Questions on the numerical methods used in Code Saturne

  • Alicia
  • Topic Author
  • Offline
  • Junior Member
  • Junior Member
More
13 years 10 months ago #4296 by Alicia
Hi again !

I don't know if this topic has its place here, but it seems I can't get an account on the forum dedicated to CS (I sent a mail a few days ago and still don't have any answer).

I have a few questions regarding the numerical methods implemented in Code Saturne :

1) How is the pressure/velocity coupling resolved (which algorithm is used) ? When should I check the option 'pseudo coupled velocity/pressure solver' ? What exactly does it change for the calculation ?

2) What does the relaxation coefficient in the steady management tab stands for ? Is it the same as under-relaxation coefficient for pressure, momentum, energy, ... ? If not, where can I specify such under relaxation coefficients ?

3) In the k-epsilon model, what exactly is the difference between two scale model and scalable wall function ? (I get that I should use scalable WF when I have a small mesh near the walls, but don't really understand what are the equations behind this model...).

4) I'm still looking for a test case (a tutorial) for a calculation with radiative transfer and/or conduction. I only found a tutorial for natural convection and radiation for Fluent, and I'm not really sure about the results... If someone could give me a detailed description of a case using these kind of heat transfer, with the results obtained with saturne, it would be perfect !! :D

I'm sorry to bother you with all those theoretical questions... !
Thank you very much for your help !

Alicia<br /><br />Post edited by: Alicia, at: 2010/05/27 19:19

Please Log in or Create an account to join the conversation.

More
13 years 10 months ago #4305 by Yvan Fournier
Hello,

Did you send a mail to This email address is being protected from spambots. You need JavaScript enabled to view it. or to another address ?
I checked that adress and registered 2 people yesterday, but I do not remember seeing your e-mail (several of us have access to that account and share the workload). The forum at code-saturne.info is off-line for the week-end, apparently due to some server maintenance, but should be back on-line Monday.

I'll try answering your other questions, but I am one of the I/O/file format/parallel algorithm specialists, and not the most knowledgeable developper regarding the numerical algorithms.

1) by default, velocity/pressure coupling is based on a SIMPLEC algorithm, with Rhie and Chow interpolation to avoid spurious modes related to the colocated velocity and pressure discretization.
I am not sure about the way the pseudo-coupled algorithm works, but it is recommended for cases where viscosity is dominant for example.

2) the relaxation parameter is a numerical parameter, similar to the one used for the unsteady algorithm (as the steady algorithm works using sub-iterations, in a matter somewhat similar to the unsteady algorithm with a time-and-space varying time step). Hopefully, the default value is reasonable, but we have much less feedback on the steady algorithm, as it is a relatively recent addition, and many of our studies require an unsteady algorithm anyways.

3) I'll have to ask colleagues for that one, and we probably need both documentation improvements and a FAQ on the forum relative to turbulence models, as questions about those come up quite often. As there is some turbulence research going on there as well, you may also look at the following Forum and wiki at the University of Manchester: cfd.mace.manchester.ac.uk/twiki/bin/view/Forum/WebHome

4) I am not sure I understood your question: do you want either conduction or radiation, or do you want radiation in any case ?

Best regards,

Yvan Fournier

Please Log in or Create an account to join the conversation.

  • Alicia
  • Topic Author
  • Offline
  • Junior Member
  • Junior Member
More
13 years 9 months ago #4309 by Alicia
Hi Yvan,

Thank you very much for your helpful answers.

For the registration on the forum, I used the link 'contact us to create an account', which (I just checked the e-mails I sent recently) use the e-mail address This email address is being protected from spambots. You need JavaScript enabled to view it.. I will try again today.

As for the test case I'm looking for, I will have to run a calculation using at the same time the thermal module, the radiative transfer module and forced convection. Is it possible to combine the 3 heat transfer modes in the same calculation with CS ? Before running this calculation, I have to check that I am able to use these different modules correctly. I already did so for the convection and radiation/convection, so I would like to have a test case either with conduction and convection or with radiation and condcution and convection. I hope I am a bit clearer here.

