I am new to fipy, so I am just getting my feet wet.  I'm quite experienced 
with Comsol, which I only mention because it has capabilities to approach 
the problem I am going to mention.  I am interested in modeling thermal 
(or electrical) conduction in a 3D region, but that region has, on a part 
of one of its surfaces a thin, highly conducting region.  To connect to 
the real world, one can envisage this being a 70 micron thick copper pad 
at the top surfaces of a 1mm thick printed circuit board (PCB), where the 
PCB material is FR4 with a much lower thermal (or electrical) conductivity 
than the copper.  There would be a heat source on the pad.  The general 
approach to handle the conduction in the thin region is to assume a 
uniform temperature across the thin dimension, integrate across that 
dimension, and end up with a 2D PDE in the transverse direction.  What 
started out as a Laplace equation then becomes more like a Helmholtz.  The 
non-derivative term is simply the sum of the fluxes out the top and bottom 
surfaces.  The bottom flux should be equal to the flux into the 3D region 
below.  Thus, in an analytical sense, it's quite straightforward to think 
about.  The question is, how to handle in Fipy?  I can think of a number 
of different ways of approaching this, but not sure which might be the 
best, not being that familiar with the nuances of Fipy.  Ideally, one 
could make the 2d region's equation a boundary condition for the 3d (in 
the correct cell faces of the 3d region of course), but I don't know if 
one could specify a constraint that has a 2D transvers Laplacian of the 
dependent variable.  Alternatively, I thought that maybe I could have two 
coupled equations, one for the 2d region and one for the 3d region, but 
then one would have a 3d mesh and the other a 2d mesh.  Is that possible?  
Finally, perhaps one could make the thin region be just one cell thick in 
the thin dimension, and simply treat it as a 3d region.  Doesn't seem like 
it would add a great deal of numerical burden with the one extra layer.  
Anyway, I am curious about how the "gurus" would approach this.  As I 
mentioned, I do this in Comsol often, but it's not quite clear what they 
are doing "under the hood" in this instance.  That is one reason why I 
like Fipy and am starting to get into it.  

_______________________________________________
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