmcf is offline   Reply With Quote

Old   May 5, 2008, 04:11
Default Well, I already did that. Sinc
  #8
kar
Senior Member
 
Kārlis Repsons
Join Date: Mar 2009
Location: Latvia
Posts: 111
Rep Power: 8
kar is on a distinguished road
Well, I already did that. Since my solver is incompressible, I used fvc::interpolate(U)&mesh.Sf() and that doesn't give me the same result for saved, converged U field as PISO loop does!
So I ask again: why there is difference and how should I calculate phi from U?
kar is offline   Reply With Quote

Old   May 7, 2008, 04:17
Default To display difference field, I
  #9
kar
Senior Member
 
Kārlis Repsons
Join Date: Mar 2009
Location: Latvia
Posts: 111
Rep Power: 8
kar is on a distinguished road
To display difference field, I used div( interpolate_lin(U)&mesh.Sf() - phi ) and result looks like a fluctuations around zero. Also boundaries look problematic for some reason I don't know...

Here's a sample:


kar is offline   Reply With Quote

Old   May 7, 2008, 04:27
Default (as Bernhard says: Warning - a
  #10
Senior Member
 
Hrvoje Jasak
Join Date: Mar 2009
Location: London, England
Posts: 1,758
Rep Power: 21
hjasak will become famous soon enough
(as Bernhard says: Warning - advertising own stuff!)

Have a look at my Thesis - the bit about pressure-velocity coupling, or try to dig out my CFD lecture material from somewhere. It will tell you what the difference between the interpolated velocity and face flux is and what to do if you insist on full consistency. For reference, this is to do with Rhie-Chow treatment.

Hrv
__________________
Hrvoje Jasak
Providing commercial FOAM/OpenFOAM and CFD Consulting: http://wikki.co.uk
hjasak is offline   Reply With Quote

Old   October 14, 2008, 06:11
Default Karlis, did you find out what
  #11
Senior Member
 
Ivan Flaminio Cozza
Join Date: Mar 2009
Location: Torino, Piemonte, Italia
Posts: 207
Rep Power: 9
ivan_cozza is on a distinguished road
Send a message via MSN to ivan_cozza
Karlis,
did you find out what happens if you choose cubicCorrection as interpolation scheme?
I'm facing a similar and strange problem:
I'm solving linearized Euler equations for a Gaussian acoustic pulse, and I tryied linear :

gradSchemes
{
default Gauss linear;
}

divSchemes
{
default Gauss linear;
// div(phi,utfold) Gauss limitedLinear 1;
}

laplacianSchemes
{
default none;
}

interpolationSchemes
{
default cubic;
}

and cubic schemes:

gradSchemes
{
default Gauss cubic;
}

divSchemes
{
default Gauss cubic;
// div(phi,utfold) Gauss limitedLinear 1;
}

laplacianSchemes
{
default none;
}

interpolationSchemes
{
default cubic;
}

and the results are the same!!
I'm wondering what I'm wrong in it...
Have a good day!
ivan_cozza 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
Discretisation sa FLUENT 0 March 8, 2007 08:19
FEM discretisation for eps-equation Sigit Main CFD Forum 3 October 26, 2000 11:49
finite volume Discretisation mehdi Main CFD Forum 1 May 31, 2000 15:29
3d discretisation of compressible NS equations sanku Main CFD Forum 1 January 30, 2000 22:01
FV discretisation on tet meshes peter grafenberger Main CFD Forum 7 October 29, 1999 06:32


All times are GMT -4. The time now is 17:43.