On May 13, 2007, at 11:03 PM, Matt Koch wrote:
I think for some reason, my e-mail below never made it to the
mailing list ([email protected]), at least it does not seem to show up
on "gmane". Perhaps it is because I used the wrong e-mail address
in the "from" field?
Looking through our logs, yes, it looks like your message was bounced
on May 3rd because it came from an alum.mit.edu address. Did you not
get a rejection message at that time? I will be happy to beat on our
list admins who should have switched to MailMan a long, long time ago.
1) I have used some meshing functions that allow extraction of the
left, right, bottom and top boundaries of a rectangular mesh.
However, if I want to work off of GMsh meshes, I need to be able
to extract boundaries (and domains) by GMsh's "Physical Line",
"Physical Surface" and/or "Physical Volume" ID numbers. Can that
be done in FiPy? In other words, how would I assign a constant
flux boundary condition to GMsh "Physical Line" 25, for example?
I don't think this is easily obtainable. A mesh imported from gmsh
does have some internal mapping tables between gmsh's indices and
ours (mesh.nodeToVertexIDs, mesh.unsortedBaseIDs,
mesh.cellFaceVertexIDs, mesh.baseFaceVertexIDs,
mesh.faceStrToFaceIDs), but I think we may ignore the "Physical ..."
objects altogether. I can't promise that any of those internal
variables is going to be useful to you, but that's where I'd start.
If that doesn't work, then I think there's presently no choice but to
identify faces by coordinates. This mechanism for identifying faces:
>>> myFaces = mesh.getFaces(filter=???)
is older than the "where=" syntax that we use now, and is very
inefficient, but it should work (fortunately, you only have to do it
once, at the beginning of a simulation). Unlike "where=", the filter
must identify faces one by one. I'm not sure if we have any examples
of its use and don't have time to look at the moment, but will post
later today if Daniel doesn't beat me to it.
2) I have seen constant flux and constant temperature boundary
conditions, but how would I assign something like the typical
convective & radiative boundary condition, i.e. -kdT/dn = h(T-T0)
+ s(T^4-T0^4), which is a combination of both constant flux and
constant temperature?
I understand why you view this as a combination of fixed flux and
fixed temperature, but it doesn't look to me like it actually *fixes*
the temperature; rather, it produces a flux if the temperature ever
deviates from the target value. If you actually imposed a FixedValue
boundary condition on temperature, then it *could not* deviate from
the target value at the boundary and the flux would always be zero by
definition. If I'm misunderstanding the problem (quite possible),
please correct me.
This *should* be as simple as setting a FixedFlux boundary condition
equal to (h(T-T0) + s(T^4-T0^4)) / k. Unfortunately, our FixedFlux
boundary conditions don't do the right thing for anything other than
zero flux. (If you only have a diffusion term, with no convection,
then it might work, but I wouldn't count on it).
The way we've been fudging this in our own codes is to produce a
SourceTerm at the boundaries that does the right thing. We expect to
automate this process in the next release of FiPy, but in the
interim, this is what I would do:
>>> boundary = FaceVariable(mesh=mesh, value=0)
>>> faceX = mesh.getFaceCenters()[...,0]
>>> faceY = mesh.getFaceCenters()[...,1]
>>> faceZ = mesh.getFaceCenters()[...,2]
>>> boundary.setValue(1, where=(faceX == ???SOMETHING DESCRIBING
X ON THE BOUNDARY???) & (faceY == ???SOMETHING DESCRIBING Y ON THE
BOUNDARY???) ...)
Do that last step on as many different boundaries as you like, or
come up with a single `where` statement that describes all of them.
>>> deltaT = T.getArithmeticFaceValue() - T0
>>> boundaryFlux = VectorFaceVariable(mesh=mesh, value=boundary
* (deltaT * h / k + deltaT ** 4 * s / k))
>>> heatFluxEqn = TransientTerm() == DiffusionTerm() + ??? +
boundaryFlux.getDivergence()
Because the boundary condition is nonlinear in T, it will probably be
necessary to sweep to convergence.
3) I understand that the FIPY_VIEWER environment variable can be
set to the preferred viewer. However, is there a way I could
invoke my favorite viewer straight from the Python script, without
having to rely on this environment variable?
If, for example, you always wanted to use matplotlib, irrespective of
dimensionality (presently you are setting FIPY_VIEWER to
"matplotlib"), then you would do
>>> from fipy.viewers import matplotlibViewer
>>> viewer = matplotlibViewer.make(vars=...
On the other hand, if you wanted to use matplotlib for 1D plots, but
gist for 2D plots, then you would do
>>> from fipy import Matplotlib1DViewer
>>> viewer1 = Matplotlib1DViewer(vars=(my1Dvar1, my1Dvar2, ...
and
>>> from fipy import Gist2DViewer
>>> viewer2 = Gist2DViewer(vars=(my2Dvar1, my2Dvar2, ...
