Loading 2D axi data to 3D

Numerical methods and mathematical models of Elmer
Post Reply
adamo
Posts: 19
Joined: 02 Feb 2017, 00:19
Antispam: Yes

Loading 2D axi data to 3D

Post by adamo »

Dear Elmer community,

I am dealing with the following problem; I managed to solve a rather complex conjugate heat transfer model in 2D axial symmetry, which gives me the radial and axial flow velocities as well as the spatial temperature distribution. Now, I would like to trace particles in these fields, using the Particle Tracking solver in full 3D (particles coming from one side, intersect the flow and exit the domain).

Therefore, my question is: is it possible to use a result of 2D axisymmetrical simulation and convert the data to 3D geometry, in which I will then trace the particles? I can easily create a 3D mesh of the same geometry by revolving and meshing it in SALOME But I don't know how to "revolve" the 2D data and map it onto the 3D mesh.

Thank you very much for any advice

Adam
mzenker
Posts: 1999
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Loading 2D axi data to 3D

Post by mzenker »

Hi Adamo,

I would try to do the rotation in ParaView. There is a "rotational extrusion" filter which you could check out (I have never used it myself, though). If you have difficulties, the ParaView mailing list and/or their forum is there to help.
To get the values into Elmer, you could store them in an ASCII file and use the include command, as discussed here: viewtopic.php?f=3&t=4249

HTH,
Matthias
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Loading 2D axi data to 3D

Post by raback »

Hi Adamo,

In Elmer there is currently no dedicated module to map 2D results to 3D. There is the opposite the module ProjectToPlane.F90 (ch. 58 in Models Manual) but it is of course of not much use here. Mapping from 2D to 3D would infact be easier but it just hasn't been done. Hence I would agree with Matthias that this would be most conveniently done in Paraview.

-Peter
adamo
Posts: 19
Joined: 02 Feb 2017, 00:19
Antispam: Yes

Re: Loading 2D axi data to 3D

Post by adamo »

Hi Matthias and Peter,

thank you for the answers. However, I think you are referring to the Rotational Extrusion filter in ParaView https://www.paraview.org/ParaView/Doc/N ... usion.html . This filter, however, "forms a surface by rotating the input about the Z axis" so I do not get a volume data from it.

Nonetheless, I have devised a simple python script that converts my 2D cylindrical velocities to revolved 3D cartesian (x,y,z,Vx,Vy,Vz,rho). Is there a way to import this scatter velocity data to the ParticleTracking solver and let the solver use them as the background field? I either need to load them as an initial condition to a Navier-Stokes solver which will then not run or do I specify the environment velocity for the Particle Tracker somewhere else?

Thank you very much

Adam
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Loading 2D axi data to 3D

Post by raback »

Hi Adam,

The challenge is that the only supported way how Elmer reads in fields accurately is the "Restart File" option which is married with a particular mesh. So if you have generated 3D velocity values yourself it is difficult to use them as such.

You could take a value table in style [x y z f] in using DataToField module. This fits the data so your data would not need to be even conforming. Using this for the whole vector field might be cumbersome. I don't remember whether this supports vectors/several components.

The Particle machinery expects that you can give the name of a vector field where the different components are [v_x v_y v_z].

-Peter
adamo
Posts: 19
Joined: 02 Feb 2017, 00:19
Antispam: Yes

Re: Loading 2D axi data to 3D

Post by adamo »

I see, thanks. I did try with the DataToField solver which I did not manage to get to work (probably some silly mistake since it keeps saying WARNING:: LocateParticleInMeshOctree: Could not locate particle in the mesh!) but even if I did, it would not help me in loading a vector field.

So if I may have another follow-up question - I can get VTK or ASCII data in the points of the 3D mesh that I want to use for the calculation (using paraView). Would it be easier to load such a VTK or ASCII file, i.e. no interpolation would be needed on Elmer side, I just need to find the nearest point (identical point up to machine tolerance)

Thank you

Adam
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Loading 2D axi data to 3D

Post by raback »

Hi Adam,

I did some minor tweaking of the code such that one can use a 2D axisymmetric case to have an initial guess in 3D cartesian coordinates.

There is now an example as test case "NonconformingRestart4" that demonstrates how this could be done for a scalar field. I recognize there are still some steps needed for what you intend to do but I think this feature may have more general appeal and was hanging reasonably low.

-Peter

PS. If you're using precompiled version of Elmer you need to wait for the nightly builds.
annier
Posts: 1168
Joined: 27 Aug 2013, 13:51
Antispam: Yes

Re: Loading 2D axi data to 3D

Post by annier »

mzenker wrote: 26 Jul 2018, 18:35
I would try to do the rotation in ParaView. There is a "rotational extrusion" filter which you could check out (I have never used it myself, though). If you have difficulties, the ParaView mailing list and/or their forum is there to help.
To get the values into Elmer, you could store them in an ASCII file and use the include command, as discussed here: viewtopic.php?f=3&t=4249
Matthias,
I have read here (https://www.paraview.org/ParaView/Doc/N ... usion.html) that the rotation to 3D is possible around z axis only (of paraview). So, if that(z-axis of paraview coordinates) is the case it can be rotated , but for a random rotation from 2D to 3D, something needs to be done in the postprocessing solvers of ElmerSolver to convert 2D data to 3D within Elmer's domain , and visualize the 3D image in paraview.

Yours Sincerely,
Anil Kunwar
Anil Kunwar
Faculty of Mechanical Engineering, Silesian University of Technology, Gliwice
mzenker
Posts: 1999
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Loading 2D axi data to 3D

Post by mzenker »

... or else the ParaView rotation could be extended. Since ParaView filters can be programmed in Python, that could be the easier way...

Matthias
adamo
Posts: 19
Joined: 02 Feb 2017, 00:19
Antispam: Yes

Re: Loading 2D axi data to 3D

Post by adamo »

Thank you very much for implementing this feature Peter! I'll give it a try and I hope it won't be useful just to me :-).

As for the DataToField solver, I just had some routine mistakes there but it seems to be working fine as well. I think I will now try to reverse-engineer it to be able to import vectors and will post it here if I get any success :-).

Adam
Post Reply