[badoumba]

Hi everyone,

In my path of learning CFD, I've seen these 2 videos which seem to require capabilities that OpenFaom doesn't provide.

https://www.youtube.com/watch?v=RH1p...DCFF4CC5DC34A0
I read some pdf about implementing an anisotropic adaptive mesh refinement in OpenFoam, is it different from dynamicRefineFvMesh?

https://www.youtube.com/watch?v=PCj-82oYgUs
Moving objects is very limited in OpenFoam. I've seen similar things in XFlow if I am right about the process (https://www.youtube.com/watch?v=fRduyb4hA0s). I had a project which would require this but put it aside because of OpenFoam limitations in this area. However, this seems related to the same process as above in a way...

If anybody has any experience with this (these) capabilitie(s), I would be happy to know a little more.

Bye!

[mprinkey]

OpenFOAM was developed as a fully unstructured flow solver with grid motion and adaptivity included within that framework. Star and FLUENT are similar. Refinement/coarsening can occur in any way and remeshing and mapping the solution (as well as load re-balancing parallel runs) become a big issue. There are some slip-interface/non-conformal mesh techniques that can handle, say spinning turbo-machinery parts or passing trains or cars. Those are pretty efficient compared to arbitrary movement and remeshing.

There are other solvers that designed specifically with grid adaption and boundary motion. These are either Lattice-Boltzmann (aka, LB) usually on staircased Cartesian meshes, Finite-Volume methods usually on oct-tree refined Cartesian meshes, or some form of Smoothed Particle Hydrodynamics (SPH) which are meshless by design.

Gerris is flow package that is build around the oct-tree FVM approach. It is a well-designed open source code.

http://gfs.sourceforge.net/wiki/index.php/Main_Page

Marsha Berger et al, have a good paper outlining the technique as applied to an analysis of the space shuttle disaster:

https://www.nas.nasa.gov/publication...a2004_1232.pdf

That paper really delves into the details of how to make an efficient parallel adaptive CFD code.

If you google Lattice Boltzmann, you will find many (MANY) reference and applications. Grid generation and mesh motion is really trivial for them.

I don't know much about SPH. I've read enough about them to understand their shortcomings and that is usually when I lose interest.

I hope some of this helps.

EDIT: Sorry, I forgot to mention immersed boundary methods. I know a little about them:

http://folk.ntnu.no/skoge/prost/proc...apers/268b.pdf

These are techniques to overlay moving boundaries using some modification of the fluid transport equations under/near those immersed boundaries.

[badoumba]

Well Michael, I have enough to read for the week I think Many thanks

I used to work with Houdini from Side FX. This 3D software is famous for its dynamic simulations. Everything is based on discrete particles. I guess we are talking about CFD-DEM model if I am right. The solutions are pretty quick to render and according to DEM decription, the solid part is computed with Newton laws and the flow part by Navier-Stokes equations. In this simulations, we can have objects interacting with fluids. Are these solutions not accurate at all? Fluid scenes are very impressive though, even with real foam calculation. Why do we prefer working with a background mesh which then prevents us from flexibility when comes motion?

[mprinkey]

The DEM method is something different. That is a multiphase flow model for particles that are much smaller than the grid spacing. There IS a spatial grid in DEM for the fluid phase. You can read up on DEM at mfix.netl.doe.gov.

I think Houdini (and most rendering engine "liquid models") are based on something like smoothed particle hydrodynamics. They are fully discrete as particles and map nicely onto GPUs.

Usually, CFD work done in service of physics, chemistry, or engineering problems have simulation requirements that are not always satisfied by present mesh-free methods. Here is a nice overview of limitations:

https://astrobites.org/2011/10/08/th...hydrodynamics/

SPH is certainly an interesting research topic and I look forward to reading new applications and capabilities, but honestly, mesh-free methods like SPH got a very late start relative to mesh-based methods (FVM, FEM, FDM). So, some of the reason for mesh-based methods popularity are just first-mover advantage. But there is a lot of strong theory and decades of experience backing the tried-and-true mesh-based schemes. And, honestly, most engineered systems are NOT SUPPOSED to move that much, or at least, should move predictably. So, some of this is application area demands are just not there (yet?) for robust, efficient arbitrary boundary motion.

Most of the interest is arbitrary motion is on the computer graphics side (offline rendered animations, computer games) so they can get realistic flags flapping or animated hair blowing in the wind. In those cases, the hydrodynamics just has to be "good enough" to look like a wake behind a boat or a smoke plume, etc. And SPH is good enough now for most of those applications.

[badoumba]

I understand your point of view.

Consider this: https://www.youtube.com/watch?v=DUix21L2SzI
I guess pressure has been rendered for each position and the rendered pictures simply frame sequenced. No turbulences displayed here, the sequence rendering would not have worked of course as any turbulence would not have been linked to the previous calculated one. Maybe Lattice-Boltzmann could have come in rescue.

I am relatively newbie in CFD and so still very curious about the possible field of applications. This case illustrates an important point I guess. Mechanical cases still represent the majority of what we have to solve with CFD simulations but open source reduced the costs and opens new non-industrial market segments (my few current projects are with companies who could not afford a 20.000$ simulation) for new kind of challenges where things are often more organic and dynamic!

Thanks again for all your links

[saddy]

i am also looking for openfoam which can refine non hexahedral meshes.
i am looking for dynamicrefinefvmesh which can refine tetrahedral meshes like show in this video
https://www.youtube.com/watch?v=RH1p...DCFF4CC5DC34A0

i also a link to similar work done in a thesis by saumitra vinay joshi
https://www5.in.tum.de/pub/Joshi2016_Thesis.pdf go to page 51

if someone has this custom DynamicRefineFvMesh library kindly share at the forum
because openfoam is yet to include these features