getDP question

[Christophe Geuzaine]

Uwe Pahner wrote:
> 
> Dear Christophe,
> 
> I am using getDP to solve a transient, coupled magneto-thermal problem.
> Maybe Francois Henrotte contacted you about this. Anyhow, he is on
> holidays, so I cannot ask him a question on getDP (on which I don't find an
> answer on the getDP homepage).
> 
> Attached you find the .pro files related to this model. I am running a
> transient analysis, where in each time step a coupled problem is solved
> (resolution 'TRA' in 1cond.pro). The model is a wire that is heated up
> using three consecutive induction heating devices. The numerical problem I
> am having is that using different time-schemes (Theta = 0.75 ... 1.0) and
> time-steps (DTime = 0.025 .. 0.1 s) results in very different solutions.
> The strange thing is that the divergence of the different solutions starts
> after the wire has passed the first oven. In other words, after a fixed
> time. I want to investigate the problem further, changing some of the
> parameters. But this would always require very long analysis runs, as I
> always need to compute the first oven again. I just want to solve once
> until time T1, and then restart with different TimeLoop settings from there
> onwards.
> 
> My question:
> 
> Is there a possibility to restart the transient loop from an already
> existing solution even when changing some of the TimeLoop settings? There
> is the "-restart" option when starting up getDP. But I was assuming that
> this only works if I don't change the TimeLoop settings. Do you have any
> suggestion?

Hello Uwe,

No, you should be able to use the '-restart' flag, even if the
parameters of the time scheme are changed. With '-restart', GetDP just
reads the last solution (for first order time schemes) and uses it as a
starting point for the upcoming computation. In other words, the
parameters used in the time stepping algorithm are those which are read
in the '.pro' file when the '-restart' is performed.

If it does not work, tell me on which platform you are working. I'll
send you a new version of the code so that we will be able to track down
the problem on a recent version (the handling of time has changed quite
a lot in the code since the last few months).

My best,

-- 
Christophe Geuzaine

Tel: +32-(0)4-366.37.10    mailto:Christophe.Geuzaine at ulg.ac.be
Fax: +32-(0)4-366.29.10    http://www.geuz.org