For the information on the turbulence models, I will check the wiki at the university of Manchester.

Thanks again for the help,

Alicia

Please Log in or Create an account to join the conversation.

  • Alicia
  • Topic Author
  • Offline
  • Junior Member
  • Junior Member
More
13 years 9 months ago #4310 by Alicia
Hi again,

I came up with a few more questions, and didn't want to post a new topic !

First, did someone in the code saturne team already run calculation using the GPU (which allows to use the processors of the graphic card for the calculation, thus reducing significantly the time needed for the calculation) ? What shall I do to do so (I assume I have to modify some routines or functions, which ones) ?

Second, not related to the numerical method, I have to model an autoclave. The configuration of the problem is a cylinder wrapped in an 'air vein'. So there is an enveloppe in which there is an air circulation, the air is injected in the cylinder through a grille at an end of the cylinder, and at the other end of the cylinder there is a heating device and a turbine which sucks air and create the circulation. (I'm not sure if I'm clear, and if my English is correct on these technical terms...).

Basically all I want to know is how I can model the turbine (which is basically just an outlet of air velocity I think) ? Should I model it like a fan or specifying a velocity at the outlet is enough (considering that the outlet is in this case fully inside the domain) ?
Should the grille be present in the model geometry, or is there a way to set a specific inlet shaped like a grille (I know that in some softwares it is possible to do so) ? I personnally think that I should include the grille in the geometry and create the mesh around it, but maybe I'm wrong.

Thank you very much for your time and your answers !!

Regards,

Alicia

Please Log in or Create an account to join the conversation.

More
13 years 9 months ago #4313 by Yvan Fournier
Hello,

First, I checked, and your account on our forum was created by my colleague David Monfort last week, so you probably received an e-mail from a server giving you your account info. If you do not find it, we can reset your password so that you receive a new e-mail.

Code_Saturne is not currently able to use GPU computing, and nothing is planned in this regard yet. When we look at recnt bibliography, GPU computing often leads to speedups in the range of 100 for dense or regular patterns, but only 3 to 5 for unstrucutured Matrix-vector products for example. As we use an unstructured data representation (whether the meshes we use are initially structured or not), and the calculation goes through many stages (complex to optimize) the potential gain actually seems quite limited given the current state-of-the art.
Some compilers such as recent PGI compilers are able to generate code for GPUS directly from C/Fortran code such as ours, so it would be easy to test this with Code_Saturne given the right hardware, but I am not sure there would be a large gain (though if anyone wants to check, we can help them with compile options and are interested in the feedback).

I am not quite sure of your geometry without a diagram, but using a simple outlet and inlet would seem sufficient to me, unless you need to model rotation effects due to the fan's presence.
A fan model exists in Code_Saturne, but is not well documented (see the code-saturne.info forum for recent discussion on the use of this functionality).

As for using at the same time the thermal module, the radiative transfer module and forced convection, I do not believe there should be any problem. I know a colleague is working on radiative transfer coupling with the SYRTHES thermal code, as coupling with SYRTHES currently only functions with that code's conduction module and not yet directly with its transparent radiative module, but using Code_Saturne's semitransparent radiative module with conduction and convection (which are part of the base model) should not be an issue.

Best regards,

Please Log in or Create an account to join the conversation.

  • Alicia
  • Topic Author
  • Offline
  • Junior Member
  • Junior Member
More
13 years 9 months ago #4314 by Alicia
Hello Yvan,

Thanks again for your answers.

I indeed received the confirmation e-mail for my account last week, but it went directly in the spam box, and I never check if there is any lost e-mail there...

For my case, I will try with the basic inlet and outlet and let you know how it works. As for the coupling with Syrthes, I have to try first with basic test cases as I've never used it.

Regards,

Alicia

Please Log in or Create an account to join the conversation.

Moderators: catux
Time to create page: 0.138 seconds
Powered by Kunena Forum