2nd resolution Wave_t_ex

[Christophe Geuzaine]

Lin Ji wrote:
> 
>   First, thank you for your quick response. I still want to use the 2nd
> resolution since it is much faster. But, there is one more thing I want to
> clarify. When you say to impose the spatial kind of distribution I want in
> the formulation, does it mean I have to define two functions 'dfdt1[]' and
> 'dfdt2[]' and two formulations for the first two resolutions, one ('Wave_t')
> use the 'dfdt1[]' function defined in both time and space, the other
> ('Wave_t_ex') use the 'dfdt2[]' function defined only in space? In this case,
> dfdt1[] would be dfdt2[]*TimeFct[]. Am I right?

Yes. You could also simply change the definition of dfdt[].

>   Also if I set the 'Re_Use_ILU' flag to 1, will it affect the accuracy of
> the solution? What is the trade off for speed up the computation?

There is no tradeoff on the accuracy. The ILU is only used as a
preconditionner for the iterative solver, so that all you could get is
more and more solver iterations during the time stepping. The accuracy
of the iterative solver is specified with the 'Stopping_Test' parameter.

>   Finally, if you have time, can you say more about using higher order (2nd
> order) interpolation? I guess I must have made some mistake because the
> result I get is wildly out of range at the end of the wave propagation. In

Did you change the number of integration points accordingly ? If you did
and it still deos not work, you can send me the files.

> the email you tell me how to add the 2nd order interpolation, you mentioned
> not to forget specifying the constraint on the 2nd order. Can you tell me how
> to do that? What I did is just to copy the constraint 'u' which means no
> constraint.

In your problem, you don't impose any strong constraint on your function
space (i.e. set some interpolation coefficients to certain values), so
the 'Constraint' field is useless. BTW, for the moment, you impose your
condition weakly (through an integration in the weak form of your
equations). Is it really what you want ?

Christophe

-- 
Christophe Geuzaine

Tel: 32 (0) 4 366 37 10    http://geuz.org
Fax: 32 (0) 4 366 29 10    mailto:Christophe.Geuzaine at ulg.ac.be