CFD codes for industrial CFD

Hi all,

I am currently doing an internship in which I have to create some sort of service (creating geometry, meshing my geometry, solving flows, post processing) for flows around buildings.
Thus I need tools and softwares able to create complex geometries and to compute mesh with an important number of volumes from complex geometries. I am currently using Salome 4.1.4 to create my geometry, and then GMSH 2.3.1 to mesh it, but so far it is not very efficient (I have some problems meshing curve surface or volume for instance in gmsh). To solve flows I am using Code_Saturne, which seems to be efficient but quite slow (I am running under a 6GB ram machine, CAELinux with Ubuntu 8.04, code_saturne 1.3.3), and Paraview 3.3.0 for post-processing (I am so far happy with it).

I just wanted to know if someone there is using such big meshes and such complex geometries, and which tools are performant to do that.

Thanx very much in advance for your answers,

Alicia

Hi !!

First : Thank you for your quick reply !!

I finally managed to mesh my volume with gmsh (it was just a box with a cylinder in it) : when meshing the 3D volume, I got this error message : Error : No tetrahedra in volume 1
After a while, I tried to first create a surface (for this time my geometry was just an extrusion of a face), then mesh it, and then extrude the face and mesh it. It worked just fine.

With Saturne, I managed to run a calculation with a lower precision of the solver (10-6 as you said), and a pretty big mesh (around 2e6 volume cells) in an acceptable time (a bit more than 2 hours for 100 time steps).

My problem is now when loading my results file in Paraview. I can actually load the .case file, but I don't have access to any of the data after loading the file... I can only display these : Solid color and vtkCompositeIndex, which are only the boundaries of my domain.
I don't have such a problem with the same case but a smaller mesh (something like 50 000 volume cells). I don't understand the problem here...

Unfortunately I won't be able to post some of my results here, all my work is confidential... !! I hope it won't be a problem for you to help me ! But when I have more time I'll try some other cases at home and post the results.

have a good evening all,

Alicia

Yeah, definitely ditch gmsh. It's way too slow.

Hi everyone !

Thanks for the tip.

When running the calculation, I get the message Normal Simulation Finish, my results files seem ok.
To try to find the problem, I did exactly the same steps with a small mesh and with a big one. So I have a geometry, which I mesh a first time. I obtain the small mesh. Then I refine this mesh twice and got the big mesh.
I run the calculation with exactly the same parameters : same precision for the solver, same time step, same number of time steps, same physical parameters... I only changed the mesh at the input.
In both cases, I got the message normal simulation finish, and have the same files in both of the directories RESU.
Then I launch Paraview, and open the file CHR.case in Little Endian. I load the data, and in the Information tab of the object browser, I got exactly the same infos : size of the mesh, structure of data, range for values, time steps saved, size of the domain, ....
The same difference is that with the small mesh, I can display all my data, and with the big one I can choose between Solid Color and vtkCompositeIndex. I attach here a screenshot for both cases.
I tried, as yu said, to launch paraview from a terminal, but didn't get any error message. My only guess is that or paraview doesn't manage well big amounts of data (which seems a bit crazy because all i read about this software says that it's it aim...) or that there is something to do to get the data when there are a lot of them that I don't know about...
Here are the screenshot. I hope someone will be able to help me on this !

Alicia