Re: oscillations when solving nonlinear system

Am Wed, 05 Oct 2005 06:21:08 -0400 schrieb Torsten Hennig:

>>Hi Torsten,
>>
>>thanks for the advise. Unfortunately I am a nooby in >using such codes. Is
>>there also a version for c?
>>
>>Thanks,
>>Andreas
>
>
> Hi Andreas,
>
> SUNDIALS is a suite of solvers for solving such
> systems of differential-algebraic equations and is
> written in C, but it is maybe too complex for your
> purpose:
>
> http://www.llnl.gov/CASC/sundials
>
>
> There is also a C-interface to the RADAU-code,
> available under
>
> http://www.unige.ch/~hairer/software.html
>
>
> Best wishes
> Torsten.

Well the C-interface doesn´t work on my machine. I don´t know why...
But I have a question concerning the derivation of the jacobian.
I think there are 2 ways:
1. Solve the system at every point in z direction
2. Solve the system for all points in z direction

The implementation of the first way is pretty easy, but I think it's
slow.
The second should be more fast, but the derivation of Jac is complicated.
When I descretize equation (1):
F1=a \frac{\partial \phi}{\partial z}+b(\phi-\psi)=0
to
F1=a/(\Delta z) {\phi^j-\phi^j-1}+b(\phi^j-\psi^j)=0.
Do I have to derive dF1/dphi^(j-1)?

Andreas
.

Relevant Pages