On Tue, Jun 11, 2013 at 9:52 AM, Kundar <[email protected]> wrote:

> **
>  Dear fipy users,
>
> I cooks my fourth order diffusion term as:
> DiffusionTerm(coeff=(Gamma1, Gamma2))
> where Gamma1 and Gamma2 are tensors, or Gamma1 is a tensor and Gamma2 is
> a scalar.
> When I throw it into fipy, it runs. As the manual says that it does not
> support the anisotropic high order terms, I'm wondering whether the result
> is reliable or fake.
>

Thanks for your email. I wrote this part of the code and I am only sure
that the anisotropic terms are correct for regular diffusion (not higher
order diffusion). I have never tried running a problem with the higher
order (biharmonic terms) and anisotropy. When I wrote this I had the
feeling that something was being overlooked, but I can't remember now. I
think the only sensible way forward it to devise a few test cases and check
the solution carefully (if you come up with some tests and FiPy doesn't
work then I will try and fix it if possible).



> How to check the accuracy?
> If the above representation is incorrect, please show me an
> alterantive method in fipy, or help me to put this fourth order diffusion
> term into a right form. Thank you very much.
>

An alternative way is to just use coupled equations in FiPy with
anisotropy, but I can't be sure that anisotropy and coupled equations works
either, it is untried. If you are unsure about coupling equations, see


http://www.ctcms.nist.gov/fipy/examples/phase/generated/examples.phase.polyxtalCoupled.htmland

and


http://www.ctcms.nist.gov/fipy/examples/cahnHilliard/generated/examples.cahnHilliard.mesh2DCoupled.html

The coupled code and the higher order code should give the same results
though if FiPy is internally consistent. You (me or someone) needs to come
up with some nice test cases. Do you know of any papers with test cases for
this?

-- 
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