Sorry I was not clear in my original question. Thanks for the suggestions.  

Kris

-----Original Message-----
From: Jonathan Guyer <[email protected]>
Sender: <[email protected]>
Date: Tue, 26 Mar 2013 08:57:04 
To: FIPY<[email protected]>
Reply-To: <[email protected]>
Subject: Re: branching grid


On Mar 25, 2013, at 7:17 PM, Kristopher Kuhlman wrote:

> I am familiar with domain decomposition methods. This seems a quite abstract 
> way to tackle the problem.  I don't want to artificially introduce iteration 
> into a linear diffusion problem.  It is really just a mesh-generation issue.

Are you trying to solve on a physically branched domain, like 
http://en.wikipedia.org/wiki/Bronchus or 
http://en.wikipedia.org/wiki/Exhaust_manifold, or are you looking for an 
abstract solution method for loosely associated domains. I originally thought 
you wanted the latter, and expect Benny did, too.

If, instead, you are interested in solutions on a physically branched domain, 
then you should mesh the domain you want. Depending on its complexity, you may 
need to use a CAD tool to generate the abstract geometry and then a meshing 
tool to discretize that geometry. All of that is beyond the scope of FiPy. If 
you can generate the mesh in Gmsh, or import it in a format that Gmsh 
understands, then FiPy should be able to use it.

> I was thinking I could use a 2D mesh and a tensor with very low permeability 
> in one direction, which would essentially enforce 1D flow in the branch 
> region and then make the tensor low permeability in the other direction on 
> the branches, to enforce 1D flow in the other direction along the trunk.

If the mesh is branched, then the mesh will enforce the effective 1D-ness of 
any branch. 

Rather than trying to do something strange to one of FiPy's existing meshes, I 
think it's more straightforward to just generate the mesh you want. If simple 
enough, you can do, e.g.,

>>> mesh = (fp.Grid2D(nx=10, ny=1) + (fp.Grid2D(nx=1, ny=5) + (fp.Grid2D(nx=3, 
>>> ny=1) + [[1], [2]]) + [[2], [1]]))

otherwise Gmsh or some other meshing tool.


> Based on other threads, it appears FiPy can't have spatially dependent 
> properties AND use tensors.  

What threads are those? I am not aware of any situation where any FiPy 
coefficients cannot be spatially dependent. If they exist, then they are bugs.


_______________________________________________
fipy mailing list
[email protected]
http://www.ctcms.nist.gov/fipy
  [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]

_______________________________________________
fipy mailing list
[email protected]
http://www.ctcms.nist.gov/fipy
  [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]

Reply via email to