Thank you for the insight.

The coefficient for the diffusion term is

thermal conductivity / (density * specific heat)

where thermal conductivity and specific heat are both non linearly
temperature dependent. Thermal conductivity is approximated by
10^(a(u)/b(u)) where a and b are polynomials in u.

Specific heat is approximated by a piece-wise defined cubic spline in u/100.

So, just to be sure, would you recommend using the chain rule as given in
your first reply? (DiffusionTerm(coeff=A) + A.grad.dot(u.grad))

Thanks again,
Chris




Chris Jones

*Project Electrical Engineer*

[email protected] | 603-601-3340




*ANTAYA* *SCIENCE & TECHNOLOGY*

7A Merrill Industrial Drive

Hampton,  NH  03842

603-601-7474


This message may contain information that is confidential, privileged or
otherwise protected from disclosure. If you are not the intended recipient,
you are hereby notified that any use, disclosure, dissemination,
distribution, or copying of this message, or any attachments, is strictly
prohibited. If you have received this message in error, please advise the
sender by reply e-mail, and delete the message and any attachments. Thank
you.

On Thu, Jan 8, 2015 at 2:05 PM, Guyer, Jonathan E. Dr. <
[email protected]> wrote:

>
> On Jan 8, 2015, at 10:30 AM, Christopher Jones <[email protected]> wrote:
>
> > I am modelling nonlinear heat conduction using time steps and sweeps.
> >
> > Basic equation is:
> >
> > du/dt = A(u) * d2u/dx2 + S(x, t)
> >
> > where u(x, t)
> >
> > I have functions written for A(u), the DiffusionTerm coefficient, and
> S(x, t), the ImplicitSourceTerm. I implement them with the following
> snippet:
> >
> > eq = TransientTerm() == DiffusionTerm(coeff=A) +
> ImplicitSourceTerm(coeff=S)
>
>
> DiffusionTerm represents d/dx(A(u) * du/dx) (or, more rigorously,
> nabla\cdot(A(u) \nabla u)).
>
> If your functional form *really* is A(u) * d2u/dx2, then you need to run
> the chain rule to obtain \nabla\cdot(A(u) \nabla u) - \nabla A(u) \cdot
> \nabla u, which equates to DiffusionTerm(coeff=A) + A.grad.dot(u.grad). It
> is rare, however for this to be needed. The form of DiffusionTerm (with A
> inside the first derivative) is the one almost invariably encountered.
>
> > S is really the whole source term, not the coefficient, so have I coded
> this correctly?
>
> Raymond has already pointed out the linearity of ImplicitSourceTerm. If
> your source is linear in u, then using +ImplicitSourceTerm(coeff=S1) is
> more efficient than using +S1*u. If your source isn't linear in u, then
> don't worry about it and write the source explicitly.
>
> > I have declared A as a FaceVariable (because diffusion happens from cell
> to cell, and S as a CellVariable, is this correct?
>
> Yes.
>
> > Since S depends on t as well, I created a Variable t. I then created a
> small function and included it with S with this snippet:
> >
> > B = 50 * numerix.exp(-t)
> >
> > S.setValue(B)
>
> >
> > After sweeping, I increment the elapsed time and then t, with
> >
> > t.setValue(elapsedTime)
> >
> > I can print B's value to see that it does decay.
>
> .setValue() sets the instantaneous value of S. S won't change with further
> changes in t.
>
> Skip B and just write
>
> S = 50 * numerix.exp(-t)
>
>
>
> _______________________________________________
> 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