On Fri, Feb 24, 2012 at 9:12 AM, wang yunbo <[email protected]> wrote:

> Hi Dr. Wheeler,
>
> I thought I still need fipy because I need facegrad and divergence. Do
> they all have counterparts in lsmlib? I have a couple of other questions.
>

Sure. It's not hard to construct curvature with numpy indexing if you need
to double check.


> 1.Computedistance2d is similar to calcdistancefunction() in fipy, is it?
>

It's exactly the same when order is set to 1 in lsmlib (I've checked).
Fipy's distance function is only first order. Actually, I've just replaced
the native distance function calculations in fipy with lsmlib's and all the
tests pass. See <http://matforge.org/fipy/browser/branches/lsmlib>. The
lsmlib wrapper stuff is now in <
http://matforge.org/fipy/browser/branches/lsmlib/fipy/tools/lsmlib>.

How often is it required to use this function to update? As often as
> possible?
>

That's a tricky question. I would recommend reading Sethian's level set
book. In theory, for a suitable velocity one would only need to calculate
the distance function once, in practice it is more than that due to
numerical issues. Calculating the distance function too frequently does
actually introduce error into the position of the interface so one does
have to be somewhat careful.


> 2. Some times I can only calculate the extension velocity where distance
> variable has a value of zero ( a very thin layer).
>

Why?

What is the best way to update distance variable in this situation?
>

That doesn't make much sense to me.

-- 
Daniel Wheeler
_______________________________________________
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